fix(ci): write browser coverage without each script's source (#432)

Every raw browser coverage file repeated the full source of each mapped
script, about 84% of its bytes, and the frontend coverage merge ran out of
Node's heap at 1,884 files. c8 reads the source from the mapped file on disk,
so the entries keep only the URL, script id and functions.

.github/workflows/viewerjs-check.yml

@@ -27,6 +27,15 @@ concurrency:
 # coverage shards 7m34s/6m04s/5m22s (tests 6m38s/4m53s/4m04s), merge 54s, firefox 7m57s/5m21s (tests
 # 7m08s/4m34s), webkit 8m16s/5m03s (tests 7m26s/3m58s); a cache miss adds up to 5m to npm ci, which
 # the 20-minute limits absorb.
+# Hangs (#425): a job killed at its limit keeps no log and skips its upload steps, so each browser
+# test step has its own timeout-minutes (18) below the job's 20. At the step limit, SIGTERM makes
+# tools/vendor/running-tests-reporter.mjs print the tests still running and hand over to Playwright's
+# interrupt, which writes the blob report; the always() uploads then run.
+# Measured on main (runs 40725289094, 40719375442, 40700439399): setup before the test step <= 52s,
+# uploads <= 35s; test steps up to 14m42s (webkit shard 1/2) and 13m05s for coverage (shard 1/3:
+# 12m53s, node units and the viewer suite before 11.8m of workflows), so coverage is the tight one
+# in its 18 minutes. This holds only when the npm cache hits: a cold cache adds up to 5m to npm ci,
+# and then the job limit can still end the job before the step limit does.
 jobs:
   frontend-coverage:
     name: frontend coverage (shard ${{ matrix.shard }}/3)
@@ -79,6 +88,7 @@ jobs:
       - if: matrix.shard == 3
         run: npm run --prefix tools/vendor test
       - name: Page workflows with coverage, shard ${{ matrix.shard }}/3
+        timeout-minutes: 18
         run: npm run --prefix tools/vendor test:coverage:shard -- ${{ matrix.shard }}/3
       - name: Upload raw coverage
         uses: actions/upload-artifact@v7
@@ -87,6 +97,7 @@ jobs:
           path: coverage/js/raw/
           if-no-files-found: error
           retention-days: 3
+      # Traces and the blob report (test-results/blob/) of a failed, timed-out or cancelled step.
       - name: Upload Chromium and touch traces
         if: always()
         uses: actions/upload-artifact@v7
@@ -211,7 +222,9 @@ jobs:
           key: npm-${{ runner.os }}-playwright-v1.64.1-${{ hashFiles('tools/vendor/package-lock.json') }}
       - run: npm ci --prefix tools/vendor --ignore-scripts
       - name: Page workflows in ${{ matrix.browser }}, shard ${{ matrix.shard }}/2
+        timeout-minutes: 18
         run: npm run --prefix tools/vendor test:workflows -- --project ${{ matrix.browser }} --shard ${{ matrix.shard }}/2
+      # Traces and the blob report (test-results/blob/) of a failed, timed-out or cancelled step.
       - name: Upload browser traces
         if: always()
         uses: actions/upload-artifact@v7

README.md

@@ -312,6 +312,12 @@ Both nodes support the same metadata fields:
 
 Both nodes will automatically save prompts to the database and link any generated images to the prompt.
 
+- **Save to library** (`save_to_library`, on by default; "Save to library" in the Focused, Split and Pair editors):
+  turn it off while iterating and PromptManager ignores that node's runs. The prompt is not saved, the run is not
+  counted and its images are not linked (a scan skips them too), while the node's output is unchanged. Workflows
+  saved before this switch existed load with it on. If any PromptManager node feeding a sampler's positive input
+  has it off, none of that run's images are linked, because the image cannot show which sampler produced it.
+
 ### 🖼️ Using the Output-Wide Gallery
 
 The comprehensive gallery system provides complete access to all your ComfyUI output:

TESTING.md

@@ -262,6 +262,31 @@ watchdog dump, is a failure like any other: read the dump, find the blocking
 call, fix it. Never raise a limit or rerun to get green; a limit changes only
 with new measurements written next to it.
 
+Browser jobs (#425): GitHub keeps no log of a job killed at its limit and skips
+its upload steps, so the test step in each browser job in `viewerjs-check.yml`
+has its own `timeout-minutes` (18, below the job's 20). At the step limit
+GitHub sends SIGTERM, which would end Playwright without any report; under `CI`
+`tools/vendor/running-tests-reporter.mjs` catches the first one, prints
+`Still running when the run ended (SIGTERM): N` and one line per unfinished
+test (path relative to the suite's `rootDir`), then sends Playwright its own
+SIGINT, so the workers stop, the run ends interrupted and the blob report is
+written to `tools/vendor/test-results/blob/`. In the coverage job npm signals
+only `frontend-coverage.mjs`, which forwards SIGTERM and SIGINT to the child it
+is running (started in its own process group, so a group-wide signal still
+arrives once). Delivery is proven locally in the pinned image for a SIGTERM to
+the whole process group (`timeout -s TERM`). A SIGTERM to npm alone
+(`timeout --foreground`) reaches nothing: npm forwards it to the `sh -c` that
+runs the script, and that shell exits without passing it on. Delivery in CI,
+where the step runs through `docker exec`, is not yet observed; the next real
+step timeout will show it. That directory and the traces
+upload with `if: always()`; `npx playwright merge-reports` opens the blob.
+Each test keeps the 30 s default and there is no `globalTimeout`. Two limits
+of this: it assumes the npm cache hits (a cold cache adds up to 5 minutes to
+`npm ci`, and the job limit can then still win), and coverage shard 1 (12m53s
+measured) is the tight one in its 18-minute step. Local runs keep the `list`
+reporter only; reproduce with `CI=1 timeout -s TERM 40 npm run --prefix
+tools/vendor test:workflows -- --project=firefox`.
+
 ### After a push
 
 Follow the run for the commit you pushed, never an older one:
@@ -391,6 +416,10 @@ A spec performance budget gets two tests:
   `test_filter_budget` (Images filters, spec D12: p95 < 150 ms at 50,000 files
   and 20,000 links).
 
+A hot SQLite read that returns many rows aggregates them into one step
+(`json_group_array`) instead of stepping once per row, so it does not wait for
+the GIL once per row while ComfyUI generates (#354, `get_image_link_matches`).
+
 The `perf budget` job in `.github/workflows/test.yml` is the reference
 measurement: `ubuntu-latest`, one Python, no coverage, only the wall-clock tests.
 It runs on every push to `main` and on `workflow_dispatch`, never on pull
@@ -659,7 +688,7 @@ The shared inline and collection inspector reads the current rooted file's suppo
 embedded records, exact generation facts and recorded or inferred prompt evidence.
 It renders in two concept surfaces (#249 D20–D22, D38). The Images side panel shows
 only the concept content: cover, title, filename, the evidence badge with a category
-chip, the + and − prompt blocks with copy and send, tag chips (the file's own image
+chip naming its origin (D146), the + and − prompt blocks with copy and send, tag chips (the file's own image
 tags first, D88, then the linked tags; image tags and the linked positive's own tags
 filter Images, D21), the six facts, and a footer with "Send to ComfyUI" and
 "Prompt history" only when prompt roles resolved or a link is saved; the sticky panel
@@ -1340,7 +1369,7 @@ How the earlier workspace Images filters moved (old → new). The legacy `/galle
 | Folder | "Refine this view" → Folder (D14); shown in the scope chip |
 | Media (All media / Images / Videos) | "Refine this view" → Media (D14) |
 | Sort: Newest first / Oldest first / Filename A–Z / Filename Z–A | toolbar sort: Recently added / Oldest first / Name A–Z / Name Z–A, plus Highest rated (D13, D14) |
-| (none) | toolbar category and "Refine this view" → Tags, through the linked positive (D13, D14) |
+| (none) | toolbar category (the effective category, D141) and "Refine this view" → Tags, through the linked positive (D13, D14) |
 | Refresh | Import → "Scan an output folder" → Rescan (D15) |
 | Grid / List / Masonry, Comfortable / Large / Compact | the same choices in the concept filter row |
 | Card image → side panel; preview icon → viewer | image → viewer; Info → side panel (D19) |
@@ -1370,22 +1399,18 @@ Recorded deviations on this page (spec §8):
   inspector's own "Download metadata JSON" envelopes in listing order); the button is
   `aria-disabled` and ignores a second press while it runs, the selection is kept, and
   the toast names the unread images (none read: no download); leaving Images aborts the
-  reads with no download and no toast. Set category opens the
-  concept's `simpleInput` dialog ("Categorize selected images", `input#action-input`
-  labelled Category, Cancel / Apply) and posts the selection's distinct linked positive
-  ids to `/prompt_manager/bulk/category` (images have no category; the filter matches
-  the linked positive's). With no linked image checked it posts nothing and opens no
-  dialog; an empty value keeps the dialog open; the field stops at the server's 100
-  characters; success reloads the categories and the listing together (clearing the
-  selection, focus to the heading, and, as any replaced collection does, the inspector
-  moves to the first listed image); failure keeps the selection. The write carries no
-  page signal: leaving Images while it runs lets it finish without a toast, and the return
-  loads the collection afresh instead of the saved visit. Three deviations from the
-  concept's `simpleInput`: the field starts empty (the concept pre-fills its demo
-  "Landscapes"), the action reads "Apply" (not "Apply in prototype"), and the
-  nothing-linked check runs on the Set category press, so no dialog opens. The route
-  rejects anything but 1-1000 positive integer ids within SQLite's INTEGER range
-  (deduplicated) and a category that fails `validate_category`. `images selection bar matches
+  reads with no download and no toast. Set category (amended by D140, #332) sets the
+  checked images' own categories: it opens the concept's `simpleInput` dialog
+  ("Categorize selected images", `input#action-input` labelled Category) and writes
+  `PUT`/`DELETE /prompt_manager/images/category`, never a prompt (see D140). Prompt
+  categories are changed in the Prompt library editor and on the legacy admin page's
+  "Set Category"; `POST /prompt_manager/bulk/category` is unchanged and serves that
+  legacy page only. The write carries no page signal: leaving Images while it runs lets
+  it finish without a toast, and the return loads the collection afresh instead of the
+  saved visit. Its success reloads the categories and the listing together (clearing
+  the selection, focus to the heading, and, as any replaced collection does, the
+  inspector moves to the first listed image); a failure keeps the selection, except a
+  changed source (409), which reloads the listing as a success does. `images selection bar matches
   the concept` compares the bar and the checked card with one card checked; the omitted
   actions flow again from the concept's own boxes, checked against the concept page with
   them removed at runtime.
@@ -1423,7 +1448,8 @@ Recorded deviations on this page (spec §8):
   `overscroll-behavior: contain`.
 - **D35**: counts pluralize ("1 positive prompt", "1 image", "1 more image loaded").
 - **D36**: the category list is the Library's, sorted alphabetically (case-insensitive)
-  by the page, whatever order the server sends.
+  by the page, whatever order the server sends (amended by D141: Images adds its
+  image categories).
 - **D37**: production loads 40 images per page (the concept 12), so the continuation
   and toast use 40; after Escape production returns focus to the opening image.
 - **D44** (#249 PR E): Filters reads "Filters · N" while N of Source, Folder, Media and
@@ -1451,6 +1477,9 @@ Recorded deviations on this page (spec §8):
   filters, while recipe, recorded-run and embedded tags stay plain `span.chip`.
   Amended by D88 (#249 PR C): the file's own current image tags lead the same row
   and are filters too; a linked tag already shown (case-insensitive) is left out.
+  Amended by D146 (#332): the category chip also names its origin in visible text
+  after the category (" · this image", " · prompt", " · recipe"), and an image's own
+  category shows for an unlinked or ambiguous file too.
 - **D22**: an empty positive reads "No positive prompt (explicitly empty)" (the concept
   applies its negative wording to both roles), and a confirmed `p.prompt-text` is
   focusable (`tabindex=0`, marked focus ring) so long prompts stay keyboard-readable.
@@ -1509,15 +1538,97 @@ Recorded deviations on this page (spec §8):
 - **D39** (ruled 2026-10-21: fallback): a recipe-linked image shows its recipe's own
   category when set, otherwise the linked positive prompt's, and its recipe's own tags
   when it has any, otherwise the positive's; each field falls back on its own and the
-  two tag sets are never merged. Set category (#259) writes the positive's category and
-  the Images category and Tags filters match the positive's, so the panel now agrees
-  with them whenever the recipe adds nothing. The inspection response names each
+  two tag sets are never merged. The Images Tags filter matches the positive's tags.
+  The category filter matches the same inherited category as the panel (D142, #332), and
+  an image's own category overrides both (D141, D146). The inspection response names each
   field's source (`library.category_scope`, `library.tags_scope`: `pair`, `positive` or
-  null); the category chip's visually hidden prefix (D21) says "Recipe category: " or
-  "Positive prompt category: " from it, and tag chips are Tags filter buttons only when
+  null, or `image` for an own category, #332); the category chip's visually hidden
+  prefix (D21) says "Recipe category: ", "Positive prompt category: " or "Image category: "
+  from it, and its visible origin (D146) " · recipe", " · prompt" or " · this image", and tag chips are Tags filter buttons only when
   they are the positive's (#260 D2). History's media inspector uses the same rule. The
   concept shows the prompt's category and tags for every image and the concept
   fixture's recipes have neither, so the inspector comparison needs no mapping for it.
+- **D141** (#332, amends D13 and D36): the Images category filter matches each image's
+  effective category: its own category (Images Set category, PR 3) when it has one,
+  else its inherited one (D142), else none. An image whose own category differs never
+  matches its prompt's category, and an unlinked or ambiguous image with an own
+  category is listed under it (the members-only gate now applies to the Tags filter
+  only). "Select all N matching" and saved views use the same value. The caption's
+  positive-prompt count still counts the linked positives of the images shown. With a
+  category set the listing reads the own categories once per source (a primary-key
+  range); a failed read is the existing 500 `listing-failed`, never an unfiltered page.
+  The link read then keeps only links whose raw recipe or positive category is the
+  value (a superset of those that inherit it; Python decides), and reads the links of
+  files whose own category is the value by path, so their rating and positive count.
+  `test_filter_budget` seeds 2,000 own categories on its 50,000-file fixture.
+  The option list adds the image categories of listed files
+  (`GET /prompt_manager/images/categories`, PR 3), merged with the Library's and sorted
+  (D36); if only that read fails, the prompt categories show and the categories alert is
+  raised. `images category options and Set category dialog match the concept` sets one
+  image's own category ("Portraits") through the real route and compares the toolbar
+  with it sorted into the concept's options; unmapped, only the options differ.
+- **D142** (#332 MC4, amends D39's filter note): the inherited category is one rule,
+  `inherited_category` in `utils/inspection_response.py`: a linked recipe's own
+  category when it is not blank, else the linked positive's. Before #332 the filter
+  matched the positive's even when the chip showed the recipe's. The link reads select
+  the raw recipe and positive categories and never decide inheritance in SQL
+  (`tests/test_api_workspace_image_categories.py` checks the filter and the chip agree
+  on blank values stored as spaces, tabs, U+00A0 and BLOB). The concept has no recipe
+  categories, so this is behavior only.
+- **D145** (#332, DC4): History's media inspector shows an image's own category with
+  the chip rule of the side panel and viewer. Only the inspection request reads it, once,
+  inside the History snapshot: the file is the symlink-resolved output folder plus the
+  receipt's subfolder and filename, keyed with `image_category_key` over the configured
+  output sources, as Images keys it. A temp output or a file no source holds has none;
+  a failed read hides the chip (DC5). The outputs list and media requests read nothing
+  new. The chip's hidden prefix reads "Image category: " for an own category; the
+  visible origin text and its comparison are D146 (PR 3). `history media inspector
+  category chip matches the concept` (history-concept.spec) sets an own category on the
+  Quality baseline's newest output, opens its Recorded run and "View current file", and
+  compares the viewer's chip with the concept viewer's category chip (app.js a402) laid
+  out at runtime with the D146 origin and that category; its place is not compared (what
+  precedes it differs), only its size, tokens, text, prefix and origin. Unmapped, the two
+  differ only in the text, the prefix, the width and `span.chip-source`.
+- **D140** (#332, amends D8): Set category sets the checked images' own categories, so
+  every checked file is a target (no nothing-linked check). More than 500 checked (or
+  "all N matching" over 500) toasts "Select at most 500 images." and opens nothing.
+  The dialog is the concept's `simpleInput` with three recorded differences: the field
+  starts empty (the concept pre-fills its demo "Landscapes") and suggests the toolbar's
+  current categories (`datalist#workspace-input-suggestions`, `openInputDialog`'s
+  `suggestions`); "Use prompt category", a plain button between Cancel and Apply
+  (`secondary`), closes it and opens the "Use prompt category" confirmation ("Remove the
+  image category from the 1 selected image? / … the N selected images? Each one then
+  shows its prompt's category, or none if it has no linked prompt. This cannot be
+  undone.", Cancel and a danger Remove, built like saved views' delete confirmation);
+  and the action reads "Apply" (D8). Apply sends `PUT /prompt_manager/images/category`
+  `{items | filter, category}`, Remove `DELETE` `{items | filter}`; Cancel sends
+  nothing. An empty value keeps the dialog open and the field stops at 100 characters.
+  Toasts: "Category “X” set on 1 image." / "… on N images." (unchanged files count as
+  set) plus " N could not be found." for unlisted files; "Removed the image category
+  from N images." or "None of the selected images had its own category."; 404 "None of
+  the selected images could be found. Refresh the images and try again."; 409 "The
+  sources changed. Refresh the images and try again." (the listing reloads); anything
+  else "Category could not be set. Try again.". `images category options and Set
+  category dialog match the concept` (images-concept.spec) checks a card, opens the
+  dialog in both documents and compares it (title, body tree, field, input, suggestions,
+  actions and every button's box and tokens) with the concept's own dialog changed at
+  runtime as recorded: field emptied, the datalist added after the field, a clone of
+  its Cancel reading "Use prompt category" after Cancel, the primary renamed "Apply".
+  Unmapped, the dialog differs only in the input's value and list, the body's tree, the
+  suggestions and the actions.
+- **D146** (#332, amends D21 and D39): the category chip names its origin in visible
+  text. It is `span.chip` > `span.ca-visually-hidden` (the D21 prefix) + the category +
+  `span.chip-source` (" · this image", " · prompt" or " · recipe"; muted
+  `--ca-muted`, `white-space: nowrap`, so at 390px the origin never splits; the chip may wrap
+  before it or inside a long category). It shows when the inspection names a non-empty category and either the
+  scope is `image` (any link status, so an unlinked or ambiguous file with an own
+  category shows it) or the file is linked. The concept's chip shows the bare name
+  (app.js a402). The inspector comparison (inspector-concept.spec) measures each chip's
+  `span.chip-source` (text, font, color, `white-space`, box) and lays the concept's own
+  chip out with the recorded rule and " · prompt" added at runtime (`withChipSource` in
+  `concept-inventory.mjs`), in the side panel and the viewer's details; unmapped they
+  differ only in the chip's text, width and `span.chip-source`. Its mutation check
+  removes the origin. History's chip (D145) is compared the same way.
 - **D42**: the concept never shows a toast and a page alert together. Production keeps
   the concept's fixed toast (centred, 22px from the bottom, 12px at 700px and below)
   and, when an Import → Rescan toast leaves the Images discovery alert
@@ -1542,7 +1653,8 @@ applies D3 (the `.file-inspector` class of the production-only rules), D20 (the
 viewer's plain tag chips, laid out by the concept page itself with its chips made
 plain at runtime), the D20/D21 side-panel scroll (`overflow: auto`,
 `max-height: calc(100dvh - 40px)` above 700px), D21 (the category's hidden scope
-prefix) and D22 (focusable prompt text). `inspector comparison detects mutations`
+prefix), D22 (focusable prompt text) and D146 (the chip's visible " · prompt", laid out
+by the concept page itself). `inspector comparison detects mutations`
 changes `.prompt-heading .role-label` letter spacing and removes `.detail-footer`.
 
 Manual check (quick), after a browser refresh on Images: hover an image (it zooms);
@@ -1877,11 +1989,13 @@ ranges include cards that bounded rendering has unmounted.
   Over the limit, the client toasts "Narrow the view to at most 10,000 images." and the
   server answers 400 `too-many-images`. No match is 400 `no-images`. Auto-tag in this
   mode does not offer Review each image, because the review walks images in the
-  browser. Set category posts `{images_filter, category}` to `/prompt_manager/bulk/category`.
-  The server sets the category on the distinct linked positives and answers
-  `prompt_count` and `skipped`. A listing whose `positive_count` is 0 toasts the
-  nothing-linked sentence and opens no dialog, and over 10,000 matches it toasts the
-  limit sentence. Export metadata reads every image in
+  browser. Set category (D140, #332) sends `{filter, category}` to
+  `PUT /prompt_manager/images/category` (Use prompt category `{filter}` to `DELETE`),
+  which sets or removes each matching image's own category; over 500 matches the client
+  toasts "Select at most 500 images." and opens no dialog, and the server answers 400
+  `too-many-images`. (`POST /prompt_manager/bulk/category` still accepts
+  `{images_filter, category}` and sets the linked positives' category, but no workspace
+  control sends it any more.) Export metadata reads every image in
   the browser, so it is `aria-disabled` in this mode, and its `title` says why.
 
 Tests: `tests/browser/workspace-images.spec.mjs` ("Shift+click selects a range across
@@ -4624,8 +4738,18 @@ Filters (`mode=workspace`; anything invalid is 400 "Invalid gallery filters"):
   rating 5 to 1, then linked but unrated, then unlinked and ambiguous; ties break
   newest first. A signed cursor covers every filter and the sort, so reusing it
   with a change is 400.
-- One link read (`get_image_link_matches`, streamed and bounded to the listed
-  paths) serves the filters, the rating sort and the count.
+- One link read (`get_image_link_matches`) serves the filters, the rating sort
+  and the count. SQLite returns its rows as one JSON value in one step (#354):
+  Python's `sqlite3` releases and retakes the GIL around every step, and while
+  ComfyUI generates each retake can wait on the busy thread, so a read that
+  stepped once per linked row took seconds. The result keeps only the listed
+  paths; peak memory grows with the database's matching linked rows (about
+  530 bytes each, freed on return). When the JSON step fails (a SQLite without
+  JSON functions, a BLOB value, a JSON text over SQLite's length limit) the read
+  streams the rows as before and logs one warning per process; a missing JSON
+  function stops the attempt for the rest of the process. An interrupted read
+  (Move library) raises as before. `tests/test_database_link_matches.py` proves both reads
+  equal on every filter and on odd values.
 
 `positive_count` is the number of distinct linked positives over the full matching
 set, not the page:
@@ -4739,6 +4863,21 @@ category; without one it shows the category its link supplies.
   appears at the same path shows it. The real-host check is the last test in
   `tests/comfyui/image-attribution.spec.mjs`: PUT, list, DELETE on a generated
   output.
+- Interface (PR 3, a browser refresh; no restart beyond PR 1-2's): the Images
+  selection bar's Set category writes these routes (D140, which amends D8); the toolbar
+  category options add the listed images' own categories (D141); the inspector's
+  category chip names its origin, " · this image", " · prompt" or " · recipe" (D146), in
+  the side panel, the viewer and History's media inspector (D145); Tags & categories
+  counts images on each category row and lists image-only values (D143), and Manage
+  category previews and renames image rows too (D144). Prompt categories are changed in
+  the Prompt library editor and on the legacy admin page's "Set Category". Concept
+  comparisons, at 1600x1100 and 390x844 in both themes: `images-concept.spec.mjs` (the
+  Set category dialog and the options), `inspector-concept.spec.mjs` (the chip),
+  `history-concept.spec.mjs` (History's chip) and `vocabulary-visual.spec.mjs` (rows and
+  preview, with `vocabulary-canonical-comparison.mjs`). Behavior: `workspace-images`,
+  `full-inspector`, `workspace-history` and `workspace-vocabulary` specs, and
+  `tests/js/workspace-images-category.test.js`, `workspace-file-inspector.test.js`,
+  `workspace-vocabulary-state.test.js`, `workspace-toolbar.test.js`.
 
 Image annotations schema (#249, `database/annotation_schema.py`): the group
 `media_annotation_metadata` (one row: schema version, `image_index_revision`),
@@ -5269,7 +5408,7 @@ which used to save `null` under a link.
   Primitive or through `PreviewAny`, for PromptManager, PromptManagerText, Focused, Split
   and the Pair (positive, negative, both, neither linked, and a Primitive negative). A
   wildcard Primitive on PromptManager (queued as written) and Focused (one choice) saves
-  and counts what was queued. The linked text is saved, counted once, recorded as
+  and counts what was queued. The linked text is saved, counted once per job, recorded as
   the run's raw positive/negative, and after a scan the run's image links to it. Text
   only known at run time (`PreviewAny`) cannot be read from the image; the case only
   requires that it is never linked to another prompt (its attribution from the recorded
@@ -5283,6 +5422,39 @@ PM_COMFY_ROOT=... PM_TEST_PYTHON=... npm run --prefix tools/vendor test:comfyui
 PM_COMFY_ROOT=... PM_TEST_PYTHON=... npm run --prefix tools/vendor test:executions -- execution-text-link
 ```
 
+## Cached runtime-linked runs (#347)
+
+Text only known at run time (a `PreviewAny` or LLM link, a list node) is counted by the
+node when it executes; the queue hook cannot read it. When ComfyUI serves such a node from
+its output cache, `utils/cached_runs.py` counts the job instead: every execution records
+a receipt (digests of its STRING outputs and hashes of the text it stores, per run key and
+queue item), and on `execution_cached` the receipt whose digests equal the cache entry's
+text is replayed once. The role (positive, negative, which Pair outputs are used) and what
+the hook already saw are recomputed from the job's own graph, and the node and replay
+decide through the same `own_counts`.
+
+- Fail closed: no receipt in this process (after a restart, an external cache provider),
+  changed cached text, more than 64 list values, a node the job's outputs do not need, or
+  a host without the queue/executor interfaces: nothing is counted, as before #347.
+- A cached node counts at job start like the hook: a job failing later still counts it.
+  A node that executes after its replay in the same job (RAM-pressure freed the entry)
+  skips its own count but still consumes the pending first-use catch-up.
+- No backfill: earlier cached jobs keep their counts. The history page's "tracked runs"
+  can stay below the Library count for these prompts (#403).
+- Tests: `tests/test_cached_runs.py` (whole jobs with the real nodes, a real database,
+  fake host caches); `tests/comfyui/execution-text-link.spec.mjs` "#347" cases on the
+  managed host: PreviewAny into each node kind, 3 jobs changing only the sampler seed,
+  count 3 in both renderers; in one renderer the Classic, LRU, RAM-pressure and no-cache
+  modes, A, B, A from the owned `PMCycledText` under LRU, typed text, a downstream
+  failure, a "run selected output" job that does not need the node, and a rewire to the
+  negative input. The RAM-pressure eviction path cannot be forced on a real host and is
+  covered by the Python tests.
+
+```sh
+python3 -m unittest discover -s tests -p "test_cached_runs.py"
+PM_COMFY_ROOT=... PM_TEST_PYTHON=... npm run --prefix tools/vendor test:executions -- execution-text-link --grep "#347"
+```
+
 ## Durable execution ledger
 
 `npm run --prefix tools/vendor test:executions` runs 52 real execution cases
@@ -5702,6 +5874,20 @@ rotates the epoch, so a pre-restore `expected_epoch` is refused with `storage-ch
 Owner rows and their timestamps never change on a tag command; a category apply only
 touches rows carrying the exact source.
 
+Image categories (#332 §7, MC9, MC10, DC7): each `GET categories` row also carries
+`image_count`, the stored `image_categories` rows holding that exact value (files
+missing on disk included). A value prompts and images share appears once, ordered by
+its oldest positive; a value only images hold appears once after every prompt-backed
+value (`first_pid` 9223372036854775807 in the keyset cursor), ordered by value, with
+`prompt_count` 0. `/library/options` keeps its own query and never lists image-only
+values. Category preview and apply report `impact.images` (rows holding the source),
+the target exists when an image row holds it, an image-only value can be renamed, the
+apply renames the image rows in the same transaction (file keys and rowids kept, so a
+merge cannot collide), and the impact token hashes the sorted image rowids, so an image
+row gaining or losing the value between preview and apply is 409
+`vocabulary-conflict` with nothing written. Every image-category statement reads
+`idx_image_categories_category` (`test_category_reads_use_the_image_category_index`).
+
 Two legacy routes change additively: `GET /tags` now lists unused registry tags, and
 `/tags/suggest` without context appends unused names at count 0. Unused tags become
 deletable only through the native Delete tag (vocabulary UI PR), not on the legacy
@@ -5719,13 +5905,18 @@ Deploying this backend needs a ComfyUI restart (Python and route changes).
 longer hands off to the legacy tag manager. That page stays reachable from Overview's
 "Open existing tag manager" card for tag search and sort, the untagged entry and
 multi-tag AND/OR browsing; bulk prompt actions (delete, tag, category) stay on the
-legacy admin page. The screen consumes only the `/prompt_manager/vocabulary/*` routes
+legacy admin page. A prompt's category is changed in the Prompt library editor or with
+the legacy admin page's "Set Category" (`POST /prompt_manager/bulk/category`, unchanged);
+the Images selection bar's Set category sets images' own categories only (D140, #332),
+and Manage category renames a value on prompts, recipes and images together (D144). The screen consumes only the `/prompt_manager/vocabulary/*` routes
 above and adds no route, so PR B needs no ComfyUI restart: a browser refresh is enough.
 
 Screen: a `+ Create tag` heading action, a Tag vocabulary table (columns Tag, Prompts,
 Images counting the files that carry the tag (`image_count`, #249 C2), and Actions with Browse and
 Rename / merge), and a Categories panel whose category rows end with the Manage
-category row (Review). Lists load 100 rows at a time, append when scrolled into view
+category row (Review). A category row reads "N positive prompts", "N positive prompts ·
+M images" when images hold the value as their own category too, or "M images" for a
+value only images hold (D143, #332). Lists load 100 rows at a time, append when scrolled into view
 and keep `Load more tags` / `Load more categories` while more remain. Prompts counts
 are the Library Browse totals.
 
@@ -5768,7 +5959,34 @@ Recorded deviations from the concept (maintainer acceptance): bounded lists with
 Load more; title "Tags"; Manage categories adds a Category select and Preview before
 Apply; "Apply" instead of "Apply in prototype"; Delete tag… inside Rename or merge;
 Browse sets the exact filter; singular "1 positive prompt"; the production list appends
-fixture-only tags after the concept's sample tags. Paired concept/production captures
+fixture-only tags after the concept's sample tags.
+
+- **D143** (#332): category rows count the images whose own category is the value
+  (`image_count`, the stored rows, missing files included): "N positive prompts · M
+  images" beside prompts, "M images" for an image-only value, which follows every
+  prompt-backed row (by value) before Manage category. Browse on an image-only row opens
+  Images with exactly that Category filter (its Library would be empty), as a fresh
+  Images visit that becomes the Images return state, the way Browse seeds the Library's; every other
+  row opens the Library as before. Manage category's select lists image-only values
+  last, so past 1,000 categories they are the first left out (the dialog says the list is
+  capped).
+- **D144** (#332): Manage category's preview counts the images holding the value ("“X”
+  is used by N positive prompts and M images. Apply renames it to “Y”."; D9's sentence
+  with "M images" added when above 0), an image-only value can be renamed, and Apply
+  renames the image rows in the same transaction; an image row gaining or losing the
+  value between Preview and Apply is 409 `vocabulary-conflict` with nothing written.
+- `category rows and the Manage category preview match the concept with image
+  categories` (vocabulary-visual.spec, 1600x1100 and 390x844, both themes) seeds the
+  fixture with `image_categories: true` ("Landscapes" on two images, "Portraits" on one
+  image only, rows keyed as Images keys them) and compares the page with D143 mapped
+  (`mapVocabularyConcept(reference, imageCategories)`, whose row model is checked against
+  the concept page with those rows added at runtime) and the Manage dialog after
+  Preview with D9 and D144 (`mapVocabularyDialog`'s `preview`). Before the preview is
+  measured the pointer leaves the pressed Apply and focus returns to ×, the state the
+  opening comparison in workspace-vocabulary.spec measures. Unmapped, the page differs
+  in the Landscapes row's text and the added Portraits row, the dialog only in the
+  sentence. The default fixture seeds no image category, so the other vocabulary
+  comparisons are unchanged. Paired concept/production captures
 for both themes and widths are attached as test artifacts; no PNG baseline is added
 until acceptance. Not compared: hover states and the dialog focus ring (the page-level
 keyboard ring and link ring offsets are asserted). No coverage floor was lowered;

__init__.py

@@ -195,6 +195,26 @@ except Exception as e:
             f"Failed to register usage hook: {e}"
         )
 
+# A cached node counts what its last execution of the same text counted (#347).
+try:
+    from .py.config import server_instance
+    from .utils.cached_runs import register_cached_runs
+
+    register_cached_runs(server_instance, _db_factory)
+except Exception as e:
+    try:
+        from .utils.logging_config import get_logger
+
+        get_logger("prompt_manager.init").error(
+            f"Failed to register cached-run counting: {e}"
+        )
+    except Exception:
+        import logging
+
+        logging.getLogger("prompt_manager.init").error(
+            f"Failed to register cached-run counting: {e}"
+        )
+
 
 def _history_db_factory() -> "PromptDatabase":
     """Initialize history on its worker even if monitoring initialization failed."""

database/duplicate_copies.py

@@ -9,6 +9,8 @@ IMMEDIATE`` transaction, so no row ever names a deleted file:
   already links the kept copy: then its saved-pair links join that record
   and it is deleted;
 - its favorite;
+- its collection memberships (#331), joined onto the kept copy's as their
+  union under the same keys a favorite moves by;
 - its image tags (``media_image_tags`` of its byte record), joined onto the
   kept copy's byte record as their union, in order;
 - its own image category (#332), copied to the kept copy when that copy has
@@ -54,6 +56,15 @@ INSERT_FAVORITE_SQL = "INSERT OR IGNORE INTO image_favorites VALUES (?, ?, ?)"
 DELETE_FAVORITE_SQL = (
     "DELETE FROM image_favorites WHERE root_key = ? AND relative_path = ?"
 )
+COPY_MEMBERSHIPS_SQL = (
+    "INSERT OR IGNORE INTO image_collection_items "
+    "(collection_id, root_key, relative_path, added_at) "
+    "SELECT collection_id, ?, ?, added_at FROM image_collection_items "
+    "WHERE root_key = ? AND relative_path = ?"
+)
+DELETE_MEMBERSHIPS_SQL = (
+    "DELETE FROM image_collection_items WHERE root_key = ? AND relative_path = ?"
+)
 COPY_BYTE_SQL = (
     f"SELECT b.id FROM {FILES_TABLE} f JOIN {BYTES_TABLE} b "  # nosec B608
     "ON b.file_id=f.id WHERE f.locator=? AND b.sha256=?"
@@ -77,6 +88,8 @@ FILE_LINK_SQL = (
     DELETE_LINK_SQL,
     FAVORITE_SQL,
     DELETE_FAVORITE_SQL,
+    COPY_MEMBERSHIPS_SQL,
+    DELETE_MEMBERSHIPS_SQL,
     COPY_BYTE_SQL,
     COPY_TAG_COUNT_SQL,
     PENDING_COUNT_SQL,
@@ -140,14 +153,17 @@ class DuplicateCopyOperations:
         by), ``targets`` (``{removed spelling: kept spelling}``, the spelling
         a moved record takes; else ``kept_path``), ``favorites`` (``(root key,
         removed key, kept root key, kept key)`` per listed root holding the
-        file), ``removed_locator``,
-        ``kept_locator``, ``sha256`` and the kept copy's ``kept_size`` and
-        ``kept_mtime_ns``; ``categories``: ``(removed file key, kept file
-        key)`` pairs of the image categories (#332). Returns ``{moved,
-        merged, favorites, tags, review, categories}`` (``categories``: the
-        image categories copied to the kept copy).
+        file; they move its favorite and its collection memberships),
+        ``removed_locator``, ``kept_locator``, ``sha256`` and the kept copy's
+        ``kept_size`` and ``kept_mtime_ns``; ``categories``: ``(removed file
+        key, kept file key)`` pairs of the image categories (#332). Returns
+        ``{moved, merged, favorites, tags, review, collections, categories}``
+        (``collections``: the memberships inserted on the kept copy;
+        ``categories``: the image categories copied to the kept copy).
         """
-        counts = dict.fromkeys(("moved", "merged", "favorites", "tags", "review"), 0)
+        counts = dict.fromkeys(
+            ("moved", "merged", "favorites", "tags", "review", "collections"), 0
+        )
         kept_keys = [normalize_image_path(path) for path in move["kept"]]
         with self._pair_transaction() as conn:
             for spelled in dict.fromkeys(move["removed"]):
@@ -180,6 +196,7 @@ class DuplicateCopyOperations:
                     counts["moved"] += 1
             for favorite in move.get("favorites", ()):
                 counts["favorites"] += _move_favorite(conn, *favorite)
+                counts["collections"] += _move_memberships(conn, *favorite)
             counts["tags"] = self._move_copy_tags(conn, move)
             counts["categories"] = sum(
                 copy_image_category(conn, removed, kept)
@@ -224,3 +241,12 @@ def _move_favorite(conn, root_key, removed_key, kept_root, kept_key):
     conn.execute(INSERT_FAVORITE_SQL, (kept_root, kept_key, row[0]))
     conn.execute(DELETE_FAVORITE_SQL, (root_key, removed_key))
     return 1
+
+
+def _move_memberships(conn, root_key, removed_key, kept_root, kept_key):
+    """Copy a removed copy's memberships to the kept copy; the rows inserted."""
+    inserted = conn.execute(
+        COPY_MEMBERSHIPS_SQL, (kept_root, kept_key, root_key, removed_key)
+    ).rowcount
+    conn.execute(DELETE_MEMBERSHIPS_SQL, (root_key, removed_key))
+    return inserted

database/image_collections.py

@@ -0,0 +1,490 @@
+"""Collections (#331): named, hand-picked groups of image files.
+
+A member names its file exactly as a favorite does (``image_favorites.py``):
+the source folder's real, case-normalized path (``root_key``) and the path
+below it (``relative_path``), both server-side only. A member whose file is
+moved, renamed or deleted outside PromptManager is kept and simply not listed;
+nothing follows a file to a new place.
+
+Names are trimmed with each run of whitespace collapsed to one space, 1 to 100
+characters without control characters; ``name_key`` (NFC of the case-folded
+NFC name) is UNIQUE, so names are unique ignoring case and Unicode form.
+
+The two tables install like ``image_favorites`` and ``saved_views``: optional
+in old backups, required once a database is current, created empty with no
+backfill. Startup checks only the structure, so a value saved under older
+rules never stops the library from opening; restore checks the values too,
+but never recomputes ``name_key`` (case tables differ between Python
+versions).
+"""
+
+import re
+import sqlite3
+import unicodedata
+from typing import Optional
+
+from .image_favorites import MAX_KEY_LENGTH
+
+COLLECTIONS_TABLE = "image_collections"
+ITEMS_TABLE = "image_collection_items"
+FILE_INDEX = "idx_image_collection_items_file"
+TABLES = (COLLECTIONS_TABLE, ITEMS_TABLE)
+MAX_NAME_LENGTH = 100
+_TIMESTAMP_BYTES = 64
+# A name key's byte bound: case folding expands a character to at most
+# three, each at most 4 UTF-8 bytes.
+_MAX_KEY_BYTES = MAX_NAME_LENGTH * 3 * 4
+_WHITESPACE = re.compile(r"\s+")
+
+_TABLE_DDL = {
+    COLLECTIONS_TABLE: (
+        f"CREATE TABLE {COLLECTIONS_TABLE} (id INTEGER PRIMARY KEY AUTOINCREMENT, "
+        "name TEXT NOT NULL, name_key TEXT NOT NULL UNIQUE, "
+        "created_at TEXT NOT NULL)"
+    ),
+    ITEMS_TABLE: (
+        f"CREATE TABLE {ITEMS_TABLE} (collection_id INTEGER NOT NULL REFERENCES "
+        f"{COLLECTIONS_TABLE}(id) ON DELETE CASCADE, root_key TEXT NOT NULL, "
+        "relative_path TEXT NOT NULL, added_at TEXT NOT NULL, "
+        "PRIMARY KEY (collection_id, root_key, relative_path))"
+    ),
+}
+_INDEX_DDL = f"CREATE INDEX {FILE_INDEX} ON {ITEMS_TABLE}(root_key, relative_path)"
+# (name, declared type, not null, default, primary key position)
+_COLUMNS = {
+    COLLECTIONS_TABLE: (
+        ("id", "INTEGER", 0, None, 1),
+        ("name", "TEXT", 1, None, 0),
+        ("name_key", "TEXT", 1, None, 0),
+        ("created_at", "TEXT", 1, None, 0),
+    ),
+    ITEMS_TABLE: (
+        ("collection_id", "INTEGER", 1, None, 1),
+        ("root_key", "TEXT", 1, None, 2),
+        ("relative_path", "TEXT", 1, None, 3),
+        ("added_at", "TEXT", 1, None, 0),
+    ),
+}
+# (column, parent table, parent column, on delete)
+_FOREIGN_KEYS = {
+    COLLECTIONS_TABLE: set(),
+    ITEMS_TABLE: {("collection_id", COLLECTIONS_TABLE, "id", "CASCADE")},
+}
+
+
+class CollectionInvalid(ValueError):
+    """A name, id or target breaks the write rules (a 400)."""
+
+
+def _has_control(text: str) -> bool:
+    """Any Unicode control character (category Cc: C0, DEL and C1)."""
+    return any(unicodedata.category(char) == "Cc" for char in text)
+
+
+def normalize_collection_name(name: object) -> str:
+    """The stored spelling of *name*: trimmed, whitespace collapsed, checked.
+
+    Raises ``CollectionInvalid`` unless the result is 1 to 100 characters
+    without control characters.
+    """
+    if not isinstance(name, str):
+        raise CollectionInvalid("Name must be a string")
+    collapsed = _WHITESPACE.sub(" ", name).strip()
+    if not 1 <= len(collapsed) <= MAX_NAME_LENGTH:
+        raise CollectionInvalid(f"Name must be 1 to {MAX_NAME_LENGTH} characters")
+    if _has_control(collapsed):
+        raise CollectionInvalid("Name contains control characters")
+    return collapsed
+
+
+def collection_name_key(name: str) -> str:
+    """The uniqueness key of a name: NFC of the case-folded NFC spelling."""
+    return unicodedata.normalize("NFC", unicodedata.normalize("NFC", name).casefold())
+
+
+def _compact(sql: str) -> str:
+    return re.sub(r"\s+", "", sql).lower()
+
+
+def _expected_catalog() -> set:
+    objects = {("table", name, _compact(sql)) for name, sql in _TABLE_DDL.items()}
+    objects.add(("index", FILE_INDEX, _compact(_INDEX_DDL)))
+    objects |= {("index", f"sqlite_autoindex_{table}_1", "") for table in TABLES}
+    return objects
+
+
+def _catalog(conn: sqlite3.Connection) -> list:
+    # sqlite_sequence (AUTOINCREMENT) is shared with other tables and ignored.
+    return conn.execute(
+        "SELECT type,name,sql FROM sqlite_master "
+        "WHERE name IN (?,?,?) OR tbl_name IN (?,?)",
+        (*TABLES, FILE_INDEX, *TABLES),
+    ).fetchall()
+
+
+def _structure_error(conn: sqlite3.Connection) -> Optional[str]:
+    for table, expected in _COLUMNS.items():
+        columns = tuple(
+            (row[1], row[2].upper(), row[3], row[4], row[5])
+            for row in conn.execute(f"PRAGMA table_info({table})")
+        )
+        if columns != expected:
+            return f"Image collections table {table} has invalid columns"
+        foreign_keys = {
+            (row[3], row[2], row[4], row[6])
+            for row in conn.execute(f"PRAGMA foreign_key_list({table})")
+        }
+        if foreign_keys != _FOREIGN_KEYS[table]:
+            return f"Image collections table {table} has invalid foreign keys"
+    return None
+
+
+def _decoded_too_long(conn, sql: str, limit: int) -> bool:
+    """Whether a byte-bounded candidate row decodes to over *limit* characters."""
+    for row in conn.execute(sql):
+        for raw in row:
+            if len(bytes(raw).decode("utf-8", "replace")) > limit:
+                return True
+    return False
+
+
+def _value_error(conn: sqlite3.Connection) -> Optional[str]:
+    # Only this module's constants are interpolated into the SQL below. Bytes
+    # bound characters from above (UTF-8 is at most 4 bytes each); the few
+    # candidates over the character bound are measured decoded in Python.
+    if conn.execute(
+        f"SELECT 1 FROM {COLLECTIONS_TABLE} WHERE "  # nosec B608
+        "typeof(id) != 'integer' OR typeof(name) != 'text' "
+        "OR typeof(name_key) != 'text' OR typeof(created_at) != 'text' "
+        f"OR length(CAST(name AS BLOB)) > {MAX_NAME_LENGTH * 4} "
+        "OR length(CAST(name_key AS BLOB)) = 0 "
+        f"OR length(CAST(name_key AS BLOB)) > {_MAX_KEY_BYTES} "
+        f"OR length(CAST(created_at AS BLOB)) > {_TIMESTAMP_BYTES} LIMIT 1"
+    ).fetchone():
+        return "Image collections contains invalid values"
+    # The key is not recomputed: Unicode case tables differ between Python
+    # versions, and a key written by another version must stay restorable.
+    # Its UNIQUE constraint keeps names distinct; writes compute it.
+    for (name,) in conn.execute(f"SELECT name FROM {COLLECTIONS_TABLE}"):  # nosec B608
+        if (
+            not 1 <= len(name) <= MAX_NAME_LENGTH
+            or _has_control(name)
+            or name != name.strip()
+        ):
+            return "Image collections contains invalid names"
+    limit = MAX_KEY_LENGTH * 4
+    if conn.execute(
+        f"SELECT 1 FROM {ITEMS_TABLE} WHERE "  # nosec B608
+        "typeof(collection_id) != 'integer' OR typeof(root_key) != 'text' "
+        "OR typeof(relative_path) != 'text' OR typeof(added_at) != 'text' "
+        f"OR length(CAST(root_key AS BLOB)) > {limit} "
+        f"OR length(CAST(relative_path AS BLOB)) > {limit} "
+        f"OR length(CAST(added_at AS BLOB)) > {_TIMESTAMP_BYTES} LIMIT 1"
+    ).fetchone() or _decoded_too_long(
+        conn,
+        f"SELECT CAST(root_key AS BLOB), CAST(relative_path AS BLOB) "  # nosec B608
+        f"FROM {ITEMS_TABLE} WHERE length(CAST(root_key AS BLOB)) > {MAX_KEY_LENGTH} "
+        f"OR length(CAST(relative_path AS BLOB)) > {MAX_KEY_LENGTH}",
+        MAX_KEY_LENGTH,
+    ):
+        return "Image collections contains invalid members"
+    if conn.execute(f"PRAGMA foreign_key_check({ITEMS_TABLE})").fetchone():
+        return "Image collections contains members of a missing collection"
+    return None
+
+
+def image_collections_schema_error(
+    conn: sqlite3.Connection,
+    *,
+    allow_missing: bool = True,
+    check_values: bool = True,
+) -> Optional[str]:
+    """Reject partial or unsupported collections schemas and invalid rows.
+
+    Reads only: nothing is ever repaired. A database without either table is
+    acceptable for backups that predate them unless ``allow_missing`` is
+    false. ``check_values=False`` (live startup) checks the structure only.
+    """
+    objects = _catalog(conn)
+    if not objects:
+        return None if allow_missing else "Missing image collections schema"
+    present = {name for kind, name, _sql in objects if kind == "table"}
+    if present != set(TABLES):
+        return "Incomplete image collections schema"
+    actual = {(kind, name, _compact(sql or "")) for kind, name, sql in objects}
+    if actual != _expected_catalog():
+        return "Image collections has an invalid schema definition"
+    error = _structure_error(conn)
+    if error or not check_values:
+        return error
+    return _value_error(conn)
+
+
+def migrate_image_collections(conn: sqlite3.Connection) -> None:
+    """Install both tables atomically; validate before and after creating them."""
+    conn.execute("SAVEPOINT image_collections_schema")
+    try:
+        error = image_collections_schema_error(conn, check_values=False)
+        if error:
+            raise sqlite3.DatabaseError(error)
+        for sql in _TABLE_DDL.values():
+            conn.execute(sql.replace("CREATE TABLE", "CREATE TABLE IF NOT EXISTS", 1))
+        conn.execute(_INDEX_DDL.replace("INDEX", "INDEX IF NOT EXISTS", 1))
+        error = image_collections_schema_error(
+            conn, allow_missing=False, check_values=False
+        )
+        if error:
+            raise sqlite3.DatabaseError(error)
+        conn.execute("RELEASE image_collections_schema")
+    except BaseException:
+        conn.execute("ROLLBACK TO image_collections_schema")
+        conn.execute("RELEASE image_collections_schema")
+        raise
+
+
+class CollectionMissing(LookupError):
+    """No collection has the id (a 404 ``collection-not-found``)."""
+
+
+class CollectionNameTaken(ValueError):
+    """Another collection already has the name's key (a 409 ``name-taken``)."""
+
+    def __init__(self, collection_id: int):
+        super().__init__("Another collection has this name")
+        self.collection_id = collection_id
+
+
+class CollectionFilesMissing(LookupError):
+    """A new name with nothing to insert (a 404 ``files-not-found``)."""
+
+
+LIST_SQL = f"SELECT id, name FROM {COLLECTIONS_TABLE} ORDER BY name_key, id"
+GET_SQL = f"SELECT id, name FROM {COLLECTIONS_TABLE} WHERE id = ?"
+BY_KEY_SQL = f"SELECT id, name FROM {COLLECTIONS_TABLE} WHERE name_key = ?"
+MEMBERS_SQL = (
+    f"SELECT relative_path FROM {ITEMS_TABLE} "
+    "WHERE collection_id = ? AND root_key = ?"
+)
+ROOT_MEMBERS_SQL = (
+    f"SELECT collection_id, relative_path FROM {ITEMS_TABLE} WHERE root_key = ?"
+)
+COUNTS_SQL = (
+    f"SELECT c.id, (SELECT COUNT(*) FROM {ITEMS_TABLE} i "
+    f"WHERE i.collection_id = c.id) FROM {COLLECTIONS_TABLE} c"
+)
+FILE_SQL = (
+    f"SELECT collection_id FROM {ITEMS_TABLE} "
+    "WHERE root_key = ? AND relative_path = ? ORDER BY collection_id"
+)
+# Every read, for the EXPLAIN QUERY PLAN check: none scans the items table and
+# only the two whole-list reads (LIST_SQL, COUNTS_SQL) scan the collections.
+READ_SQL = (
+    LIST_SQL,
+    GET_SQL,
+    BY_KEY_SQL,
+    MEMBERS_SQL,
+    ROOT_MEMBERS_SQL,
+    COUNTS_SQL,
+    FILE_SQL,
+)
+_INSERT_ITEM_SQL = f"INSERT OR IGNORE INTO {ITEMS_TABLE} VALUES (?, ?, ?, ?)"
+_DELETE_ITEM_SQL = (
+    f"DELETE FROM {ITEMS_TABLE} "
+    "WHERE collection_id = ? AND root_key = ? AND relative_path = ?"
+)
+
+
+def _collection_id(value: object) -> int:
+    if type(value) is not int or not 1 <= value <= 999_999_999:
+        raise CollectionInvalid("Collection id must be a positive integer")
+    return value
+
+
+def _items(items: object) -> list:
+    """Distinct ``(root key, relative key)`` pairs, in order."""
+    if not isinstance(items, (list, tuple)):
+        raise CollectionInvalid("Items must be a list")
+    found = []
+    for item in items:
+        if (
+            not isinstance(item, (list, tuple))
+            or len(item) != 2
+            or not all(
+                isinstance(part, str) and 1 <= len(part) <= MAX_KEY_LENGTH
+                for part in item
+            )
+        ):
+            raise CollectionInvalid("Each item is a root key and a relative key")
+        found.append((item[0], item[1]))
+    return list(dict.fromkeys(found))
+
+
+def _row(row) -> dict:
+    return {"id": row[0], "name": row[1]}
+
+
+class ImageCollectionOperations:
+    """PromptDatabase mixin: collections and their members.
+
+    Each write runs in its own ``BEGIN IMMEDIATE`` transaction
+    (``_pair_transaction``), so two tabs creating one name are serialized.
+    """
+
+    def list_collections(self) -> list[dict]:
+        """Every collection ``{id, name}`` in name-key order, then id."""
+        rows = self.model.get_connection().execute(LIST_SQL)
+        return [_row(row) for row in rows]
+
+    def get_collection(self, collection_id: int) -> Optional[dict]:
+        """``{id, name}`` of one collection, or None."""
+        row = (
+            self.model.get_connection()
+            .execute(GET_SQL, (_collection_id(collection_id),))
+            .fetchone()
+        )
+        return _row(row) if row else None
+
+    def collection_exists(self, collection_id: int) -> bool:
+        return self.get_collection(collection_id) is not None
+
+    def collection_members(self, collection_id: int, root_key: str) -> set[str]:
+        """The relative keys stored for one collection under one source."""
+        rows = self.model.get_connection().execute(
+            MEMBERS_SQL, (_collection_id(collection_id), root_key)
+        )
+        return {path for (path,) in rows}
+
+    def root_collection_members(self, root_key: str) -> dict[int, set[str]]:
+        """``{collection id: relative keys}`` stored under one source."""
+        found: dict[int, set[str]] = {}
+        for collection_id, path in self.model.get_connection().execute(
+            ROOT_MEMBERS_SQL, (root_key,)
+        ):
+            found.setdefault(collection_id, set()).add(path)
+        return found
+
+    def collection_count_snapshot(self, root_keys: list) -> tuple[dict, dict]:
+        """``({root key: {id: relative keys}}, {id: stored rows})`` read in
+        one read transaction, so a write between the reads cannot skew them.
+        """
+        conn = self.model.get_connection()
+        if conn.in_transaction:
+            raise RuntimeError("Collection counts require their own transaction")
+        conn.execute("BEGIN")
+        try:
+            members = {}
+            for root_key in dict.fromkeys(root_keys):
+                found: dict[int, set[str]] = {}
+                for collection_id, path in conn.execute(ROOT_MEMBERS_SQL, (root_key,)):
+                    found.setdefault(collection_id, set()).add(path)
+                members[root_key] = found
+            stored = {cid: count for cid, count in conn.execute(COUNTS_SQL)}
+        finally:
+            conn.rollback()  # it only read
+        return members, stored
+
+    def collection_counts(self) -> dict[int, int]:
+        """``{collection id: stored rows}`` for every collection."""
+        rows = self.model.get_connection().execute(COUNTS_SQL)
+        return {collection_id: count for collection_id, count in rows}
+
+    def file_collections(self, root_key: str, relative_path: str) -> list[int]:
+        """Ids of the collections holding one file."""
+        rows = self.model.get_connection().execute(FILE_SQL, (root_key, relative_path))
+        return [collection_id for (collection_id,) in rows]
+
+    def add_to_collection(
+        self, *, items: object, collection_id: object = None, name: object = None
+    ) -> dict:
+        """Add files to an existing id or a name; create the name when needed.
+
+        A name whose key exists uses that collection; otherwise a collection
+        is created only when at least one item is inserted. Returns
+        ``{collection, created, added, already}``.
+        """
+        from .models import utc_now_iso  # models imports this module's schema
+
+        if (collection_id is None) == (name is None):
+            raise CollectionInvalid("Give either a collection id or a name")
+        pairs = _items(items)
+        if name is not None:
+            spelled = normalize_collection_name(name)
+        else:
+            collection_id = _collection_id(collection_id)
+        now = utc_now_iso()
+        with self._pair_transaction() as conn:
+            created = False
+            if name is not None:
+                row = conn.execute(
+                    BY_KEY_SQL, (collection_name_key(spelled),)
+                ).fetchone()
+                if row is None:
+                    if not pairs:
+                        raise CollectionFilesMissing("Nothing to add")
+                    cursor = conn.execute(
+                        f"INSERT INTO {COLLECTIONS_TABLE} "  # nosec B608
+                        "(name, name_key, created_at) VALUES (?, ?, ?)",
+                        (spelled, collection_name_key(spelled), now),
+                    )
+                    row, created = (cursor.lastrowid, spelled), True
+            else:
+                row = conn.execute(GET_SQL, (collection_id,)).fetchone()
+                if row is None:
+                    raise CollectionMissing("No such collection")
+            collection = _row(row)
+            added = sum(
+                conn.execute(
+                    _INSERT_ITEM_SQL, (collection["id"], root, path, now)
+                ).rowcount
+                for root, path in pairs
+            )
+        return {
+            "collection": collection,
+            "created": created,
+            "added": added,
+            "already": len(pairs) - added,
+        }
+
+    def remove_from_collection(self, collection_id: object, items: object) -> dict:
+        """Remove files from one collection; ``{collection, removed}``."""
+        collection_id = _collection_id(collection_id)
+        pairs = _items(items)
+        with self._pair_transaction() as conn:
+            row = conn.execute(GET_SQL, (collection_id,)).fetchone()
+            if row is None:
+                raise CollectionMissing("No such collection")
+            removed = sum(
+                conn.execute(_DELETE_ITEM_SQL, (collection_id, root, path)).rowcount
+                for root, path in pairs
+            )
+        return {"collection": _row(row), "removed": removed}
+
+    def rename_collection(self, collection_id: object, name: object) -> dict:
+        """The renamed ``{id, name}``; a case variant of its own name is fine."""
+        collection_id = _collection_id(collection_id)
+        spelled = normalize_collection_name(name)
+        key = collection_name_key(spelled)
+        with self._pair_transaction() as conn:
+            if conn.execute(GET_SQL, (collection_id,)).fetchone() is None:
+                raise CollectionMissing("No such collection")
+            taken = conn.execute(BY_KEY_SQL, (key,)).fetchone()
+            if taken is not None and taken[0] != collection_id:
+                raise CollectionNameTaken(taken[0])
+            conn.execute(
+                f"UPDATE {COLLECTIONS_TABLE} SET name = ?, name_key = ? "  # nosec B608
+                "WHERE id = ?",
+                (spelled, key, collection_id),
+            )
+        return {"id": collection_id, "name": spelled}
+
+    def delete_collection(self, collection_id: object) -> bool:
+        """Delete one collection and its members (never a file). True."""
+        collection_id = _collection_id(collection_id)
+        with self._pair_transaction() as conn:
+            cursor = conn.execute(
+                f"DELETE FROM {COLLECTIONS_TABLE} WHERE id = ?",  # nosec B608
+                (collection_id,),
+            )
+            if cursor.rowcount == 0:
+                raise CollectionMissing("No such collection")
+        return True

database/models.py

@@ -22,6 +22,10 @@ from .managed_connection import (
 from . import restore_stage
 from .library_schema import library_schema_error, migrate_library_state
 from .image_favorites import image_favorites_schema_error, migrate_image_favorites
+from .image_collections import (
+    image_collections_schema_error,
+    migrate_image_collections,
+)
 from .annotation_schema import annotation_schema_error, migrate_annotation_state
 from .autotag_schema import (
     autotag_schema_error,
@@ -338,6 +342,7 @@ class PromptModel:
         migrate_library_state(conn)
         migrate_image_favorites(conn)
         migrate_saved_views(conn)
+        migrate_image_collections(conn)
         # A review group without its annotation group is refused before the
         # annotation install could make it look valid (catalog only; the rows
         # are validated once, in migrate_review_state).
@@ -1159,8 +1164,8 @@ class PromptModel:
             path: File to inspect
             require_library_state: Also require the tables every current
                 database has (Library state, image favorites, image
-                categories, saved views, image annotations and the review
-                queue), as a migrated
+                categories, saved views, collections, image annotations and
+                the review queue), as a migrated
                 restore stage must; old backups may lack them.
             documents: Read every workflow document and refuse a damaged
                 one. A restore checks its upload without it and repairs
@@ -1234,6 +1239,11 @@ class PromptModel:
             )
             if views_error:
                 return False, views_error
+            collections_error = image_collections_schema_error(
+                conn, allow_missing=not require_library_state
+            )
+            if collections_error:
+                return False, collections_error
             annotation_error = annotation_schema_error(
                 conn, allow_missing=not require_library_state
             )

database/operations.py

@@ -7,12 +7,18 @@ import json
 import math
 import datetime
 import os
-from typing import Optional, List, Dict, Any, Set, Tuple
+from typing import Any, Dict, Iterable, List, Optional, Sequence, Set, Tuple
+
+try:
+    from ..utils.inspection_response import inherited_category
+except ImportError:
+    from utils.inspection_response import inherited_category
 
 from .models import PromptModel, normalize_image_path, utc_now_iso
 from .pairs import PairDependencyError, PairOperations  # noqa: F401 (re-exported)
 from .library import LibraryOperations, titled
 from .image_favorites import ImageFavoriteOperations
+from .image_collections import ImageCollectionOperations
 from .annotations import AnnotationOperations
 from .review import ReviewOperations
 from .review_decisions import ReviewDecisionOperations
@@ -238,12 +244,17 @@ LEFT JOIN prompt_library_state pls ON pls.prompt_id = pp.positive_prompt_id
 WHERE gi.image_path IN ({marks}) OR gi.file_path IN ({marks}){fallback}"""
 # One image row describes a card only when no other row claims the same file
 # (as in get_image_link_summaries, several rows are ambiguous). The pair scope
-# and the linked positive follow _LINK_SUMMARY_SQL; category, rating, tags and
-# the searchable text and titles come from that positive (spec D12/D13). Every
+# and the linked positive follow _LINK_SUMMARY_SQL; rating, tags and the
+# searchable text and titles come from that positive (spec D12/D13). The raw
+# recipe and positive categories are read for the inherited category, which
+# Python decides (``inherited_category``, #332 MC4); SQL never filters by
+# category. Every
 # writer since the 3.2.4 migration stores file_path, so it is the path key.
 # The prompt-level conditions are uncorrelated IN sets SQLite builds once, and
 # the CASE evaluates the per-row ambiguity probe only for admitted rows.
-_LINK_FILTER_SQL = """SELECT gi.file_path, p.rating, {hit}, p.id
+_LINK_FILTER_SQL = """SELECT gi.file_path AS file_path, p.rating AS rating,
+       {hit} AS matched, p.id AS positive, pp.id IS NOT NULL AS has_pair,
+       pp.category AS pair_category, p.category AS category
 FROM generated_images gi
 LEFT JOIN prompt_pairs pp ON pp.id = (
     SELECT CASE WHEN COUNT(DISTINCT ip.pair_id) = 1 THEN MIN(ip.pair_id) END
@@ -252,10 +263,31 @@ JOIN prompts p ON p.id = COALESCE(pp.positive_prompt_id, gi.prompt_id)
 WHERE gi.file_path IS NOT NULL AND CASE WHEN {admit} THEN NOT EXISTS (
     SELECT 1 FROM generated_images other
     WHERE other.file_path = gi.file_path AND other.id <> gi.id) ELSE 0 END"""
+# _LINK_FILTER_SQL's rows as one JSON array, built by SQLite in one step (#354).
+# Python's sqlite3 releases and retakes the GIL around every step, so a read
+# that steps once per linked row waits on a busy ComfyUI thread thousands of
+# times. A BLOB value (no writer stores one) makes the value NULL (or
+# json_array refuses it), and the read streams instead. The raw categories are
+# put in only for a category read (#332), so plain reads stay as small as #354
+# measured them.
+_LINK_MATCHES_JSON_SQL = """SELECT CASE
+    WHEN max({blob}) THEN NULL
+    ELSE json_group_array(json_array({columns})) END
+FROM ({select})"""
+_LINK_PLAIN_COLUMNS = ("file_path", "rating", "matched", "positive")
+_LINK_CATEGORY_COLUMNS = _LINK_PLAIN_COLUMNS + ("has_pair", "pair_category", "category")
+# The columns a BLOB can reach: matched, positive and has_pair are computed.
+_LINK_STORED_COLUMNS = ("file_path", "rating", "pair_category", "category")
+# Set when this SQLite has no JSON functions: every later read streams.
+_one_step_unsupported = False
+# One warning per process when a read falls back to streaming.
+_one_step_warned = False
 # The linked positive of each given path (one unambiguous link, pair scope as
-# in _LINK_FILTER_SQL) with its category; {hit} as in link_filter_query. Used
-# for the few files whose current image tags meet a Tags filter (#249 B3).
-_LINK_FACTS_SQL = """SELECT gi.file_path, p.rating, {hit}, p.id, p.category
+# in _LINK_FILTER_SQL) with the raw categories of its inherited category; {hit}
+# as in link_filter_query. Used for the few files whose current image tags meet
+# a Tags filter (#249 B3).
+_LINK_FACTS_SQL = """SELECT gi.file_path, p.rating, {hit}, p.id,
+pp.id IS NOT NULL, pp.category, p.category
 FROM generated_images gi
 LEFT JOIN prompt_pairs pp ON pp.id = (
     SELECT CASE WHEN COUNT(DISTINCT ip.pair_id) = 1 THEN MIN(ip.pair_id) END
@@ -284,8 +316,24 @@ _LINK_FILTER_TAG = (
 )
 
 
+# One unambiguously linked image's facts for the Images filters, a plain tuple
+# (hit, rating, positive id, inherited category): built per linked row, so no
+# NamedTuple. The category is the inherited one (#332 MC4: the recipe's when
+# set, else the positive's), never the image's own category.
+LinkMatch = Tuple[bool, Optional[int], int, Any]
+
+
+def _link_match(rating, matched, positive, has_pair, pair_category, category):
+    inherited, _scope = inherited_category(category, pair_category, bool(has_pair))
+    return (bool(matched), rating, positive, inherited)
+
+
 def link_filter_query(
-    q: str, category: Optional[str], tags: List[str], *, every_link: bool = False
+    q: str,
+    tags: List[str],
+    *,
+    every_link: bool = False,
+    category: Optional[str] = None,
 ) -> Tuple[str, Dict[str, Any]]:
     """The workspace filter SELECT and its named parameters (all bound).
 
@@ -299,7 +347,9 @@ def link_filter_query(
         params["pattern"] = f"%{escape_like(q)}%"
     admitted = []
     if category is not None:
-        admitted.append("p.category = :category")
+        # Only narrows: an inherited category is one of the two raw values;
+        # which one is decided in Python (inherited_category, #332 MC4).
+        admitted.append("(p.category = :category OR pp.category = :category)")
         params["category"] = category
     for index, tag in enumerate(tags):
         admitted.append(_LINK_FILTER_TAG.format(index=index))
@@ -310,6 +360,50 @@ def link_filter_query(
     return sql, params
 
 
+def link_matches_json_query(sql: str, categories: bool = False) -> str:
+    """*sql* from link_filter_query, wrapped to return its rows as one JSON value.
+
+    With *categories* each row also carries the raw values the inherited
+    category is decided from; otherwise only path, rating, hit and positive.
+    """
+    columns = _LINK_CATEGORY_COLUMNS if categories else _LINK_PLAIN_COLUMNS
+    blob = " OR ".join(
+        f"typeof({name}) = 'blob'" for name in _LINK_STORED_COLUMNS if name in columns
+    )
+    return _LINK_MATCHES_JSON_SQL.format(
+        blob=blob, columns=", ".join(columns), select=sql
+    )
+
+
+def _link_matches_from_rows(
+    rows: Iterable[Sequence[Any]], categories: bool, wanted: Optional[Set[str]]
+) -> Dict[str, LinkMatch]:
+    """LinkMatch per wanted path from link rows (one-step or streamed).
+
+    Without *categories* the inherited category is None, as the caller then
+    never compares it.
+    """
+    matches: Dict[str, LinkMatch] = {}
+    for path, rating, matched, positive, *raw in rows:
+        if wanted is None or path in wanted:
+            if categories:
+                matches[path] = _link_match(rating, matched, positive, *raw[:3])
+            else:
+                matches[path] = (bool(matched), rating, positive, None)
+    return matches
+
+
+def _streamed_link_matches(
+    conn: sqlite3.Connection,
+    sql: str,
+    params: Dict[str, Any],
+    wanted: Optional[Set[str]],
+    categories: bool = False,
+) -> Dict[str, LinkMatch]:
+    """The link read stepped once per row (the fallback of the one-step read)."""
+    return _link_matches_from_rows(conn.execute(sql, params), categories, wanted)
+
+
 _ASCII_FOLD = str.maketrans("ABCDEFGHIJKLMNOPQRSTUVWXYZ", "abcdefghijklmnopqrstuvwxyz")
 
 
@@ -360,6 +454,7 @@ class PromptDatabase(
     ExecutionOutputOperations,
     HistoryReadOperations,
     ImageFavoriteOperations,
+    ImageCollectionOperations,
     AnnotationOperations,
     ReviewOperations,
     ReviewDecisionOperations,
@@ -2543,38 +2638,94 @@ class PromptDatabase(
     def get_image_link_matches(
         self,
         q: str,
-        category: Optional[str],
         tags: List[str],
         *,
         every_link: bool = False,
         wanted: Optional[Set[str]] = None,
-    ) -> Dict[str, Tuple[bool, Optional[int], int]]:
-        """Map normalized paths of unambiguously linked images to
-        (hit, rating, positive id).
+        category: Optional[str] = None,
+    ) -> Dict[str, LinkMatch]:
+        """Map normalized paths of unambiguously linked images to a LinkMatch
+        (hit, rating, positive id, inherited category).
 
-        Rows are limited to images whose linked positive has *category* and
-        every tag in *tags* (exact names). ``hit`` says whether *q* occurs in
+        Rows are limited to images whose linked positive has every tag in
+        *tags* (exact names). The category filter is applied by the caller to
+        the effective category (#332). ``hit`` says whether *q* occurs in
         the positive's text, Library title or a tag name; a pair's own title
         names its negative variant and is not searched.
-        Without category or tags only hits are returned, unless *every_link*
+        Without tags only hits are returned, unless *every_link*
         asks for every linked image (the rating sort and the positive count).
-        The positive id is the card's linked positive. One SELECT whose rows
-        are streamed; with *wanted* (the listing's normalized paths) only those
-        paths are kept, so memory is bounded by the listing, not the database.
+        The positive id is the card's linked positive. With *category* only
+        links whose raw recipe or positive category is that value are read (a
+        superset of the links that inherit it; the caller compares the
+        inherited value) and the inherited category is decided per row;
+        otherwise it is None, which keeps the plain reads cheap.
+
+        One SELECT, returned as one JSON value (#354): SQLite steps once, so
+        the read takes the GIL back a constant two or three times instead of
+        once per linked row while ComfyUI generates. Peak memory therefore
+        grows with the database's matching linked rows (about 530 bytes each,
+        more with the raw categories of a category read, freed on return); the
+        result is bounded by *wanted* (the listing's normalized paths) when
+        given. Where the JSON step fails (no JSON functions, a BLOB value, a
+        JSON text over SQLite's length limit) the rows are streamed as before.
         """
-        sql, params = link_filter_query(q, category, tags, every_link=every_link)
-        matches: Dict[str, Tuple[bool, Optional[int], int]] = {}
+        sql, params = link_filter_query(
+            q, tags, every_link=every_link, category=category
+        )
+        categories = category is not None
         with self.model.get_connection() as conn:
-            for path, rating, matched, positive in conn.execute(sql, params):
-                if wanted is None or path in wanted:
-                    matches[path] = (bool(matched), rating, positive)
-        return matches
+            rows = None
+            if not _one_step_unsupported:
+                rows = self._one_step_link_rows(conn, sql, params, categories)
+            if rows is None:
+                return _streamed_link_matches(conn, sql, params, wanted, categories)
+        return _link_matches_from_rows(rows, categories, wanted)
+
+    def _one_step_link_rows(
+        self,
+        conn: sqlite3.Connection,
+        sql: str,
+        params: Dict[str, Any],
+        categories: bool = False,
+    ) -> Optional[List[List[Any]]]:
+        """*sql*'s rows from one JSON value, or None when this read must stream.
+
+        An interrupted read (Move library retiring the connection) is raised
+        as before, never retried row by row.
+        """
+        global _one_step_unsupported, _one_step_warned
+        error: Optional[str] = None
+        try:
+            cursor = conn.execute(link_matches_json_query(sql, categories), params)
+            try:
+                ((document,),) = cursor.fetchall()
+            finally:
+                cursor.close()  # resets the statement: its snapshot ends now
+        except (sqlite3.OperationalError, sqlite3.DataError) as failure:
+            # DataError: the JSON text is longer than SQLITE_LIMIT_LENGTH.
+            error = str(failure)
+            if error == "interrupted":
+                raise
+            if error.startswith("no such function"):
+                _one_step_unsupported = True
+        else:
+            if document is not None:
+                return json.loads(document)
+            error = "a BLOB value"
+        if not _one_step_warned:
+            _one_step_warned = True
+            self.logger.warning(
+                "Reading image links row by row (slower while ComfyUI "
+                "generates); the one-step read failed: %s",
+                error,
+            )
+        return None
 
     def get_image_link_facts(
         self, paths: List[str], q: str = ""
-    ) -> Dict[str, Tuple[bool, Optional[int], int, Optional[str], Tuple[str, ...]]]:
+    ) -> Dict[str, Tuple[bool, Optional[int], int, Any, Tuple[str, ...]]]:
         """Map each normalized path with one unambiguous link to its positive's
-        (hit, rating, positive id, category, tag names).
+        (hit, rating, positive id, inherited category, tag names).
 
         The link is resolved as in get_image_link_matches, and ``hit`` says
         whether *q* occurs in the positive's text, Library title or a tag
@@ -2585,7 +2736,7 @@ class PromptDatabase(
         if q:
             hit = _LINK_FILTER_HIT
             params["pattern"] = f"%{escape_like(q)}%"
-        rows: Dict[str, Tuple[bool, Optional[int], int, Optional[str]]] = {}
+        rows: Dict[str, LinkMatch] = {}
         names: Dict[int, List[str]] = {}
         unique = list(dict.fromkeys(paths))
         with self.model.get_connection() as conn:
@@ -2594,10 +2745,8 @@ class PromptDatabase(
                 bound = {f"path{i}": path for i, path in enumerate(chunk)}
                 marks = ",".join(f":{key}" for key in bound)
                 sql = _LINK_FACTS_SQL.format(hit=hit, marks=marks)
-                for path, rating, matched, positive, category in conn.execute(
-                    sql, {**params, **bound}
-                ):
-                    rows[path] = (bool(matched), rating, positive, category)
+                for path, *row in conn.execute(sql, {**params, **bound}):
+                    rows[path] = _link_match(*row)
             positives = sorted({row[2] for row in rows.values()})
             for start in range(0, len(positives), _FACTS_CHUNK):
                 chunk = positives[start : start + _FACTS_CHUNK]
@@ -2610,20 +2759,16 @@ class PromptDatabase(
             path: (*row, tuple(names.get(row[2], ()))) for path, row in rows.items()
         }
 
-    def get_image_paths_matching(
-        self, q: str, category: Optional[str], tags: List[str]
-    ) -> Set[str]:
-        """Normalized paths whose linked positive matches *q*, *category* and *tags*.
+    def get_image_paths_matching(self, q: str, tags: List[str]) -> Set[str]:
+        """Normalized paths whose linked positive matches *q* and *tags*.
 
         See get_image_link_matches; an empty *q* matches every linked image
-        the category and tags admit.
+        the tags admit.
         """
         return {
             path
-            for path, (hit, _rating, _positive) in self.get_image_link_matches(
-                q, category, tags
-            ).items()
-            if hit or not q
+            for path, match in self.get_image_link_matches(q, tags).items()
+            if match[0] or not q
         }
 
     def get_prompt_id_for_image(self, image_path: str) -> Optional[int]:

database/restore_stage.py

@@ -59,6 +59,7 @@ SUMMARY_KEYS = (
     "image_links",
     "runs",
     "saved_views",
+    "collections",
 )
 
 
@@ -144,7 +145,8 @@ def library_summary(conn: sqlite3.Connection) -> Dict[str, int]:
     ``positives`` are the Library's prompts (``library_counts``'s
     ``positive_prompts``); ``pairs`` the active saved pairs and
     ``archived_pairs`` the archived ones; ``image_links`` the indexed image
-    rows; ``runs`` the recorded executions.
+    rows; ``runs`` the recorded executions; ``collections`` the image
+    collections (#331).
     """
     from .library import _ELIGIBLE
 
@@ -169,6 +171,7 @@ def library_summary(conn: sqlite3.Connection) -> Dict[str, int]:
         "image_links": _count(conn, "generated_images"),
         "runs": _count(conn, "execution_runs"),
         "saved_views": _count(conn, "saved_views"),
+        "collections": _count(conn, "image_collections"),
     }
 
 
@@ -180,6 +183,7 @@ _FINGERPRINT_SQL = (
     ("prompt_pairs", "SELECT MAX(id), MAX(updated_at) FROM prompt_pairs"),
     ("execution_runs", "SELECT COUNT(*), MAX(rowid) FROM execution_runs"),
     ("saved_views", "SELECT MAX(rowid) FROM saved_views"),
+    ("image_collections", "SELECT MAX(rowid) FROM image_collections"),
 )
 
 
@@ -190,9 +194,9 @@ def live_fingerprint(conn: sqlite3.Connection):
     tables a restore replaces, or when a prompt, pair, run or saved view is
     the newest edited (counts plus the newest row and edit of each table);
     the apply refuses with ``plan-changed`` then. It does not see every
-    edit (a tag or favorite change on an older row); those are kept in the
-    restore's safety backup. A read transaction the caller already has is
-    used and left open; one opened here is ended here.
+    edit (a tag, favorite or collection membership change on an older row);
+    those are kept in the restore's safety backup. A read transaction the
+    caller already has is used and left open; one opened here is ended here.
     """
     import json
 

database/vocabulary.py

@@ -13,7 +13,9 @@ same connection before commit. Owner rows and their timestamps never change.
 
 Category consolidation is previewed in one ``library_snapshot`` and applied in
 one ``_pair_transaction`` that recomputes the owner projection and requires the
-preview's ``impact_token``; only rows carrying the exact source change.
+preview's ``impact_token``; only rows carrying the exact source change. Image
+rows holding the source (``image_categories``, #332) are owners too: they are
+counted, hashed into the token and renamed with the prompts and pairs.
 """
 
 import sqlite3
@@ -97,16 +99,30 @@ def _validate_tag_id(tag_id: object) -> None:
         raise ValueError("Invalid tag ID")
 
 
-# The /library/options universe: nonblank values on eligible parents or on
-# active pairs, grouped exactly (whitespace and case variants stay distinct).
-_CATEGORY_GROUPS = (
+# Category values in Tags & categories (#332 §7): the prompt-backed values
+# (nonblank values on eligible parents or on active pairs, grouped exactly, so
+# whitespace and case variants stay distinct; /library/options lists the same
+# prompt-backed set with its own query) plus the values only image rows hold.
+# A prompt-backed value is ordered by its oldest owning positive; an image-only
+# value sorts after all of them (first_pid is SQLite's INTEGER maximum, above
+# any prompt id), then by value. ``image_count`` counts stored rows, including
+# files missing on disk. No FULL OUTER JOIN, so it runs on SQLite 3.31+.
+_IMAGE_ONLY_FIRST_PID = 2**63 - 1
+_CATEGORY_ROWS = (
     "WITH owners(value,pid) AS ("
     "SELECT p.category,p.id FROM prompts p WHERE p.category IS NOT NULL "
     "AND TRIM(p.category)!='' AND " + _ELIGIBLE + " "
     "UNION SELECT pp.category,pp.positive_prompt_id FROM prompt_pairs pp "
     "WHERE pp.archived=0 AND pp.category IS NOT NULL AND TRIM(pp.category)!=''),"
-    "grouped AS (SELECT value,MIN(pid) AS first_pid,"
-    "COUNT(DISTINCT pid) AS prompt_count FROM owners GROUP BY value) "
+    "image_values(value,image_count) AS ("
+    "SELECT category,COUNT(*) FROM image_categories GROUP BY category),"
+    "grouped(value,first_pid,prompt_count) AS ("
+    "SELECT value,MIN(pid),COUNT(DISTINCT pid) FROM owners GROUP BY value "
+    "UNION ALL SELECT value," + str(_IMAGE_ONLY_FIRST_PID) + ",0 FROM image_values "
+    "WHERE value NOT IN (SELECT value FROM owners)),"
+    "counted AS (SELECT g.value,g.first_pid,g.prompt_count,"
+    "COALESCE(i.image_count,0) AS image_count FROM grouped g "
+    "LEFT JOIN image_values i ON i.value=g.value) "
 )
 
 
@@ -117,15 +133,23 @@ _CATEGORY_COMPONENTS = (
     "WHERE p.category=? ORDER BY p.id"
 )
 _CATEGORY_PAIRS = "SELECT id,archived FROM prompt_pairs WHERE category=? ORDER BY id"
+# Image rows holding one exact value (#332 DC7), an index search.
+_IMAGE_CATEGORY_ROWIDS = (
+    "SELECT rowid FROM image_categories WHERE category=? ORDER BY rowid"
+)
 _CATEGORY_EXISTS = (
     "SELECT EXISTS(SELECT 1 FROM prompts WHERE category=:v)"
     " OR EXISTS(SELECT 1 FROM prompt_pairs WHERE category=:v)"
+    " OR EXISTS(SELECT 1 FROM image_categories WHERE category=:v)"
 )
 
-# Pairs after components, so a failing pair update rolls both back.
+# Pairs after components and image rows last, so a failing update rolls every
+# earlier one back. An image row is keyed by its file, never by its value, so
+# merging into a value other files already hold cannot conflict.
 _CATEGORY_UPDATES = (
     "UPDATE prompts SET category=?,updated_at=? WHERE category=?",
     "UPDATE prompt_pairs SET category=?,updated_at=? WHERE category=?",
+    "UPDATE image_categories SET category=?,updated_at=? WHERE category=?",
 )
 
 
@@ -136,8 +160,15 @@ def category_impact_token(
     target_exists: bool,
     components: list,
     pairs: list,
+    *,
+    images: list = (),
 ) -> str:
-    """Hash a category projection like ``library_edit_token`` (no secret)."""
+    """Hash a category projection like ``library_edit_token`` (no secret).
+
+    ``images`` are the rowids of the image rows holding ``source`` (#332 DC7).
+    With none the preimage has no ``images`` key, so a prompt-only token keeps
+    the value it had before image categories existed.
+    """
     preimage = {
         "epoch": epoch,
         "source": source,
@@ -146,6 +177,8 @@ def category_impact_token(
         "components": sorted([int(i), bool(flag)] for i, flag in components),
         "pairs": sorted([int(i), bool(flag)] for i, flag in pairs),
     }
+    if images:
+        preimage["images"] = sorted(int(i) for i in images)
     return canonical_token(_CATEGORY_DOMAIN, preimage)
 
 
@@ -325,7 +358,8 @@ class VocabularyOperations:
         pairs = [
             [row[0], bool(row[1])] for row in conn.execute(_CATEGORY_PAIRS, (source,))
         ]
-        if not components and not pairs:
+        images = [row[0] for row in conn.execute(_IMAGE_CATEGORY_ROWIDS, (source,))]
+        if not components and not pairs and not images:
             raise LibraryCommandError("vocabulary-conflict", epoch)
         target_exists = bool(
             conn.execute(_CATEGORY_EXISTS, {"v": target}).fetchone()[0]
@@ -339,9 +373,10 @@ class VocabularyOperations:
                 "other_prompts": len(components) - eligible,
                 "negative_variants": len(pairs) - archived,
                 "archived_negative_variants": archived,
+                "images": len(images),
             },
             "impact_token": category_impact_token(
-                epoch, source, target, target_exists, components, pairs
+                epoch, source, target, target_exists, components, pairs, images=images
             ),
         }
 
@@ -365,7 +400,8 @@ class VocabularyOperations:
     def apply_vocabulary_category(
         self, *, expected_epoch: str, source: str, target: str, impact_token: str
     ) -> dict:
-        """Set the exact ``source`` to ``target`` on every component and pair.
+        """Set the exact ``source`` to ``target`` on every component, pair
+        and image row.
 
         The recomputed projection must still match the preview's token; a
         mismatch is ``vocabulary-conflict`` with nothing written.
@@ -422,15 +458,15 @@ class VocabularyOperations:
         _limit(limit)
         key = _category_after(after)
         total = conn.execute(
-            _CATEGORY_GROUPS + "SELECT COUNT(*) FROM grouped"
+            _CATEGORY_ROWS + "SELECT COUNT(*) FROM counted"
         ).fetchone()[0]
         where, params = "", []
         if key is not None:
             where = " WHERE first_pid>? OR (first_pid=? AND value>?)"
             params = [key[0], key[0], key[1]]
         rows = conn.execute(
-            _CATEGORY_GROUPS
-            + "SELECT value,first_pid,prompt_count FROM grouped"
+            _CATEGORY_ROWS
+            + "SELECT value,first_pid,prompt_count,image_count FROM counted"
             + where
             + " ORDER BY first_pid,value LIMIT ?",
             [*params, limit + 1],
@@ -440,7 +476,11 @@ class VocabularyOperations:
         return {
             "library_epoch": epoch,
             "categories": [
-                {"value": row["value"], "prompt_count": row["prompt_count"]}
+                {
+                    "value": row["value"],
+                    "prompt_count": row["prompt_count"],
+                    "image_count": row["image_count"],
+                }
                 for row in rows
             ],
             "total": total,

prompt_manager.py

@@ -26,13 +26,13 @@ except ImportError:
     InputTypeDict = dict
 
 try:
-    from .prompt_manager_base import PromptManagerBase
+    from .prompt_manager_base import SAVE_TO_LIBRARY_INPUT, PromptManagerBase
 except ImportError:
     import os
     import sys
 
     sys.path.insert(0, os.path.dirname(os.path.abspath(__file__)))
-    from prompt_manager_base import PromptManagerBase
+    from prompt_manager_base import SAVE_TO_LIBRARY_INPUT, PromptManagerBase
 
 
 class PromptManager(PromptManagerBase, ComfyNodeABC):
@@ -115,6 +115,7 @@ class PromptManager(PromptManagerBase, ComfyNodeABC):
                         )
                     },
                 ),
+                "save_to_library": SAVE_TO_LIBRARY_INPUT,
             },
             "hidden": {
                 # The queued graph and this node's id: used to tell the positive
@@ -152,6 +153,7 @@ class PromptManager(PromptManagerBase, ComfyNodeABC):
         append_text: str = "",
         prompt: Optional[Dict[str, Any]] = None,
         unique_id: Optional[str] = None,
+        save_to_library: bool = True,
     ) -> Tuple[Any]:
         """
         Encode the text prompt and save it to the database.
@@ -164,6 +166,7 @@ class PromptManager(PromptManagerBase, ComfyNodeABC):
             search_text: Text to search for in past prompts
             prepend_text: Text to prepend to the main prompt
             append_text: Text to append to the main prompt
+            save_to_library: Off: nothing is saved, counted or tracked (#90)
 
         Returns:
             Tuple containing the conditioning for the diffusion model and the final
@@ -205,7 +208,9 @@ class PromptManager(PromptManagerBase, ComfyNodeABC):
         # Save prompt to database and set execution context for gallery tracking
         prompt_id = None
         extended_tags = []
-        if storage_text and storage_text.strip():
+        if not save_to_library:
+            self._skip_not_saved(prompt, unique_id)
+        elif storage_text and storage_text.strip():
             self.logger.debug(f"Processing prompt text: {storage_text[:100]}...")
 
             # Add prepend/append info to tags if they exist
@@ -249,6 +254,11 @@ class PromptManager(PromptManagerBase, ComfyNodeABC):
         tokens = clip.tokenize(encoding_text)
         conditioning = clip.encode_from_tokens_scheduled(tokens)
 
+        self.logger.info(f"CLIP encoding completed, text: {repr(encoding_text)[:80]}")
+        result = (conditioning, encoding_text)
+        if not save_to_library:
+            return result  # #90: no metadata registration, no history text
+
         # Register with ComfyUI integration for standard metadata compatibility
         node_id = f"promptmanager_{int(time.time() * 1000)}"
         self.comfyui_integration.register_prompt(
@@ -262,12 +272,10 @@ class PromptManager(PromptManagerBase, ComfyNodeABC):
                 "append_text": append_text.strip() if append_text else None,
             },
         )
-
-        self.logger.info(f"CLIP encoding completed, text: {repr(encoding_text)[:80]}")
-        result = (conditioning, encoding_text)
         record_runtime_context(
             type(self).__name__, text, prepend_text, append_text, encoding_text
         )
+        self._record_cached_item(unique_id, (encoding_text,), storage_text)
         return result
 
     @classmethod

prompt_manager_base.py

@@ -23,9 +23,16 @@ try:
     from .utils.image_monitor import get_image_monitor
     from .utils.prompt_tracker import get_prompt_tracker
     from .utils.hashing import generate_prompt_hash
-    from .utils.prompt_graph import resolve_text, run_prompt_nodes
+    from .utils.prompt_graph import resolve_text, run_prompt_nodes, save_setting
     from .utils.run_keys import executing_run_id, prompt_node_key
     from .utils.usage_tracking import PENDING_FIRST_USE
+    from .utils.cached_runs import (
+        CACHED_RUNS,
+        Role,
+        Seen,
+        own_counts,
+        stored_facts,
+    )
 except ImportError:
     import sys
 
@@ -35,9 +42,32 @@ except ImportError:
     from utils.image_monitor import get_image_monitor
     from utils.prompt_tracker import get_prompt_tracker
     from utils.hashing import generate_prompt_hash
-    from utils.prompt_graph import resolve_text, run_prompt_nodes
+    from utils.prompt_graph import resolve_text, run_prompt_nodes, save_setting
     from utils.run_keys import executing_run_id, prompt_node_key
     from utils.usage_tracking import PENDING_FIRST_USE
+    from utils.cached_runs import (
+        CACHED_RUNS,
+        Role,
+        Seen,
+        own_counts,
+        stored_facts,
+    )
+
+
+# The optional input every PromptManager node declares last (#90): declared
+# last so no widget value of a workflow saved before it shifts.
+SAVE_TO_LIBRARY_INPUT = (
+    "BOOLEAN",
+    {
+        "default": True,
+        "label_on": "save",
+        "label_off": "skip",
+        "tooltip": (
+            "Save this node's prompt to the library, count its runs and link its "
+            "images. Off: PromptManager ignores the run; the output is unchanged."
+        ),
+    },
+)
 
 
 class PromptManagerBase:
@@ -71,8 +101,10 @@ class PromptManagerBase:
         """Save the prompt to the SQLite database.
 
         Runs are normally counted by the queue hook (utils/usage_tracking.py), so a
-        save only counts one when asked to (batch items whose text the hook could not
-        see) or when the hook queued this prompt before it existed.
+        save only counts one when asked to (text the hook could not see, unless
+        this job already counted the node from ComfyUI's cache) or when the hook
+        queued this prompt before it existed. The two are independent: the
+        pending catch-up is written even when ``count_run`` is False.
 
         Args:
             text: The prompt text
@@ -140,7 +172,9 @@ class PromptManagerBase:
         resolved = resolve_text(prompt_graph, node_id)
         return {
             "is_run_prompt": node_id in run_prompt_nodes(prompt_graph),
+            # The hook skips a switch it cannot read (#90): the node counts
             "seen_at_queue": resolved is not None
+            and save_setting(prompt_graph, node_id) is True
             and generate_prompt_hash(resolved) == generate_prompt_hash(text),
             "known": True,
         }
@@ -167,8 +201,19 @@ class PromptManagerBase:
         batch_item = (
             role["known"] and role["is_run_prompt"] and not role["seen_at_queue"]
         )
+        counts = own_counts(
+            type(self).__name__,
+            stored_facts(text),
+            Role(
+                frozenset({"text"} if role["known"] and role["is_run_prompt"] else ())
+            ),
+            Seen(text=role["seen_at_queue"]),
+        )
         prompt_id = self._save_prompt_to_database(
-            text=text, category=category, tags=tags, count_run=batch_item
+            text=text,
+            category=category,
+            tags=tags,
+            count_run=bool(counts) and not self._replayed(unique_id),
         )
         if role["is_run_prompt"] and not prompt_id:
             # The save failed: an older entry of this key must not stand in (#342)
@@ -182,6 +227,34 @@ class PromptManagerBase:
             )
         return prompt_id
 
+    def _cached_runs(self) -> Any:
+        return getattr(self, "cached_runs", CACHED_RUNS)
+
+    def _replayed(self, unique_id: Any) -> bool:
+        """This job already counted the node from ComfyUI's cache (#347).
+
+        A RAM-pressure cache can free the entry after ``execution_cached``, so
+        the node executes in a job that replayed it: its own count is skipped.
+        """
+        try:
+            return unique_id is not None and self._cached_runs().replayed(unique_id)
+        except Exception as error:
+            self.logger.debug(f"Could not check the cached-run replay: {error}")
+            return False
+
+    def _record_cached_item(
+        self, unique_id: Any, outputs: tuple, text: Any, negative_text: Any = None
+    ) -> None:
+        """Remember this execution's outputs and stored text for cached jobs (#347)."""
+        if unique_id is None:
+            return
+        try:
+            self._cached_runs().record_item(
+                unique_id, outputs, stored_facts(text, negative_text)
+            )
+        except Exception as error:
+            self.logger.debug(f"Could not record the cached-run receipt: {error}")
+
     @staticmethod
     def _run_identity(prompt_graph: Any, unique_id: Any) -> Dict[str, Any]:
         """The tracker's proof that an image belongs to this execution (#342)."""
@@ -190,6 +263,24 @@ class PromptManagerBase:
             "run_id": executing_run_id(),
         }
 
+    def _skip_not_saved(self, prompt_graph: Any = None, unique_id: Any = None) -> None:
+        """``save_to_library`` is off: nothing is saved, counted or tracked (#90).
+
+        A literal off makes the node no run prompt (``utils.prompt_graph``), so
+        no image looks up its run key. A switch wired from a node the graph
+        cannot read keeps the node a run prompt: its run key is marked like an
+        empty editor's, so an earlier run's entry of that key never links this
+        run's images.
+        """
+        message = (
+            "Save to library is off for this node: its prompt is not saved, the "
+            "run is not counted and its images are not linked"
+        )
+        if isinstance(prompt_graph, dict) and unique_id is not None:
+            self._mark_untracked(prompt_graph, unique_id, message)
+        else:
+            self.logger.info(message)
+
     def _skip_empty_prompt(self, prompt_graph: Any, unique_id: Any) -> None:
         """The ``text`` editor is empty: save and track nothing (#342).
 

prompt_manager_pair.py

@@ -12,9 +12,21 @@ else:
 
 try:
     from .prompt_manager import IO, InputTypeDict, PromptManagerFocused
+    from .utils.cached_runs import (
+        PAIR,
+        Role,
+        Seen,
+        own_counts,
+        pair_components,
+        stored_facts,
+    )
     from .utils.hashing import generate_prompt_hash
     from .utils.pair_execution import record_pair_execution
-    from .utils.prompt_graph import intrinsic_pair_nodes, run_prompt_components
+    from .utils.prompt_graph import (
+        intrinsic_pair_nodes,
+        run_prompt_components,
+        save_setting,
+    )
     from .utils.prompt_pair_resolver import resolve_text_input
     from .utils.usage_tracking import (
         PENDING_FIRST_USE,
@@ -23,9 +35,21 @@ try:
     )
 except ImportError:
     from prompt_manager import IO, InputTypeDict, PromptManagerFocused
+    from utils.cached_runs import (
+        PAIR,
+        Role,
+        Seen,
+        own_counts,
+        pair_components,
+        stored_facts,
+    )
     from utils.hashing import generate_prompt_hash
     from utils.pair_execution import record_pair_execution
-    from utils.prompt_graph import intrinsic_pair_nodes, run_prompt_components
+    from utils.prompt_graph import (
+        intrinsic_pair_nodes,
+        run_prompt_components,
+        save_setting,
+    )
     from utils.prompt_pair_resolver import resolve_text_input
     from utils.usage_tracking import (
         PENDING_FIRST_USE,
@@ -74,8 +98,13 @@ class PromptManagerPair(PromptManagerFocused):
         prompt: Optional[Dict[str, Any]] = None,
         unique_id: Optional[str] = None,
         extra_pnginfo: Optional[Dict[str, Any]] = None,
+        save_to_library: bool = True,
     ) -> Tuple[Any, Any, str, str]:
-        """Encode both sides once; database and tracking failures never block CLIP."""
+        """Encode both sides once; database and tracking failures never block CLIP.
+
+        With ``save_to_library`` off nothing is saved, counted, tracked or
+        recorded (#90); the four outputs are the same.
+        """
         if clip is None:
             raise RuntimeError("ERROR: clip input is invalid: None")
         if not isinstance(text, str) or not isinstance(negative_text, str):
@@ -93,6 +122,10 @@ class PromptManagerPair(PromptManagerFocused):
         negative_conditioning = clip.encode_from_tokens_scheduled(
             clip.tokenize(negative_text)
         )
+        result = positive_conditioning, negative_conditioning, positive, negative_text
+        if not save_to_library:
+            self._skip_not_saved(prompt, unique_id)
+            return result
         record_pair_execution(
             extra_pnginfo,
             unique_id,
@@ -153,7 +186,6 @@ class PromptManagerPair(PromptManagerFocused):
             )
         except Exception as error:
             self.logger.warning(f"Could not register paired metadata: {error}")
-        result = positive_conditioning, negative_conditioning, positive, negative_text
         record_runtime_context(
             type(self).__name__,
             text,
@@ -165,6 +197,9 @@ class PromptManagerPair(PromptManagerFocused):
         record_runtime_context(
             type(self).__name__, negative_text, "", "", negative_text, role="negative"
         )
+        self._record_cached_item(
+            unique_id, (positive, negative_text), text, negative_text
+        )
         return result
 
     def _save_and_track_pair(
@@ -200,6 +235,9 @@ class PromptManagerPair(PromptManagerFocused):
         resolved_negative = (
             resolve_text_input(prompt, node_id, "negative_text") if known else None
         )
+        if known and save_setting(prompt, node_id) is not True:
+            # The hook skips a switch it cannot read (#90): the node counts
+            resolved_positive = resolved_negative = None
         positive_seen = resolved_positive is not None and generate_prompt_hash(
             resolved_positive
         ) == generate_prompt_hash(text)
@@ -209,36 +247,35 @@ class PromptManagerPair(PromptManagerFocused):
             and generate_prompt_hash(resolved_negative)
             == generate_prompt_hash(negative_text)
         )
+        negative_seen = resolved_negative is not None and generate_prompt_hash(
+            resolved_negative
+        ) == generate_prompt_hash(negative_text)
         positive_pending = getattr(self, "pending_first_use", PENDING_FIRST_USE)
         pair_pending = getattr(self, "pending_pair_first_use", PENDING_PAIR_FIRST_USE)
-        counted = set()
-        for field, component_text, resolved_text, component_id in (
-            ("text", text, resolved_positive, prompt_id),
-            (
-                "negative_text",
-                negative_text,
-                resolved_negative,
-                pair["negative_prompt_id"],
-            ),
-        ):
-            component_hash = generate_prompt_hash(component_text)
-            if (
-                field not in used_fields
-                or not component_id
-                or component_hash in counted
-            ):
+        facts = stored_facts(text, negative_text)
+        role = Role(frozenset(used_fields), pair_used)
+        # Own counts are skipped when this job already counted the node from
+        # ComfyUI's cache (#347); the pending first-use catch-up never is.
+        counts = (
+            []
+            if self._replayed(unique_id)
+            else own_counts(
+                PAIR, facts, role, Seen(positive_seen, negative_seen, pair_seen)
+            )
+        )
+        component_ids = {"text": prompt_id, "negative_text": pair["negative_prompt_id"]}
+        for field, component_hash in pair_components(facts, role):
+            component_id = component_ids[field]
+            if not component_id:
                 continue
-            counted.add(component_hash)
-            seen_at_queue = (
-                resolved_text is not None
-                and generate_prompt_hash(resolved_text) == component_hash
+            runs = positive_pending.consume(component_hash) + int(
+                ("prompt", component_hash) in counts
             )
-            runs = positive_pending.consume(component_hash) + int(not seen_at_queue)
             if runs:
                 self.db.record_prompt_use(component_id, times=runs)
         if pair_used:
             runs = pair_pending.consume(pair_run_key(text, negative_text)) + int(
-                not pair_seen
+                ("pair", facts.positive_hash, facts.negative_hash) in counts
             )
             if runs:
                 self.db.record_prompt_pair_use(pair_id, times=runs)

prompt_manager_text.py

@@ -26,13 +26,13 @@ except ImportError:
     InputTypeDict = dict
 
 try:
-    from .prompt_manager_base import PromptManagerBase
+    from .prompt_manager_base import SAVE_TO_LIBRARY_INPUT, PromptManagerBase
 except ImportError:
     import os
     import sys
 
     sys.path.insert(0, os.path.dirname(os.path.abspath(__file__)))
-    from prompt_manager_base import PromptManagerBase
+    from prompt_manager_base import SAVE_TO_LIBRARY_INPUT, PromptManagerBase
 
 
 class PromptManagerText(PromptManagerBase, ComfyNodeABC):
@@ -115,6 +115,7 @@ class PromptManagerText(PromptManagerBase, ComfyNodeABC):
                         ),
                     },
                 ),
+                "save_to_library": SAVE_TO_LIBRARY_INPUT,
             },
             "hidden": {
                 # The queued graph and this node's id: used to tell the positive
@@ -149,6 +150,7 @@ class PromptManagerText(PromptManagerBase, ComfyNodeABC):
         append_text: str = "",
         prompt: Optional[Dict[str, Any]] = None,
         unique_id: Optional[str] = None,
+        save_to_library: bool = True,
     ) -> Tuple[str]:
         """
         Process the text prompt and save it to the database.
@@ -160,6 +162,7 @@ class PromptManagerText(PromptManagerBase, ComfyNodeABC):
             search_text: Text to search for in past prompts
             prepend_text: Text to prepend to the main prompt
             append_text: Text to append to the main prompt
+            save_to_library: Off: nothing is saved, counted or tracked (#90)
 
         Returns:
             Tuple containing the final processed text string
@@ -184,7 +187,9 @@ class PromptManagerText(PromptManagerBase, ComfyNodeABC):
         # Save prompt to database and set execution context for gallery tracking
         prompt_id = None
         extended_tags = []
-        if storage_text and storage_text.strip():
+        if not save_to_library:
+            self._skip_not_saved(prompt, unique_id)
+        elif storage_text and storage_text.strip():
             self.logger.debug(f"Processing prompt text: {storage_text[:100]}...")
 
             # Add prepend/append info to tags if they exist
@@ -220,6 +225,11 @@ class PromptManagerText(PromptManagerBase, ComfyNodeABC):
         else:
             self._skip_empty_prompt(prompt, unique_id)
 
+        self.logger.debug(f"Text processing completed: {final_text[:100]}...")
+        result = (final_text,)
+        if not save_to_library:
+            return result  # #90: no metadata registration, no history text
+
         # Register with ComfyUI integration for standard metadata compatibility
         node_id = f"promptmanagertext_{int(time.time() * 1000)}"
         self.comfyui_integration.register_prompt(
@@ -233,12 +243,10 @@ class PromptManagerText(PromptManagerBase, ComfyNodeABC):
                 "append_text": append_text.strip() if append_text else None,
             },
         )
-
-        self.logger.debug(f"Text processing completed: {final_text[:100]}...")
-        result = (final_text,)
         record_runtime_context(
             type(self).__name__, text, prepend_text, append_text, final_text
         )
+        self._record_cached_item(unique_id, (final_text,), storage_text)
         return result
 
     @classmethod

py/api/__init__.py

@@ -48,6 +48,7 @@ from .review import ReviewRoutesMixin
 from .review_recording import ReviewRecordingMixin
 from .review_events import ReviewEventsMixin
 from .saved_views import SavedViewRoutesMixin
+from .collections import CollectionRoutesMixin
 from .history import HistoryRoutesMixin
 from .images import ImageRoutesMixin
 from .gallery_listing import WALK_STOPPED, RootListing, WalkOutcome
@@ -315,6 +316,7 @@ class PromptManagerAPI(
     ReviewRecordingMixin,
     ReviewEventsMixin,
     SavedViewRoutesMixin,
+    CollectionRoutesMixin,
     HistoryRoutesMixin,
     PromptRoutesMixin,
     ImageRoutesMixin,
@@ -686,6 +688,7 @@ class PromptManagerAPI(
         self._register_review_routes(routes)
         self._register_review_event_routes(routes)
         self._register_saved_view_routes(routes)
+        self._register_collection_routes(routes)
         self._register_history_routes(routes)
         self._register_prompt_routes(routes)
         self._register_image_routes(routes)

py/api/admin.py

@@ -52,6 +52,7 @@ try:
         extract_generation_params,
         parse_page_json,
     )
+    from ...utils.prompt_graph import metadata_opted_out, side_opted_out
     from ...utils.prompt_pair_resolver import resolve_prompt_pair, sampler_less_prompt
     from ...utils.review_capture import reviewable
 except ImportError:
@@ -67,6 +68,7 @@ except ImportError:
         extract_generation_params,
         parse_page_json,
     )
+    from utils.prompt_graph import metadata_opted_out, side_opted_out
     from utils.prompt_pair_resolver import resolve_prompt_pair, sampler_less_prompt
     from utils.review_capture import reviewable
 
@@ -260,6 +262,11 @@ SCAN_DETAILS = {
     "failed": "Error: the prompt could not be stored",
     "error": "Error: the file could not be processed (see the server log)",
     "found": "Metadata found, but no prompt text to store",
+    "not_saved": "Skipped · its PromptManager node does not save to the library",
+    "not_saved_wired": (
+        "Skipped · its Save to library switch is wired from another node, so only "
+        "the run itself can link it"
+    ),
     None: "No prompt metadata in the file",
 }
 
@@ -1720,10 +1727,34 @@ class AdminRoutesMixin:
                 except (ValueError, TypeError):
                     # Preserve malformed metadata for conservative resolver rejection.
                     pass
+        if metadata_opted_out(canonical):
+            # Save to library was off for the run's prompt: no prompt, link,
+            # recorded-run fallback or Needs review item (#90)
+            return "not_saved"
+        if metadata_opted_out(canonical, unreadable=True):
+            # The switch is wired from a node the file cannot show: only the
+            # live run (the node saw the value) can link it, never a scan
+            return "not_saved_wired"
         if not canonical.get("prompt") and not canonical.get("parameters"):
             return self._link_recorded_run(media_file, canonical, detailed, history)
         _with_file_size(canonical, media_file)
         resolution = resolve_prompt_pair(canonical)
+        if resolution["state"] == "resolved" and side_opted_out(
+            canonical.get("prompt"), resolution["negative"]
+        ):
+            # Only the negative's node is off: store the positive alone (#90)
+            prompt_text = (
+                resolution["positive"]
+                .get("storage_text", resolution["positive"]["text"])
+                .strip()
+            )
+            status = self._store_scanned_prompt(media_file, canonical, prompt_text)
+            if not detailed or status not in SCAN_TITLED_DETAILS:
+                return status
+            return {
+                "status": status,
+                "title": self._scan_title(media_file, prompt_text),
+            }
         if blank_positive(resolution):
             # Only prepend/append text: no stored prompt can own it (#342)
             resolution = {**resolution, "state": "unresolved"}

py/api/annotation_suggest.py

@@ -86,6 +86,23 @@ def unavailable_message(model_name: str) -> str:
     return f"{model_name} is not downloaded. {UNAVAILABLE_HINT}"
 
 
+def model_refusal(status: dict, model_type: str, model_name: str):
+    """Why *model_type* cannot run, from ``get_models_status()``, else None.
+
+    Not downloaded first (``unavailable_message``), then a backend package
+    that cannot be imported (#424). Both are a 409 ``model-unavailable``.
+    """
+    entry = status.get(model_type) or {}
+    if not entry.get("downloaded"):
+        return unavailable_message(model_name)
+    missing = entry.get("missing_packages") or []
+    if missing:
+        from ..autotag import missing_packages_message
+
+        return missing_packages_message(model_type, [str(name) for name in missing])
+    return None
+
+
 def run_suggestions(service, capture_path: str, model_type: str, options: dict):
     """Blocking: load (or swap to) *model_type*, then suggest for the capture.
 

py/api/annotations.py

@@ -21,9 +21,9 @@ from aiohttp import web
 from .annotation_suggest import (
     SUGGEST_FIELDS,
     OptionError,
+    model_refusal,
     run_suggestions,
     suggestion_rows,
-    unavailable_message,
     validate_options,
 )
 from .autotag_routes import model_busy_response
@@ -595,7 +595,7 @@ class AnnotationRoutesMixin:
         it, including a legacy batch), never downloads (``model-unavailable``)
         and never writes.
         """
-        from ..autotag import MODELS, get_autotag_service
+        from ..autotag import MODELS, MissingBackendPackages, get_autotag_service
         from ..config import MAX_SUGGESTION_LIMIT, AutotagConfig
 
         data, error = await self._annotation_body(request, SUGGEST_FIELDS)
@@ -633,9 +633,10 @@ class AnnotationRoutesMixin:
                 return error
             service = get_autotag_service()
             status = await self._run_in_executor(service.get_models_status)
-            if not status.get(model_type, {}).get("downloaded"):
-                name = MODELS.get(model_type, {}).get("name", model_type)
-                return _coded("model-unavailable", unavailable_message(name))
+            name = MODELS.get(model_type, {}).get("name", model_type)
+            refusal = model_refusal(status, model_type, name)
+            if refusal is not None:
+                return _coded("model-unavailable", refusal)
             owner = model_gate.new_owner("image-annotations")
             if not model_gate.try_acquire(owner):
                 return model_busy_response()
@@ -654,6 +655,9 @@ class AnnotationRoutesMixin:
                     options,
                     on_skipped=functools.partial(store.release_use, token),
                 )
+            except MissingBackendPackages as exc:
+                # A package gone since the status read (load_model logged it).
+                return _coded("model-unavailable", str(exc))
             except Exception:
                 self.logger.exception("Tag suggestion failed")
                 return _error(

py/api/autotag_routes.py

@@ -102,6 +102,17 @@ def size_label(size: str) -> str:
     return f"{match.group(1)} {match.group(2)}" if match else str(size or "")
 
 
+def package_entry(flags: dict) -> dict:
+    """The public package fields of one status entry (#424): the pip names
+    that cannot be imported, the install command and note (strings or null)."""
+    missing = flags.get("missing_packages") or []
+    text = {
+        key: flags[key] if isinstance(flags.get(key), str) else None
+        for key in ("install_command", "install_note")
+    }
+    return {"missing_packages": [str(name) for name in missing], **text}
+
+
 def public_models(status: dict, loaded_type: Optional[str]) -> dict:
     """Each engine model's public entry: names and sizes from ``MODELS``, the
     flags from ``status``; never a path (``model_path`` is dropped)."""
@@ -122,6 +133,7 @@ def public_models(status: dict, loaded_type: Optional[str]) -> dict:
         for key in ("model_exists", "mmproj_exists"):
             if key in flags:
                 entry[key] = bool(flags[key])
+        entry.update(package_entry(flags))
         models[model_type] = entry
     return models
 
@@ -750,6 +762,7 @@ class AutotagRoutesMixin:
                     "downloaded": bool(
                         (status.get(model_type) or {}).get("downloaded")
                     ),
+                    **package_entry(status.get(model_type) or {}),
                 },
                 "existing": existing,
                 "review_required": existing == "replace_after_review",
@@ -1073,8 +1086,8 @@ class AutotagRoutesMixin:
         return self.db.current_image_tags({locator: fingerprint}).get(locator, [])
 
     async def _suggest_single(self, target, model_type, options, base):
-        from ..autotag import MODELS, get_autotag_service
-        from .annotation_suggest import unavailable_message
+        from ..autotag import MODELS, MissingBackendPackages, get_autotag_service
+        from .annotation_suggest import model_refusal
         from .annotations import _pinned_suggestions, _settled, _submit
 
         try:
@@ -1084,8 +1097,8 @@ class AutotagRoutesMixin:
 
         service = get_autotag_service()
         status = await self._run_in_executor(service.get_models_status)
-        if not (status.get(model_type) or {}).get("downloaded"):
-            message = unavailable_message(MODELS[model_type]["name"])
+        message = model_refusal(status, model_type, MODELS[model_type]["name"])
+        if message is not None:
             return self._coded_error(409, "model-unavailable", message)
         owner = model_gate.new_owner("legacy-single")
         if not model_gate.try_acquire(owner):
@@ -1131,6 +1144,9 @@ class AutotagRoutesMixin:
                     options,
                     on_skipped=functools.partial(store.release_use, capture.token),
                 )
+            except MissingBackendPackages as exc:
+                # A package gone since the status read (load_model logged it).
+                return self._coded_error(409, "model-unavailable", str(exc))
             finally:
                 store.release(capture.token)
         finally:

py/api/collections.py

@@ -0,0 +1,350 @@
+"""Collections (#331): named, hand-picked groups of images.
+
+Routes under ``/prompt_manager/collections``. A file is named by its listing
+identity ``{root_index, root_scope, path}`` (``media_target.py``) and stored
+under the favorites' keys (``_favorite_root_key``, ``_favorite_relative_key``),
+which never leave the server. A stale scope (the sources were reordered) is a
+409 ``source-changed`` before anything is written. Responses are
+``{"success": bool, ...}``; an unexpected failure is a logged 500 with a
+generic message, never the exception text.
+"""
+
+import hmac
+import re
+from functools import wraps
+from typing import Callable
+
+from aiohttp import web
+
+from .images import (
+    _favorite_relative_key,
+    _favorite_root_key,
+    _forbidden,
+    _json_body_or_error,
+    collection_not_found,
+)
+from .media_target import media_target
+from .prompts import bad_request
+
+try:
+    from ...database.image_collections import (
+        CollectionFilesMissing,
+        CollectionInvalid,
+        CollectionMissing,
+        CollectionNameTaken,
+        normalize_collection_name,
+    )
+    from ...database.image_favorites import MAX_KEY_LENGTH
+except ImportError:
+    from database.image_collections import (
+        CollectionFilesMissing,
+        CollectionInvalid,
+        CollectionMissing,
+        CollectionNameTaken,
+        normalize_collection_name,
+    )
+    from database.image_favorites import MAX_KEY_LENGTH
+
+BASE = "/prompt_manager/collections"
+# The most files one add or remove names (the bulk tagging cap, MC12).
+MAX_ITEMS = 500
+MAX_ID = 999_999_999
+_PATH_ID = re.compile(r"[1-9][0-9]{0,8}")
+_ROOT_INDEX = re.compile(r"[0-9]{1,9}")
+_ADD_TARGETS = (frozenset({"collection_id", "items"}), frozenset({"name", "items"}))
+_REMOVE_FIELDS = frozenset({"collection_id", "items"})
+_RENAME_FIELDS = frozenset({"name"})
+_MEDIA_MESSAGE = "Only media files can be added to a collection"
+READ_FAILED = "Collections could not be read."
+CHANGE_FAILED = "The collection could not be changed."
+SOURCES_CHANGED = "The sources changed. Refresh the images and try again."
+FILES_NOT_FOUND = (
+    "None of the selected images could be found. Refresh the images and try again."
+)
+NAME_TAKEN = "Another collection already has this name."
+
+
+class _Refused(Exception):
+    """A request refused before any write; carries its response."""
+
+    def __init__(self, response: web.Response):
+        super().__init__(response.status)
+        self.response = response
+
+
+def _coded(status: int, code: str, message: str, **extra) -> web.Response:
+    return web.json_response(
+        {"success": False, "code": code, "error": message, **extra},
+        status=status,
+        headers={"Cache-Control": "no-store"},
+    )
+
+
+def _collection_response(failure: str) -> Callable:
+    """Map the database's refusals to their status; anything else to a 500."""
+
+    def decorate(handler: Callable) -> Callable:
+        @wraps(handler)
+        async def wrapped(self, request: web.Request) -> web.Response:
+            try:
+                return await handler(self, request)
+            except _Refused as refused:
+                return refused.response
+            except CollectionInvalid as exc:
+                return bad_request(str(exc))
+            except CollectionMissing:
+                return collection_not_found()
+            except CollectionFilesMissing:
+                return _coded(404, "files-not-found", FILES_NOT_FOUND)
+            except CollectionNameTaken as exc:
+                return _coded(
+                    409, "name-taken", NAME_TAKEN, collection_id=exc.collection_id
+                )
+            except Exception:
+                self.logger.exception("Collection request failed")
+                return web.json_response(
+                    {"success": False, "error": failure}, status=500
+                )
+
+        return wrapped
+
+    return decorate
+
+
+def _path_id(request: web.Request) -> int:
+    raw = request.match_info["collection_id"]
+    if not raw.isascii() or _PATH_ID.fullmatch(raw) is None:
+        raise _Refused(bad_request("Invalid collection id"))
+    return int(raw)
+
+
+def _body_id(value: object) -> int:
+    if type(value) is not int or not 1 <= value <= MAX_ID:
+        raise _Refused(bad_request("Invalid collection id"))
+    return value
+
+
+async def _body(request: web.Request, allowed: tuple) -> dict:
+    data, error = await _json_body_or_error(request)
+    if error is not None:
+        raise _Refused(error)
+    if frozenset(data) not in allowed:
+        names = " or ".join(", ".join(sorted(fields)) for fields in allowed)
+        raise _Refused(bad_request(f"Request must contain exactly {names}"))
+    return data
+
+
+class CollectionRoutesMixin:
+    """Mixin providing the collection routes (#331)."""
+
+    def _register_collection_routes(self, routes: web.RouteTableDef) -> None:
+        # The fixed paths go first: "/items" must never match "/{collection_id}".
+        routes.get(BASE)(self.list_collections)
+        routes.post(BASE + "/items")(self.add_collection_items)
+        routes.delete(BASE + "/items")(self.remove_collection_items)
+        routes.get(BASE + "/membership")(self.collection_membership)
+        routes.patch(BASE + "/{collection_id}")(self.rename_collection)
+        routes.delete(BASE + "/{collection_id}")(self.delete_collection)
+
+    # -- reads -------------------------------------------------------------
+
+    @_collection_response(READ_FAILED)
+    async def list_collections(self, request: web.Request) -> web.Response:
+        """Every collection with its listed ``count`` and its ``missing`` rows.
+
+        ``count`` is the members listed under every configured source (one
+        primary-key range read per source, matched against its cached
+        listing); ``missing`` the stored rows beyond it. When a scan or a read
+        fails both are null for every collection and the names still return.
+        """
+        collections = await self._run_in_executor(self.db.list_collections)
+        counts = None
+        if collections:
+            try:
+                counts = await self._collection_counts()
+            except Exception:
+                self.logger.warning(
+                    "Collection counts could not be read", exc_info=True
+                )
+        rows = []
+        for collection in collections:
+            listed, stored = (None, None) if counts is None else counts
+            count = None if listed is None else listed.get(collection["id"], 0)
+            stored_rows = None if stored is None else stored.get(collection["id"], 0)
+            rows.append(
+                {
+                    **collection,
+                    "count": count,
+                    # Clamped: never below zero, whatever the snapshot saw.
+                    "missing": None if count is None else max(0, stored_rows - count),
+                }
+            )
+        return web.json_response({"success": True, "collections": rows})
+
+    async def _collection_counts(self) -> tuple[dict, dict]:
+        """``({id: listed members}, {id: stored rows})`` over every source.
+
+        The scans come from the gallery cache; the root keys (realpath), the
+        listings and one read snapshot of every member and stored count run
+        in the executor, so a write between two reads cannot skew them.
+        """
+        roots = list(self._get_all_output_dirs())
+        root_keys = await self._run_in_executor(
+            lambda: [_favorite_root_key(root) for root in roots]
+        )
+        distinct = {}  # one folder configured twice counts once
+        for root, root_key in zip(roots, root_keys):
+            distinct.setdefault(root_key, root)
+        scans = {
+            root_key: await self._get_gallery_files(root)
+            for root_key, root in distinct.items()
+        }
+        return await self._run_in_executor(self._count_members, distinct, scans)
+
+    def _count_members(self, roots: dict, scans: dict) -> tuple[dict, dict]:
+        """Blocking: match one snapshot's members against each listing."""
+        members, stored = self.db.collection_count_snapshot(list(roots))
+        listed: dict[int, int] = {}
+        for root_key, root in roots.items():
+            listing = self._favorite_listing(root, scans[root_key])
+            for collection_id, keys in members.get(root_key, {}).items():
+                found = sum(1 for key in keys if key in listing)
+                listed[collection_id] = listed.get(collection_id, 0) + found
+        return listed, stored
+
+    @_collection_response(READ_FAILED)
+    async def collection_membership(self, request: web.Request) -> web.Response:
+        """GET ``?root_index&root_scope&path``: the ids holding that file."""
+        query = request.query
+        if len(query) != 3 or set(query) != {"root_index", "root_scope", "path"}:
+            # len() counts a repeated key (?path=a&path=b) twice.
+            return bad_request("Query must name root_index, root_scope and path once")
+        raw_index = query["root_index"]
+        if not raw_index.isascii() or _ROOT_INDEX.fullmatch(raw_index) is None:
+            return bad_request("Invalid gallery root")
+        roots = list(self._get_all_output_dirs())
+        target = {
+            "root_index": int(raw_index),
+            "root_scope": query["root_scope"],
+            "path": query["path"],
+        }
+        root_key, key = await self._run_in_executor(
+            self._collection_key, target, roots, {}
+        )
+        ids = await self._run_in_executor(self.db.file_collections, root_key, key)
+        return web.json_response({"success": True, "collection_ids": ids})
+
+    # -- writes ------------------------------------------------------------
+
+    @_collection_response(CHANGE_FAILED)
+    async def add_collection_items(self, request: web.Request) -> web.Response:
+        """POST ``{collection_id | name, items}``: add the listed files.
+
+        Only files the source's listing holds are added; the rest count in
+        ``not_found``. A new name with nothing to add creates nothing (404
+        ``files-not-found``).
+        """
+        data = await _body(request, _ADD_TARGETS)
+        target = {}
+        if "name" in data:
+            target["name"] = normalize_collection_name(data["name"])
+        else:
+            target["collection_id"] = _body_id(data["collection_id"])
+        roots = list(self._get_all_output_dirs())
+        keys = await self._run_in_executor(self._collection_items, data["items"], roots)
+        found, not_found = [], 0
+        listings = {}
+        for index, root_key, key in keys:
+            if index not in listings:
+                root = roots[index]
+                files = await self._get_gallery_files(root)
+                listings[index] = await self._run_in_executor(
+                    self._favorite_listing, root, files
+                )
+            if key in listings[index]:
+                found.append((root_key, key))
+            else:
+                not_found += 1
+        result = await self._run_in_executor(
+            lambda: self.db.add_to_collection(items=found, **target)
+        )
+        return web.json_response({"success": True, **result, "not_found": not_found})
+
+    @_collection_response(CHANGE_FAILED)
+    async def remove_collection_items(self, request: web.Request) -> web.Response:
+        """DELETE ``{collection_id, items}``: listed or not, every named file."""
+        data = await _body(request, (_REMOVE_FIELDS,))
+        collection_id = _body_id(data["collection_id"])
+        roots = list(self._get_all_output_dirs())
+        keys = await self._run_in_executor(self._collection_items, data["items"], roots)
+        result = await self._run_in_executor(
+            self.db.remove_from_collection,
+            collection_id,
+            [(root_key, key) for _index, root_key, key in keys],
+        )
+        return web.json_response({"success": True, **result})
+
+    @_collection_response(CHANGE_FAILED)
+    async def rename_collection(self, request: web.Request) -> web.Response:
+        """PATCH ``{name}``: a case variant of its own name is allowed."""
+        collection_id = _path_id(request)
+        data = await _body(request, (_RENAME_FIELDS,))
+        name = normalize_collection_name(data["name"])
+        collection = await self._run_in_executor(
+            self.db.rename_collection, collection_id, name
+        )
+        return web.json_response({"success": True, "collection": collection})
+
+    @_collection_response(CHANGE_FAILED)
+    async def delete_collection(self, request: web.Request) -> web.Response:
+        """DELETE: the collection and its members; files are never touched."""
+        collection_id = _path_id(request)
+        await self._run_in_executor(self.db.delete_collection, collection_id)
+        return web.json_response({"success": True, "deleted": True})
+
+    # -- request steps -----------------------------------------------------
+
+    def _collection_items(self, items: object, roots: list) -> list[tuple]:
+        """``(index, root key, relative key)`` per distinct file, in order.
+
+        Blocking (realpath per root, once per request): run in the executor.
+        Every item is checked before anything is written: 400 for a malformed
+        shape, 403 for an unsafe path, 409 for a stale scope. Two spellings
+        that fold to one key (Windows) count once.
+        """
+        if not isinstance(items, list) or not 1 <= len(items) <= MAX_ITEMS:
+            raise _Refused(bad_request(f"items must list 1 to {MAX_ITEMS} images"))
+        keys, sources = {}, {}
+        for item in items:
+            index = item.get("root_index") if isinstance(item, dict) else None
+            root_key, key = self._collection_key(item, roots, sources)
+            keys.setdefault((root_key, key), index)
+        return [(index, root_key, key) for (root_key, key), index in keys.items()]
+
+    def _collection_key(
+        self, item: object, roots: list, sources: dict
+    ) -> tuple[str, str]:
+        """The stored keys of one target, or raise its refusal (blocking).
+
+        *sources* caches ``{index: (scope, root key)}`` for one request, so
+        each root's realpath runs once however many items name it.
+        """
+        target, error = media_target(item, media_message=_MEDIA_MESSAGE)
+        if error is not None:
+            raise _Refused(error)
+        index, scope, relative = target
+        if index >= len(roots):
+            raise _Refused(_coded(409, "source-changed", SOURCES_CHANGED))
+        if index not in sources:
+            sources[index] = (
+                self._gallery_root_scope(roots[index], index),
+                _favorite_root_key(roots[index]),
+            )
+        current, root_key = sources[index]
+        if not hmac.compare_digest(scope, current):
+            raise _Refused(_coded(409, "source-changed", SOURCES_CHANGED))
+        key = _favorite_relative_key(relative)
+        if max(len(root_key), len(key)) > MAX_KEY_LENGTH:
+            # A stored key past the schema bound would fail the next restore.
+            raise _Refused(_forbidden("Path is too long to add to a collection"))
+        return root_key, key

py/api/duplicate_files.py

@@ -588,7 +588,9 @@ class DuplicateFilesMixin:
         job.update(total=len(targets), status_line="Removing duplicate files")
         spellings = await self._run_in_executor(self._source_spellings)
         removed, failure, keepers = 0, None, {}
-        totals = dict.fromkeys(("moved", "merged", "favorites", "tags", "review"), 0)
+        totals = dict.fromkeys(
+            ("moved", "merged", "favorites", "tags", "review", "collections"), 0
+        )
         totals["categories"] = 0  # image categories kept on the kept copies (#332)
         for done, target in enumerate(targets, start=1):
             if failure is not None:
@@ -632,6 +634,11 @@ class DuplicateFilesMixin:
             job.log_line(
                 f"Moved {_plural(totals['favorites'], 'favorite')} to the kept copies"
             )
+        if totals["collections"]:
+            job.log_line(
+                f"Moved {_plural(totals['collections'], 'collection membership')} "
+                "to the kept copies"
+            )
         if totals["tags"]:
             job.log_line(
                 f"Added {_plural(totals['tags'], 'image tag')} to the kept copies"

py/api/duplicate_spellings.py

@@ -104,7 +104,8 @@ def copy_move(target, kept_stat, spellings):
 
     A record found under a folder spelling moves to the kept copy under
     the same spelling when that folder holds it too, else to its real
-    path. Its favorite moves under every listed root holding the file.
+    path. Its favorite and its collection memberships move under every
+    listed root holding the file.
     """
     removed = spellings.of(target["abs"])
     kept = spellings.of(target["keeper_abs"])

py/api/history.py

@@ -5,6 +5,7 @@ import base64
 import binascii
 from functools import wraps
 import json
+import os
 import re
 import sqlite3
 from typing import Callable
@@ -27,6 +28,7 @@ from .history_current import (
     snapshot_current_output_scope,
 )
 from .inspection_service import CurrentFileChanged, inspect_current_file
+from .images import image_category_key
 
 try:
     from ...database.execution_history import HistoryEpochChanged
@@ -39,7 +41,11 @@ try:
         WindowPosition,
     )
     from ...utils.history_read_fence import HistoryReadFence
-    from ...utils.inspection_response import encode_history_inspection_response
+    from ...utils.inspection_response import (
+        OWN_CATEGORY_UNREAD,
+        apply_own_category,
+        encode_history_inspection_response,
+    )
 except ImportError:
     from database.execution_history import HistoryEpochChanged
     from database.history_gallery import read_history_gallery
@@ -51,7 +57,11 @@ except ImportError:
         WindowPosition,
     )
     from utils.history_read_fence import HistoryReadFence
-    from utils.inspection_response import encode_history_inspection_response
+    from utils.inspection_response import (
+        OWN_CATEGORY_UNREAD,
+        apply_own_category,
+        encode_history_inspection_response,
+    )
 
 
 _RUN_MAX_BYTES = 8 * 1024 * 1024
@@ -677,6 +687,32 @@ class HistoryRoutesMixin:
     ) -> web.Response:
         return await self._history_current(request, "inspection")
 
+    def _history_own_category(self, reference: dict, scope):
+        """The recorded output's own category (#332), inside the snapshot.
+
+        The path is the symlink-resolved output folder plus the receipt's
+        subfolder and filename, keyed by ``image_category_key`` over the
+        configured output sources as Images keys it. A temp output or a path
+        no source holds has none. A failed read is OWN_CATEGORY_UNREAD.
+        """
+        if (
+            reference.get("folder_type") != "output"
+            or scope.canonical_output is None
+            or candidate_locator(reference, scope) is None
+        ):
+            return None
+        path = os.path.join(
+            str(scope.canonical_output), reference["subfolder"], reference["filename"]
+        )
+        try:
+            key = image_category_key(self._get_all_output_dirs(), path)
+            return None if key is None else self.db.image_category(*key)
+        except Exception:
+            self.logger.warning(
+                "History image category could not be read", exc_info=True
+            )
+            return OWN_CATEGORY_UNREAD
+
     async def _history_current(
         self, request: web.Request, kind: str
     ) -> web.StreamResponse:
@@ -701,6 +737,7 @@ class HistoryRoutesMixin:
                     if scope.current_scope != current_scope:
                         raise _current_stale()
                     result = db.get_history_occurrence(run_id, node, ordinal)
+                    own = None
                     if result is not None:
                         receipt = result["receipt"]
                         path = candidate_locator(receipt["reference"], scope)
@@ -708,6 +745,10 @@ class HistoryRoutesMixin:
                             db.model.get_connection(),
                             (path,) if path is not None else (),
                         )
+                        if kind == "inspection":
+                            own = self._history_own_category(
+                                receipt["reference"], scope
+                            )
             finally:
                 db.model.close()
             if result is None:
@@ -726,9 +767,16 @@ class HistoryRoutesMixin:
                         resolved.canonical_target,
                         source_label=resolved.public["source_label"],
                         saved=saved.get("saved"),
-                        library=saved.get(
-                            "library",
-                            {"status": "unavailable" if saved_unavailable else "none"},
+                        library=apply_own_category(
+                            saved.get(
+                                "library",
+                                {
+                                    "status": (
+                                        "unavailable" if saved_unavailable else "none"
+                                    )
+                                },
+                            ),
+                            own,
                         ),
                         recorded=recorded_output_lookup(
                             db, close=True, reference=receipt["reference"]

py/api/image_categories.py

@@ -29,9 +29,11 @@ from .prompts import bad_request
 
 try:
     from ...database.image_categories import MAX_CATEGORY_LENGTH, MAX_KEY_LENGTH
+    from ...database.models import normalize_image_path
     from ...utils.validators import validate_category
 except ImportError:
     from database.image_categories import MAX_CATEGORY_LENGTH, MAX_KEY_LENGTH
+    from database.models import normalize_image_path
     from utils.validators import validate_category
 
 CATEGORY_MEDIA_MESSAGE = "Only media files can have a category"
@@ -143,7 +145,7 @@ class ImageCategoryRoutesMixin:
     # -- reading ---------------------------------------------------------------
 
     def _own_category_map(self, sources, listings) -> dict:
-        """``{root / listed path: own category}`` for the listed files (blocking).
+        """``{normalized listed path: own category}`` for the listed files (blocking).
 
         One primary-key range read per source; each stored row is matched to
         the listing of every selected root that holds its source, so the work
@@ -162,19 +164,30 @@ class ImageCategoryRoutesMixin:
         for _index, root, files in listings:
             listing = self._favorite_listing(root, files)
             root_key = _favorite_root_key(root)
+            # normalize_image_path(root / listed) without a call per file: the
+            # listed spellings are clean relative POSIX paths. Windows abspath
+            # also drops a component's trailing dots and spaces, so it keeps
+            # the call there.
+            base = os.path.join(normalize_image_path(str(root)), "")
+
+            def normalized(listed, base=base, root=root):
+                if os.name == "nt":
+                    return normalize_image_path(str(root / listed))
+                return base + os.path.normcase(listed.replace("/", os.sep))
+
             nested = [
                 item for item in _nested_sources(root, sources) if item[1] != root_key
             ]
             for relative, category in by_source.get(root_key, {}).items():
                 listed = listing.get(relative)
                 if listed is not None and _source_prefix(relative, nested) is None:
-                    found[root / listed] = category
+                    found[normalized(listed)] = category
             for prefix, source_key in nested:
                 for relative, category in by_source.get(source_key, {}).items():
                     full = f"{prefix}/{relative}"
                     listed = listing.get(full)
                     if listed is not None and _source_prefix(full, nested) == prefix:
-                        found[root / listed] = category
+                        found[normalized(listed)] = category
         return found
 
     async def _source_listings(self, roots) -> list[tuple]:

py/api/image_tag_jobs.py

@@ -42,7 +42,7 @@ from .. import model_gate
 from .admin import _plural, _time_left
 from .annotation_suggest import (
     OptionError,
-    unavailable_message,
+    model_refusal,
     validate_options,
 )
 from .annotations import (
@@ -58,7 +58,9 @@ from .annotations import (
 from .autotag_routes import BATCH_CANCELLED, MODEL_BUSY_ERROR, model_busy_response
 from .images import (
     MAX_LISTING_SELECTION,
+    CollectionMissing,
     ListingSourceChanged,
+    collection_not_found,
     _current_fingerprint,
     _listing_posix,
     _public_path,
@@ -510,10 +512,8 @@ class ImageTagJobsMixin:
             )
         )
         label = MODELS[model_type]["label"]
-        if not (status.get(model_type) or {}).get("downloaded"):
-            message = (
-                f"{label} is not downloaded. Download it in System › Auto-tagging."
-            )
+        message = model_refusal(status, model_type, label)
+        if message is not None:
             return None, _refusal(409, "model-unavailable", message)
         if backup_on and short is not None:
             return None, _refusal(507, "insufficient-space", short_space_message(short))
@@ -689,6 +689,8 @@ class ImageTagJobsMixin:
             selection = await self.resolve_listing_filter(spec.filter, cap)
         except ListingSourceChanged:
             return None, _coded("source-changed")
+        except CollectionMissing:
+            return None, collection_not_found()
         except ValueError as exc:
             return None, bad_request(str(exc))
         except Exception:
@@ -788,13 +790,13 @@ class ImageTagJobsMixin:
 
     async def _prepare_tagging_model(self, job, spec: _Spec, owner):
         """The loaded service, or None (logged) when the job cannot run."""
-        from ..autotag import MODELS, get_autotag_service
+        from ..autotag import MODELS, MissingBackendPackages, get_autotag_service
 
         service = get_autotag_service()
         status = await self._run_in_executor(service.get_models_status)
-        if not status.get(spec.model_type, {}).get("downloaded"):
-            name = MODELS.get(spec.model_type, {}).get("name", spec.model_type)
-            message = unavailable_message(name)
+        name = MODELS.get(spec.model_type, {}).get("name", spec.model_type)
+        message = model_refusal(status, spec.model_type, name)
+        if message is not None:
             job.update(status_line=message)
             job.log_line(f"! {message}")
             return None
@@ -806,6 +808,11 @@ class ImageTagJobsMixin:
                 spec.model_type,
                 spec.options["use_gpu"],
             )
+        except MissingBackendPackages as exc:
+            # A package gone since the status read (load_model logged it).
+            job.update(status_line=str(exc))
+            job.log_line(f"! {exc}")
+            return None
         except Exception:
             self.logger.exception("The auto-tag job could not load its model")
             job.update(status_line=INFERENCE_FAILED)

py/api/images.py

@@ -18,7 +18,7 @@ import urllib.parse
 from bisect import bisect_left, bisect_right, insort
 from collections.abc import Mapping
 from pathlib import Path
-from typing import Callable, NamedTuple
+from typing import Callable, Iterable, NamedTuple
 
 from aiohttp import web
 from PIL import Image, UnidentifiedImageError
@@ -58,6 +58,7 @@ from .inspection_service import (
 )
 
 try:
+    from ...database.image_collections import CollectionMissing
     from ...database.image_favorites import MAX_KEY_LENGTH
     from ...database.library import display_title
     from ...database.models import normalize_image_path
@@ -66,6 +67,8 @@ try:
     from ...utils.parallel import map_parallel
     from ...utils.json_strict import dumps_strict
     from ...utils.inspection_response import (
+        OWN_CATEGORY_UNREAD,
+        apply_own_category,
         encode_inspection_response,
         library_metadata,
     )
@@ -79,6 +82,7 @@ try:
         reveal_in_file_manager,
     )
 except ImportError:
+    from database.image_collections import CollectionMissing
     from database.image_favorites import MAX_KEY_LENGTH
     from database.library import display_title
     from database.models import normalize_image_path
@@ -87,6 +91,8 @@ except ImportError:
     from utils.parallel import map_parallel
     from utils.json_strict import dumps_strict
     from utils.inspection_response import (
+        OWN_CATEGORY_UNREAD,
+        apply_own_category,
         encode_inspection_response,
         library_metadata,
     )
@@ -106,6 +112,21 @@ IMAGE_EXTENSIONS = _IMAGE_EXTENSIONS
 _CARD_RECORDS = ("prompt", "workflow")
 _PNG_SIGNATURE = b"\x89PNG\r\n\x1a\n"
 _LISTING_FAILED = "The image list could not be filtered. Check server logs for details."
+_COLLECTION_ID = re.compile(r"[1-9][0-9]{0,8}")
+COLLECTION_NOT_FOUND = "That collection no longer exists."
+
+
+def collection_not_found() -> web.Response:
+    """The 404 a write or a listing answers for an unknown collection (#331)."""
+    return web.json_response(
+        {
+            "success": False,
+            "code": "collection-not-found",
+            "error": COLLECTION_NOT_FOUND,
+        },
+        status=404,
+        headers={"Cache-Control": "no-store"},
+    )
 
 
 def _card_header(path: Path) -> tuple:
@@ -565,7 +586,8 @@ class _ImageTagMatch(NamedTuple):
 
     tags: bool  # every filter tag is in its image tags or linked positive tags
     query: bool  # ``q`` is a case-folded substring of one image tag name
-    link: tuple | None  # (hit, rating, positive id) of its unambiguous link
+    # (hit, rating, positive id, inherited category) of its unambiguous link
+    link: tuple | None
 
 
 def _current_fingerprint(path: Path) -> tuple[int, int] | None:
@@ -1105,8 +1127,9 @@ def _workspace_filters(query: Mapping) -> dict:
     the linked positive's text, title and tag names. ``category`` is exact,
     ``tags`` is a comma list that must all be present, ``sort=rating`` ranks by
     the linked positive, and ``favorites=1`` keeps only favorites (the
-    Favorites page, #249 D51). Raises ValueError for any unknown or oversized
-    value.
+    Favorites page, #249 D51). ``collection=<id>`` (1 to 9 digits) keeps only
+    that collection's members (#331); it wins over ``favorites`` (a page never
+    sets both). Raises ValueError for any unknown or oversized value.
     """
     text = query.get("q", "").strip()
     folded = text.casefold()
@@ -1120,8 +1143,10 @@ def _workspace_filters(query: Mapping) -> dict:
         {tag.strip() for tag in query.get("tags", "").split(",") if tag.strip()}
     )
     favorites = query.get("favorites", "0")
+    collection = query.get("collection", "")
     if (
         favorites not in ("0", "1")
+        or (collection != "" and _COLLECTION_ID.fullmatch(collection) is None)
         or len(text) > 500
         or len(folded) > 500
         or media not in ("all", "image", "video")
@@ -1144,7 +1169,8 @@ def _workspace_filters(query: Mapping) -> dict:
         "path": exact.casefold(),
         "category": category,
         "tags": tags,
-        "favorites": favorites == "1",
+        "favorites": favorites == "1" and not collection,
+        "collection": int(collection) if collection else None,
     }
 
 
@@ -1162,6 +1188,8 @@ LISTING_FILTER_KEYS = frozenset(
         "media",
         "sort",
         "favorites",
+        # A collection page's filter (#331): bulk actions act on its members.
+        "collection",
     }
 )
 MAX_LISTING_FILTER_VALUE = 2048
@@ -1218,22 +1246,29 @@ def _admitted_entries(
     entries: list,
     filters: dict,
     links: dict | None,
-    favorites: frozenset | None = None,
+    members: frozenset | None = None,
     image_tags: dict | None = None,
+    own_categories: dict | None = None,
 ):
     """Yield ``(position, entry, link match)`` for each entry the filters admit.
 
-    *links* maps normalized paths to (hit, rating, positive id) for every
-    unambiguously linked image (see PromptDatabase.get_image_link_matches); a
-    category or tag filter admits only those, and ``q`` admits a filename match
-    or a linked-positive hit. *favorites*, when given, admits only those listed
-    paths (``favorites=1``). *image_tags* maps normalized paths to their
-    _ImageTagMatch: a tag filter also admits a file whose current image tags
-    (with its linked positive's tags) carry every tag, and ``q`` also admits an
-    image tag name hit.
+    *links* maps normalized paths to (hit, rating, positive id, inherited
+    category) for every unambiguously linked image (see
+    PromptDatabase.get_image_link_matches); a tag filter admits only those,
+    and ``q`` admits a filename match or a linked-positive hit. A category
+    admits a file whose effective category equals it (#332 MC5): its own
+    category (*own_categories*, ``{normalized path: category}``) when it has one,
+    else the inherited one, so an unlinked file with its own category is
+    admitted. *members*, when given, admits only those listed paths (the
+    favorites of ``favorites=1`` or a ``collection``'s members). *image_tags*
+    maps normalized paths to their _ImageTagMatch: a tag filter also admits a
+    file whose current image tags (with its linked positive's tags) carry every
+    tag, and ``q`` also admits an image tag name hit.
     """
     query, media, folder = filters["query"], filters["media"], filters["folder"]
-    members_only = filters["category"] is not None or bool(filters["tags"])
+    category = filters["category"]
+    members_only = bool(filters["tags"])
+    own = own_categories or {}
     native_folder = folder.replace("/", os.sep)
     nested = native_folder + os.sep
     exact = filters.get("path", "")
@@ -1246,12 +1281,18 @@ def _admitted_entries(
             continue
         if media != "all" and (media == "video") != entry.video:
             continue
-        if favorites is not None and entry.path not in favorites:
+        if members is not None and entry.path not in members:
             continue
         match = None if links is None else links.get(entry.normalized)
         image = None if image_tags is None else image_tags.get(entry.normalized)
         if members_only and match is None and not (image and image.tags):
             continue
+        if category is not None:
+            effective = own.get(entry.normalized) if own else None
+            if effective is None and match is not None:
+                effective = match[3]
+            if effective != category:
+                continue
         if (
             query
             and query not in entry.folded
@@ -1268,8 +1309,9 @@ def _filtered_page(
     size: int,
     filters: dict,
     links: dict | None,
-    favorites: frozenset | None = None,
+    members: frozenset | None = None,
     image_tags: dict | None = None,
+    own_categories: dict | None = None,
 ) -> tuple[int, list, int | None]:
     """Count the *entries* the filters admit and take *size* from *start* on.
 
@@ -1279,7 +1321,7 @@ def _filtered_page(
     """
     total, page, positives = 0, [], set()
     for position, entry, match in _admitted_entries(
-        entries, filters, links, favorites, image_tags
+        entries, filters, links, members, image_tags, own_categories
     ):
         total += 1
         if match is not None:
@@ -1983,6 +2025,7 @@ class ImageRoutesMixin:
                         filters["category"],
                         filters["tags"],
                         filters["favorites"],
+                        filters["collection"],
                     ]
                 ).encode()
             ).hexdigest()
@@ -1993,17 +2036,19 @@ class ImageRoutesMixin:
         except ValueError as exc:
             return bad_request(str(exc))
 
-        listings = [
-            (index, root, await self._get_gallery_files(root))
-            for index, root in enumerate(output_dirs)
-            if root_index is None or index == root_index
-        ]
         try:
+            if filters["collection"] is not None and not await self._run_in_executor(
+                self.db.collection_exists, filters["collection"]
+            ):
+                raise CollectionMissing("No such collection")
+            listings = await self._workspace_listings(output_dirs, root_index)
             page, total, has_more, next_cursor, positive_count = (
                 await self._run_in_executor(
                     self._select_workspace_page, listings, filters, after, limit, scope
                 )
             )
+        except CollectionMissing:
+            return collection_not_found()
         except Exception:
             # A filter that needs the prompt links or the image tags cannot be
             # served without them. Never echo the exception: it can carry SQL
@@ -2219,9 +2264,10 @@ class ImageRoutesMixin:
         stat'ed (never every listed file), and a candidate's tags count only
         while a byte record is verified at its present ``(size, mtime_ns)``
         (spec amendment S1). Under a Tags filter the candidates' linked
-        positives are read too, so their tags and category combine.
+        positives are read too, so their tags combine; the category is checked
+        once, on the effective category, by _admitted_entries (#332).
         """
-        tags, query, category = filters["tags"], filters["query"], filters["category"]
+        tags, query = filters["tags"], filters["query"]
         wanted = set(tags)
         located = self._workspace_locators(listings)
         # The tag index answers which files carry a relevant name (#379); no
@@ -2251,11 +2297,9 @@ class ImageRoutesMixin:
             link = facts.get(entry.normalized)
             carried = set(names).union(link[4] if link else ())
             matches[entry.normalized] = _ImageTagMatch(
-                bool(wanted)
-                and wanted <= carried
-                and (category is None or (link is not None and link[3] == category)),
+                bool(wanted) and wanted <= carried,
                 bool(query) and any(query in name.casefold() for name in names),
-                link[:3] if link else None,
+                tuple(link[:4]) if link else None,
             )
         return matches
 
@@ -2264,10 +2308,16 @@ class ImageRoutesMixin:
     ) -> tuple:
         """The ordered entries and what admits them (blocking; run in executor).
 
-        Returns ``(entries, key, links, image_tags, favorites)``: the listing in
-        the filters' order, its keyset key, and the link, image-tag and favorite
-        maps _admitted_entries reads. One SELECT (get_image_link_matches) reads
-        every linked positive when *read_links*. A failed read raises when the
+        Returns ``(entries, key, links, image_tags, members, own)``: the
+        listing in the filters' order, its keyset key, and the link, image-tag,
+        member and own-category maps _admitted_entries reads (members: a
+        collection's, else the favorites, else None). A collection deleted since
+        the route checked it raises CollectionMissing. One SELECT
+        (get_image_link_matches) reads every linked positive when *read_links*;
+        under a category it keeps only links whose raw recipe or positive
+        category is that value (the inherited one is decided in Python), and
+        the links of files whose own category is that value are read by path,
+        so their rating and positive still count (#332). A failed read raises when the
         filters need it (category, tags, the rating sort) or *strict*; else the
         listing stays usable without links (a search then matches filenames
         only). A failed image-tag read raises under a Tags filter; a search then
@@ -2279,10 +2329,10 @@ class ImageRoutesMixin:
             try:
                 links = self.db.get_image_link_matches(
                     filters["text"],
-                    category,
                     tags,
                     every_link=True,
                     wanted=self._workspace_wanted(listings),
+                    category=category,
                 )
             except Exception:
                 if strict or category is not None or tags or order == "rating":
@@ -2296,6 +2346,13 @@ class ImageRoutesMixin:
                     ),
                     exc_info=True,
                 )
+        own = None
+        if category is not None:
+            # A failed read raises: an unfiltered page must never pass for the
+            # category (DC6).
+            own = self._own_category_map(self._get_all_output_dirs(), listings)
+            if links is not None:
+                links = self._own_category_links(links, own, filters)
         image_tags = None
         if filters["query"] or tags:
             try:
@@ -2339,10 +2396,37 @@ class ImageRoutesMixin:
             def key(entry):
                 return _workspace_key(entry, order)
 
-        favorites = (
-            self._workspace_favorite_paths(listings) if filters["favorites"] else None
+        return (
+            entries,
+            key,
+            links,
+            image_tags,
+            self._workspace_members(listings, filters),
+            own,
         )
-        return entries, key, links, image_tags, favorites
+
+    def _own_category_links(self, links: dict, own: dict, filters: dict) -> dict:
+        """*links* plus the links of files whose own category is the filter's.
+
+        Their inherited category differs, so the category's link read left
+        them out; read by path (get_image_link_facts), kept only when their
+        positive has every filter tag, as the link read would.
+        """
+        category, wanted = filters["category"], set(filters["tags"])
+        missing = [
+            path
+            for path, value in own.items()
+            if value == category and path not in links
+        ]
+        if not missing:
+            return links
+        facts = self.db.get_image_link_facts(missing, filters["text"])
+        extra = {
+            path: tuple(fact[:4])
+            for path, fact in facts.items()
+            if wanted <= set(fact[4])
+        }
+        return {**links, **extra} if extra else links
 
     def _select_workspace_page(
         self,
@@ -2371,7 +2455,7 @@ class ImageRoutesMixin:
             or bool(filters["tags"])
             or order == "rating"
         )
-        entries, key, links, image_tags, favorites = self._workspace_admission(
+        entries, key, links, image_tags, members, own = self._workspace_admission(
             listings, filters, read_links=needed or after is None
         )
         start = 0
@@ -2382,7 +2466,7 @@ class ImageRoutesMixin:
             else:
                 start = bisect_right(entries, after, key=key)
         total, page, positive_count = _filtered_page(
-            entries, start, limit + 1, filters, links, favorites, image_tags
+            entries, start, limit + 1, filters, links, members, image_tags, own
         )
         if not needed and after is not None:
             positive_count = _COUNT_OMITTED
@@ -2404,12 +2488,12 @@ class ImageRoutesMixin:
 
         The links are read strictly: the positives must be the listing's own.
         """
-        entries, _key, links, image_tags, favorites = self._workspace_admission(
+        entries, _key, links, image_tags, members, own = self._workspace_admission(
             listings, filters, read_links=True, strict=True
         )
         targets, positives, unlinked, scopes = [], {}, 0, {}
         for _position, entry, match in _admitted_entries(
-            entries, filters, links, favorites, image_tags
+            entries, filters, links, members, image_tags, own
         ):
             if len(targets) == cap:
                 return ListingSelection(targets, list(positives), unlinked, True)
@@ -2431,9 +2515,11 @@ class ImageRoutesMixin:
     ) -> ListingSelection:
         """The images a parse_listing_filter result admits right now (#392).
 
-        Raises ValueError for a root outside the current roots, and
+        Raises ValueError for a root outside the current roots,
         ListingSourceChanged when the filter's ``root_scope`` no longer names
-        its root (the sources were renumbered since the listing).
+        its root (the sources were renumbered since the listing), and
+        CollectionMissing when its ``collection`` no longer exists (checked
+        before any scan and again after the members are read).
         """
         output_dirs = list(self._get_all_output_dirs())
         root_index = _workspace_root(params, output_dirs)
@@ -2443,11 +2529,11 @@ class ImageRoutesMixin:
                 raise ListingSourceChanged()
             raise ValueError("Invalid gallery root scope")
         filters = _workspace_filters(params)
-        listings = [
-            (index, root, await self._get_gallery_files(root))
-            for index, root in enumerate(output_dirs)
-            if root_index is None or index == root_index
-        ]
+        if filters["collection"] is not None and not await self._run_in_executor(
+            self.db.collection_exists, filters["collection"]
+        ):
+            raise CollectionMissing("No such collection")
+        listings = await self._workspace_listings(output_dirs, root_index)
         return await self._run_in_executor(
             self._select_workspace_matches, listings, filters, cap
         )
@@ -3632,25 +3718,58 @@ class ImageRoutesMixin:
         self._favorite_listing_cache[cache_key] = (files, listing)
         return listing
 
-    def _workspace_favorite_paths(self, listings: list[tuple]) -> frozenset:
-        """The listed paths of every favorite in the selected roots (blocking).
+    def _workspace_members(self, listings: list[tuple], filters: dict):
+        """The member set the filters ask for, or None (blocking)."""
+        collection = filters.get("collection")
+        if collection is not None:
+            members = self._workspace_member_paths(
+                listings,
+                lambda root_key: self.db.collection_members(collection, root_key),
+            )
+            # Checked after the reads: a collection deleted before or while
+            # they ran is a 404, never an empty page passing for it.
+            if not self.db.collection_exists(collection):
+                raise CollectionMissing("No such collection")
+            return members
+        if filters["favorites"]:
+            return self._workspace_favorite_paths(listings)
+        return None
 
-        One primary-key read per root (get_root_favorites), matched against the
-        root's cached ``{relative key: listing spelling}``, so the work follows
-        the favorites rather than the files, and a key folds case exactly as
-        the toggle stored it (Windows). A favorite whose file is gone matches
-        nothing. A failed read raises: an unfiltered page must never pass for
-        the favorites.
+    def _workspace_favorite_paths(self, listings: list[tuple]) -> frozenset:
+        """The listed paths of every favorite in the selected roots (blocking)."""
+        return self._workspace_member_paths(listings, self.db.get_root_favorites)
+
+    def _workspace_member_paths(
+        self, listings: list[tuple], keys_for_source: Callable[[str], Iterable[str]]
+    ) -> frozenset:
+        """The listed paths of the stored keys of each selected root (blocking).
+
+        One read per root (``keys_for_source(root key)``: the favorites' or a
+        collection's primary-key range), matched against the root's cached
+        ``{relative key: listing spelling}``, so the work follows the members
+        rather than the files, and a key folds case exactly as it was stored
+        (Windows). A member whose file is gone matches nothing. A failed read
+        raises: an unfiltered page must never pass for the members.
         """
         paths = set()
         for _index, root, files in listings:
             listing = self._favorite_listing(root, files)
-            for key in self.db.get_root_favorites(_favorite_root_key(root)):
+            for key in keys_for_source(_favorite_root_key(root)):
                 listed = listing.get(key)
                 if listed is not None:
                     paths.add(root / listed)
         return frozenset(paths)
 
+    async def _workspace_listings(
+        self, output_dirs: list[Path], root_index: int | None
+    ) -> list[tuple]:
+        """``(index, root, scan)`` of every selected root (cached scans)."""
+        return [
+            (index, root, await self._get_gallery_files(root))
+            for index, root in enumerate(output_dirs)
+            if root_index is None or index == root_index
+        ]
+
     def _workspace_favorites(self, page: list[tuple]) -> set[tuple[int, str]]:
         """``(root index, listing path)`` of the favorites among *page* entries."""
         by_root: dict[tuple[int, Path], dict[str, str]] = {}
@@ -3728,13 +3847,20 @@ class ImageRoutesMixin:
                     except Exception:
                         # The link stays known; only its title is withheld.
                         self.logger.warning("Inspection link title could not be read")
+                # An output file's own category (#332); LoRA roots have none.
+                own = (
+                    self._read_own_category(outputs, path)
+                    if index < len(outputs)
+                    else None
+                )
                 try:
                     observed = inspect_current_file(
                         path,
                         source_label=_public_path(roots[index]),
                         saved=saved,
-                        library=self._inspection_library(
-                            saved, unavailable=unavailable
+                        library=apply_own_category(
+                            self._inspection_library(saved, unavailable=unavailable),
+                            own,
                         ),
                         image_tags=self._inspection_image_tags(
                             roots[index], Path(relative).parts
@@ -3802,6 +3928,19 @@ class ImageRoutesMixin:
 
         return read
 
+    def _read_own_category(self, sources, path):
+        """One file's own category, None, or OWN_CATEGORY_UNREAD (blocking).
+
+        Keyed by ``image_category_key`` over the output sources, as Images
+        writes it. A failed read is logged and hides the category (DC5).
+        """
+        try:
+            key = image_category_key(sources, path)
+            return None if key is None else self.db.image_category(*key)
+        except Exception:
+            self.logger.warning("Image category could not be read", exc_info=True)
+            return OWN_CATEGORY_UNREAD
+
     def _inspection_library(self, saved: dict | None, *, unavailable: bool) -> dict:
         """The inspector's Library facts; the route sets the title afterwards."""
         result = {

py/api/prompts.py

@@ -1227,7 +1227,9 @@ class PromptRoutesMixin:
         """
         from .images import (
             MAX_LISTING_SELECTION,
+            CollectionMissing,
             ListingSourceChanged,
+            collection_not_found,
             parse_listing_filter,
         )
 
@@ -1252,6 +1254,8 @@ class PromptRoutesMixin:
                 },
                 status=409,
             )
+        except CollectionMissing:
+            return collection_not_found()
         except ValueError as exc:
             return refused(str(exc))
         if selection.truncated:

py/autotag.py

@@ -7,6 +7,7 @@ Refactored from standalone_tagger.py for integration with PromptManager API.
 
 import errno
 import gc
+import importlib.util
 import inspect
 import os
 import sys
@@ -155,6 +156,91 @@ MODELS = {
     },
 }
 
+# The Python modules each backend imports when it loads, with their pip names
+# (#424). None of them is a PromptManager requirement: they are optional, and
+# llama-cpp-python needs a build for the user's GPU. Nothing is ever installed.
+BACKEND_PACKAGES: Dict[str, Tuple[Tuple[str, str], ...]] = {
+    "gguf": (("llama_cpp", "llama-cpp-python"),),
+    # The default 8-bit load needs bitsandbytes, and device_map needs accelerate.
+    "hf": (
+        ("torch", "torch"),
+        ("transformers", "transformers"),
+        ("accelerate", "accelerate"),
+        ("bitsandbytes", "bitsandbytes"),
+    ),
+    "wd14-swinv2": (("onnxruntime", "onnxruntime"),),
+    "wd14-vit": (("onnxruntime", "onnxruntime"),),
+}
+_INSTALL_NOTES = {
+    "gguf": (
+        'For a CUDA build: CMAKE_ARGS="-DGGML_CUDA=on" pip install llama-cpp-python'
+    ),
+    "wd14-swinv2": "Use onnxruntime-gpu for CUDA.",
+    "wd14-vit": "Use onnxruntime-gpu for CUDA.",
+}
+_PIP_NAMES = {
+    module: pip for packages in BACKEND_PACKAGES.values() for module, pip in packages
+}
+
+
+def _importable(module: str) -> bool:
+    """Whether *module* (a top-level name) can be imported; imports nothing."""
+    if sys.modules.get(module) is not None:
+        return True
+    try:
+        return importlib.util.find_spec(module) is not None
+    except (ImportError, ValueError, OSError):
+        return False
+
+
+def missing_packages(model_type: str) -> List[str]:
+    """The pip names of the packages *model_type* needs and cannot import."""
+    return [
+        pip
+        for module, pip in BACKEND_PACKAGES.get(model_type, ())
+        if not _importable(module)
+    ]
+
+
+def package_fields(model_type: str, missing: List[str]) -> Dict[str, Any]:
+    """The status fields of *missing*: pip names, install command and note."""
+    missing = list(missing)
+    return {
+        "missing_packages": missing,
+        "install_command": "pip install " + " ".join(missing) if missing else None,
+        "install_note": _INSTALL_NOTES.get(model_type) if missing else None,
+    }
+
+
+def _joined(names: List[str]) -> str:
+    if len(names) <= 1:
+        return "".join(names)
+    return ", ".join(names[:-1]) + " and " + names[-1]
+
+
+def missing_packages_message(model_type: str, missing: List[str]) -> str:
+    """The one sentence every refusal of a model with missing packages uses."""
+    label = MODELS.get(model_type, {}).get("label", model_type)
+    fields = package_fields(model_type, missing)
+    pronoun = "it" if len(missing) == 1 else "them"
+    message = (
+        f"{label} needs {_joined(missing)}. Install {pronoun} into ComfyUI's "
+        f"Python: {fields['install_command']}."
+    )
+    if fields["install_note"]:
+        message += " " + fields["install_note"]
+    return message
+
+
+class MissingBackendPackages(RuntimeError):
+    """A model whose backend packages cannot be imported (#424)."""
+
+    def __init__(self, model_type: str, missing: List[str]):
+        self.model_type = model_type
+        self.missing = list(missing)
+        super().__init__(missing_packages_message(model_type, self.missing))
+
+
 # Default prompts
 DEFAULT_PROMPT = DEFAULT_TAG_TEMPLATE
 
@@ -167,6 +253,10 @@ SYSTEM_PROMPT = (
 )
 
 
+def _none(_model_type: str) -> List[str]:
+    return []
+
+
 def _load_rgb(image_path: str) -> Image.Image:
     """Decode the image as RGB and close the file before inference.
 
@@ -203,6 +293,7 @@ class AutoTagService:
         self,
         models_dir: Optional[Path] = None,
         model_loader: Optional[ModelLoader] = None,
+        package_probe: Optional[Callable[[str], List[str]]] = None,
     ):
         """Initialize the AutoTag service.
 
@@ -213,9 +304,16 @@ class AutoTagService:
                        ``(kind, backend)`` tuple used by ``generate_tags``.
                        Defaults to ``_load_model``, which imports the ML
                        backends lazily. Tests inject a fake here.
+            package_probe: Callable ``(model_type)`` returning the pip names
+                       of the backend packages that cannot be imported.
+                       Defaults to ``missing_packages`` with the default
+                       loader; an injected loader brings its own backend.
         """
         self.logger = get_logger("autotag.service")
         self._model_loader: ModelLoader = model_loader or self._load_model
+        if package_probe is None:
+            package_probe = missing_packages if model_loader is None else _none
+        self._package_probe = package_probe
 
         # Determine models directory
         if models_dir:
@@ -320,6 +418,9 @@ class AutoTagService:
                     # Get the actual path (local or cache)
                     model_status["model_path"] = str(self._get_hf_model_path())
 
+            model_status.update(
+                package_fields(model_type, self._package_probe(model_type))
+            )
             status[model_type] = model_status
 
         return status
@@ -455,13 +556,9 @@ class AutoTagService:
 
     def _load_wd14_tagger(self, model_type: str, use_gpu: bool = True):
         """Load WD14 ONNX model and tag list."""
-        try:
-            import onnxruntime as ort
-        except ImportError:
-            raise RuntimeError(
-                "onnxruntime is required for WD14 models. "
-                "Install with: pip install onnxruntime (or onnxruntime-gpu for CUDA)"
-            )
+        # A missing onnxruntime is refused by load_model before this runs
+        # (#424); a ModuleNotFoundError here becomes the same refusal.
+        import onnxruntime as ort
         import csv
 
         config = MODELS[model_type]
@@ -694,6 +791,7 @@ class AutoTagService:
         Raises:
             ValueError: If model_type is invalid
             RuntimeError: If model not downloaded or loading fails
+            MissingBackendPackages: If its backend packages cannot be imported
         """
         if model_type not in MODELS:
             raise ValueError(f"Invalid model type: {model_type}")
@@ -705,6 +803,11 @@ class AutoTagService:
         status = self.get_models_status()
         if not status[model_type]["downloaded"]:
             raise RuntimeError(f"Model {model_type} not downloaded")
+        missing = status[model_type].get("missing_packages") or []
+        if missing:
+            refusal = MissingBackendPackages(model_type, missing)
+            self.logger.warning(str(refusal))
+            raise refusal
 
         try:
             self._tagger = self._model_loader(model_type, use_gpu)
@@ -712,6 +815,17 @@ class AutoTagService:
             self.logger.info(f"Model {model_type} loaded successfully")
             return True
 
+        except ModuleNotFoundError as e:
+            # A package find_spec found but that does not import (a broken
+            # install), or a dependency of one: the same refusal, no traceback.
+            self._tagger = None
+            self._current_model_type = None
+            module = (e.name or "").split(".")[0] or str(e)
+            refusal = MissingBackendPackages(
+                model_type, [_PIP_NAMES.get(module, module)]
+            )
+            self.logger.warning(str(refusal))
+            raise refusal from None
         except Exception as e:
             self.logger.error(f"Failed to load model {model_type}: {e}")
             self._tagger = None

tests/browser/classic-node.spec.mjs

@@ -19,7 +19,7 @@ test.beforeEach(async ({page, browserName}) => {
         await import('/extensions/promptmanager/prompt_manager.js');
         class Node {
             constructor(){
-                this.widgets=['text','category','tags','search_text'].map(name=>({name,value:'',inputEl:document.createElement('textarea')}));
+                this.widgets=[...['text','category','tags','search_text'].map(name=>({name,value:'',inputEl:document.createElement('textarea')})),{name:'save_to_library',type:'toggle',value:true}];
                 this.widgets_values=[]; this.inputs=[];
             }
             addDOMWidget(name,type,element){document.body.append(element);return {element};}
@@ -36,6 +36,14 @@ test.afterEach(async ({page,browserName}) => {
     expect(page.classicErrors).toEqual([]);
 });
 
+test('Classic loads an older workflow with Save to library on and keeps a saved off (#90)',async({page})=>{
+    // A workflow saved before the switch held the search UI's slot ('') at the switch's index
+    const loaded=await page.evaluate(()=>{const widget=node.widgets.find(w=>w.name==='save_to_library');widget.value='';node.configure({});return widget.value;});
+    expect(loaded).toBe(true);
+    const off=await page.evaluate(()=>{const widget=node.widgets.find(w=>w.name==='save_to_library');widget.value=false;node.configure({});return [widget.value,node.serialize().widgets_values[4]];});
+    expect(off).toEqual([false,false]);
+});
+
 test('Classic keeps immediate selection, search query and metadata serialization',async({page})=>{
     await page.route('**/prompt_manager/search?**', route=>route.fulfill({json:{success:true,results:[{text:'A new portrait',category:'Portrait',tags:['film grain'],rating:5}]}}));
     await page.evaluate(()=>{node.widgets.find(w=>w.name==='search_text').value='tag:"film grain" portrait';});

tests/browser/concept-inventory.mjs

@@ -187,6 +187,33 @@ export async function mutate(page,{css=null,remove=[],clone=[]}={}) {
     },token);
 }
 
+/**
+ * #332 D146: the category chip names its origin after the category, in a muted `span.chip-source` that never splits
+ * (workspace.css `.chip .chip-source`). The concept's chip shows the bare name (app.js a402), so the recorded rule and
+ * the span are added to the concept page at runtime and the browser lays the chip out: the first `.tags` row's
+ * non-badge chip under `root`. `category` also replaces the chip's text (an own category the concept never shows).
+ * Returns the undo.
+ */
+export const D146_CHIP_SOURCE='.chip .chip-source { color: var(--ca-muted); white-space: nowrap; }';
+export async function withChipSource(page,root,origin,category=null) {
+    const restoreRule=await mutate(page,{css:D146_CHIP_SOURCE});
+    const token=await page.evaluate(({root,origin,category})=>{
+        const chip=document.querySelector(root)?.querySelector('.tags')?.querySelector(':scope > .chip:not(.blue)');
+        if(!chip)throw new Error(`withChipSource: no category chip under ${root}`);
+        const text=chip.textContent,source=document.createElement('span');
+        source.className='chip-source';source.textContent=` · ${origin}`;
+        if(category!==null)chip.textContent=category;
+        chip.append(source);
+        const token=`chip-source-${Math.random().toString(36).slice(2)}`;
+        (window.__chipSources??=new Map()).set(token,()=>{chip.textContent=text;});
+        return token;
+    },{root,origin,category});
+    return async()=>{
+        await page.evaluate(token=>{window.__chipSources.get(token)();window.__chipSources.delete(token);},token);
+        await restoreRule();
+    };
+}
+
 /**
  * Inventory sections: {chrome, heading, toolbar, filterRow, bulkbar, caption, grid, cards, continuation, panel, library,
  * history, noPageScroll}. A section is null when its element is absent. The Images sections (toolbar … panel, #249 C7)
@@ -615,9 +642,12 @@ export function measureInspector(selector) {
             primary:node.classList.contains('ca-primary'),...box(node),...font(node),...edges(node),...layout(node),
             minHeight:style.minHeight,minWidth:style.minWidth,icon:iconOf(node)};
     }
+    // #332 D146: a category chip's visible origin (`span.chip-source`), null on every other chip.
     function chipOf(node) {
+        const source=node.querySelector(':scope > .chip-source');
         return {tag:node.tagName.toLowerCase(),className:classes(node),text:textOf(node),hidden:hiddenOf(node),...box(node),...font(node),...edges(node),
-            minHeight:getComputedStyle(node).minHeight};
+            minHeight:getComputedStyle(node).minHeight,
+            source:when(source,()=>({text:source.textContent,...font(source),...css(source,'whiteSpace'),...box(source)}))};
     }
     function rowOf(row) {
         return {...layout(row),...edges(row),...box(row),chips:[...row.children].map(chipOf)};

tests/browser/focused-node.spec.mjs

@@ -26,7 +26,7 @@ test.beforeEach(async({page,browserName})=>{
         window.fixtureApp={registerExtension(extension){this.extension=extension;},ui:{settings:new EventTarget()},graph:{setDirtyCanvas(){}},canvas:{}};
         await import('/extensions/promptmanager/prompt_manager.js');
         class Node {
-            constructor(){this.widgets=['text','category','tags','search_text'].map(name=>({name,type:'text',value:'',options:{},inputEl:document.createElement('textarea')}));this.inputs=[...['clip','prepend_text','append_text'].map(name=>({name,type:name==='clip'?'CLIP':'STRING',link:null})),...['text','category','tags','search_text'].map(name=>({name,type:'STRING',link:null,widget:{name}}))];this.properties={};this.size=[550,700];}
+            constructor(){this.widgets=[...['text','category','tags','search_text'].map(name=>({name,type:'text',value:'',options:{},inputEl:document.createElement('textarea')})),{name:'save_to_library',type:'toggle',value:true,options:{}}];this.inputs=[...['clip','prepend_text','append_text'].map(name=>({name,type:name==='clip'?'CLIP':'STRING',link:null})),...['text','category','tags','search_text'].map(name=>({name,type:'STRING',link:null,widget:{name}})),{name:'save_to_library',type:'BOOLEAN',link:null,widget:{name:'save_to_library'}}];this.properties={};this.size=[550,700];}
             findInputByType(type){const index=this.inputs.findIndex(input=>input.type===type&&input.link==null);return index<0?undefined:{index,slot:this.inputs[index]};}
             addWidget(type,name,value,callback,options){const widget={type,name,value,callback,options};this.widgets.push(widget);return widget;}
             addDOMWidget(name,type,element){const host=document.createElement('div');host.style.height='650px';host.className='fixture-dom-widget';host.append(element);document.body.append(host);return {element};}
@@ -41,6 +41,32 @@ test.beforeEach(async({page,browserName})=>{
 });
 test.afterEach(async({page,browserName})=>{await saveCoverage(page,browserName);expect(page.focusedErrors).toEqual([]);});
 
+test(`${variant} Save to library switch saves its value and loads older workflows on (#90)`, async ({page}) => {
+    const toggle=page.getByRole('checkbox',{name:'Save to library',exact:true});
+    await expect(toggle).toBeChecked();
+    expect(await page.evaluate(()=>node.widgets.map(widget=>widget.name))).toEqual(['text','category','tags','search_text','prepend_text','append_text','save_to_library']);
+    // The native switch is drawn in the toolbar only: the host keeps it hidden and zero-height, like the text fallbacks
+    expect(await page.evaluate(()=>{const widget=node.widgets.find(item=>item.name==='save_to_library');return {type:widget.type,hidden:widget.hidden,size:widget.computeSize()};})).toEqual({type:'hidden',hidden:true,size:[0,-4]});
+    expect(await page.evaluate(()=>node.widgets.find(item=>item.name==='text').computeSize())).toEqual([0,-4]);
+    await toggle.uncheck();
+    await expect(page.locator('.pm-focused > .pm-editor-status[role=status]')).toHaveText('Save to library is off: runs of this node are not added to the library.');
+    const saved=await page.evaluate(()=>node.serialize());
+    expect(saved.widgets_values[6]).toBe(false);
+    await page.evaluate(saved=>node.configure(saved),saved);
+    await expect(toggle).not.toBeChecked();
+    await page.getByLabel('Main prompt',{exact:true}).fill('Typed while off');
+    await expect(page.locator('.pm-focused > .pm-editor-status[role=status]')).toHaveText('Save to library is off: runs of this node are not added to the library.');
+    // A workflow saved before the switch held the editor's own slot ('') at its index
+    await page.evaluate(()=>node.configure({widgets_values:['Old text','','','','','',''],properties:{}}));
+    await expect(toggle).toBeChecked();
+    expect(await page.evaluate(()=>node.widgets.find(widget=>widget.name==='save_to_library').value)).toBe(true);
+    await toggle.check();
+    await page.getByLabel('Main prompt',{exact:true}).fill('Typed while on');
+    await expect(page.locator('.pm-focused > .pm-editor-status[role=status]')).toHaveText('Main prompt is saved when this node runs.');
+    await page.evaluate(()=>{node.inputs.find(input=>input.name==='save_to_library').link=9;node.onConnectionsChange?.();});
+    await expect(toggle).toBeDisabled();
+});
+
 test(`${variant} library autocompletes filters and searches without changing editor metadata`, async ({page}) => {
     const requests=[];
     await page.route('**/prompt_manager/categories',route=>route.fulfill({json:{success:true,categories:['Portrait']}}));
@@ -257,7 +283,7 @@ test(`${variant} restores workflow values and disclosures and protects connected
 test(`${variant} offers text as a link socket, keeps typed text while linked and through a reload (#343)`,async({page})=>{
     const sockets=()=>page.evaluate(()=>Object.fromEntries(node.inputs.map(input=>[input.name,input.widget?.name??null])));
     // The hidden text widget no longer owns the input, so the host draws it as a socket; the other widget inputs are unchanged.
-    expect(await sockets()).toEqual({clip:null,prepend_text:null,append_text:null,text:null,category:'category',tags:'tags',search_text:'search_text'});
+    expect(await sockets()).toEqual({clip:null,prepend_text:null,append_text:null,text:null,category:'category',tags:'tags',search_text:'search_text',save_to_library:'save_to_library'});
     expect(await page.evaluate(()=>node.findInputByType('STRING').slot.name)).toBe('text');
     expect(await page.evaluate(()=>node.findInputByType('CLIP').slot.name)).toBe('clip');
     const main=page.getByRole('textbox',{name:'Main prompt',exact:true});

tests/browser/full-inspector.spec.mjs

@@ -136,7 +136,8 @@ test('panel matches the concept order and copy',async({page,request})=>{
     await expect(panel.locator('.detail-inner > p.small')).toHaveText('complete.png');
     const badges=panel.locator('.detail-inner > .tags').first();
     await expect(badges.locator('span.chip.blue')).toHaveText('Workflow inferred');
-    await expect(badges.locator('span.chip:not(.blue)')).toHaveText('Recipe category: Inspection');
+    // #332 MC8: the chip names where its value came from, in visible text after the category.
+    await expect(badges.locator('span.chip:not(.blue)')).toHaveText('Recipe category: Inspection · recipe');
     await expect(badges.locator('span.chip:not(.blue) .ca-visually-hidden')).toHaveText('Recipe category: ');
     // Recipe tags are not the positive tags the Images Tags filter matches, so they stay plain chips (D21 ruling).
     await expect(panel.locator('.detail-inner > .tags').last().locator('span.chip')).toHaveText(['fixture','geometric']);
@@ -185,6 +186,53 @@ test('panel matches the concept order and copy',async({page,request})=>{
     await expect(panel.locator('.detail-inner > button')).toHaveText(['Retry prompt details']);
 });
 
+// #332 CUJ 4, CUJ 5 (MC8, D146): every category chip names its origin: the image's own, the prompt's or the recipe's.
+const originChip=scope=>scope.locator('span.chip:has(> span.chip-source)');
+async function expectOrigin(scope,{prefix,category,origin}) {
+    await expect(originChip(scope)).toHaveText(`${prefix}${category} · ${origin}`);
+    await expect(originChip(scope).locator('.ca-visually-hidden')).toHaveText(prefix);
+    expect(await originChip(scope).locator('.chip-source').evaluate(node=>node.textContent)).toBe(` · ${origin}`);
+}
+async function putOwnCategory(request,paths,category) {
+    const {images}=await(await request.get('/prompt_manager/images/output?mode=workspace&subfolder=inspector&limit=40')).json();
+    const items=paths.map(path=>images.find(image=>image.root_index===0&&image.path===path)).map(({root_index,root_scope,path})=>({root_index,root_scope,path}));
+    const response=await request.put('/prompt_manager/images/category',{data:{items,category}});await expect(response).toBeOK();
+}
+
+test('the category chip names its origin in the side panel and the viewer (#332 CUJ 4, CUJ 5)',async({page,request})=>{
+    await page.setViewportSize({width:1600,height:1100});
+    await expect(await request.post('/__test__/reset',{data:{full_inspector:true}})).toBeOK();
+    // The positive's own category, set through the legacy prompt route; complete.png's recipe keeps "Inspection".
+    const list=await(await request.get('/prompt_manager/library')).json();const id=list.items.find(item=>item.text.startsWith('Inspector portrait')).id;
+    await expect(await request.post('/prompt_manager/bulk/category',{data:{prompt_ids:[id],category:'Landscapes'}})).toBeOK();
+    await page.goto('/prompt_manager/workspace#/images');const panel=inspector(page);
+    await selectFile(page,'complete.png');await expect(panel.locator('.panel-heading h2')).toHaveText(POSITIVE_TITLE);
+    await expectOrigin(panel,{prefix:'Recipe category: ',category:'Inspection',origin:'recipe'});
+    await selectFile(page,'workflow-only.png');await expect(panel.locator('.detail-inner > p.small')).toHaveText('workflow-only.png');
+    await expectOrigin(panel,{prefix:'Positive prompt category: ',category:'Landscapes',origin:'prompt'});
+    // An unlinked file shows no category until it has its own.
+    await selectFile(page,'empty.png');await expect(panel.locator('.panel-heading h2')).toHaveText('empty.png');
+    await expect(originChip(panel)).toHaveCount(0);
+    // Own categories: one overrides the prompt's on a linked file, one sits on an unlinked file.
+    await putOwnCategory(request,['inspector/workflow-only.png'],'Portraits');await putOwnCategory(request,['inspector/empty.png'],'Scans');
+    await page.reload();
+    await selectFile(page,'workflow-only.png');await expect(panel.locator('.detail-inner > p.small')).toHaveText('workflow-only.png');
+    await expectOrigin(panel,{prefix:'Image category: ',category:'Portraits',origin:'this image'});
+    await selectFile(page,'empty.png');await expect(panel.locator('.panel-heading h2')).toHaveText('empty.png');
+    await expectOrigin(panel,{prefix:'Image category: ',category:'Scans',origin:'this image'});
+    await selectFile(page,'complete.png');await expect(panel.locator('.panel-heading h2')).toHaveText(POSITIVE_TITLE);
+    await expectOrigin(panel,{prefix:'Recipe category: ',category:'Inspection',origin:'recipe'});
+    // The viewer's full details show the same chips.
+    let details=await viewerDetails(page,'workflow-only.png');
+    await expectOrigin(details,{prefix:'Image category: ',category:'Portraits',origin:'this image'});
+    await page.keyboard.press('Escape');await expect(page.locator('dialog#viewer-dialog[open]')).toHaveCount(0);
+    details=await viewerDetails(page,'complete.png');
+    await expectOrigin(details,{prefix:'Recipe category: ',category:'Inspection',origin:'recipe'});
+    await page.keyboard.press('Escape');await expect(page.locator('dialog#viewer-dialog[open]')).toHaveCount(0);
+    details=await viewerDetails(page,'empty.png');
+    await expectOrigin(details,{prefix:'Image category: ',category:'Scans',origin:'this image'});
+});
+
 test('tag chip filters Images by that tag',async({page,request})=>{
     await expect(await request.post('/__test__/reset',{data:{full_inspector:true}})).toBeOK();
     const list=await(await request.get('/prompt_manager/library')).json();const id=list.items.find(item=>item.text.startsWith('Inspector portrait')).id;
@@ -676,6 +724,9 @@ test('collection viewer uses real public transforms and resets on navigation',as
 async function slideshowFixture(page,request,preferences={videoAutoplay:false,videoMute:true,videoLoop:false}) {
     await expect(await request.post('/__test__/reset',{data:{full_inspector:true,include_video:true}})).toBeOK();
     await page.goto('/prompt_manager/workspace#/images');
+    // Install the fake clock before the fixture wraps setTimeout: the wrapped timer (which the viewer session captures)
+    // must be the fake one, so page.clock.runFor alone decides when the 3500 ms slideshow deadline fires.
+    await page.clock.install();
     await page.evaluate(async preferences=>{
         if(preferences) localStorage.setItem('gallerySettings',JSON.stringify(preferences));
         const [{createMediaViewer},{createInspectionController},{createScope},{createPromptSender}]=await Promise.all([
@@ -695,7 +746,7 @@ async function slideshowFixture(page,request,preferences={videoAutoplay:false,vi
     },preferences);
     await expect(mediaDialog(page).getByRole('button',{name:'Zoom in',exact:true})).toHaveAttribute('aria-disabled','false');
     await expect(mediaDialog(page).locator('#viewer-details > div > p.small')).toHaveText('complete.png');
-    await page.clock.install();await page.clock.pauseAt(new Date(Date.now()+1000));
+    await page.clock.pauseAt(new Date(Date.now()+1000));
 }
 const slideshowButton=page=>mediaDialog(page).locator('#viewer-toolbar').getByRole('button',{name:/^(Play|Pause|Stop) slideshow$/});
 const timerCount=page=>page.evaluate(()=>window.slideshowTimers.size);

tests/browser/history-concept.spec.mjs

@@ -1,7 +1,7 @@
 import {writeFile} from 'node:fs/promises';
 import {test,expect} from './fixtures.mjs';
 import {captureCanonicalState} from './history-canonical-comparison.mjs';
-import {SIZES,THEMES,MOTIONS,seedConcept,openConcept,openProduction,inventoryOf,diffInventory,mutate,stripExact} from './concept-inventory.mjs';
+import {SIZES,THEMES,MOTIONS,seedConcept,openConcept,openProduction,inventoryOf,diffInventory,mutate,stripExact,measureInspector,withChipSource} from './concept-inventory.mjs';
 
 // Task F5 of #249: the production Prompt history (#/library/<id>) against the immutable concept's (#prompt/alpine,
 // app.js renderHistory a587–622, the dialogs a1282–1304, style.css 1059–1179 and 1546–1567), on the concept-shaped real
@@ -272,6 +272,62 @@ for(const [width,height] of SIZES)for(const theme of THEMES)test(`history geomet
     } finally {await host.evaluate(()=>window.conceptHost.dispose());await host.close();}
 });
 
+/**
+ * D145 (#332): the History media inspector (a recorded run's current file, `recorded-run.mjs`) shows the file's own
+ * category with the chip rule of the Images side panel and viewer (D146). The concept has no recorded-run inspector, so
+ * its chip is the concept viewer's own category chip (app.js a402, imageDetails(i, true)) with the recorded D146 origin
+ * and the own category added at runtime (`withChipSource`). The chip's own place depends on what precedes it in either
+ * inspector, so its box is compared by size only.
+ */
+const OWN_CATEGORY='Portraits';
+const chipOnly=chip=>{const {left,top,source,...rest}=chip;return {...rest,source:source&&(({left,top,...own})=>own)(source)};};
+const categoryChipOf=async(page,root)=>(await page.evaluate(measureInspector,root)).badges.chips.find(chip=>chip.source);
+
+async function conceptViewerChip(page,theme) {
+    await openConcept(page,'images',theme);await page.mouse.move(0,0);
+    await page.locator('.gallery-body > .image-grid > .media-card').first().locator('.image-open').click();
+    await expect(page.locator('#viewer-dialog[open] #viewer-details .panel-heading h2')).toHaveText(TITLE);
+    const restore=await withChipSource(page,'#viewer-details','this image',OWN_CATEGORY);await settle(page);
+    const chip=await categoryChipOf(page,'#viewer-details');
+    await restore();
+    return chip;
+}
+
+/** Recorded run for the baseline's newest run, its one reported output opened in the viewer. */
+async function productionRunChip(page,theme,positiveId) {
+    await openProductionHistory(page,theme,'no-preference',positiveId);
+    await page.getByRole('button',{name:'Show repeat run for Quality baseline',exact:true}).click();
+    const run=page.getByRole('dialog',{name:'Recorded run',exact:true});
+    await expect(run).toBeVisible();
+    const view=run.getByRole('button',{name:'View current file',exact:true});
+    await expect(view).toBeEnabled();await view.click();
+    const details=page.locator('#viewer-dialog[open] #viewer-details');
+    await expect(details.locator('.chip-source')).toHaveText(' · this image');
+    await page.mouse.move(0,0);await settle(page);
+    return categoryChipOf(page,'#viewer-dialog[open] #viewer-details');
+}
+
+async function setOwnCategory(request,path,category) {
+    const listing=await (await request.get('/prompt_manager/images/output?mode=workspace&limit=40')).json();
+    const root_scope=listing.images.find(item=>item.root_index===0).root_scope;
+    const response=await request.put('/prompt_manager/images/category',{data:{items:[{root_index:0,root_scope,path}],category}});
+    await expect(response).toBeOK();
+}
+
+for(const [width,height] of SIZES)for(const theme of THEMES)test(`history media inspector category chip matches the concept ${width}x${height} ${theme} (D145)`,async({page,request},testInfo)=>{
+    await page.setViewportSize({width,height});
+    const {manifest}=testInfo;
+    // The baseline's newest image: the output of the run its dot opens (the fixture records one run per image).
+    const baseline=manifest.images.find(image=>image.prompt==='alpine'&&image.variant===0);
+    await setOwnCategory(request,baseline.path,OWN_CATEGORY);
+    const reference=await conceptViewerChip(page,theme);
+    const actual=await productionRunChip(page,theme,manifest.prompts[0].positive_id);
+    await captureCanonicalState(page,testInfo,`production-history-run-chip-${width}-${theme}`,'#viewer-dialog[open]');
+    expect.soft(reference,'concept chip').toMatchObject({tag:'span',className:'chip',text:`${OWN_CATEGORY} · this image`,hidden:null});
+    // D21: the hidden prefix names the scope; D146 as above.
+    expect.soft(diffInventory({...chipOnly(reference),hidden:'Image category: '},chipOnly(actual)),'chip differences after D21, D145, D146').toEqual([]);
+});
+
 // The comparison is only worth something if it notices a change (spec §7 mutation check): one CSS mutation (the variant
 // dot's width) and one DOM mutation (Compare negatives removed), each applied after an exact match.
 test('history comparison detects mutations',async({page},testInfo)=>{

tests/browser/images-concept.spec.mjs

@@ -85,15 +85,17 @@ function listInfo(card,actionsWidth) {
 /** Recorded Images deviations only (spec §8); everything else in the concept inventory must equal production. */
 const nocase=(a,b)=>{const x=a.toLowerCase(),y=b.toLowerCase();return x<y?-1:x>y?1:a<b?-1:a>b?1:0;};
 
-export function mapImagesConcept(reference,{manifest,now=new Date(manifest.now*1000)}) {
+export function mapImagesConcept(reference,{manifest,now=new Date(manifest.now*1000),imageCategories=[]}) {
     const paths=new Map(manifest.images.map(image=>[image.filename,image.path]));
     // D14: the sort appends production's two sorts. Save view is the concept's own (saved views, #249 D16 retired).
     const toolbar={...reference.toolbar};
     toolbar.items=toolbar.items.map(item=>item.id==='sort'?{...item,options:[...item.options,'Oldest first','Name Z–A']}:item);
     // D36: the category list is the Library's real list, sorted alphabetically (database/library.py `ORDER BY category
     // COLLATE NOCASE, category`); the concept lists its samples in prompt order (app.js 348). Only the order is mapped.
+    // D141 (#332): the list also holds the listed images' own categories (`GET /prompt_manager/images/categories`),
+    // sorted in with the Library's.
     toolbar.items=toolbar.items.map(item=>item.id==='category-filter'
-        ?{...item,options:[item.options[0],...item.options.slice(1).sort(nocase)]}:item);
+        ?{...item,options:[item.options[0],...[...item.options.slice(1),...imageCategories].sort(nocase)]}:item);
     // D29: no "Illustrative previews" sample tag.
     const filterRow=withoutItems(reference.filterRow,item=>String(item.className).includes('sample-tag'));
     filterRow.items=filterRow.items.map(item=>{
@@ -163,9 +165,9 @@ const pick=inventory=>stripExact(inventory);
 // The concept's own sample categories, in its first-appearance order (app.js 348): what D36 reorders.
 const CONCEPT_CATEGORIES=['All categories','Landscapes','Architecture','Environments','Illustration'];
 
-function compare(reference,actual,manifest,sections=SECTIONS) {
+function compare(reference,actual,manifest,sections=SECTIONS,imageCategories=[]) {
     const only=inventory=>Object.fromEntries(sections.map(key=>[key,inventory[key]]));
-    return diffInventory(pick(only(mapImagesConcept(reference,{manifest}))),pick(only(actual)));
+    return diffInventory(pick(only(mapImagesConcept(reference,{manifest,imageCategories}))),pick(only(actual)));
 }
 
 /** The comparison as a gate: throws, naming every differing path, unless production equals the mapped concept. */
@@ -470,6 +472,117 @@ for(const [width,height] of SIZES)for(const theme of THEMES)test(`images selecti
     expect(compare(oracle,await inventoryOf(page,'#workspace-main'),manifest,['bulkbar']).map(item=>item.path)).toContain('bulkbar.bar.paddingTop');
 });
 
+/** #332: an own category on one listed image, set through the real route (PUT /prompt_manager/images/category). */
+const IMAGE_CATEGORY='Portraits';
+async function setImageCategory(request,path,category) {
+    const listing=await (await request.get('/prompt_manager/images/output?mode=workspace&limit=40')).json();
+    const root_scope=listing.images.find(item=>item.root_index===0).root_scope;
+    const response=await request.put('/prompt_manager/images/category',{data:{items:[{root_index:0,root_scope,path}],category}});
+    await expect(response).toBeOK();
+    expect(await response.json()).toMatchObject({success:true,category,updated:1});
+}
+
+/**
+ * Runs in the page: the open Set category dialog (the concept's `simpleInput`, app.js 1384-1390 and 1138): its title,
+ * the body's tree, the field, the input and its suggestions (`datalist`), and the actions with every button's box,
+ * all relative to the dialog, with the concept's geometry and tokens.
+ */
+function measureCategoryDialog() {
+    const dialog=document.querySelector('dialog#modal[open]');
+    const css=(node,...keys)=>node?Object.fromEntries(keys.map(key=>[key,getComputedStyle(node)[key]])):null;
+    const name=node=>[node.localName,...[...node.classList].sort()].join('.');
+    const text=node=>node?.textContent.trim()??null;
+    const origin=dialog.getBoundingClientRect();
+    const box=node=>{const rect=node.getBoundingClientRect();return {left:Math.round(rect.left-origin.left),top:Math.round(rect.top-origin.top),
+        width:Math.round(rect.width),height:Math.round(rect.height)};};
+    const body=dialog.querySelector('#modal-body');
+    const field=body.querySelector(':scope > .field');
+    const label=field?.querySelector(':scope > label'),input=field?.querySelector(':scope > input');
+    const list=body.querySelector(':scope > datalist');
+    const actions=body.querySelector(':scope > .dialog-actions');
+    return {
+        title:text(dialog.querySelector('#modal-title')),
+        tree:[...body.children].map(name),
+        field:field&&{tree:[...field.children].map(name),...css(field,'display','flexDirection','rowGap','marginBottom'),...box(field)},
+        label:label&&{text:text(label),for:label.htmlFor,...css(label,'fontSize','fontWeight','color')},
+        input:input&&{id:input.id,type:input.type,value:input.value,list:input.getAttribute('list'),
+            ...css(input,'height','paddingLeft','fontSize','color','backgroundColor','borderTopColor','borderTopLeftRadius'),...box(input)},
+        suggestions:list&&{id:list.id,shown:list.getClientRects().length>0,options:[...list.options].map(option=>option.value)},
+        actions:{tree:[...actions.children].map(name),...css(actions,'display','justifyContent','columnGap','rowGap','flexWrap','marginTop'),...box(actions),
+            buttons:[...actions.children].map(button=>({text:text(button),className:button.className,
+                ...css(button,'backgroundColor','color','fontSize','paddingTop','paddingLeft','borderTopColor','minHeight'),...box(button)}))},
+        width:Math.round(origin.width),height:Math.round(origin.height),
+    };
+}
+
+/**
+ * The concept's own open dialog as production records it, changed at runtime so the browser lays it out: D140 (#332)
+ * empties the field (the concept pre-fills its demo "Landscapes"), lists the toolbar's categories as the field's
+ * suggestions (`datalist#workspace-input-suggestions`) and adds "Use prompt category", a plain button of the concept's
+ * own Cancel, between Cancel and the primary action; D8 names the primary action "Apply" ("Apply in prototype").
+ */
+async function categoryDialogAsRecorded(page,suggestions) {
+    await mutate(page,{clone:[{selector:'dialog#modal[open] .dialog-actions > button:first-child',text:'Use prompt category'}]});
+    await page.evaluate(suggestions=>{
+        const body=document.querySelector('dialog#modal[open] #modal-body');
+        const input=body.querySelector('#action-input');input.value='';input.setAttribute('list','workspace-input-suggestions');
+        const list=document.createElement('datalist');list.id='workspace-input-suggestions';
+        for(const value of suggestions){const option=document.createElement('option');option.value=value;list.append(option);}
+        body.querySelector(':scope > .field').after(list);
+        body.querySelector('.dialog-actions > .ca-primary').textContent='Apply';
+    },suggestions);
+    await settle(page);
+    return page.evaluate(measureCategoryDialog);
+}
+
+async function openCategoryDialog(page,scope) {
+    await page.locator(scope).locator(CARDS).first().locator('.image-overlays > label > input[type="checkbox"]').check();
+    await expect(page.locator('#bulkbar > .bulkbar > strong')).toHaveText('1 selected');
+    await page.locator('#bulkbar').getByRole('button',{name:'Set category',exact:true}).click();
+    await expect(page.locator('dialog#modal[open] #modal-title')).toHaveText('Categorize selected images');
+    await page.mouse.move(0,0);await settle(page);
+    return page.evaluate(measureCategoryDialog);
+}
+
+// D140 and D141 (#332): with one image's own category set, the toolbar's category options and the Set category dialog
+// against the concept's (app.js 348 and 1384-1390). The dialog is compared with the concept page itself changed at
+// runtime as recorded (`categoryDialogAsRecorded`); unmapped, it differs only in the recorded parts.
+for(const [width,height] of SIZES)for(const theme of THEMES)test(`images category options and Set category dialog match the concept ${width}x${height} ${theme}`,async({page,request},testInfo)=>{
+    await page.setViewportSize({width,height});
+    const manifest=testInfo.manifest;
+    await setImageCategory(request,manifest.images.at(-1).path,IMAGE_CATEGORY);
+    await openConceptLoaded(page,theme,'no-preference');
+    const reference=await inventoryOf(page,'#main');
+    const referenceDialog=await openCategoryDialog(page,'#main');
+    await captureCanonicalState(page,testInfo,`concept-images-category-${width}-${theme}`,'dialog#modal');
+    const suggestions=[...CONCEPT_CATEGORIES.slice(1),IMAGE_CATEGORY].sort(nocase);
+    const oracle=await categoryDialogAsRecorded(page,suggestions);
+    await page.keyboard.press('Escape');await expect(page.locator('dialog#modal[open]')).toHaveCount(0);
+
+    await openProductionLoaded(page,theme,'no-preference');
+    await expect(page.locator('#category-filter option')).toHaveText(['All categories',...suggestions]);
+    const actual=await inventoryOf(page,'#workspace-main');
+    const actualDialog=await openCategoryDialog(page,'#workspace-main');
+    await captureCanonicalState(page,testInfo,`production-images-category-${width}-${theme}`,'dialog#modal');
+    const path=testInfo.outputPath('images-category-geometry.json');
+    await writeFile(path,JSON.stringify({referenceDialog,oracle,actualDialog,reference:reference.toolbar,actual:actual.toolbar},null,2));
+    await testInfo.attach('images-category-geometry',{path,contentType:'application/json'});
+
+    // The concept itself: its sample categories and its pre-filled simpleInput with Cancel / Apply in prototype.
+    expect.soft(reference.toolbar.items.find(item=>item.id==='category-filter').options,'concept categories').toEqual(CONCEPT_CATEGORIES);
+    expect.soft(referenceDialog,'concept dialog').toMatchObject({title:'Categorize selected images',label:{text:'Category'},
+        input:{id:'action-input',value:'Landscapes',list:null},suggestions:null});
+    expect.soft(referenceDialog.actions.buttons.map(button=>button.text),'concept actions').toEqual(['Cancel','Apply in prototype']);
+    // D141: the toolbar options add the image's own category, sorted in (D36).
+    expect.soft(compare(reference,actual,manifest,['toolbar'],[IMAGE_CATEGORY]),'toolbar differences after D14, D36, D141').toEqual([]);
+    // D140 and D8: the dialog equals the concept page laid out as recorded.
+    expect.soft(diffInventory(oracle,actualDialog),'dialog differences after D8, D140').toEqual([]);
+    // Unmapped, the concept differs only in the field's value and suggestions, the body's datalist and the actions.
+    const recorded=/^(input\.(value|list)|tree\.|suggestions|actions\.)/;
+    expect.soft(diffInventory(referenceDialog,actualDialog).map(item=>item.path).filter(item=>!recorded.test(item)),'unrecorded dialog differences').toEqual([]);
+    expect.soft(actualDialog.suggestions,'suggestions').toEqual({id:'workspace-input-suggestions',shown:false,options:suggestions});
+});
+
 // The comparison is only worth something if it notices a change (spec §7 mutation check): one CSS mutation (the
 // `.media-info` padding) and one DOM mutation (the Inspect buttons removed), each applied after an exact match.
 test('images comparison detects mutations',async({page},testInfo)=>{

tests/browser/inspector-concept.spec.mjs

@@ -1,7 +1,7 @@
 import {writeFile} from 'node:fs/promises';
 import {test,expect} from './fixtures.mjs';
 import {captureCanonicalState} from './history-canonical-comparison.mjs';
-import {SIZES,THEMES,seedConcept,openConcept,openProduction,inspectorOf,diffInventory,mutate,countCopy,saveImageTags} from './concept-inventory.mjs';
+import {SIZES,THEMES,seedConcept,openConcept,openProduction,inspectorOf,diffInventory,mutate,countCopy,saveImageTags,withChipSource} from './concept-inventory.mjs';
 
 // Task D4 of #249: the Images side panel (app.js imageDetails a400–414) and the viewer's full details (imageDetails(i,
 // true)) against the immutable concept, on the concept-shaped real fixture, at 1600x1100 and 390x844 in both themes.
@@ -13,6 +13,8 @@ test.use({locale:'en-US',timezoneId:'UTC'});
 
 const CARDS='.gallery-body > .image-grid > .media-card';
 const TITLE='Alpine observatory';
+// D146 (#332): alpine_00001.png's recipe has no category, so its chip shows the positive's, " · prompt" (D39).
+const ORIGIN='prompt';
 const FIXTURE_IMAGES=24;
 // The concept's sticky panel never scrolls; production's scrolls inside the window above 700px (D20/D21 side panel).
 const PANEL_INSET=40;
@@ -38,14 +40,18 @@ function withFocusablePrompts(inventory) {
     return {...inventory,controls,sections:inventory.sections.map(section=>({...section,text:{...section.text,tabIndex:0}}))};
 }
 
-/** D21: the category chip names its scope in a visually hidden prefix; here the linked positive's (D39 fallback). */
+/**
+ * D21: the category chip names its scope in a visually hidden prefix; here the linked positive's (D39 fallback). Its
+ * visible origin (D146, #332: " · prompt") is the concept page's own chip with `span.chip-source` added at runtime
+ * (`withChipSource`), so `reference.panel` and `viewerPlain` already carry it.
+ */
 function withCategoryScope(inventory) {
     const [badge,category,...rest]=inventory.badges.chips;
     return {...inventory,badges:{...inventory.badges,chips:[badge,{...category,hidden:'Positive prompt category: '},...rest]}};
 }
 
 /**
- * Recorded inspector deviations only (spec §8, TESTING.md D3, D20–D22, D34, D38 and D117); everything else must
+ * Recorded inspector deviations only (spec §8, TESTING.md D3, D20–D22, D34, D38, D117 and D146); everything else must
  * equal production. `reference.viewerPlain` is the concept's own full details
  * with its tag chips made plain (D20), D117's summary button added and, under a coarse pointer, D34's targets, all at
  * runtime: the browser lays out what follows them, as the Images spec's removed toolbar items do.
@@ -112,6 +118,7 @@ async function openProductionPanel(page,theme) {
  */
 async function plainViewerChips(page) {
     const restoreTargets=await mutate(page,{css:D34_COARSE_TARGETS});
+    const restoreSource=await withChipSource(page,'#viewer-details',ORIGIN);
     await page.evaluate(()=>{
         window.__plainChips=[...document.querySelectorAll('#viewer-details .tags > button.chip')].map(button=>{
             const span=document.createElement('span');span.className='chip';span.textContent=button.textContent;
@@ -126,7 +133,7 @@ async function plainViewerChips(page) {
     return async()=>{
         await page.evaluate(()=>{for(const [span,button] of window.__plainChips)span.replaceWith(button);delete window.__plainChips;
             window.__summaryButton.remove();delete window.__summaryButton;});
-        await restoreTargets();
+        await restoreSource();await restoreTargets();
     };
 }
 
@@ -162,7 +169,11 @@ async function viewerOf(page,concept,mutation=null) {
     return {viewerDetails,viewerPlain,fileControls};
 }
 async function inventory(page,concept=false) {
+    // D146: the concept panel's category chip with its visible origin, laid out by the concept page itself.
+    const restoreSource=concept?await withChipSource(page,'#image-inspector',ORIGIN):null;
+    if(concept)await settle(page);
     const {panel}=await inspectorOf(page);
+    await restoreSource?.();
     const {viewerDetails,viewerPlain,fileControls}=await viewerOf(page,concept);
     return concept?{panel,viewerDetails,viewerPlain,fileControls}:{panel,viewerDetails,fileControls};
 }
@@ -288,7 +299,8 @@ for(const [width,height] of SIZES)for(const theme of THEMES)test(`inspector pane
     expect.soft(panel.sections.map(section=>section.label.color)[0],'concept role colors').not.toBe(panel.sections[1].label.color);
     expect.soft(panel.facts,'concept facts').toMatchObject({columns:2,columnGap:'12px',rowGap:'12px',fontSize:width>700?'11px':'12px'});
     expect.soft(panel.footer,'concept footer').toMatchObject({paddingTop:'13px',paddingRight:'17px',paddingBottom:'13px',paddingLeft:'17px'});
-    expect.soft(panel.badges.chips.map(chip=>chip.text),'concept badges').toEqual(['Execution recorded','Landscapes']);
+    // The concept's bare "Landscapes" with the D146 origin added at runtime (`withChipSource`).
+    expect.soft(panel.badges.chips.map(chip=>chip.text),'concept badges').toEqual(['Execution recorded','Landscapes · prompt']);
     expect.soft(reference.viewerDetails.structure.inner.slice(-2),'concept full details').toEqual(['details','details']);
 
     expect.soft(diff,'differences after the recorded deviations').toEqual([]);
@@ -322,8 +334,10 @@ test('a recipe-linked image without recipe metadata shows its positive prompt me
     await expect(panelOf(page).locator('.panel-heading h2')).toHaveText('Quiet geometry');
     const rows=panelOf(page).locator('.detail-inner > .tags');
     await expect(rows).toHaveCount(2);
-    await expect(rows.first().locator('.chip')).toHaveText(['Execution recorded','Positive prompt category: Architecture']);
+    await expect(rows.first().locator('.chip')).toHaveText(['Execution recorded','Positive prompt category: Architecture · prompt']);
     await expect(rows.first().locator('.chip .ca-visually-hidden')).toHaveText('Positive prompt category: ');
+    // D146 (#332): the visible origin names where the value came from.
+    await expect(rows.first().locator('.chip .chip-source')).toHaveText(' · prompt');
     await expect(rows.nth(1).locator('button.chip')).toHaveText(['minimal','warm_light','editorial']);
     await expect(rows.nth(1).locator('span.chip')).toHaveCount(0);
     await expect(panelOf(page).locator('.detail-footer button')).toHaveText(['Send to ComfyUI','Prompt history']);
@@ -400,6 +414,13 @@ test('inspector comparison detects mutations',async({page})=>{
     expect(()=>assertInspectorMatch(reference,withoutSummary,viewport)).toThrow(/viewerDetails\.structure\.inner/);
     assertInspectorMatch(reference,await viewerMutated(),viewport);
 
+    // D146: the visible origin is part of the mapped concept, so losing it is a difference.
+    const restoreSource=await mutate(page,{remove:['#image-inspector .chip-source']});
+    const sourceMutated=await measure();
+    expect(()=>assertInspectorMatch(reference,sourceMutated,viewport)).toThrow(/panel\.badges\.chips\.1\.source/);
+    await restoreSource();await settle(page);
+    assertInspectorMatch(reference,await measure(),viewport);
+
     await mutate(page,{remove:['#image-inspector > .detail-footer']});
     const domMutated=await measure();
     expect(()=>assertInspectorMatch(reference,domMutated,viewport)).toThrow(/panel\.footer/);

tests/browser/pair-node.spec.mjs

@@ -17,7 +17,7 @@ test.beforeEach(async({page,browserName})=>{
         window.fixtureApp={registerExtension(extension){this.extension=extension;},ui:{settings},graph:{_nodes:[],setDirtyCanvas(){}},canvas:{},graphToPrompt:async()=>({})};
         await import('/extensions/promptmanager/prompt_manager.js');
         class Node {
-            constructor(){this.id=1;this.type='PromptManagerPair';this.widgets=['text','negative_text','category','tags','search_text'].map(name=>({name,type:'text',value:'',options:{},inputEl:document.createElement('textarea')}));this.inputs=[...['clip','prepend_text','append_text'].map(name=>({name,type:name==='clip'?'CLIP':'STRING',link:null})),...['text','negative_text','category','tags','search_text'].map(name=>({name,type:'STRING',link:null,widget:{name}}))];this.properties={};this.size=[1100,700];}
+            constructor(){this.id=1;this.type='PromptManagerPair';this.widgets=[...['text','negative_text','category','tags','search_text'].map(name=>({name,type:'text',value:'',options:{},inputEl:document.createElement('textarea')})),{name:'save_to_library',type:'toggle',value:true,options:{}}];this.inputs=[...['clip','prepend_text','append_text'].map(name=>({name,type:name==='clip'?'CLIP':'STRING',link:null})),...['text','negative_text','category','tags','search_text'].map(name=>({name,type:'STRING',link:null,widget:{name}})),{name:'save_to_library',type:'BOOLEAN',link:null,widget:{name:'save_to_library'}}];this.properties={};this.size=[1100,700];}
             findInputByType(type){const index=this.inputs.findIndex(input=>input.type===type&&input.link==null);return index<0?undefined:{index,slot:this.inputs[index]};}
             addWidget(type,name,value,callback,options){const widget={type,name,value,callback,options};this.widgets.push(widget);return widget;}
             addDOMWidget(name,type,element){const host=document.createElement('div');host.style.cssText='height:700px;width:1100px;max-width:100%';host.className='fixture-dom-widget';host.append(element);document.body.append(host);return {element};}
@@ -34,6 +34,18 @@ test.afterEach(async({page,browserName})=>{await saveCoverage(page,browserName);
 async function selectPair(page){await page.getByRole('button',{name:'Recent',exact:true}).click();await page.locator('.pm-editor-result').click();}
 async function draft(page,positive='Draft positive',negative='Draft negative'){await page.getByRole('textbox',{name:'Positive prompt',exact:true}).fill(positive);await page.getByRole('textbox',{name:'Negative prompt',exact:true}).fill(negative);}
 
+test('paired editor Save to library switch keeps the widget order and loads older workflows on (#90)',async({page})=>{
+    await expect(page.locator('.pm-pair')).toBeVisible();
+    const toggle=page.getByRole('checkbox',{name:'Save to library',exact:true});
+    await expect(toggle).toBeChecked();
+    expect(await page.evaluate(()=>node.widgets.map(widget=>widget.name))).toEqual(['text','negative_text','category','tags','search_text','prepend_text','append_text','save_to_library']);
+    await toggle.uncheck();
+    expect(await page.evaluate(()=>node.serialize().widgets_values[7])).toBe(false);
+    await expect(page.locator('.pm-pair > .pm-editor-status[role=status]').last()).toHaveText('Save to library is off: runs of this node are not added to the library.');
+    await page.evaluate(()=>node.configure({widgets_values:['P','N','','','','','',''],properties:{}}));
+    await expect(toggle).toBeChecked();
+});
+
 test('paired editor previews both texts before explicitly replacing both',async({page})=>{
     await expect(page.locator('.pm-pair')).toBeVisible();
     await draft(page);

tests/browser/server.py

@@ -75,6 +75,8 @@ class TestModel:
     loaded = None
     # Models the Tag image dialog workflows mark as not downloaded.
     unavailable: set = set()
+    # Pip names per model whose backend packages "cannot be imported" (#424).
+    missing_packages: dict = {}
     # Scored names the concept comparisons answer instead of the fixed list
     # (``/__test__/tagging/suggestions``); None keeps the fixed list.
     suggestions: list | None = None
@@ -82,10 +84,13 @@ class TestModel:
     models_config: dict = {}
 
     def get_models_status(self) -> dict:
+        from py.autotag import package_fields
+
         return {
             name: {
                 "downloaded": name not in self.unavailable,
                 "loaded": self.loaded == name,
+                **package_fields(name, self.missing_packages.get(name, [])),
             }
             for name in self.models_config
         }
@@ -262,6 +267,8 @@ def create_app(root: Path) -> web.Application:
         with target.db.model.get_connection() as conn:
             for table in (
                 "saved_views",
+                "image_collection_items",
+                "image_collections",
                 "image_favorites",
                 "media_review_decisions",
                 "media_review_items",

tests/browser/vocabulary-canonical-comparison.mjs

@@ -1,6 +1,7 @@
 // Concept-vs-production inventory for the native Tags & categories screen (concept renderTags, app.js:719-729).
 // pageInventoryOf runs in the page; dialogs reuse the editor's inventoryOf. The concept is never rewritten:
 // mapVocabularyConcept / mapVocabularyDialog apply only the recorded deviations, each named with its ruling.
+import {diffInventory} from './library-editor-canonical-comparison.mjs';
 export {inventoryOf, diffInventory} from './library-editor-canonical-comparison.mjs';
 
 export function pageInventoryOf(main) {
@@ -61,23 +62,51 @@ export function pageInventoryOf(main) {
     };
 }
 
+const countOf=(count,noun)=>`${count} ${noun}${count===1?'':'s'}`;
+
+/**
+ * D143 (#332): a category row counts the images whose own category it is: "N positive prompts · M images" when
+ * prompts hold it too, "M images" for a value only images hold. Image-only rows follow every prompt-backed row, before
+ * the closing Manage category row, each drawn as the concept's last category row with its own text (vocabulary-visual
+ * checks this against the concept page with those rows added at runtime). `images` is `{value: count}`; `only` lists
+ * the image-only values in their order.
+ */
+export function withImageCounts(operations,{images={},only=[]}={}) {
+    const rows=operations.map(row=>images[row.h3.text]&&!only.includes(row.h3.text)
+        ?{...row,p:{...row.p,text:`${row.p.text} · ${countOf(images[row.h3.text],'image')}`}}:row);
+    const manage=rows.at(-1),last=rows.at(-2);
+    const added=only.map(value=>({...last,h3:{...last.h3,text:value},p:{...last.p,text:countOf(images[value],'image')}}));
+    return [...rows.slice(0,-1),...added,manage];
+}
+
 // Recorded deviations only (plan "Mock inventory" + decisions D1-D15 + ledger rulings). Everything else must be equal.
-export function mapVocabularyConcept(reference) {
+// `imageCategories` (#332 D143) is the fixture's own image categories, none by default.
+export function mapVocabularyConcept(reference,imageCategories={}) {
+    const operations=reference.operations.map(row=>({...row,p:{...row.p,
+        // Ledger ruling: "1 positive prompts" -> singular, consistent with the shared plural helper (#238/#243).
+        text:row.p.text.replace(/^1 positive prompts$/,'1 positive prompt')}}));
     return {...reference,
         table:{...reference.table,rows:reference.table.rows.map(row=>({...row,
             // Data, not a deviation: the concept tags each sample image with its prompt's tags; the vocabulary fixture
             // tags no image, so every Images count is 0 (#249 C2 retired #245 D3).
             images:'0'}))},
-        operations:reference.operations.map(row=>({...row,p:{...row.p,
-            // Ledger ruling: "1 positive prompts" -> singular, consistent with the shared plural helper (#238/#243).
-            text:row.p.text.replace(/^1 positive prompts$/,'1 positive prompt')}})),
+        operations:withImageCounts(operations,imageCategories),
     };
 }
 
+/** The page comparison: concept rows are the concept's sample tags; production appends fixture-only tags (D4). */
+export function compareVocabularyPage(reference,actual,imageCategories={}) {
+    const expected=mapVocabularyConcept(reference,imageCategories);
+    const shared=expected.table.rows.length;
+    return {diff:diffInventory(expected,{...actual,table:{...actual.table,rows:actual.table.rows.slice(0,shared)}}),
+        extra:actual.table.rows.slice(shared).map(row=>row.tag)};
+}
+
 // Concept dialogs are simpleInput (app.js:1138-1143). donors are concept inventories for controls the concept
 // shows elsewhere: a `.field select` and a `p.small` (Filters dialog, app.js:1351) and a danger dialog action
 // (Archive… in the editor, app.js:976), so added controls are held to concept tokens, never to production values.
-export function mapVocabularyDialog(mode,reference,{filters,editor,categories}) {
+// `preview` is the Manage categories sentence once Preview has been pressed (D9; D144, #332, adds its images).
+export function mapVocabularyDialog(mode,reference,{filters,editor,categories,preview=null}) {
     const fields=reference.fields.map(item=>({...item,
         // Inventory rows 27-29: the name is `required` (blank Apply is blocked natively).
         required:true}));
@@ -98,9 +127,9 @@ export function mapVocabularyDialog(mode,reference,{filters,editor,categories})
                 // D9 / Task 8: the proposed name offers a category datalist (no trailing commas: one value).
                 ...fields.map(item=>({...item,listed:true,commaOptions:false}))],
             // D9: the preview sentence is a `p.small`, empty and hidden until Preview.
-            small:[{...filters.small[0],text:''}],
+            small:[{...filters.small[0],text:preview??''}],
             // D9: the primary is `Preview` until a preview is shown, then `Apply`.
-            actions:{...reference.actions,buttons:buttons.map(item=>item.text==='Apply'?{...item,text:'Preview'}:item)}};
+            actions:{...reference.actions,buttons:preview===null?buttons.map(item=>item.text==='Apply'?{...item,text:'Preview'}:item):buttons}};
     }
     return {...reference,fields,actions:{...reference.actions,buttons}};
 }

tests/browser/vocabulary-visual.spec.mjs

@@ -1,11 +1,14 @@
+import {writeFile} from 'node:fs/promises';
 import {test,expect} from './fixtures.mjs';
 import {captureCanonicalState} from './history-canonical-comparison.mjs';
+import {pageInventoryOf,inventoryOf,diffInventory,mapVocabularyConcept,mapVocabularyDialog,compareVocabularyPage} from './vocabulary-canonical-comparison.mjs';
 
 // Artifact-only captures of Tags & categories states that have no concept counterpart: the Manage categories
 // preview, loading and 1,000-category cap notes, tag conflict and restore notes, the Delete tag confirmation (D14),
 // empty lists (D12), a failed read with Retry, and Load more (D5). The page and the three concept dialogs are captured
 // as concept/production pairs by workspace-vocabulary.spec.mjs. There are deliberately no pixel assertions or
-// baselines: the maintainer accepts these captures visually before any baseline is added.
+// baselines: the maintainer accepts these captures visually before any baseline is added. The last test (#332 D143,
+// D144) compares the category rows with image counts and the Manage category preview against the concept.
 test.use({locale:'en-US',timezoneId:'UTC',reducedMotion:'reduce'});
 const route='/prompt_manager/workspace#/tags';
 const main=page=>page.locator('#workspace-main');
@@ -101,3 +104,87 @@ for(const [width,height] of [[1600,1100],[390,844]])for(const theme of ['dark','
     await testInfo.attach(`production-load-more-${width}-${theme}`,{path,contentType:'image/png'});
     await expect(rows(page)).toHaveCount(100);
 });
+
+// D143 and D144 (#332): with own image categories seeded (`image_categories`: "Landscapes" on two images, "Portraits"
+// on one image only), the Categories panel and the Manage category preview against the concept (renderTags, app.js
+// 719-729; the Manage dialog is its simpleInput, 1138-1143). The page is compared with D143 mapped, a model checked
+// against the concept page with those rows added at runtime; the preview with D9's sentence and D144's images.
+const concept='/__test__/canonical/workspace';
+const IMAGE_CATEGORIES=Object.freeze({images:{Landscapes:2,Portraits:1},only:['Portraits']});
+// D9's sentence; D144 adds the images holding the value ("and 2 images").
+const PREVIEW='“Landscapes” is used by 3 positive prompts and 2 images. Apply renames it to “Scenery”.';
+const pageOf=(page,production)=>page.locator(production?'#workspace-main':'#main').evaluate(pageInventoryOf);
+async function openConceptRoute(page,route,theme) {
+    await page.goto('about:blank');await page.goto(`${concept}#${route}`);
+    await page.getByRole('combobox',{name:'Theme',exact:true}).selectOption(theme);
+}
+/** The concept page as D143 records it: image counts after the prompt counts, image-only rows before Manage category. */
+function conceptWithImageCounts(page) {
+    return page.evaluate(({images,only})=>{
+        const surface=document.querySelectorAll('#main .two-col > section.surface')[1];
+        const rows=[...surface.querySelectorAll('.operation')];
+        const count=value=>`${images[value]} image${images[value]===1?'':'s'}`;
+        for(const row of rows) {
+            const value=row.querySelector('h3').textContent;
+            if(images[value]&&!only.includes(value))row.querySelector('p').textContent+=` · ${count(value)}`;
+        }
+        const manage=rows.at(-1),last=rows.at(-2);
+        for(const value of only) {
+            const row=last.cloneNode(true);row.querySelector('h3').textContent=value;row.querySelector('p').textContent=count(value);manage.before(row);
+        }
+    },IMAGE_CATEGORIES);
+}
+/** Concept controls the Manage dialog adds (as in workspace-vocabulary.spec.mjs): the Filters field select and p.small. */
+async function conceptDonors(page,theme,categories) {
+    await openConceptRoute(page,'images',theme);
+    await page.locator('#main').getByRole('button',{name:'Filters',exact:true}).click();await expect(page.locator('#modal')).toBeVisible();
+    const filters=await page.locator('#modal').evaluate(inventoryOf);
+    await openConceptRoute(page,'prompt/alpine',theme);
+    await page.getByRole('button',{name:'Edit metadata',exact:true}).click();await expect(page.locator('#modal')).toBeVisible();
+    const editor=await page.locator('#modal').evaluate(inventoryOf);
+    return {filters,editor,categories};
+}
+
+for(const [width,height] of [[1600,1100],[390,844]])for(const theme of ['dark','light'])test(`category rows and the Manage category preview match the concept with image categories ${width}x${height} ${theme} (D143, D144)`,async({page,request},testInfo)=>{
+    await page.setViewportSize({width,height});
+    const manifest=await seed(request,{image_categories:true});
+    expect(manifest.image_categories).toEqual(IMAGE_CATEGORIES.images);
+    await openConceptRoute(page,'tags',theme);await expect(page.locator('#main h1')).toHaveText('Tags & categories');
+    await page.mouse.move(0,0);
+    const reference=await pageOf(page,false);
+    await conceptWithImageCounts(page);
+    const oracle=await pageOf(page,false);
+    await captureCanonicalState(page,testInfo,`concept-image-categories-${width}-${theme}`,'#main');
+    const donors=await conceptDonors(page,theme,['Landscapes','Architecture','Environments','Illustration','Portraits']);
+    await openConceptRoute(page,'tags',theme);
+    await page.locator('#main').getByRole('button',{name:'Review',exact:true}).click();await expect(page.locator('#modal')).toBeVisible();
+    const referenceDialog=await page.locator('#modal').evaluate(inventoryOf);
+
+    await open(page,theme,manifest.tags.length);
+    await expect(categoryOperation(page,'Portraits')).toHaveCount(1);
+    // The pointer still rests where the concept's Review was pressed: park it so no row is hovered.
+    await page.mouse.move(0,0);
+    const actual=await pageOf(page,true);
+    await captureCanonicalState(page,testInfo,`production-image-categories-${width}-${theme}`,'#workspace-main');
+    const box=await openDialog(page,categoryOperation(page,'Manage category').getByRole('button',{name:'Review',exact:true}));
+    await expect(box.getByLabel('Category',{exact:true})).toHaveValue('Landscapes');
+    await box.getByLabel('Proposed new category name',{exact:true}).fill('Scenery');await button(box,'Preview').click();
+    await expect(box.locator('p.small')).toHaveText(PREVIEW);
+    // The pressed primary keeps the pointer and focus; the concept is measured as it opens (focus on ×, no hover),
+    // which workspace-vocabulary.spec.mjs compares, so this state is restored before measuring the preview.
+    await page.mouse.move(0,0);await box.locator('.dialog-header button').focus();
+    const actualDialog=await box.evaluate(inventoryOf);
+    await captureCanonicalState(page,testInfo,`production-image-categories-preview-${width}-${theme}`);
+    const path=testInfo.outputPath('vocabulary-image-categories.json');
+    await writeFile(path,JSON.stringify({reference,oracle,actual,referenceDialog,actualDialog},null,2));
+    await testInfo.attach('vocabulary-image-categories',{path,contentType:'application/json'});
+
+    // The D143 model reproduces the concept page with the rows added at runtime.
+    expect.soft(diffInventory(mapVocabularyConcept(reference,IMAGE_CATEGORIES).operations,mapVocabularyConcept(oracle).operations),'D143 model').toEqual([]);
+    const {diff,extra}=compareVocabularyPage(reference,actual,IMAGE_CATEGORIES);
+    expect.soft(actual.operations.map(row=>[row.h3.text,row.p.text]).slice(0,-1),'category rows').toEqual([['Landscapes','3 positive prompts · 2 images'],
+        ['Architecture','1 positive prompt'],['Environments','1 positive prompt'],['Illustration','1 positive prompt'],['Portraits','1 image']]);
+    expect.soft(diff,'page differences after the recorded deviations and D143').toEqual([]);
+    expect.soft(extra,'fixture-only rows').toEqual([manifest.pair_only_tag,manifest.archived_only_tag,manifest.negative_only_tag,manifest.unused_tag]);
+    expect.soft(diffInventory(mapVocabularyDialog('manage',referenceDialog,{...donors,preview:PREVIEW}),actualDialog),'preview differences after D9 and D144').toEqual([]);
+});

tests/browser/workspace-history.spec.mjs

@@ -495,6 +495,30 @@ test('the inline current file names its prompts apart from the frozen run fields
     expect(await exact(dialog(page).getByRole('region',{name:'Positive prompt',exact:true}).locator('p.prompt-text'))).toBe(before);
 });
 
+// #332 CUJ 5 (MC8, §5.3): History's current-file inspector reads the file's own category with the Images key rule, so
+// the inline current file and its viewer show "Portraits · this image"; current.png has two saved links, so it inherits
+// nothing and shows no chip before.
+test('the current file shows its own category with its origin in the recorded run and its viewer',async({page,request})=>{
+    const chip=scope=>scope.locator('span.chip:has(> span.chip-source)');
+    const current=dialog(page).getByRole('region',{name:'Current reported file',exact:true});
+    const viewCurrent=async()=>{await dialog(page).locator('[data-receipt]').first().getByRole('button',{name:'View current file',exact:true}).click();
+        const details=page.locator('dialog#viewer-dialog #viewer-details');await expect(details.locator('.panel-heading')).toBeVisible();return details;};
+    await open(page,page.fixture.detail_id);await marker(page,page.fixture.runs.shared).click();await expect(dialog(page).locator('[data-receipt]')).toHaveCount(5);
+    let details=await viewCurrent();await expect(details.locator('.panel-heading h2')).toBeVisible();await expect(chip(details)).toHaveCount(0);
+    await page.keyboard.press('Escape');await expect(current).toBeVisible();await expect(chip(current)).toHaveCount(0);
+    await dialog(page).getByRole('button',{name:'Close recorded run',exact:true}).click();
+    const {images}=await(await request.get('/prompt_manager/images/output?mode=workspace&subfolder=history&limit=40')).json();
+    const file=images.find(image=>image.root_index===0&&image.path==='history/current.png');
+    const put=await request.put('/prompt_manager/images/category',{data:{items:[{root_index:0,root_scope:file.root_scope,path:file.path}],category:'Portraits'}});
+    await expect(put).toBeOK();expect((await put.json()).updated).toBe(1);
+    await open(page,page.fixture.detail_id);await marker(page,page.fixture.runs.shared).click();await expect(dialog(page).locator('[data-receipt]')).toHaveCount(5);
+    details=await viewCurrent();
+    await expect(chip(details)).toHaveText('Image category: Portraits · this image');
+    await page.keyboard.press('Escape');await expect(current).toBeVisible();
+    await expect(chip(current)).toHaveText('Image category: Portraits · this image');
+    expect(await chip(current).locator('.chip-source').evaluate(node=>node.textContent)).toBe(' · this image');
+});
+
 // LI-4 (D117): the recorded history's current-file viewer is the shared full inspector, so it copies the legacy summary.
 test('the current file viewer copies its settings summary',async({page})=>{
     await page.addInitScript(()=>{window.__copied=[];Object.defineProperty(navigator,'clipboard',{configurable:true,value:{writeText:async text=>{window.__copied.push(text);}}});});

tests/browser/workspace-images.spec.mjs

@@ -1988,70 +1988,286 @@ test('Export metadata downloads nothing when no image can be read', async ({ pag
     await expect(bulkbar(page).locator('.bulkbar > strong')).toHaveText('2 selected');
 });
 
-// #249 D8 follow-up (image selection, Task 3): Set category sets the linked positives' category (app.js simpleInput).
+// #332 (D140, MC6, MC7): Set category writes each selected image's own category (PUT /prompt_manager/images/category);
+// "Use prompt category" removes it after a confirmation (DELETE). It never writes a prompt: Images no longer posts
+// /prompt_manager/bulk/category (#259's prompt action is retired from Images).
 const categoryButton = page => bulkbar(page).getByRole('button', { name: 'Set category', exact: true });
 const categoryDialog = page => page.getByRole('dialog', { name: 'Categorize selected images', exact: true });
-const categoryRoute = '**/prompt_manager/bulk/category';
-/** Unlink the input root's card by deleting its prompt; returns the output card's positive id. */
-async function unlinkInputCard(request) {
-    const { images } = await (await request.get('/prompt_manager/images/output?mode=workspace')).json();
-    const byKey = Object.fromEntries(images.map(image => [`${image.root_index}:${image.path}`, image]));
-    const [output, input] = [byKey['0:shared/same.png'], byKey['1:shared/same.png']];
-    expect([output.linked.status, input.linked.status]).toEqual(['linked', 'linked']);
-    await expect(await request.delete(`/prompt_manager/delete/${input.linked.positive_id}`)).toBeOK();
-    return output.linked.positive_id;
+const removeDialog = page => page.getByRole('dialog', { name: 'Use prompt category', exact: true });
+const categoryFilter = page => page.getByRole('combobox', { name: 'Filter by category', exact: true });
+const categoryRoute = '**/prompt_manager/images/category';
+const IMAGE_CATEGORY = '/prompt_manager/images/category';
+const cardOf = (page, key) => page.locator(`article[data-image-key="${key}"]`);
+const keyOf = image => `${image.root_index}:${image.path}`;
+const cardKeys = page => cards(page).evaluateAll(nodes => nodes.map(node => node.dataset.imageKey).sort());
+// The four outputs of "Cat portrait" (CUJ 1's batch); the input root's same.png is linked to "Dog landscape".
+const BATCH = ['0:bulk/0000.png', '0:bulk/0001.png', '0:bulk/0002.png', '0:shared/same.png'];
+/** Every category write the page sends, Images' own route and #259's prompt route alike. */
+function categoryWrites(page) {
+    const writes = [];
+    page.on('request', item => {
+        const path = new URL(item.url()).pathname;
+        if (item.method() === 'GET' || !['/prompt_manager/images/category', '/prompt_manager/bulk/category'].includes(path)) return;
+        writes.push({ method: item.method(), path, body: item.postDataJSON() });
+    });
+    return writes;
+}
+async function listedImages(request, query = '') {
+    const response = await request.get(`/prompt_manager/images/output?mode=workspace&limit=40${query}`);
+    await expect(response).toBeOK();
+    return (await response.json()).images;
+}
+async function targetsOf(request, keys) {
+    const images = await listedImages(request);
+    return keys.map(key => {
+        const image = images.find(item => keyOf(item) === key);
+        return { root_index: image.root_index, root_scope: image.root_scope, path: image.path };
+    });
+}
+async function putOwnCategory(request, keys, category) {
+    const response = await request.put(IMAGE_CATEGORY, { data: { items: await targetsOf(request, keys), category } });
+    await expect(response).toBeOK();
+}
+const filteredKeys = async (request, category) => (await listedImages(request, `&category=${encodeURIComponent(category)}`)).map(keyOf).sort();
+async function positiveOf(request, id) {
+    const response = await request.get(`/prompt_manager/library/${id}`);
+    await expect(response).toBeOK();
+    const { positive } = await response.json();
+    return { id: positive.id, text: positive.text, category: positive.category };
+}
+/** CUJ 1's Given: four images linked to the positive "Cat portrait", whose category is "Landscapes". */
+async function seedBatch(request) {
+    await expect(await request.post('/__test__/reset', { data: { extra_images: 3 } })).toBeOK();
+    const response = await request.post('/prompt_manager/save', { data: { text: 'Cat portrait', category: 'Landscapes' } });
+    await expect(response).toBeOK();
+    const id = (await response.json()).prompt_id;
+    expect(await filteredKeys(request, 'Landscapes')).toEqual(BATCH);
+    return id;
+}
+async function openSetCategory(page) {
+    await categoryButton(page).click();
+    await expect(categoryDialog(page)).toBeVisible();
+    return categoryDialog(page);
+}
+async function applyCategory(page, category) {
+    const dialog = await openSetCategory(page);
+    await dialog.getByLabel('Category', { exact: true }).fill(category);
+    await dialog.getByRole('button', { name: 'Apply', exact: true }).click();
+    await expect(dialog).toBeHidden();
 }
 
-test('Set category applies the category to the linked positives and the category filter then matches', async ({ page, request }) => {
-    const outputPositive = await unlinkInputCard(request);
+test('Set category overrides one image of a batch and never writes its prompt (#332 CUJ 1)', async ({ page, request }) => {
+    const positiveId = await seedBatch(request);
+    const before = await positiveOf(request, positiveId);
+    const writes = categoryWrites(page);
     await page.goto(workspace);
-    await checkBoth(page);
-    const opener = categoryButton(page);
-    await opener.click();
-    const dialog = categoryDialog(page);
-    await expect(dialog).toBeVisible();
-    // The concept's simpleInput: one field and Cancel / Apply (primary).
+    await expect(cards(page)).toHaveCount(5);
+    await checkbox(cardOf(page, '0:bulk/0001.png'), '0001.png').check();
+    const dialog = await openSetCategory(page);
+    // The concept's simpleInput plus "Use prompt category" between Cancel and Apply (primary).
     await expect(dialog.locator('#modal-body > .field > label[for="action-input"]')).toHaveText('Category');
     await expect(dialog.locator('#modal-body > .field > input#action-input')).toHaveValue('');
-    await expect(dialog.locator('#modal-body > .dialog-actions > button')).toHaveText(['Cancel', 'Apply']);
+    await expect(dialog.locator('#modal-body > .dialog-actions > button')).toHaveText(['Cancel', 'Use prompt category', 'Apply']);
     await expect(dialog.getByRole('button', { name: 'Apply', exact: true })).toHaveClass(/ca-primary/);
     await expect(dialog.getByLabel('Category', { exact: true })).toBeFocused();
+    await expect(dialog.getByLabel('Category', { exact: true })).toHaveAttribute('maxlength', '100');
     // An empty value applies nothing and keeps the dialog open.
-    const posts = [];
-    page.on('request', item => { if (item.method() === 'POST' && item.url().endsWith('/prompt_manager/bulk/category')) posts.push(item.postDataJSON()); });
     await dialog.getByRole('button', { name: 'Apply', exact: true }).click();
     await expect(dialog).toBeVisible();
-    await dialog.getByLabel('Category', { exact: true }).fill('  Landscapes ');
+    await dialog.getByLabel('Category', { exact: true }).fill('  Portraits ');
     const listings = [];
     page.on('request', item => { if (imagePages(new URL(item.url()))) listings.push(item.url()); });
     await dialog.getByRole('button', { name: 'Apply', exact: true }).click();
     await expect(dialog).toBeHidden();
-    await expect(toast(page)).toHaveText('Category set on 1 prompt. 1 selected image had no linked prompt and was skipped.');
-    expect(posts).toEqual([{ prompt_ids: [outputPositive], category: 'Landscapes' }]);
+    await expect(toast(page)).toHaveText('Category “Portraits” set on 1 image.');
+    expect(writes).toEqual([{ method: 'PUT', path: IMAGE_CATEGORY,
+        body: { items: await targetsOf(request, ['0:bulk/0001.png']), category: 'Portraits' } }]);
     // The listing and the category options reload; the reload clears the selection.
     await expect.poll(() => listings.length).toBeGreaterThan(0);
     await expect(bulkbar(page)).toBeEmpty();
-    await expect(page.locator('#workspace-main h1')).toBeFocused();
-    await expect(cards(page).getByRole('checkbox', { checked: true })).toHaveCount(0);
-    const filter = page.getByRole('combobox', { name: 'Filter by category', exact: true });
-    await expect(filter.locator('option')).toContainText(['Landscapes']);
-    await filter.selectOption('Landscapes');
+    await expect(categoryFilter(page).locator('option')).toHaveText(['All categories', 'Landscapes', 'Portraits']);
+    await categoryFilter(page).selectOption('Portraits');
     await expect(cards(page)).toHaveCount(1);
-    await expect(cards(page)).toHaveAttribute('data-image-key', '0:shared/same.png');
+    await expect(cards(page)).toHaveAttribute('data-image-key', '0:bulk/0001.png');
+    await categoryFilter(page).selectOption('Landscapes');
+    await expect(cards(page)).toHaveCount(3);
+    expect(await cardKeys(page)).toEqual(BATCH.filter(key => key !== '0:bulk/0001.png'));
+    // The prompt keeps its category: only the image carries "Portraits".
+    expect(await positiveOf(request, positiveId)).toEqual(before);
+    expect(before.category).toBe('Landscapes');
+    expect(writes.filter(write => write.path === '/prompt_manager/bulk/category')).toEqual([]);
 });
 
-test('Set category without a linked image posts nothing and opens no dialog', async ({ page, request }) => {
-    await unlinkInputCard(request);
+test('Set category categorizes unlinked images and lists the value in the toolbar and the suggestions (#332 CUJ 2)', async ({ page, request }) => {
+    // Unlink both images by deleting their prompts.
+    for (const image of await listedImages(request)) {
+        await expect(await request.delete(`/prompt_manager/delete/${image.linked.positive_id}`)).toBeOK();
+    }
+    expect((await listedImages(request)).filter(image => image.linked.status === 'linked')).toEqual([]);
+    const writes = categoryWrites(page);
     await page.goto(workspace);
+    await checkBoth(page);
+    await applyCategory(page, 'Scans');
+    await expect(toast(page)).toHaveText('Category “Scans” set on 2 images.');
+    expect(writes).toEqual([{ method: 'PUT', path: IMAGE_CATEGORY,
+        body: { items: await targetsOf(request, ['0:shared/same.png', '1:shared/same.png']), category: 'Scans' } }]);
+    await expect(categoryFilter(page).locator('option')).toHaveText(['All categories', 'Scans']);
+    await categoryFilter(page).selectOption('Scans');
     await expect(cards(page)).toHaveCount(2);
-    let posted = 0;
-    await page.route(categoryRoute, route => { posted += 1; return route.continue(); });
-    await checkbox(cards(page).last(), 'same.png').check();
-    await categoryButton(page).click();
-    await expect(toast(page)).toHaveText("None of the selected images is linked to a saved prompt, so there is no prompt category to set.");
+    expect(await cardKeys(page)).toEqual(['0:shared/same.png', '1:shared/same.png']);
+    // The field suggests the toolbar's categories through its own datalist.
+    await checkbox(cards(page).first(), 'same.png').check();
+    const dialog = await openSetCategory(page);
+    await expect(dialog.getByLabel('Category', { exact: true })).toHaveAttribute('list', 'workspace-input-suggestions');
+    expect(await dialog.locator('datalist#workspace-input-suggestions option').evaluateAll(nodes => nodes.map(node => node.value))).toEqual(['Scans']);
+    await dialog.getByRole('button', { name: 'Cancel', exact: true }).click();
     await expect(page.locator('dialog[open]')).toHaveCount(0);
-    expect(posted).toBe(0);
-    await expect(bulkbar(page).locator('.bulkbar > strong')).toHaveText('1 selected');
+    expect(writes).toHaveLength(1);
+});
+
+test('Use prompt category removes the own category only after Remove in its confirmation (#332 CUJ 3)', async ({ page, request }) => {
+    const positiveId = await seedBatch(request);
+    await putOwnCategory(request, ['0:bulk/0001.png'], 'Portraits');
+    const writes = categoryWrites(page);
+    await page.goto(workspace);
+    await expect(cards(page)).toHaveCount(5);
+    await expect(categoryFilter(page).locator('option')).toHaveText(['All categories', 'Landscapes', 'Portraits']);
+    await checkbox(cardOf(page, '0:bulk/0001.png'), '0001.png').check();
+    // Cancel in the confirmation sends nothing and keeps the selection.
+    await (await openSetCategory(page)).getByRole('button', { name: 'Use prompt category', exact: true }).click();
+    await expect(categoryDialog(page)).toHaveCount(0);
+    const confirm = removeDialog(page);
+    await expect(confirm).toBeVisible();
+    await expect(confirm.locator('p').first()).toHaveText("Remove the image category from the 1 selected image? Each one then shows its prompt's category, or none if it has no linked prompt. This cannot be undone.");
+    await expect(confirm.locator('.dialog-actions > button')).toHaveText(['Cancel', 'Remove']);
+    await expect(confirm.getByRole('button', { name: 'Remove', exact: true })).toHaveClass(/danger/);
+    await expect(confirm.getByRole('button', { name: 'Cancel', exact: true })).toBeFocused();
+    await confirm.getByRole('button', { name: 'Cancel', exact: true }).click();
+    await expect(page.locator('dialog[open]')).toHaveCount(0);
+    await expect(selectedCount(page)).toHaveText('1 selected');
+    expect(writes).toEqual([]);
+    expect(await filteredKeys(request, 'Portraits')).toEqual(['0:bulk/0001.png']);
+    // Remove sends exactly one DELETE naming the selected image.
+    await (await openSetCategory(page)).getByRole('button', { name: 'Use prompt category', exact: true }).click();
+    await removeDialog(page).getByRole('button', { name: 'Remove', exact: true }).click();
+    await expect(toast(page)).toHaveText('Removed the image category from 1 image.');
+    expect(writes).toEqual([{ method: 'DELETE', path: IMAGE_CATEGORY, body: { items: await targetsOf(request, ['0:bulk/0001.png']) } }]);
+    await expect(bulkbar(page)).toBeEmpty();
+    // The image shows its prompt's category again: "Landscapes" lists all four, and "Portraits" is gone.
+    await expect(categoryFilter(page).locator('option')).toHaveText(['All categories', 'Landscapes']);
+    await categoryFilter(page).selectOption('Landscapes');
+    await expect(cards(page)).toHaveCount(4);
+    expect(await cardKeys(page)).toEqual(BATCH);
+    expect((await positiveOf(request, positiveId)).category).toBe('Landscapes');
+});
+
+test('a saved view filtered by a category lists an image categorized after it was saved (#332 CUJ 8)', async ({ page, request }) => {
+    await seedBatch(request);
+    await putOwnCategory(request, ['0:bulk/0001.png'], 'Portraits');
+    await page.goto(workspace);
+    await categoryFilter(page).selectOption('Portraits');
+    await expect(cards(page)).toHaveCount(1);
+    await page.locator('#workspace-main search.toolbar').getByRole('button', { name: 'Save view', exact: true }).click();
+    const save = page.getByRole('dialog', { name: 'Save this view', exact: true });
+    await save.getByRole('textbox', { name: 'View name', exact: true }).fill('Portraits');
+    await save.getByRole('button', { name: 'Save', exact: true }).click();
+    await expect(toast(page)).toHaveText('Saved view added to the sidebar.');
+    // Another image gets "Portraits" through Set category.
+    await categoryFilter(page).selectOption('');
+    await expect(cards(page)).toHaveCount(5);
+    await checkbox(cardOf(page, '0:bulk/0002.png'), '0002.png').check();
+    await applyCategory(page, 'Portraits');
+    await expect(toast(page)).toHaveText('Category “Portraits” set on 1 image.');
+    // Reopening the view lists both.
+    await page.getByRole('navigation', { name: 'Saved views', exact: true }).getByRole('link', { name: 'Portraits', exact: true }).click();
+    await expect(page.locator('#workspace-main h1')).toHaveText('Portraits');
+    await expect(categoryFilter(page)).toHaveValue('Portraits');
+    await expect(cards(page)).toHaveCount(2);
+    expect(await cardKeys(page)).toEqual(['0:bulk/0001.png', '0:bulk/0002.png']);
+});
+
+test('at 390px in light theme the Set category dialog and the category chip fit (#332 CUJ 10)', async ({ page, request }) => {
+    const LONG = 'Portraits taken on a phone at the end of a long walk through the old harbour at dusk';
+    await page.addInitScript(() => { try { localStorage.setItem('pm_workspace_theme', 'light'); } catch { /* the default theme */ } });
+    await page.setViewportSize({ width: 390, height: 844 });
+    await seedBatch(request);
+    await putOwnCategory(request, ['0:bulk/0001.png'], LONG);
+    await page.goto(workspace);
+    await expect(cards(page)).toHaveCount(5);
+    await expect(page.locator('html')).toHaveCSS('color-scheme', 'light');
+    const card = cardOf(page, '0:bulk/0001.png');
+    // The chip wraps inside the side panel; its origin never splits.
+    await card.getByRole('button', { name: 'Inspect 0001.png', exact: true }).click();
+    const chip = details(page).locator('span.chip', { has: page.locator('span.chip-source') });
+    await expect(chip).toHaveText(`Image category: ${LONG} · this image`);
+    const fit = await chip.evaluate(node => {
+        const panel = node.closest('aside, [role="complementary"]').getBoundingClientRect(), box = node.getBoundingClientRect();
+        const source = node.querySelector('.chip-source');
+        const line = parseFloat(getComputedStyle(source).lineHeight) || source.getBoundingClientRect().height;
+        return { inside: box.left >= panel.left - 0.5 && box.right <= panel.right + 0.5, sourceLines: source.getClientRects().length,
+            wrapped: box.height > line * 1.5, pageScroll: document.documentElement.scrollWidth - window.innerWidth };
+    });
+    expect(fit).toEqual({ inside: true, sourceLines: 1, wrapped: true, pageScroll: 0 });
+    // The dialog fits the width with no horizontal scroll, and its three buttons wrap without overlap.
+    await checkbox(card, '0001.png').check();
+    const dialog = await openSetCategory(page);
+    const geometry = await dialog.evaluate(node => {
+        const box = node.getBoundingClientRect();
+        const buttons = [...node.querySelectorAll('.dialog-actions > button')].map(button => button.getBoundingClientRect())
+            .map(({ left, right, top, bottom }) => ({ left, right, top, bottom }));
+        const overlaps = buttons.some((a, i) => buttons.some((b, j) => j > i && a.left < b.right && b.left < a.right && a.top < b.bottom && b.top < a.bottom));
+        const scrollers = [node, ...node.querySelectorAll('*')].filter(item => item.scrollWidth > item.clientWidth + 1 && getComputedStyle(item).overflowX !== 'visible');
+        return { fits: box.left >= 0 && box.right <= window.innerWidth, scrollers: scrollers.length, count: buttons.length, overlaps,
+            inside: buttons.every(b => b.left >= box.left && b.right <= box.right), pageScroll: document.documentElement.scrollWidth - window.innerWidth };
+    });
+    expect(geometry).toEqual({ fits: true, scrollers: 0, count: 3, overlaps: false, inside: true, pageScroll: 0 });
+});
+
+test('Set category of more than 500 images says so before any dialog or request (#332)', async ({ page, request }) => {
+    await page.route(url => url.pathname === '/prompt_manager/images/output' && url.searchParams.get('limit') !== '1', async route => {
+        const response = await route.fetch();
+        const body = await response.json();
+        await route.fulfill({ response, json: { ...body, total: 501 } });
+    });
+    await bulkReset(page, request);
+    await cardAt(page, 0).getByRole('button', { name: /^Open / }).focus();
+    await page.keyboard.press('ControlOrMeta+a');
+    await bulkbar(page).getByRole('button', { name: 'Select all 501 matching', exact: true }).click();
+    await expect(selectedCount(page)).toHaveText('All 501 matching selected');
+    const writes = categoryWrites(page);
+    await categoryButton(page).click();
+    await expect(toast(page)).toHaveText('Select at most 500 images.');
+    await expect(page.locator('dialog[open]')).toHaveCount(0);
+    expect(writes).toEqual([]);
+    await expect(selectedCount(page)).toHaveText('All 501 matching selected');
+});
+
+test('a changed source says so and reloads the listing (#332)', async ({ page }) => {
+    await page.goto(workspace);
+    await checkBoth(page);
+    await page.route(categoryRoute, route => route.fulfill({ status: 409,
+        json: { success: false, code: 'source-changed', error: 'The output sources changed. Refresh the images and try again.' } }));
+    const reloaded = page.waitForRequest(item => imagePages(new URL(item.url())));
+    await applyCategory(page, 'Landscapes');
+    await expect(toast(page)).toHaveText('The sources changed. Refresh the images and try again.');
+    await reloaded;
+    await expect(cards(page)).toHaveCount(2);
+    // The reload hides the bar with Set category in it: focus moves to the page heading, as after a success.
+    await expect(page.locator('#workspace-main h1')).toBeFocused();
+});
+
+test('a recipe-linked image is listed under its recipe\'s category (#332 D142)', async ({ page, request }) => {
+    await expect(await request.post('/__test__/reset', { data: { native_library: true } })).toBeOK();
+    await page.goto(workspace);
+    await expect(cards(page)).toHaveCount(3);
+    await expect(categoryFilter(page).locator('option')).toHaveText(['All categories', 'Animals', 'Experiments', 'Portraits']);
+    // The output same.png is linked to recipe "Portrait variant 1" (Portraits), the input copy to variant 2 (Experiments);
+    // only the image without a recipe shows its positive's "Animals".
+    for (const [category, keys] of [['Portraits', ['0:shared/same.png']], ['Experiments', ['1:shared/same.png']], ['Animals', ['0:legacy-unassigned.png']]]) {
+        await categoryFilter(page).selectOption(category);
+        await expect(cards(page)).toHaveCount(keys.length);
+        expect(await cardKeys(page)).toEqual(keys);
+    }
 });
 
 test('Set category reports a failure, keeps the selection and ignores a second press while it runs', async ({ page }) => {
@@ -2059,9 +2275,7 @@ test('Set category reports a failure, keeps the selection and ignores a second p
     await checkBoth(page);
     const held = [];
     await page.route(categoryRoute, route => { held.push(route); });
-    await categoryButton(page).click();
-    await categoryDialog(page).getByLabel('Category', { exact: true }).fill('Landscapes');
-    await categoryDialog(page).getByRole('button', { name: 'Apply', exact: true }).click();
+    await applyCategory(page, 'Landscapes');
     await expect.poll(() => held.length).toBe(1);
     await expect(categoryButton(page)).toHaveAttribute('aria-disabled', 'true');
     // Playwright waits on an aria-disabled control, so the second press is a dispatched click.
@@ -2098,14 +2312,12 @@ test('leaving Images during Set category finishes the write, and the return relo
     await checkBoth(page);
     const held = [];
     await page.route(categoryRoute, route => { held.push(route); });
-    await categoryButton(page).click();
-    await categoryDialog(page).getByLabel('Category', { exact: true }).fill('Landscapes');
-    await categoryDialog(page).getByRole('button', { name: 'Apply', exact: true }).click();
+    await applyCategory(page, 'Landscapes');
     await expect.poll(() => held.length).toBe(1);
     await page.evaluate(() => { window.location.hash = '/overview'; });
     await expect(page.locator('#workspace-main h1')).toHaveText('Overview');
     // The write is not aborted with the page: the server still receives and answers it.
-    const answered = page.waitForResponse(response => response.url().endsWith('/prompt_manager/bulk/category'));
+    const answered = page.waitForResponse(response => new URL(response.url()).pathname === IMAGE_CATEGORY);
     await held[0].continue();
     expect((await answered).ok()).toBe(true);
     await page.evaluate(() => new Promise(resolve => setTimeout(resolve, 150)));
@@ -2116,23 +2328,25 @@ test('leaving Images during Set category finishes the write, and the return relo
     await expect(cards(page)).toHaveCount(2);
     await expect(bulkbar(page)).toBeEmpty();
     await expect(cards(page).getByRole('checkbox', { checked: true })).toHaveCount(0);
-    await expect(page.getByRole('combobox', { name: 'Filter by category', exact: true }).locator('option')).toContainText(['Landscapes']);
+    await expect(categoryFilter(page).locator('option')).toContainText(['Landscapes']);
 });
 
-test('Set category limits the field to the server length, shows the server count, and a rejection keeps the selection', async ({ page }) => {
+test('Set category shows the server counts and coded refusals, and a rejection keeps the selection', async ({ page }) => {
     await page.goto(workspace);
     await checkBoth(page);
     const answers = [
         { status: 400, json: { success: false, error: 'Category cannot exceed 100 characters' } },
-        { status: 200, json: { success: true, updated_count: 0 } },
+        { status: 404, json: { success: false, code: 'files-not-found', error: 'None of the images could be found.' } },
+        { status: 200, json: { success: true, category: 'Landscapes', updated: 1, unchanged: 0, not_found: 1 } },
     ];
     await page.route(categoryRoute, route => route.fulfill(answers.shift()));
-    for (const expected of ['Category could not be set. Try again.', 'Category set on 0 prompts.']) {
-        await categoryButton(page).click();
-        const field = categoryDialog(page).getByLabel('Category', { exact: true });
-        await expect(field).toHaveAttribute('maxlength', '100');
-        await field.fill('Landscapes');
-        await categoryDialog(page).getByRole('button', { name: 'Apply', exact: true }).click();
+    for (const expected of ['Category could not be set. Try again.',
+        'None of the selected images could be found. Refresh the images and try again.',
+        'Category “Landscapes” set on 1 image. 1 could not be found.']) {
+        const dialog = await openSetCategory(page);
+        await expect(dialog.getByLabel('Category', { exact: true })).toHaveAttribute('maxlength', '100');
+        await dialog.getByLabel('Category', { exact: true }).fill('Landscapes');
+        await dialog.getByRole('button', { name: 'Apply', exact: true }).click();
         await expect(toast(page)).toHaveText(expected);
         if (answers.length) await expect(bulkbar(page).locator('.bulkbar > strong')).toHaveText('2 selected');
     }
@@ -2381,24 +2595,20 @@ test('all matching sets the category through the listing filter and survives a r
     // Export metadata waits for loaded images: a press does nothing.
     const exportButton = bulkbar(page).getByRole('button', { name: 'Export metadata', exact: true });
     await expect(exportButton).toHaveAttribute('title', 'Export metadata reads each image in the browser, so it works on loaded images only.');
-    // No match linked (a stubbed answer): the nothing-linked sentence, and the selection stays.
-    await page.route('**/prompt_manager/bulk/category', route => route.fulfill({ json: { success: true, updated_count: 0, prompt_count: 0, skipped: 202 } }), { times: 1 });
-    const categorize = page.getByRole('dialog', { name: 'Categorize selected images', exact: true });
-    await bulkbar(page).getByRole('button', { name: 'Set category', exact: true }).click();
-    await categorize.getByRole('textbox', { name: 'Category', exact: true }).fill('Bulk392');
-    await categorize.getByRole('button', { name: 'Apply', exact: true }).click();
-    await expect(toast(page)).toHaveText("None of the selected images is linked to a saved prompt, so there is no prompt category to set.");
+    // No match listed any more (a stubbed answer): the not-found sentence, and the selection stays.
+    await page.route(categoryRoute, route => route.fulfill({ status: 404, json: { success: false, code: 'files-not-found', error: 'None of the images could be found.' } }), { times: 1 });
+    await applyCategory(page, 'Bulk392');
+    await expect(toast(page)).toHaveText('None of the selected images could be found. Refresh the images and try again.');
     await expect(selectedCount(page)).toHaveText('All 202 matching selected');
-    // The real write: one request with the filter, every matching image's positive changes.
-    await bulkbar(page).getByRole('button', { name: 'Set category', exact: true }).click();
-    await categorize.getByRole('textbox', { name: 'Category', exact: true }).fill('Bulk392');
-    const posted = page.waitForRequest(item => item.method() === 'POST' && new URL(item.url()).pathname === '/prompt_manager/bulk/category');
-    await categorize.getByRole('button', { name: 'Apply', exact: true }).click();
-    expect((await posted).postDataJSON()).toEqual({ images_filter: { sort: 'newest' }, category: 'Bulk392' });
-    await expect(toast(page)).toHaveText(/^Category set on \d+ prompts?\.$/);
+    // The real write: one PUT with the listing filter sets every matching image's own category; no prompt changes.
+    const writes = categoryWrites(page);
+    await applyCategory(page, 'Bulk392');
+    await expect(toast(page)).toHaveText('Category “Bulk392” set on 202 images.');
+    expect(writes).toEqual([{ method: 'PUT', path: IMAGE_CATEGORY, body: { filter: { sort: 'newest' }, category: 'Bulk392' } }]);
     await expect(bulkbar(page)).toBeEmpty();
     const filtered = await (await request.get('/prompt_manager/images/output?mode=workspace&limit=1&category=Bulk392')).json();
     expect(filtered.total).toBe(202);
+    expect((await (await request.get('/prompt_manager/library/options')).json()).categories).not.toContain('Bulk392');
     // A return visit keeps "all matching" as the loaded images, and the offer shows again.
     await cardAt(page, 0).getByRole('button', { name: /^Open / }).focus();
     await page.keyboard.press('ControlOrMeta+a');
@@ -2426,10 +2636,12 @@ test('all matching over 10,000 images says so before any dialog or request', asy
     await bulkbar(page).getByRole('button', { name: 'Select all 20,002 matching', exact: true }).click();
     await expect(selectedCount(page)).toHaveText('All 20,002 matching selected');
     let posted = 0;
-    page.on('request', item => { if (item.method() === 'POST') posted += 1; });
-    for (const action of ['Set category', 'Add tags', 'Auto-tag selected']) {
+    page.on('request', item => { if (item.method() !== 'GET') posted += 1; });
+    // Set category names at most 500 images (#332); the tagging jobs take up to 10,000.
+    for (const [action, sentence] of [['Set category', 'Select at most 500 images.'], ['Add tags', 'Narrow the view to at most 10,000 images.'],
+        ['Auto-tag selected', 'Narrow the view to at most 10,000 images.']]) {
         await bulkbar(page).getByRole('button', { name: action, exact: true }).click();
-        await expect(toast(page)).toHaveText('Narrow the view to at most 10,000 images.');
+        await expect(toast(page)).toHaveText(sentence);
         await expect(page.locator('dialog[open]')).toHaveCount(0);
     }
     expect(posted).toBe(0);

tests/browser/workspace-system-tagging.spec.mjs

@@ -123,6 +123,32 @@ test('model options are the four real engines with size and the description show
     await expect(actionOf(page, 'Unload')).toBeDisabled();
 });
 
+const NEEDS_LLAMA = 'GGUF JoyCaption needs llama-cpp-python. Install it into ComfyUI\'s Python: pip install llama-cpp-python. '
+    + 'For a CUDA build: CMAKE_ARGS="-DGGML_CUDA=on" pip install llama-cpp-python';
+
+test('a downloaded model whose backend package is missing says so and is not offered as ready (#424)', async ({ page, request }) => {
+    await expect(await request.post('/__test__/tagging/models', { data: { unavailable: ['wd14-vit'],
+        missing_packages: { gguf: ['llama-cpp-python'], 'wd14-vit': ['onnxruntime'] } } })).toBeOK();
+    await openTagging(page);
+    const line = modelRow(page).locator('p').first();
+    await page.locator('#tag-model').selectOption('gguf');
+    await expect(line).toHaveText('Downloaded. Needs llama-cpp-python: pip install llama-cpp-python. '
+        + 'For a CUDA build: CMAKE_ARGS="-DGGML_CUDA=on" pip install llama-cpp-python');
+    // Files and packages are separate: Download (again) stays offered.
+    await expect(actionOf(page, 'Download again')).not.toHaveAttribute('aria-disabled', 'true');
+    // Review explains instead of reviewing; nothing is previewed.
+    await expect(reviewButton(page)).toHaveAttribute('aria-disabled', 'true');
+    await reviewButton(page).dispatchEvent('click');
+    await expect(toast(page)).toHaveText(NEEDS_LLAMA);
+    expect(page.modelCalls.filter(call => call.endsWith('/autotag/batch/preview'))).toEqual([]);
+    // Not downloaded reads differently and still names what it will need.
+    await page.locator('#tag-model').selectOption('wd14-vit');
+    await expect(line).toHaveText('Not downloaded. Download size ~400 MB. Also needs onnxruntime.');
+    await page.locator('#tag-model').selectOption('wd14-swinv2');
+    await expect(line).toHaveText('Downloaded. Not loaded.');
+    await expect(reviewButton(page)).not.toHaveAttribute('aria-disabled', 'true');
+});
+
 test('entering the section makes no model call besides status', async ({ page }) => {
     await openTagging(page);
     await page.locator('#workspace-main .section-nav').getByRole('button', { name: 'Integrations', exact: true }).click();

tests/browser/workspace-tagging-bulk.spec.mjs

@@ -186,6 +186,26 @@ test('a model that is not downloaded keeps Start disabled and says why', async (
     expect(posted).toBe(0);
 });
 
+const NEEDS_LLAMA = 'GGUF JoyCaption needs llama-cpp-python. Install it into ComfyUI\'s Python: pip install llama-cpp-python. '
+    + 'For a CUDA build: CMAKE_ARGS="-DGGML_CUDA=on" pip install llama-cpp-python';
+
+test('a model missing its backend package keeps Start disabled and names the install command (#424)', async ({ page, request }) => {
+    await expect(await request.post('/__test__/tagging/models', { data: { unavailable: [],
+        missing_packages: { gguf: ['llama-cpp-python'] } } })).toBeOK();
+    await select(page, FIRST);
+    await barButton(page, 'Auto-tag selected').click();
+    const status = modal(page).locator('#tag-generation-status');
+    const start = actionButtons(page).last();
+    await modal(page).getByRole('combobox', { name: 'Tagging model', exact: true }).selectOption('gguf');
+    await expect(status).toHaveText(NEEDS_LLAMA);
+    await expect(start).toHaveAttribute('aria-disabled', 'true');
+    await modal(page).getByRole('combobox', { name: 'Tagging model', exact: true }).selectOption('wd14-vit');
+    await expect(status).toHaveText('');
+    await expect(start).not.toHaveAttribute('aria-disabled', 'true');
+    await actionButtons(page).first().click();
+    await expect(modal(page)).toHaveCount(0);
+});
+
 test('Suggestions field prefills from the setting and caps chips at the entered count', async ({ page, request }) => {
     await expect(await request.post('/__test__/tagging/defaults', { data: { general: 0.5, character: 0.9, limit: 7 } })).toBeOK();
     await select(page, FIRST);

tests/browser/workspace-tagging.spec.mjs

@@ -260,6 +260,18 @@ test('unavailable model shows inline status and never downloads', async ({ page,
     await expect(dialog(page).locator('.dialog-actions .ca-primary')).toBeFocused();
 });
 
+test('a missing backend package is the server\'s install sentence, no traceback page (#424)', async ({ page, request }) => {
+    await expect(await request.post('/__test__/tagging/models', { data: { unavailable: [],
+        missing_packages: { 'wd14-swinv2': ['onnxruntime'] } } })).toBeOK();
+    await openTag(page);
+    const answer = await generated(page);
+    expect(answer.status()).toBe(409);
+    expect(await answer.json()).toMatchObject({ success: false, code: 'model-unavailable' });
+    await expect(status(page)).toHaveText('WD14 SwinV2 needs onnxruntime. Install it into ComfyUI\'s Python: '
+        + 'pip install onnxruntime. Use onnxruntime-gpu for CUDA.');
+    await expect(chips(page)).toHaveCount(0);
+});
+
 test('video tag is manual only', async ({ page, request }) => {
     await openTag(page, VIDEO);
     const preview = dialog(page).locator('video.tag-image-preview');

tests/browser/workspace-vocabulary.spec.mjs

@@ -383,6 +383,62 @@ test('manage categories previews real impact before applying',async({page,reques
     expect(after.prompts.filter(row=>row[2]==='Scenery')).toHaveLength(3);
 });
 
+// #332 MC9, MC10: image categories (their own, set on Images) count beside the prompts. The vocabulary fixture has no
+// image files, so these tests add two committed files to the output folder and set their categories through the API.
+async function imageCategories(request,category) {
+    for(const name of ['f02.png','f03.png']) await expect(await request.post('/__test__/add-local-inspect-file',{data:{name}})).toBeOK();
+    const listing=await(await request.get('/prompt_manager/images/output?mode=workspace&subfolder=local-inspect&limit=40')).json();
+    const items=listing.images.map(({root_index,root_scope,path})=>({root_index,root_scope,path}));expect(items).toHaveLength(2);
+    const response=await request.put('/prompt_manager/images/category',{data:{items,category}});await expect(response).toBeOK();
+    expect((await response.json()).updated).toBe(2);
+}
+const ownCategories=async request=>(await(await request.get('/prompt_manager/images/categories')).json()).categories;
+const IMAGE_KEYS=['0:local-inspect/f02.png','0:local-inspect/f03.png'];
+
+test('manage categories renames the images\' own categories too and the preview counts them (#332 CUJ 6)',async({page,request})=>{
+    await imageCategories(request,'Landscapes');
+    const writes=trackWrites(page);await openPage(page);
+    await expect(categoryRow(page,'Landscapes').locator('p')).toHaveText('3 positive prompts · 2 images');
+    const box=await openManage(page);
+    await expect(field(box,'Category')).toHaveValue('Landscapes');
+    await field(box,'Proposed new category name').fill('Landscape');await button(box,'Preview').click();
+    await expect(box.locator('p.small')).toHaveText('“Landscapes” is used by 3 positive prompts and 2 images. Apply renames it to “Landscape”.');
+    await button(box,'Apply').click();
+    await expect(box).toHaveCount(0);await expect(toast(page)).toHaveText('Category renamed.');
+    expect(writes.at(-1)).toMatchObject({url:'/prompt_manager/vocabulary/categories/apply',body:{source:'Landscapes',target:'Landscape'}});
+    await expect(categoryRow(page,'Landscapes')).toHaveCount(0);
+    await expect(categoryRow(page,'Landscape').locator('p')).toHaveText('3 positive prompts · 2 images');
+    // The prompts and both images carry the new name; nothing keeps the old one.
+    expect((await snapshot(request)).prompts.filter(row=>row[2]==='Landscape')).toHaveLength(3);
+    expect(await ownCategories(request)).toEqual(['Landscape']);
+    for(const [category,total] of [['Landscape',2],['Landscapes',0]]) {
+        const listing=await(await request.get(`/prompt_manager/images/output?mode=workspace&limit=40&category=${category}`)).json();
+        expect(listing.images.map(image=>`${image.root_index}:${image.path}`).sort()).toEqual(total?IMAGE_KEYS:[]);
+    }
+});
+
+test('a category only images hold is listed after the prompt-backed ones and Browse opens Images (#332 CUJ 7)',async({page,request})=>{
+    await imageCategories(request,'Scans');
+    await openPage(page);
+    const operations=surfaces(page).nth(1).locator('.operation');
+    await expect(operations.locator('h3')).toHaveText([...CATEGORIES.map(([name])=>name),'Scans','Manage category']);
+    await expect(categoryRow(page,'Scans').locator('p')).toHaveText('2 images');
+    for(const [name,count] of CATEGORIES) await expect(categoryRow(page,name).locator('p')).toHaveText(`${count} positive prompt${count===1?'':'s'}`);
+    // Manage category can rename it: its select lists it last.
+    const box=await openManage(page);
+    await expect(field(box,'Category').locator('option')).toHaveText([...CATEGORIES.map(([name])=>name),'Scans']);
+    await button(box,'Cancel').click();await expect(box).toHaveCount(0);
+    // Browse opens Images filtered by it (its Library would be empty).
+    await categoryRow(page,'Scans').getByRole('button',{name:'Browse',exact:true}).click();
+    await expect(page).toHaveURL(/#\/images$/);
+    await expect(main(page).getByRole('heading',{level:1})).toHaveText('Images');
+    await expect(page.getByRole('combobox',{name:'Filter by category',exact:true})).toHaveValue('Scans');
+    const cards=page.locator('article[data-image-key]');await expect(cards).toHaveCount(2);
+    expect((await cards.evaluateAll(nodes=>nodes.map(node=>node.dataset.imageKey))).sort()).toEqual(IMAGE_KEYS);
+    await expect(page.getByRole('searchbox',{name:'Search workspace',exact:true})).toHaveValue('');
+    await expect(page.locator('#workspace-main search.toolbar').getByRole('button',{name:'Filters',exact:true})).toBeVisible();
+});
+
 test('an edit during a pending preview needs a new Preview; Apply never sends the old values',async({page,request})=>{
     const writes=trackWrites(page);await openPage(page);const before=await snapshot(request);
     let release;const held=new Promise(resolve=>{release=resolve;});

tests/browser/workspace_tagging_fixture.py

@@ -41,6 +41,18 @@ FILES = (
     ("loose/unlinked.png", "unlinked"),
 )
 MODELS = ("wd14-swinv2", "wd14-vit", "gguf", "hf")
+
+
+def _valid_missing(value) -> bool:
+    """``{"<model>": ["<pip name>", ...]}`` with known models and string names."""
+    return isinstance(value, dict) and all(
+        name in MODELS
+        and isinstance(names, list)
+        and all(isinstance(pip, str) and pip for pip in names)
+        for name, names in value.items()
+    )
+
+
 # Unlinked copies enough to virtualize the Images grid (it renders all up to 120).
 MAX_EXTRA_IMAGES = 160
 _TABLES = (
@@ -217,6 +229,7 @@ def register_tagging_commands(
         if not 0 <= extra <= MAX_EXTRA_IMAGES:
             raise web.HTTPBadRequest()
         model.unavailable = set()
+        model.missing_packages = {}
         model.suggestions = None
         model.loaded = None
         restore_download()
@@ -235,17 +248,25 @@ def register_tagging_commands(
 
     @routes.post("/__test__/tagging/models")
     async def models(request: web.Request) -> web.Response:
-        """Mark fake models as not downloaded (``{"unavailable": [...]}``)."""
+        """Mark fake models as not downloaded (``{"unavailable": [...]}``) and,
+        optionally, as missing backend packages (``"missing_packages":
+        {"<model>": ["<pip name>", ...]}``, #424)."""
         data = await request.json() if request.can_read_body else None
         if (
             request.query
             or not isinstance(data, dict)
-            or set(data) != {"unavailable"}
+            or "unavailable" not in data
+            or set(data) - {"unavailable", "missing_packages"}
             or not isinstance(data["unavailable"], list)
             or any(name not in MODELS for name in data["unavailable"])
+            or not _valid_missing(data.get("missing_packages", {}))
         ):
             raise web.HTTPBadRequest()
         model.unavailable = set(data["unavailable"])
+        model.missing_packages = {
+            name: list(names)
+            for name, names in data.get("missing_packages", {}).items()
+        }
         return web.json_response({"success": True})
 
     @routes.post("/__test__/tagging/loaded")

tests/browser/workspace_vocabulary_fixture.py

@@ -72,6 +72,10 @@ CONCEPT_RECIPES = (
         "negative": "oversaturated, flat lighting, watermark, text",
     },
 )
+# #332 D143/D144 (opt-in, ``image_categories``): own image categories, one on
+# images only and one shared with prompts. Stored rows count whether or not the
+# files exist, so the keys name files the fixture never writes.
+IMAGE_CATEGORY_ROWS = (("Landscapes", 2), ("Portraits", 1))
 PAIR_ONLY_TAG = "pair_only_detail"
 ARCHIVED_ONLY_TAG = "archived_pair_detail"
 NEGATIVE_ONLY_TAG = "legacy_negative_detail"
@@ -112,8 +116,26 @@ _SNAPSHOT = {
 }
 
 
+def _seed_image_categories(api: PromptManagerAPI) -> dict:
+    """Own image category rows keyed as Images keys them; ``{value: count}``."""
+    from py.api.images import image_category_key
+
+    sources = list(api._get_all_output_dirs())
+    for category, count in IMAGE_CATEGORY_ROWS:
+        keys = [
+            image_category_key(sources, str(Path(sources[0]) / f"{category}_{n}.png"))
+            for n in range(count)
+        ]
+        api.db.set_image_categories(keys, category)
+    return dict(IMAGE_CATEGORY_ROWS)
+
+
 def seed_vocabulary(
-    api: PromptManagerAPI, root: Path, extra_tags: int = 0, extra_categories: int = 0
+    api: PromptManagerAPI,
+    root: Path,
+    extra_tags: int = 0,
+    extra_categories: int = 0,
+    image_categories: bool = False,
 ) -> dict:
     """Replace this server's disposable rows with the concept vocabulary."""
     db = api.db
@@ -167,6 +189,7 @@ def seed_vocabulary(
                 "category": f"bulk_category_{number:04}",
             },
         )
+    images = _seed_image_categories(api) if image_categories else {}
     with db.model.get_connection() as conn:
         tags = [row[0] for row in conn.execute("SELECT name FROM tags ORDER BY id")]
     api.invalidate_gallery_cache()
@@ -180,6 +203,7 @@ def seed_vocabulary(
         "archived_only_tag": ARCHIVED_ONLY_TAG,
         "negative_only_tag": NEGATIVE_ONLY_TAG,
         "unused_tag": UNUSED_TAG,
+        "image_categories": images,
     }
 
 
@@ -200,7 +224,14 @@ def register_vocabulary_commands(
     @routes.post("/__test__/vocabulary/reset")
     async def reset(request: web.Request) -> web.Response:
         data = await request.json() if request.can_read_body else {}
-        if not isinstance(data, dict) or set(data) - {"extra_tags", "extra_categories"}:
+        if not isinstance(data, dict) or set(data) - {
+            "extra_tags",
+            "extra_categories",
+            "image_categories",
+        }:
+            raise web.HTTPBadRequest()
+        images = data.get("image_categories", False)
+        if not isinstance(images, bool):
             raise web.HTTPBadRequest()
         extra = data.get("extra_tags", 0)
         categories = data.get("extra_categories", 0)
@@ -215,7 +246,7 @@ def register_vocabulary_commands(
         if request.query:
             raise web.HTTPBadRequest()
         manifest = await api._run_in_executor(
-            seed_vocabulary, api, root, extra, categories
+            seed_vocabulary, api, root, extra, categories, images
         )
         return web.json_response({"success": True, **manifest})
 

tests/comfyui/execution-text-link.spec.mjs

@@ -11,13 +11,13 @@ import { installHostHooks } from './host-fixture.mjs';
 
 // The worker's owned host must be up before the shared host hooks open the page.
 test.beforeEach(async ({ execution }) => { expect(execution.databasePath).toBeTruthy(); });
-installHostHooks(test, ['PromptManager', 'PromptManagerText', 'PromptManagerFocused', 'PromptManagerPair', 'PMHistoryInputs']);
+installHostHooks(test, ['PromptManager', 'PromptManagerText', 'PromptManagerFocused', 'PromptManagerPair', 'PMHistoryInputs', 'PMCycledText']);
 
 // `primitive`: a core string node the queue-time resolver follows. `legacy`: the frontend's virtual Primitive node,
 // whose value is queued as literal text. `preview`: PrimitiveString -> PreviewAny, the shape of the maintainer's
 // TextGenerate -> PreviewAny chain, whose text is only known when the node runs.
-function build(page, { kind, positive, negative, prefix }) {
-    return page.evaluate(async ({ kind, positive, negative, prefix }) => {
+function build(page, { kind, positive, negative, prefix, saveToLibrary }) {
+    return page.evaluate(async ({ kind, positive, negative, prefix, saveToLibrary }) => {
         const { app } = await import('/scripts/app.js');
         app.graph.clear();
         const add = (type, x, y) => { const node = LiteGraph.createNode(type); app.graph.add(node); node.pos = [x, y]; return node; };
@@ -28,6 +28,7 @@ function build(page, { kind, positive, negative, prefix }) {
         const set = (name, value) => { const widget = prompt.widgets.find(item => item.name === name); widget.value = value; if (widget.inputEl) widget.inputEl.value = value; };
         set('text', positive.typed);
         if (negative) set('negative_text', negative.typed);
+        if (saveToLibrary !== undefined) prompt.widgets.find(item => item.name === 'save_to_library').value = saveToLibrary;
         const drive = (name, side, y) => {
             if (!side.via) return;
             if (side.via === 'legacy') {
@@ -37,6 +38,7 @@ function build(page, { kind, positive, negative, prefix }) {
                 primitive.widgets[0].value = side.text;
                 return;
             }
+            if (side.via === 'cycled') { connect(add('PMCycledText', 0, y), 0, prompt, name); return; }
             const text = add('PrimitiveString', 0, y);
             text.widgets.find(widget => widget.name === 'value').value = side.text;
             let source = text;
@@ -72,12 +74,12 @@ function build(page, { kind, positive, negative, prefix }) {
         await prompt._pmTextSerializerReady;
         const { output, workflow } = await app.graphToPrompt();
         return { output, workflow, promptNode: String(prompt.id) };
-    }, { kind, positive, negative, prefix });
+    }, { kind, positive, negative, prefix, saveToLibrary });
 }
 
-async function run(execution, page, testCase) {
+async function run(execution, page, testCase, built) {
     const prefix = `textlink/${randomUUID()}`;
-    const { output, workflow, promptNode } = await build(page, { ...testCase, prefix });
+    const { output, workflow, promptNode } = built ?? await build(page, { ...testCase, prefix });
     const id = randomUUID();
     const response = await execution.request.post('/prompt', { data: { prompt: output, prompt_id: id, extra_data: { extra_pnginfo: { workflow } } } });
     expect(response.ok(), await response.text()).toBe(true);
@@ -193,3 +195,205 @@ for (const [label, linked, negativeVia] of [['positive', ['text'], 'primitive'],
         await expectImagesLinked(execution, evidence, sides.text.via === 'preview');
     });
 }
+
+// #90: with Save to library off the run still produces its image, but PromptManager saves, counts and links nothing,
+// and the run's history records no prompt text for it.
+for (const kind of ['PromptManager', 'PromptManagerText', 'PromptManagerFocused', 'PromptManagerSplit', 'PromptManagerPair']) {
+    test(`${kind} with Save to library off saves, counts and links nothing (#90)`, async ({ execution, page }) => {
+        test.setTimeout(90000);
+        const text = unique(`Not saved ${kind}`);
+        const negative = kind === 'PromptManagerPair' ? { typed: unique('Not saved negative') } : undefined;
+        const prefix = `notsaved/${randomUUID()}`;
+        const { output, workflow, promptNode } = await build(page, { kind, positive: { typed: text }, negative, prefix, saveToLibrary: false });
+        expect(output[promptNode].inputs.save_to_library).toBe(false);
+        const id = randomUUID();
+        const response = await execution.request.post('/prompt', { data: { prompt: output, prompt_id: id, extra_data: { extra_pnginfo: { workflow } } } });
+        expect(response.ok(), await response.text()).toBe(true);
+        let item;
+        await expect.poll(async () => {
+            item = (await (await execution.request.get(`/history/${id}`)).json())[id];
+            return item?.status?.status_str ?? null;
+        }, { timeout: 20000 }).toBe('success');
+        let result;
+        await expect.poll(async () => {
+            result = await readExecutionRuns(execution.databasePath, id);
+            const run = result.rows[0];
+            return Boolean(run && result.captures[run.run_id]);
+        }, { timeout: 10000, message: 'the run reaches SQLite' }).toBe(true);
+        const run = result.rows[0];
+        for (const context of result.captures[run.run_id].contexts || []) {
+            expect(context.positive_prompt_id ?? null).toBeNull();
+            expect(context.pair_id ?? null).toBeNull();
+        }
+        expect((result.usage[run.run_id] || []).every(entry => entry.run_count === 0)).toBe(true);
+        const files = Object.values(item.outputs).flatMap(value => value.images || []).map(image => image.filename);
+        expect(files.length).toBeGreaterThan(0);
+        await scan(execution);
+        expect(Object.values(await linkedPrompts(execution, files))).toEqual(files.map(() => null));
+        for (const query of [text, negative?.typed].filter(Boolean)) {
+            const found = await (await execution.request.get(`/prompt_manager/search?text=${encodeURIComponent(query)}`)).json();
+            expect(found.results ?? [], query).toEqual([]);
+        }
+    });
+}
+
+// #347: a cached PromptManager node counts every job, as an executing one does. Jobs after the first change only
+// the sampler seed, so ComfyUI serves the PromptManager node (and its PreviewAny source) from its output cache.
+const COUNTS = `import json, sqlite3, sys, pathlib
+db = sqlite3.connect(pathlib.Path(sys.argv[1]).resolve().as_uri() + "?mode=ro", uri=True)
+one = lambda sql, key: (db.execute(sql, (int(key),)).fetchone() or [None])[0] if key != "-" else None
+print(json.dumps({"positive": one("SELECT run_count FROM prompts WHERE id=?", sys.argv[2]),
+    "negative": one("SELECT run_count FROM prompts WHERE id=?", sys.argv[3]),
+    "pair": one("SELECT run_count FROM prompt_pairs WHERE id=?", sys.argv[4])}))`;
+async function runCounts(execution, context) {
+    const key = value => value == null ? '-' : String(value);
+    const { stdout } = await promisify(execFile)(process.env.PM_TEST_PYTHON || 'python',
+        ['-c', COUNTS, execution.databasePath, key(context.positive_prompt_id), key(context.negative_prompt_id), key(context.pair_id)], { timeout: 5000 });
+    return JSON.parse(stdout);
+}
+
+async function setCache(execution, mode) {
+    expect((await execution.request.post('/pm_execution_fixture/cache', { data: { mode } })).ok()).toBe(true);
+}
+
+// Queue an already built workflow again with another sampler seed; returns the job's history item and cached nodes.
+async function queueAgain(execution, first, { seed, edit, targets, status = 'success' } = {}) {
+    const output = structuredClone(first.output);
+    const sampler = Object.values(output).find(node => node.class_type === 'KSampler');
+    sampler.inputs.seed = seed;
+    edit?.(output);
+    const id = randomUUID();
+    const data = { prompt: output, prompt_id: id, extra_data: { extra_pnginfo: { workflow: first.workflow } } };
+    if (targets) data.partial_execution_targets = targets;
+    const response = await execution.request.post('/prompt', { data });
+    expect(response.ok(), await response.text()).toBe(true);
+    let item;
+    await expect.poll(async () => {
+        item = (await (await execution.request.get(`/history/${id}`)).json())[id];
+        return item?.status?.status_str ?? null;
+    }, { timeout: 20000 }).toBe(status);
+    const cached = item.status.messages.find(([event]) => event === 'execution_cached')?.[1]?.nodes ?? [];
+    return { id, item, cached };
+}
+
+// The recorded context of one job, once its capture reached SQLite.
+async function capturedContext(execution, id) {
+    let result;
+    await expect.poll(async () => {
+        result = await readExecutionRuns(execution.databasePath, id);
+        const row = result.rows[0];
+        return row ? result.captures[row.run_id].state : null;
+    }, { timeout: 10000 }).toMatch(/^(complete|partial)$/);
+    return result.captures[result.rows[0].run_id].contexts[0];
+}
+
+async function firstRun(execution, page, testCase) {
+    const prefix = `textlink/${randomUUID()}`;
+    const built = await build(page, { ...testCase, prefix });
+    const evidence = await run(execution, page, testCase, built);
+    return { ...evidence, workflow: built.workflow };
+}
+
+for (const kind of ['PromptManager', 'PromptManagerFocused', 'PromptManagerSplit', 'PromptManagerText', 'PromptManagerPair']) {
+    test(`${kind} counts each cached re-run of text linked from PreviewAny (#347)`, async ({ execution, page }) => {
+        test.setTimeout(90000);
+        await setCache(execution, 'RAM_PRESSURE');
+        const pair = kind === 'PromptManagerPair';
+        const testCase = { kind, positive: { typed: 'Typed under the link', text: unique(`Cached ${kind}`), via: 'preview' },
+            negative: pair ? { typed: unique('Typed negative') } : undefined };
+        const first = await firstRun(execution, page, testCase);
+        for (const seed of [2, 3]) {
+            const again = await queueAgain(execution, first, { seed });
+            expect(again.cached, 'ComfyUI served the PromptManager node from its cache').toContain(first.promptNode);
+        }
+        const counts = await runCounts(execution, first.context);
+        expect(counts.positive).toBe(3);
+        if (pair) expect(counts).toMatchObject({ negative: 0, pair: 3 });
+    });
+}
+
+test.describe('cached runs in one renderer (#347)', () => {
+    // Cache behavior does not depend on the renderer: one project runs these.
+    test.beforeEach(({}, testInfo) => { test.skip(testInfo.project.name !== 'classic-start', 'covered in classic-start'); });
+    const linked = () => ({ kind: 'PromptManager', positive: { typed: 'Typed under the link', text: unique('Cached'), via: 'preview' } });
+
+    for (const mode of ['CLASSIC', 'LRU', 'RAM_PRESSURE', 'NONE']) {
+        test(`counts 3 jobs under the ${mode} cache`, async ({ execution, page }) => {
+            test.setTimeout(90000);
+            await setCache(execution, mode);
+            try {
+                const first = await firstRun(execution, page, linked());
+                for (const seed of [2, 3]) {
+                    const again = await queueAgain(execution, first, { seed });
+                    if (mode === 'NONE') expect(again.cached).not.toContain(first.promptNode);
+                    else expect(again.cached).toContain(first.promptNode);
+                }
+                expect((await runCounts(execution, first.context)).positive).toBe(3);
+            } finally { await setCache(execution, 'RAM_PRESSURE'); }
+        });
+    }
+
+    test('alternating upstream values A, B, A under LRU each count their own prompt', async ({ execution, page }) => {
+        test.setTimeout(90000);
+        await setCache(execution, 'LRU');
+        try {
+            const [a, b] = [unique('Cycled A'), unique('Cycled B')];
+            const cycle = async value => expect((await execution.request.post('/pm_execution_fixture/cycled', { data: { value } })).ok()).toBe(true);
+            await cycle(a);
+            const first = await firstRun(execution, page, { kind: 'PromptManager', positive: { typed: 'Typed under the link', text: a, via: 'cycled' } });
+            expect(first.context.positive.raw).toBe(a);
+            await cycle(b);
+            const second = await queueAgain(execution, first, { seed: 2 });
+            expect(second.cached).not.toContain(first.promptNode);
+            await cycle(a);
+            const third = await queueAgain(execution, first, { seed: 3 });
+            expect(third.cached, 'LRU kept A').toContain(first.promptNode);
+            expect((await runCounts(execution, first.context)).positive).toBe(2);
+            const bContext = await capturedContext(execution, second.id);
+            expect(bContext.positive.raw).toBe(b);
+            expect((await runCounts(execution, bContext)).positive).toBe(1);
+        } finally { await setCache(execution, 'RAM_PRESSURE'); }
+    });
+
+    test('typed text served from the cache still counts once per job', async ({ execution, page }) => {
+        test.setTimeout(90000);
+        await setCache(execution, 'RAM_PRESSURE');
+        const first = await firstRun(execution, page, { kind: 'PromptManager', positive: { typed: unique('Typed cached') } });
+        for (const seed of [2, 3]) expect((await queueAgain(execution, first, { seed })).cached).toContain(first.promptNode);
+        expect((await runCounts(execution, first.context)).positive).toBe(3);
+    });
+
+    test('a job failing downstream of a cached node counts it once; a job that does not need it, never', async ({ execution, page }) => {
+        test.setTimeout(90000);
+        await setCache(execution, 'RAM_PRESSURE');
+        const first = await firstRun(execution, page, linked());
+        const failing = await queueAgain(execution, first, { seed: 2, status: 'error', edit: output => {
+            output.pm_failure = { class_type: 'PMHistoryOutput', inputs: { value: [Object.keys(output).find(id => output[id].class_type === 'KSampler'), 0], mode: 'failure' } };
+        } });
+        expect(failing.cached).toContain(first.promptNode);
+        expect((await runCounts(execution, first.context)).positive).toBe(2);
+        // "Run selected output": only an unrelated output node; ComfyUI still reports the cached PromptManager node
+        const unrelated = await queueAgain(execution, first, { seed: 2, targets: ['pm_unrelated'], edit: output => {
+            output.pm_unrelated = { class_type: 'PMHistoryCached', inputs: { value: unique('unrelated') } };
+        } });
+        expect(unrelated.cached).toContain(first.promptNode);
+        expect((await runCounts(execution, first.context)).positive).toBe(2);
+    });
+
+    test('a cached node rewired to the negative input does not count the positive prompt', async ({ execution, page }) => {
+        test.setTimeout(90000);
+        await setCache(execution, 'RAM_PRESSURE');
+        const first = await firstRun(execution, page, linked());
+        const rewired = await queueAgain(execution, first, { seed: 2, edit: output => {
+            const sampler = Object.values(output).find(node => node.class_type === 'KSampler');
+            [sampler.inputs.positive, sampler.inputs.negative] = [sampler.inputs.negative, sampler.inputs.positive];
+        } });
+        expect(rewired.cached).toContain(first.promptNode);
+        expect((await runCounts(execution, first.context)).positive).toBe(1);
+        // The history capture of a cached job is unchanged: effective text from the cache, no stored-prompt hint
+        const cachedContext = await capturedContext(execution, (await queueAgain(execution, first, { seed: 3 })).id);
+        expect(cachedContext.positive.effective).toBe(first.context.positive.effective);
+        expect(cachedContext.positive_prompt_id).toBeNull();
+        expect((await runCounts(execution, first.context)).positive).toBe(2);
+    });
+});

tests/comfyui/fixtures/ledger_nodes.py

@@ -201,6 +201,39 @@ NODE_CLASS_MAPPINGS.update(
     {"PMHistoryImages": PMHistoryImages, "PMHistoryImageOutput": PMHistoryImageOutput}
 )
 
+# #347: text whose value (and cache signature) follows a value the test sets, so
+# one queued graph can serve A, then B, then A again (each cached under LRU).
+_CYCLED = {"value": "cycled"}
+
+
+class PMCycledText:
+    @classmethod
+    def INPUT_TYPES(cls) -> dict:
+        return {"required": {}}
+
+    RETURN_TYPES = ("STRING",)
+    FUNCTION = "execute"
+    CATEGORY = "testing/execution-history"
+
+    @classmethod
+    def IS_CHANGED(cls, **kwargs: object) -> str:
+        return _CYCLED["value"]
+
+    def execute(self) -> tuple:
+        return (_CYCLED["value"],)
+
+
+NODE_CLASS_MAPPINGS["PMCycledText"] = PMCycledText
+
+
+@PromptServer.instance.routes.post("/pm_execution_fixture/cycled")
+async def cycled_text(request: web.Request) -> web.Response:
+    value = (await request.json())["value"]
+    if type(value) is not str:
+        raise web.HTTPBadRequest()
+    _CYCLED["value"] = value
+    return web.json_response({"value": value})
+
 
 @PromptServer.instance.routes.post("/pm_execution_fixture/metadata")
 async def metadata(request: web.Request) -> web.Response:

tests/comfyui/text-link.spec.mjs

@@ -370,6 +370,7 @@ test('workflows saved before #343 load unchanged and their linked text keeps wor
         const nodes = Object.fromEntries(app.graph._nodes.map(node => [node.id, node]));
         const reloaded = app.graph.serialize();
         return {
+            saves: Object.fromEntries(Object.entries(output).filter(([, node]) => node.class_type.startsWith('PromptManager')).map(([id, node]) => [id, node.inputs.save_to_library])),
             values: Object.fromEntries(reloaded.nodes.map(node => [node.id, node.widgets_values])),
             links: Object.fromEntries(Object.entries(output).map(([id, node]) => [id, { text: node.inputs.text, negative_text: node.inputs.negative_text }])),
             widgetBound: Object.fromEntries(Object.values(nodes).filter(node => node.type.startsWith('PromptManager')).map(node => [node.id, node.inputs.filter(input => ['text', 'negative_text'].includes(input.name) && input.widget).map(input => input.name)])),
@@ -377,8 +378,16 @@ test('workflows saved before #343 load unchanged and their linked text keeps wor
     }, workflow);
     for (const node of workflow.nodes) {
         // Every saved widget value comes back as it was saved, except the typed text main lost under a link (null).
-        if (node.widgets_values?.[0] !== null) expect(result.values[node.id], `${node.type} ${node.id}`).toEqual(node.widgets_values);
+        // #90: the last saved value is the editor's own slot (''), where the Save to library switch is now: it loads on.
+        if (node.widgets_values?.[0] === null) continue;
+        if (!node.type.startsWith('PromptManager')) { expect(result.values[node.id], `${node.type} ${node.id}`).toEqual(node.widgets_values); continue; }
+        const saved = node.widgets_values.length - 1;
+        expect(result.values[node.id].slice(0, saved), `${node.type} ${node.id}`).toEqual(node.widgets_values.slice(0, saved));
+        expect(result.values[node.id][saved], `${node.type} ${node.id} save_to_library`).toBe(true);
     }
+    // Every PromptManager node of the old workflow queues with Save to library on.
+    for (const save of Object.values(result.saves)) expect(save).toBe(true);
+    expect(Object.keys(result.saves).length).toBeGreaterThan(0);
     expect(result.links).toMatchObject({
         1: { text: 'Saved PromptManager text' }, 2: { text: 'Saved PromptManagerText text' }, 3: { text: 'Saved PromptManagerFocused text' },
         4: { text: 'Saved PromptManagerSplit text' }, 5: { text: 'Saved PromptManagerPair text', negative_text: 'Saved pair negative' },

tests/fixtures/legacy_contract_snapshot.json

@@ -1,5 +1,7 @@
 {
   "routes": [
+    "DELETE /prompt_manager/collections/items",
+    "DELETE /prompt_manager/collections/{collection_id}",
     "DELETE /prompt_manager/delete/{prompt_id}",
     "DELETE /prompt_manager/image-annotations/capture/{token}",
     "DELETE /prompt_manager/images/category",
@@ -15,6 +17,8 @@
     "GET /prompt_manager/backup",
     "GET /prompt_manager/backup/status",
     "GET /prompt_manager/categories",
+    "GET /prompt_manager/collections",
+    "GET /prompt_manager/collections/membership",
     "GET /prompt_manager/diagnostics",
     "GET /prompt_manager/export",
     "GET /prompt_manager/gallery",
@@ -92,6 +96,7 @@
     "GET /prompt_manager/vocabulary/tags/{tag_id}/impact",
     "GET /prompt_manager/web",
     "GET /prompt_manager/workspace",
+    "PATCH /prompt_manager/collections/{collection_id}",
     "PATCH /prompt_manager/library/{positive_id}",
     "PATCH /prompt_manager/pairs/{pair_id}",
     "PATCH /prompt_manager/saved-views/{view_id}",
@@ -106,6 +111,7 @@
     "POST /prompt_manager/bulk/delete",
     "POST /prompt_manager/bulk/tags",
     "POST /prompt_manager/cleanup",
+    "POST /prompt_manager/collections/items",
     "POST /prompt_manager/delete_duplicate_images",
     "POST /prompt_manager/diagnostics/test-link",
     "POST /prompt_manager/gallery/rescan",

tests/js/save-to-library.test.js

@@ -0,0 +1,68 @@
+// #90: the node's save_to_library switch (web/js/save-to-library.js).
+const test = require('node:test');
+const assert = require('node:assert/strict');
+const source = import('../../web/js/save-to-library.js');
+
+function classicNode(saved) {
+    const names = ['text', 'category', 'tags', 'search_text', 'prepend_text', 'append_text', 'save_to_library', 'prompt_manager_search_ui'];
+    const widgets = names.map(name => ({name, value: name === 'save_to_library' ? true : ''}));
+    const node = {widgets, widgets_values: saved ? [...saved] : undefined};
+    // What a host configure does: each saved value goes to the widget at its index.
+    (saved || []).forEach((value, index) => { if (widgets[index]) widgets[index].value = value; });
+    return node;
+}
+
+test('an older workflow stored the editor slot where the switch is now: it loads on', async () => {
+    const {normalizeSaveWidget, savesToLibrary} = await source;
+    const node = classicNode(['a cat', 'c', 't', '', '', '', '']);
+    assert.equal(node.widgets[6].value, '', 'the host put the old DOM slot value into the switch');
+    assert.equal(normalizeSaveWidget(node), true);
+    assert.equal(node.widgets[6].value, true);
+    assert.equal(node.widgets_values[6], true);
+    assert.equal(savesToLibrary(node), true);
+    assert.deepEqual(node.widgets_values.slice(0, 6), ['a cat', 'c', 't', '', '', ''], 'no earlier value moved');
+});
+
+test('null, undefined and missing slots load on; saved booleans are kept', async () => {
+    const {normalizeSaveWidget, savesToLibrary} = await source;
+    for (const value of [null, undefined, 0, 'false']) {
+        const node = classicNode(['a', '', '', '', '', '', value]);
+        normalizeSaveWidget(node);
+        assert.equal(savesToLibrary(node), true, String(value));
+    }
+    const short = classicNode(['a', '', '', '', '', '']);
+    assert.equal(normalizeSaveWidget(short), false);
+    assert.equal(savesToLibrary(short), true);
+    const off = classicNode(['a', '', '', '', '', '', false]);
+    assert.equal(normalizeSaveWidget(off), false);
+    assert.equal(savesToLibrary(off), false);
+});
+
+test('a node without the widget (an older backend) saves', async () => {
+    const {normalizeSaveWidget, savesToLibrary} = await source;
+    assert.equal(savesToLibrary({widgets: [{name: 'text', value: ''}]}), true);
+    assert.equal(normalizeSaveWidget({widgets: []}), false);
+    assert.equal(savesToLibrary(undefined), true);
+});
+
+test('the switch moves after the editor fallbacks so every frontend has one order', async () => {
+    const {placeSaveWidgetLast} = await source;
+    const widgets = ['text', 'category', 'tags', 'search_text', 'save_to_library', 'prepend_text', 'append_text'].map(name => ({name}));
+    const node = {widgets};
+    assert.equal(placeSaveWidgetLast(node), true);
+    assert.deepEqual(node.widgets.map(widget => widget.name), ['text', 'category', 'tags', 'search_text', 'prepend_text', 'append_text', 'save_to_library']);
+    assert.equal(placeSaveWidgetLast(node), false, 'already last');
+    assert.equal(placeSaveWidgetLast({widgets: [{name: 'text'}]}), false);
+});
+
+test('setting the switch updates the widget, its saved value and notifies the host', async () => {
+    const {setSaveToLibrary, savesToLibrary} = await source;
+    const calls = [];
+    const node = classicNode(['a', '', '', '', '', '', true]);
+    node.widgets[6].callback = value => calls.push(value);
+    setSaveToLibrary(node, false);
+    assert.equal(savesToLibrary(node), false);
+    assert.equal(node.widgets_values[6], false);
+    assert.deepEqual(calls, [false]);
+    setSaveToLibrary({widgets: []}, false);
+});

tests/js/workspace-file-inspector.test.js

@@ -383,3 +383,32 @@ test('the D117 wrap rule targets only the settings summary button', async () =>
     assert.deepEqual(rules.map(line => line.trim()), ['.ca-tools #viewer-details .settings-summary { white-space: normal; /* D117 */ }']);
     assert.doesNotMatch(css, /#viewer-details > div > button/);
 });
+
+test('category chip names its origin visibly and shows an own category for any link status (#332 D146)', async () => {
+    const base = await localRecord(localPng('a.png'));
+    const chipTexts = record => {
+        const {f, done} = renderFull(record);
+        try {
+            return f.walk(f.container, n => /(^| )chip( |$)/.test(n.className ?? '')).map(n => n.textContent);
+        } finally { done(); }
+    };
+    const cases = [
+        [{status:'none', category:'Scans', category_scope:'image'}, 'Image category: Scans · this image'],
+        [{status:'ambiguous', category:'Scans', category_scope:'image'}, 'Image category: Scans · this image'],
+        [{status:'linked', category:'Portraits', category_scope:'image'}, 'Image category: Portraits · this image'],
+        [{status:'linked', category:'Landscapes', category_scope:'positive'}, 'Positive prompt category: Landscapes · prompt'],
+        [{status:'linked', category:'Recipes', category_scope:'pair'}, 'Recipe category: Recipes · recipe'],
+    ];
+    for (const [library, text] of cases) {
+        const chips = chipTexts({...base, library:{...base.library, tags:[], tags_scope:null, ...library}});
+        assert.ok(chips.includes(text), `${JSON.stringify(library)} -> ${chips}`);
+    }
+    for (const library of [
+        {status:'none', category:'Inherited', category_scope:'positive'},
+        {status:'linked', category:null, category_scope:null},
+        {status:'none', category:null, category_scope:null},
+    ]) {
+        const chips = chipTexts({...base, library:{...base.library, tags:[], tags_scope:null, ...library}});
+        assert.equal(chips.some(chip => /category: /.test(chip)), false, JSON.stringify(library));
+    }
+});

tests/js/workspace-images-category.test.js

@@ -1,53 +1,61 @@
 const test = require('node:test');
 const assert = require('node:assert/strict');
+
+// #332 D140: Set category writes each selected image's own category; Use prompt category removes it.
 const source = import('../../web/js/workspace/images-category.mjs');
 
-const linked = (id, extra = {}) => ({path:`p${id}.png`, linked:{status:'linked', positive_id:id, ...extra}});
-
-test('targets are the distinct linked positives, in selection order, and the rest count as skipped', async () => {
-    const {categoryTargets} = await source;
-    const images = [
-        linked(7), linked(3), linked(7),
-        {path:'u.png', linked:{status:'unlinked', positive_id:null}},
-        {path:'a.png', linked:{status:'ambiguous', positive_id:4}},
-        {path:'v.png', linked:{status:'unavailable', positive_id:null}},
-        {path:'none.png'},
-        {path:'bad.png', linked:{status:'linked', positive_id:'9'}},
-        {path:'zero.png', linked:{status:'linked', positive_id:0}},
-        {path:'frac.png', linked:{status:'linked', positive_id:1.5}},
-        null,
-    ];
-    assert.deepEqual(categoryTargets(images), {promptIds:[7, 3], skipped:8});
-    assert.deepEqual(categoryTargets([]), {promptIds:[], skipped:0});
-    assert.deepEqual(categoryTargets([linked(2)]), {promptIds:[2], skipped:0});
-});
-
-test('the outcome toast counts the prompts and the skipped images', async () => {
-    const {categoryOutcome} = await source;
-    assert.equal(categoryOutcome(3, 0), 'Category set on 3 prompts.');
-    assert.equal(categoryOutcome(1, 0), 'Category set on 1 prompt.');
-    assert.equal(categoryOutcome(2, 4), 'Category set on 2 prompts. 4 selected images had no linked prompt and were skipped.');
-    assert.equal(categoryOutcome(1, 1), 'Category set on 1 prompt. 1 selected image had no linked prompt and was skipped.');
-});
-
-test('the dialog and failure copy is the plan\'s', async () => {
-    const {CATEGORY_DIALOG, CATEGORY_NONE_LINKED, CATEGORY_FAILED} = await source;
-    assert.deepEqual(CATEGORY_DIALOG, {title:'Categorize selected images', label:'Category', action:'Apply'});
-    assert.equal(CATEGORY_NONE_LINKED, "None of the selected images is linked to a saved prompt, so there is no prompt category to set.");
+test('the dialog keeps the concept\'s simpleInput words and names the routes', async () => {
+    const {CATEGORY_DIALOG, USE_PROMPT_CATEGORY, REMOVE_DIALOG, CATEGORY_URL, IMAGE_CATEGORIES_URL, CATEGORY_FAILED,
+        CATEGORY_MAX_LENGTH, MAX_CATEGORY_IMAGES} = await source;
+    assert.deepEqual({...CATEGORY_DIALOG}, {title:'Categorize selected images', label:'Category', action:'Apply'});
+    assert.equal(USE_PROMPT_CATEGORY, 'Use prompt category');
+    assert.deepEqual({...REMOVE_DIALOG}, {title:'Use prompt category', action:'Remove'});
+    assert.equal(CATEGORY_URL, '/prompt_manager/images/category');
+    assert.equal(IMAGE_CATEGORIES_URL, '/prompt_manager/images/categories');
     assert.equal(CATEGORY_FAILED, 'Category could not be set. Try again.');
-    assert.ok(Object.isFrozen(CATEGORY_DIALOG));
+    assert.equal(CATEGORY_MAX_LENGTH, 100);
+    assert.equal(MAX_CATEGORY_IMAGES, 500);
 });
 
-test('the updated count is the server\'s when it is a count, else the number of prompts posted', async () => {
-    const {updatedCount} = await source;
-    assert.equal(updatedCount({updated_count:0}, 3), 0);
-    assert.equal(updatedCount({updated_count:2}, 3), 2);
-    for (const result of [{}, {updated_count:-1}, {updated_count:1.5}, {updated_count:'2'}, null, undefined]) {
-        assert.equal(updatedCount(result, 3), 3, JSON.stringify(result));
+test('a set counts unchanged files as set and names files that could not be found', async () => {
+    const {setOutcome} = await source;
+    assert.equal(setOutcome({updated:1, unchanged:0, not_found:0}, 'Portraits'), 'Category “Portraits” set on 1 image.');
+    assert.equal(setOutcome({updated:1, unchanged:2, not_found:0}, 'Scans'), 'Category “Scans” set on 3 images.');
+    assert.equal(setOutcome({updated:2, unchanged:0, not_found:1}, 'Scans'), 'Category “Scans” set on 2 images. 1 could not be found.');
+    assert.equal(setOutcome({updated:0, unchanged:1, not_found:3}, 'Scans'), 'Category “Scans” set on 1 image. 3 could not be found.');
+    assert.equal(setOutcome({}, 'Scans'), 'Category “Scans” set on 0 images.');
+});
+
+test('a removal says how many lost their own category, or that none had one', async () => {
+    const {clearOutcome, NONE_HAD_OWN} = await source;
+    assert.equal(clearOutcome({cleared:1}), 'Removed the image category from 1 image.');
+    assert.equal(clearOutcome({cleared:4}), 'Removed the image category from 4 images.');
+    assert.equal(clearOutcome({cleared:0}), NONE_HAD_OWN);
+    assert.equal(NONE_HAD_OWN, 'None of the selected images had its own category.');
+});
+
+test('the confirmation states the count', async () => {
+    const {removeQuestion} = await source;
+    assert.equal(removeQuestion(1), "Remove the image category from the 1 selected image? Each one then shows its prompt's " +
+        'category, or none if it has no linked prompt. This cannot be undone.');
+    assert.match(removeQuestion(1200), /^Remove the image category from the 1,200 selected images\? /);
+});
+
+test('coded failures have their own sentences, anything else the fixed one', async () => {
+    const {failureMessage, NONE_FOUND, SOURCES_CHANGED, CATEGORY_FAILED} = await source;
+    assert.equal(failureMessage({code:'files-not-found'}), NONE_FOUND);
+    assert.equal(NONE_FOUND, 'None of the selected images could be found. Refresh the images and try again.');
+    assert.equal(failureMessage({code:'source-changed'}), SOURCES_CHANGED);
+    assert.equal(SOURCES_CHANGED, 'The sources changed. Refresh the images and try again.');
+    assert.equal(failureMessage({code:'too-many-images', message:'Select at most 500 images.'}), 'Select at most 500 images.');
+    for (const error of [null, undefined, new Error('x'), {code:'other', message:'secret'}]) {
+        assert.equal(failureMessage(error), CATEGORY_FAILED);
     }
 });
 
-test('the category field is limited to the server\'s 100 characters', async () => {
-    const {CATEGORY_MAX_LENGTH} = await source;
-    assert.equal(CATEGORY_MAX_LENGTH, 100);
+test('the toolbar list merges prompt and image categories once each', async () => {
+    const {mergeCategories} = await source;
+    assert.deepEqual(mergeCategories(['Landscapes', 'Portraits'], ['Scans', 'Landscapes']), ['Landscapes', 'Portraits', 'Scans']);
+    assert.deepEqual(mergeCategories(['A', '', '  ', 7], null), ['A']);
+    assert.deepEqual(mergeCategories(undefined, ['landscapes', 'Landscapes']), ['landscapes', 'Landscapes']);
 });

tests/js/workspace-images-tagging-bulk.test.js

@@ -135,6 +135,41 @@ test('a model that is not downloaded is named with the server sentence', () => {
     assert.throws(() => modelUnavailable({}, 'nope'), TypeError);
 });
 
+test('a downloaded model whose backend package is missing names the package and the command (#424)', () => {
+    const { modelUnavailable, needsPackages } = state;
+    const gguf = { downloaded: true, name: 'GGUF (Recommended)', label: 'GGUF JoyCaption', missing_packages: ['llama-cpp-python'],
+        install_command: 'pip install llama-cpp-python',
+        install_note: 'For a CUDA build: CMAKE_ARGS="-DGGML_CUDA=on" pip install llama-cpp-python' };
+    const sentence = 'GGUF JoyCaption needs llama-cpp-python. Install it into ComfyUI\'s Python: pip install llama-cpp-python. '
+        + 'For a CUDA build: CMAKE_ARGS="-DGGML_CUDA=on" pip install llama-cpp-python';
+    assert.equal(modelUnavailable({ gguf }, 'gguf'), sentence);
+    // Not downloaded is named first, in its own sentence.
+    assert.equal(modelUnavailable({ gguf: { ...gguf, downloaded: false } }, 'gguf'),
+        'GGUF (Recommended) is not downloaded. Download it in System › Auto-tagging.');
+    assert.equal(modelUnavailable({ gguf: { ...gguf, missing_packages: [] } }, 'gguf'), '');
+    assert.equal(needsPackages('HuggingFace JoyCaption', { missing_packages: ['accelerate', 'bitsandbytes'] }),
+        'HuggingFace JoyCaption needs accelerate and bitsandbytes. Install them into ComfyUI\'s Python: pip install accelerate bitsandbytes.');
+    assert.equal(needsPackages('X', { missing_packages: ['a', 'b', 'c'], install_command: 'pip install a b c' }),
+        'X needs a, b and c. Install them into ComfyUI\'s Python: pip install a b c.');
+    assert.equal(needsPackages('X', null), '');
+    assert.equal(needsPackages('X', { missing_packages: [1, ''] }), '');
+    // Without the server's label the option's own name is used.
+    assert.ok(modelUnavailable({ 'wd14-vit': { downloaded: true, missing_packages: ['onnxruntime'] } }, 'wd14-vit')
+        .startsWith('WD14 ViT needs onnxruntime.'));
+});
+
+test('the bulk dialog keeps Start disabled while the chosen model misses a package (#424)', async () => {
+    const { createTaggingOptionsController } = state;
+    const models = { 'wd14-vit': { downloaded: true, label: 'WD14 ViT', missing_packages: ['onnxruntime'],
+        install_command: 'pip install onnxruntime' } };
+    const options = createTaggingOptionsController({ api: { get: async () => ({ success: true, models, model_type: 'wd14-vit' }) } });
+    await options.open();
+    assert.equal(options.state.model, 'wd14-vit');
+    assert.equal(options.state.canStart, false);
+    assert.equal(options.state.unavailable, 'WD14 ViT needs onnxruntime. Install it into ComfyUI\'s Python: pip install onnxruntime.');
+    options.dispose();
+});
+
 test('the bulk options start from the server defaults and wait for them before Start', async () => {
     const { createTaggingOptionsController } = state;
     let answer;

tests/js/workspace-inspection-state.test.js

@@ -278,6 +278,7 @@ test('record rejects root path media and invalid versioned fields with retryable
         data => { data.resolution.sampler_id = {}; },
         data => { data.library.tags_scope = 'recipe'; },
         data => { data.library.category_scope = 'recipe'; },
+        data => { data.library.tags_scope = 'image'; },
         data => { delete data.library.category_scope; },
         data => { delete data.library.tags_scope; },
     ]) {
@@ -942,3 +943,14 @@ test("names the server quotes (parentheses, '!*') still match their original and
     assert.equal(record.download.url, entry.url);
     assert.deepEqual(record.preview, { url: entry.thumbnail_url });
 });
+
+test('record accepts an image\'s own category on any link status (#332)', () => {
+    const toRecord = contract('toInspectionRecord');
+    for (const status of ['linked', 'none', 'ambiguous', 'unavailable']) {
+        const data = envelope();
+        Object.assign(data.library, { status, category: 'Scans', category_scope: 'image' });
+        const record = toRecord(data, image(), extractGenerationFacts);
+        assert.equal(record.library.category, 'Scans');
+        assert.equal(record.library.category_scope, 'image');
+    }
+});

tests/js/workspace-system-backups.test.js

@@ -63,9 +63,9 @@ test('status line shows only server-reported backups (D103)', async () => {
 
 const PREVIEW = Object.freeze({
     success: true, token: 'r'.repeat(43), expires_at: '2026-10-25T12:15:00+00:00',
-    summary: { prompts: 1234, positives: 900, pairs: 3, archived_pairs: 0, image_links: 56, runs: 1, saved_views: 2 },
+    summary: { prompts: 1234, positives: 900, pairs: 3, archived_pairs: 0, image_links: 56, runs: 1, saved_views: 2, collections: 1 },
     schema: { migrated: true, library_state: true }, integrity: 'ok', documents: { repaired: 0, cleared: 0, images: 0 },
-    live: { prompts: 1, positives: 1, pairs: 0, archived_pairs: 0, image_links: 0, runs: 0, saved_views: 0 }, size: 4096,
+    live: { prompts: 1, positives: 1, pairs: 0, archived_pairs: 0, image_links: 0, runs: 0, saved_views: 0, collections: 0 }, size: 4096,
     safety_backup: { destination_label: 'prompts.db.backup_<date and time>' },
 });
 

tests/js/workspace-system-tagging.test.js

@@ -482,6 +482,60 @@ test('Tag output images › Review previews the drafts, else says why; a refused
     } finally { scope.dispose(); dom.restore(); }
 });
 
+// #424: GGUF on disk without llama-cpp-python.
+const NEEDS_LLAMA = { missing_packages: ['llama-cpp-python'], install_command: 'pip install llama-cpp-python',
+    install_note: 'For a CUDA build: CMAKE_ARGS="-DGGML_CUDA=on" pip install llama-cpp-python' };
+const NEEDS_SENTENCE = 'GGUF JoyCaption needs llama-cpp-python. Install it into ComfyUI\'s Python: pip install llama-cpp-python. '
+    + 'For a CUDA build: CMAKE_ARGS="-DGGML_CUDA=on" pip install llama-cpp-python';
+
+test('a model on disk whose backend package is missing never reads as ready (#424)', async () => {
+    const { modelDescription, readStatus, batchCloseNote, STATUS_COPY } = await tagging;
+    const status = readStatus({ ...MODELS, models: { ...MODELS.models,
+        gguf: entry('GGUF JoyCaption', '~4.5 GB', { ...NEEDS_LLAMA }),
+        'wd14-vit': entry('WD14 ViT', '~400 MB', { downloaded: false, missing_packages: ['onnxruntime'],
+            install_command: 'pip install onnxruntime', install_note: 'Use onnxruntime-gpu for CUDA.' }),
+        hf: entry('HuggingFace JoyCaption', '~16 GB', { downloaded: false, model_exists: undefined,
+            missing_packages: ['accelerate', 'bitsandbytes'], install_command: 'pip install accelerate bitsandbytes' }) } });
+    const ready = modelDescription(status, 'gguf');
+    assert.notEqual(ready, STATUS_COPY.downloaded);
+    assert.equal(ready, 'Downloaded. Needs llama-cpp-python: pip install llama-cpp-python. '
+        + 'For a CUDA build: CMAKE_ARGS="-DGGML_CUDA=on" pip install llama-cpp-python');
+    // Not downloaded reads differently, and still says what it will need.
+    assert.equal(modelDescription(status, 'wd14-vit'), 'Not downloaded. Download size ~400 MB. Also needs onnxruntime.');
+    assert.equal(modelDescription(status, 'hf'), 'Not downloaded. Download size ~16 GB. Also needs accelerate and bitsandbytes.');
+    const partial = readStatus({ ...MODELS, models: { ...MODELS.models, gguf: entry('GGUF JoyCaption', '~4.5 GB',
+        { downloaded: false, model_exists: true, mmproj_exists: false, ...NEEDS_LLAMA }) } });
+    assert.equal(modelDescription(partial, 'gguf'), 'Partly downloaded. Download again to finish. Also needs llama-cpp-python.');
+    // Untrusted shapes are ignored; a missing command is built from the names.
+    const odd = readStatus({ ...MODELS, models: { ...MODELS.models, gguf: entry('GGUF JoyCaption', '~4.5 GB',
+        { missing_packages: ['llama-cpp-python', 3, ''], install_command: 7, install_note: {} }) } });
+    assert.equal(modelDescription(odd, 'gguf'), 'Downloaded. Needs llama-cpp-python: pip install llama-cpp-python.');
+    const junk = readStatus({ ...MODELS, models: { ...MODELS.models, gguf: entry('GGUF JoyCaption', '~4.5 GB',
+        { missing_packages: 'llama-cpp-python' }) } });
+    assert.equal(modelDescription(junk, 'gguf'), STATUS_COPY.downloaded);
+    // The batch review says why it cannot start, in the server's sentence.
+    assert.equal(batchCloseNote({ ...PLAN, model: { type: 'gguf', label: 'GGUF JoyCaption', downloaded: true, ...NEEDS_LLAMA } }),
+        NEEDS_SENTENCE);
+    assert.equal(batchCloseNote({ ...PLAN, model: { ...PLAN.model, downloaded: false, ...NEEDS_LLAMA } }),
+        'WD14 SwinV2 is not downloaded. Download it first.');
+});
+
+test('Review explains a missing backend package instead of reviewing (#424)', async () => {
+    const models = { ...MODELS, models: { ...MODELS.models, gguf: entry('GGUF JoyCaption', '~4.5 GB', { ...NEEDS_LLAMA }) } };
+    const { dom, scope, button, byId, calls } = await mounted({ models });
+    try {
+        byId('tag-model').value = 'gguf'; await byId('tag-model').dispatch('change');
+        const review = button('Review');
+        assert.equal(review.getAttribute('aria-disabled'), 'true');
+        assert.equal(review.dataset.reason, NEEDS_SENTENCE);
+        await review.dispatch('click');
+        assert.equal(calls.toasts.at(-1), NEEDS_SENTENCE);
+        assert.deepEqual(calls.post, []);
+        // Download stays offered: files and packages are separate.
+        assert.equal(button('Download again').getAttribute('aria-disabled'), null);
+    } finally { scope.dispose(); dom.restore(); }
+});
+
 test('settingRefusal marks only an Auto-tagging save\'s plain 400 as invalid-setting (CUJ 2a)', async () => {
     const { settingRefusal } = await import('../../web/js/workspace/system.mjs');
     const refusal = (extra = {}) => Object.assign(new Error('General-tag confidence must be between 0 and 1.'),

tests/js/workspace-toolbar.test.js

@@ -567,6 +567,43 @@ test('an input dialog with a fallback applies it for an empty value', async () =
     }
 });
 
+// #332: Set category's field suggests the categories under its own datalist id, and "Use prompt category" sits
+// between Cancel and Apply, closing the dialog before it runs.
+test('an input dialog offers suggestions and a secondary action', async () => {
+    const { openInputDialog, closeWorkspaceDialog, INPUT_SUGGESTIONS_ID } = await dialogsModule;
+    const dom = fakeDOM();
+    const scope = (await coreModule).createScope();
+    try {
+        const calls = [];
+        openInputDialog({ scope, title: 'Categorize selected images', label: 'Category', action: 'Apply', maxLength: 100,
+            suggestions: ['Landscapes', 'Scans'], secondary: { label: 'Use prompt category', onClick: () => calls.push('secondary') },
+            onApply: value => calls.push(value) });
+        assert.equal(INPUT_SUGGESTIONS_ID, 'workspace-input-suggestions');
+        const body = byId(dom, 'modal-body');
+        const [field, list, actions] = body.children;
+        const input = field.children[1];
+        assert.equal(input.getAttribute('list'), 'workspace-input-suggestions');
+        assert.equal(list.tagName.toLowerCase(), 'datalist');
+        assert.equal(list.id, 'workspace-input-suggestions');
+        assert.deepEqual(list.children.map(option => option.value), ['Landscapes', 'Scans']);
+        assert.deepEqual(actions.children.map(node => node.textContent), ['Cancel', 'Use prompt category', 'Apply']);
+        actions.children[1].click();
+        assert.deepEqual(calls, ['secondary']);
+        // No suggestions: no datalist and no list attribute.
+        openInputDialog({ scope, title: 'T', label: 'L', action: 'A', onApply() {} });
+        const [plainField, plainActions] = byId(dom, 'modal-body').children;
+        assert.equal(plainField.children[1].getAttribute('list'), null);
+        assert.deepEqual(plainActions.children.map(node => node.textContent), ['Cancel', 'A']);
+        assert.throws(() => openInputDialog({ scope, title: 'T', label: 'L', action: 'A', suggestions: [1], onApply() {} }), TypeError);
+        assert.throws(() => openInputDialog({ scope, title: 'T', label: 'L', action: 'A', secondary: { label: '' }, onApply() {} }), TypeError);
+        assert.throws(() => openInputDialog({ scope, title: 'T', label: 'L', action: 'A', secondary: { label: 'X' }, onApply() {} }), TypeError);
+    } finally {
+        closeWorkspaceDialog();
+        scope.dispose();
+        dom.restore();
+    }
+});
+
 test('dialogs restore focus to the trigger on Escape', async () => {
     const { openRefineDialog, openImportDialog, openCopyFallback, closeWorkspaceDialog } = await dialogsModule;
     const dom = fakeDOM();

tests/js/workspace-vocabulary-state.test.js

@@ -553,3 +553,16 @@ test('delete text names the tag, its real links and what stays unchanged', async
     assert.equal(deleteTagText('unused', zero),
         '“unused” is not linked to any saved prompt or image. Deleting it removes only the name.');
 });
+
+test('a category row counts prompts and images, and an image-only row browses Images (#332 D143)', async () => {
+    const {categoryRowText, categoryBrowseTarget, categoryPreviewText} = await import('../../web/js/workspace/vocabulary-state.mjs');
+    assert.equal(categoryRowText({value: 'L', prompt_count: 3}), '3 positive prompts');
+    assert.equal(categoryRowText({value: 'L', prompt_count: 1, image_count: 0}), '1 positive prompt');
+    assert.equal(categoryRowText({value: 'L', prompt_count: 3, image_count: 2}), '3 positive prompts · 2 images');
+    assert.equal(categoryRowText({value: 'S', prompt_count: 0, image_count: 1}), '1 image');
+    assert.equal(categoryBrowseTarget({value: 'L', prompt_count: 3, image_count: 2}), 'library');
+    assert.equal(categoryBrowseTarget({value: 'S', prompt_count: 0, image_count: 2}), 'images');
+    const impact = {positive_prompts: 3, other_prompts: 0, negative_variants: 0, archived_negative_variants: 0, images: 2};
+    assert.equal(categoryPreviewText({source: 'Landscapes', target: 'Landscape', target_exists: false, impact}),
+        '“Landscapes” is used by 3 positive prompts and 2 images. Apply renames it to “Landscape”.');
+});

tests/test_api_annotation_suggest.py

@@ -29,6 +29,7 @@ except ImportError:
     model_gate = None
 
 from py import config as config_module
+from py.autotag import MissingBackendPackages, package_fields
 
 AutotagConfig = getattr(config_module, "AutotagConfig", None)
 
@@ -37,6 +38,11 @@ WAIT = 30.0
 WD14 = {"general_threshold": 0.35, "character_threshold": 0.85}
 LLM = {"instructions": "Generate tags."}
 UNAVAILABLE = "Download it in System › Auto-tagging."
+NEEDS_LLAMA = (
+    "GGUF JoyCaption needs llama-cpp-python. Install it into ComfyUI's Python: "
+    'pip install llama-cpp-python. For a CUDA build: CMAKE_ARGS="-DGGML_CUDA=on" '
+    "pip install llama-cpp-python"
+)
 
 SCORED = [("sandstone", 0.98), ("desert", 0.96)] + [
     (f"tag_{i:03d}", round(0.9 - i * 0.005, 4)) for i in range(150)
@@ -59,6 +65,8 @@ class FakeTagger:
             "hf": {},
         }
         self.downloaded = {name: True for name in self.models_config}
+        # pip names of the backend packages that cannot be imported (#424)
+        self.missing = {name: [] for name in self.models_config}
         self.loaded = None
         self.load_calls = []
         self.unload_calls = 0
@@ -71,7 +79,11 @@ class FakeTagger:
 
     def get_models_status(self):
         return {
-            name: {"downloaded": self.downloaded[name], "model_path": None}
+            name: {
+                "downloaded": self.downloaded[name],
+                "model_path": None,
+                **package_fields(name, self.missing[name]),
+            }
             for name in self.models_config
         }
 
@@ -315,6 +327,37 @@ class TestModelAvailability(SuggestTestCase):
         self.assertEqual(self.tagger.calls, [])
         self.assertIsNone(model_gate.holder())
 
+    async def test_missing_backend_package_is_refused_before_the_model(self):
+        observed = await self.observed()
+        self.tagger.missing["gguf"] = ["llama-cpp-python"]
+
+        status, body = await self.suggest(self.suggest_body(observed, "gguf"))
+
+        self.assertEqual((status, body["code"]), (409, "model-unavailable"))
+        self.assertEqual(body["success"], False)
+        self.assertEqual(body["error"], NEEDS_LLAMA)
+        self.assertEqual(self.tagger.load_calls, [])
+        self.assertIsNone(model_gate.holder())
+        # Not downloaded still reads as before, whatever the packages.
+        self.tagger.downloaded["gguf"] = False
+        status, body = await self.suggest(self.suggest_body(observed, "gguf"))
+        self.assertIn("is not downloaded", body["error"])
+
+    async def test_a_package_lost_after_the_check_is_the_same_409_without_a_traceback(
+        self,
+    ):
+        observed = await self.observed()
+        self.tagger.load_error = MissingBackendPackages("gguf", ["llama-cpp-python"])
+
+        with self.assertNoLogs(self.api.logger, level="ERROR"):
+            status, body = await self.suggest(self.suggest_body(observed, "gguf"))
+
+        self.assertEqual((status, body["code"]), (409, "model-unavailable"))
+        self.assertEqual(body["error"], NEEDS_LLAMA)
+        self.assertIsNone(model_gate.holder())
+        entry = self.api._captures._entries[observed["capture"]["token"]]
+        self.assertEqual(entry.uses, 0)
+
     async def test_load_failure_is_a_generic_error_and_frees_everything(self):
         observed = await self.observed()
         self.tagger.load_error = RuntimeError(f"cannot load {self.output_dir}")

tests/test_api_autotag.py

@@ -73,6 +73,7 @@ class FakeAutotagService:
         self.status_calls = 0
         self.status_threads = []
         self.status_paths = {}
+        self.status_extra = {}
         self.suggest_calls = 0
 
     def get_models_status(self):
@@ -82,6 +83,7 @@ class FakeAutotagService:
             name: {
                 "downloaded": self.downloaded,
                 "model_path": self.status_paths.get(name),
+                **self.status_extra.get(name, {}),
             }
             for name in self.models_config
         }
@@ -1111,6 +1113,24 @@ class TestModelStatus(AutotagAPITestCase):
         leaked = [text for text in _strings(data) if _absolute(text)]
         self.assertEqual(leaked, [])
 
+    async def test_models_status_names_missing_backend_packages(self):
+        fields = {
+            "missing_packages": ["llama-cpp-python"],
+            "install_command": "pip install llama-cpp-python",
+            "install_note": "For a CUDA build: ...",
+        }
+        self.service.status_extra = {"gguf": fields}
+        models = (await self._models())["models"]
+        self.assertTrue(models["gguf"]["downloaded"])
+        self.assertEqual(
+            {key: models["gguf"][key] for key in fields},
+            fields,
+        )
+        self.assertEqual(
+            {key: models["wd14-vit"][key] for key in fields},
+            {"missing_packages": [], "install_command": None, "install_note": None},
+        )
+
     async def test_models_status_adds_label_size_label_and_loaded(self):
         self.service._loaded_type = "wd14-vit"
         data = await self._models()
@@ -1669,6 +1689,10 @@ class TestTagOutputImagesPreview(BatchJobTestCase):
                 "label": "WD14 ViT",
                 "size_label": "~400 MB",
                 "downloaded": True,
+                # #424: the backend packages it cannot import, none here.
+                "missing_packages": [],
+                "install_command": None,
+                "install_note": None,
             },
         )
         self.assertEqual(skip["existing"], "skip")

tests/test_api_collections.py

@@ -0,0 +1,601 @@
+"""Collections (#331): the routes under /prompt_manager/collections."""
+
+import json
+import ntpath
+import sqlite3
+from unittest.mock import patch
+
+try:
+    from test_api_images import ImageAPITestCase, make_png
+    from open_handles import assert_closed
+except ImportError:
+    from tests.test_api_images import ImageAPITestCase, make_png
+    from tests.open_handles import assert_closed
+
+BASE = "/prompt_manager/collections"
+ITEMS = BASE + "/items"
+MEMBERSHIP = BASE + "/membership"
+SOURCES_CHANGED = "The sources changed. Refresh the images and try again."
+
+
+class CollectionRoutesTestCase(ImageAPITestCase):
+    async def tearDownAsync(self):
+        self.api.db.close_all()
+        assert_closed(self, self._temp_db.name, self._output_tmp.name)
+        await super().tearDownAsync()
+
+    def make(self, *names, root=None):
+        for name in names:
+            make_png((root or self.output_dir) / name)
+
+    def scope(self, index=0):
+        roots = list(self.api._get_all_output_dirs())
+        return self.api._gallery_root_scope(roots[index], index)
+
+    def item(self, path, index=0, scope=None):
+        return {
+            "root_index": index,
+            "root_scope": scope or self.scope(index),
+            "path": path,
+        }
+
+    def two_sources(self):
+        second = self.comfy_dir / "second"
+        second.mkdir()
+        self.api._get_all_output_dirs = lambda: [self.output_dir, second]
+        return second
+
+    async def call(self, method, url, body=None, status=200, **params):
+        response = await self.client.request(method, url, json=body, params=params)
+        text = await response.text()
+        self.assertEqual(response.status, status, text)
+        data = json.loads(text)
+        self.assertNotIn(self.comfy_dir.as_posix(), text)
+        self.assertNotIn(str(self.comfy_dir).replace("\\", "\\\\"), text)
+        return data
+
+    async def add(self, status=200, **body):
+        return await self.call("POST", ITEMS, body, status)
+
+    async def listed(self):
+        return (await self.call("GET", BASE))["collections"]
+
+
+class TestCollectionRoutes(CollectionRoutesTestCase):
+    async def test_each_route_succeeds(self):
+        self.make("a.png", "b.png", "c.png")
+        self.assertEqual(
+            await self.call("GET", BASE), {"success": True, "collections": []}
+        )
+        made = await self.add(
+            name="  LoRa   testing ", items=[self.item("a.png"), self.item("b.png")]
+        )
+        grids = made["collection"]
+        self.assertEqual(grids["name"], "LoRa testing")
+        self.assertEqual(
+            made,
+            {
+                "success": True,
+                "collection": grids,
+                "created": True,
+                "added": 2,
+                "already": 0,
+                "not_found": 0,
+            },
+        )
+        again = await self.add(
+            collection_id=grids["id"], items=[self.item("b.png"), self.item("c.png")]
+        )
+        self.assertEqual(
+            (again["created"], again["added"], again["already"]), (False, 1, 1)
+        )
+        self.assertEqual(
+            await self.listed(),
+            [{"id": grids["id"], "name": "LoRa testing", "count": 3, "missing": 0}],
+        )
+        membership = await self.call(
+            "GET", MEMBERSHIP, root_index="0", root_scope=self.scope(), path="a.png"
+        )
+        self.assertEqual(membership, {"success": True, "collection_ids": [grids["id"]]})
+        removed = await self.call(
+            "DELETE",
+            ITEMS,
+            {"collection_id": grids["id"], "items": [self.item("a.png")]},
+        )
+        self.assertEqual(removed, {"success": True, "collection": grids, "removed": 1})
+        renamed = await self.call(
+            "PATCH", f"{BASE}/{grids['id']}", {"name": "LoRA grids"}
+        )
+        self.assertEqual(
+            renamed,
+            {"success": True, "collection": {"id": grids["id"], "name": "LoRA grids"}},
+        )
+        deleted = await self.call("DELETE", f"{BASE}/{grids['id']}")
+        self.assertEqual(deleted, {"success": True, "deleted": True})
+        self.assertEqual(await self.listed(), [])
+        # Files are never touched.
+        self.assertTrue((self.output_dir / "a.png").is_file())
+
+    async def test_names_are_unique_ignoring_case(self):
+        self.make("a.png", "b.png")
+        first = (await self.add(name="LoRa testing", items=[self.item("a.png")]))[
+            "collection"
+        ]
+        second = await self.add(name="lora TESTING", items=[self.item("b.png")])
+        self.assertEqual(second["collection"], first)
+        self.assertFalse(second["created"])
+        self.assertEqual(len(await self.listed()), 1)
+
+    async def test_bad_requests_are_400(self):
+        self.make("a.png")
+        made = await self.add(name="Set", items=[self.item("a.png")])
+        cid = made["collection"]["id"]
+        good = [self.item("a.png")]
+        for body in (
+            {"items": good},
+            {"name": "x", "collection_id": cid, "items": good},
+            {"name": "x"},
+            {"name": "x", "items": good, "extra": 1},
+            {"collection_id": "1", "items": good},
+            {"collection_id": True, "items": good},
+            {"collection_id": 0, "items": good},
+            {"collection_id": 10**10, "items": good},
+            {"name": "", "items": good},
+            {"name": "x" * 101, "items": good},
+            {"name": "bell\x07", "items": good},
+            {"name": 5, "items": good},
+            {"name": "x", "items": []},
+            {"name": "x", "items": "a.png"},
+            {"name": "x", "items": [self.item("a.png")] * 501},
+            {"name": "x", "items": [{"root_index": 0, "path": "a.png"}]},
+            {"name": "x", "items": [{**self.item("a.png"), "root_index": "0"}]},
+            {"name": "x", "items": [{**self.item("a.png"), "root_scope": "xyz"}]},
+        ):
+            with self.subTest(body=body):
+                data = await self.add(400, **body)
+                self.assertFalse(data["success"])
+        for body in ({"items": good}, {"collection_id": cid}):
+            with self.subTest(remove=body):
+                await self.call("DELETE", ITEMS, body, 400)
+        for raw in ("abc", "0", "1234567890", "-1", "1.0"):
+            with self.subTest(path_id=raw):
+                await self.call("PATCH", f"{BASE}/{raw}", {"name": "x"}, 400)
+                await self.call("DELETE", f"{BASE}/{raw}", None, 400)
+        await self.call("PATCH", f"{BASE}/{cid}", {"name": ""}, 400)
+        await self.call("PATCH", f"{BASE}/{cid}", {"label": "x"}, 400)
+        for params in (
+            {"root_index": "x", "root_scope": self.scope(), "path": "a.png"},
+            {"root_index": "0", "path": "a.png"},
+            {"root_index": "0", "root_scope": "nope", "path": "a.png"},
+        ):
+            with self.subTest(membership=params):
+                await self.call("GET", MEMBERSHIP, status=400, **params)
+        self.assertEqual((await self.listed())[0]["count"], 1)
+
+    async def test_unsafe_paths_are_403(self):
+        self.make("a.png")
+        for path in (
+            "../a.png",
+            "sub/../../a.png",
+            "C:/a.png",
+            "a\\b.png",
+            "/a.png",
+            "notes.txt",
+        ):
+            with self.subTest(path=path):
+                await self.add(403, name="Set", items=[self.item(path)])
+                await self.call(
+                    "GET",
+                    MEMBERSHIP,
+                    status=403,
+                    root_index="0",
+                    root_scope=self.scope(),
+                    path=path,
+                )
+        self.assertEqual(await self.listed(), [])
+
+    async def test_a_body_over_the_cap_is_413(self):
+        response = await self.client.post(
+            ITEMS, data=b'{"name":"x","items":[' + b" " * (1024 * 1024 + 1) + b"]}"
+        )
+        self.assertEqual(response.status, 413)
+
+    async def test_missing_collection_and_files_are_404(self):
+        self.make("a.png")
+        data = await self.add(404, collection_id=999, items=[self.item("a.png")])
+        self.assertEqual(data["code"], "collection-not-found")
+        data = await self.call(
+            "DELETE", ITEMS, {"collection_id": 999, "items": [self.item("a.png")]}, 404
+        )
+        self.assertEqual(data["code"], "collection-not-found")
+        data = await self.call("PATCH", f"{BASE}/999", {"name": "x"}, 404)
+        self.assertEqual(data["code"], "collection-not-found")
+        data = await self.call("DELETE", f"{BASE}/999", None, 404)
+        self.assertEqual(data["code"], "collection-not-found")
+        data = await self.add(404, name="New", items=[self.item("never.png")])
+        self.assertEqual(data["code"], "files-not-found")
+        self.assertEqual(await self.listed(), [])
+
+    async def test_unlisted_files_count_as_not_found(self):
+        self.make("a.png")
+        made = await self.add(
+            name="Set", items=[self.item("a.png"), self.item("later.png")]
+        )
+        self.assertEqual((made["added"], made["not_found"]), (1, 1))
+        again = await self.add(
+            collection_id=made["collection"]["id"],
+            items=[self.item("a.png"), self.item("later.png")],
+        )
+        self.assertEqual(
+            (again["added"], again["already"], again["not_found"]), (0, 1, 1)
+        )
+
+    async def test_name_taken_is_409(self):
+        self.make("a.png")
+        first = (await self.add(name="First", items=[self.item("a.png")]))["collection"]
+        second = (await self.add(name="Second", items=[self.item("a.png")]))[
+            "collection"
+        ]
+        data = await self.call(
+            "PATCH", f"{BASE}/{second['id']}", {"name": "FIRST"}, 409
+        )
+        self.assertEqual(
+            (data["code"], data["collection_id"]), ("name-taken", first["id"])
+        )
+        renamed = await self.call("PATCH", f"{BASE}/{second['id']}", {"name": "SECOND"})
+        self.assertEqual(renamed["collection"]["name"], "SECOND")
+
+    async def test_stale_scope_is_409_after_reordering_sources(self):
+        second = self.two_sources()
+        self.make("a.png")
+        self.make("b.png", root=second)
+        stale = self.item("b.png", 1)
+        made = await self.add(name="Set", items=[self.item("a.png")])
+        cid = made["collection"]["id"]
+        self.api._get_all_output_dirs = lambda: [second, self.output_dir]
+        for method, body in (
+            ("POST", {"collection_id": cid, "items": [self.item("b.png", 0), stale]}),
+            ("DELETE", {"collection_id": cid, "items": [stale]}),
+        ):
+            with self.subTest(method=method):
+                data = await self.call(method, ITEMS, body, 409)
+                self.assertEqual(data["code"], "source-changed")
+                self.assertEqual(data["error"], SOURCES_CHANGED)
+        data = await self.call(
+            "GET",
+            MEMBERSHIP,
+            status=409,
+            root_index="1",
+            root_scope=stale["root_scope"],
+            path="b.png",
+        )
+        self.assertEqual(data["code"], "source-changed")
+        # Nothing was written by the refused add; the member follows its folder.
+        listed = await self.listed()
+        self.assertEqual((listed[0]["count"], listed[0]["missing"]), (1, 0))
+        membership = await self.call(
+            "GET", MEMBERSHIP, root_index="1", root_scope=self.scope(1), path="a.png"
+        )
+        self.assertEqual(membership["collection_ids"], [cid])
+
+    async def test_counts_exclude_a_missing_member(self):
+        second = self.two_sources()
+        self.make("a.png", "b.png")
+        self.make("c.png", root=second)
+        made = await self.add(
+            name="Grids",
+            items=[self.item("a.png"), self.item("b.png"), self.item("c.png", 1)],
+        )
+        other = await self.add(name="Other", items=[self.item("c.png", 1)])
+        (self.output_dir / "b.png").unlink()
+        self.api.invalidate_gallery_cache()
+        listed = await self.listed()
+        self.assertEqual(
+            listed,
+            [
+                {
+                    "id": made["collection"]["id"],
+                    "name": "Grids",
+                    "count": 2,
+                    "missing": 1,
+                },
+                {
+                    "id": other["collection"]["id"],
+                    "name": "Other",
+                    "count": 1,
+                    "missing": 0,
+                },
+            ],
+        )
+        # The member is removable although it is no longer listed.
+        removed = await self.call(
+            "DELETE",
+            ITEMS,
+            {"collection_id": made["collection"]["id"], "items": [self.item("b.png")]},
+        )
+        self.assertEqual(removed["removed"], 1)
+
+    async def test_counts_are_null_when_a_source_scan_fails(self):
+        self.make("a.png")
+        await self.add(name="Set", items=[self.item("a.png")])
+
+        async def broken(_root):
+            raise OSError("scan failed at /secret/place")
+
+        with patch.object(self.api, "_get_gallery_files", side_effect=broken):
+            data = await self.call("GET", BASE)
+        self.assertEqual(
+            data["collections"][0] | {"id": 0},
+            {"id": 0, "name": "Set", "count": None, "missing": None},
+        )
+        self.assertNotIn("secret", json.dumps(data))
+
+    async def test_windows_case_spellings_are_one_member(self):
+        self.make("Sub/A.png")
+        with patch("py.api.images._path_flavor", ntpath):
+            made = await self.add(
+                name="Set", items=[self.item("Sub/A.png"), self.item("SUB/a.PNG")]
+            )
+            self.assertEqual((made["added"], made["already"]), (1, 0))
+            (self.output_dir / "Sub" / "A.png").rename(
+                self.output_dir / "Sub" / "a.png"
+            )
+            (self.output_dir / "Sub").rename(self.output_dir / "sub")
+            self.api.invalidate_gallery_cache()
+            listed = await self.listed()
+            self.assertEqual((listed[0]["count"], listed[0]["missing"]), (1, 0))
+            membership = await self.call(
+                "GET",
+                MEMBERSHIP,
+                root_index="0",
+                root_scope=self.scope(),
+                path="sub/a.png",
+            )
+            self.assertEqual(membership["collection_ids"], [made["collection"]["id"]])
+            response = await self.client.get(
+                "/prompt_manager/images/output",
+                params={
+                    "mode": "workspace",
+                    "collection": str(made["collection"]["id"]),
+                },
+            )
+            page = await response.json()
+            self.assertEqual([i["path"] for i in page["images"]], ["sub/a.png"])
+
+    async def test_posix_case_distinct_files_stay_two_members(self):
+        self.make("A.png", "a.png")
+        if len(list(self.output_dir.iterdir())) != 2:
+            self.skipTest("case-insensitive filesystem")
+        made = await self.add(
+            name="Set", items=[self.item("A.png"), self.item("a.png")]
+        )
+        self.assertEqual(made["added"], 2)
+
+    async def test_a_500_never_carries_the_exception_text(self):
+        self.make("a.png")
+        made = await self.add(name="Set", items=[self.item("a.png")])
+        cid = made["collection"]["id"]
+
+        def broken(*_args, **_kwargs):
+            raise sqlite3.OperationalError("disk I/O error at /secret/prompts.db")
+
+        cases = (
+            (
+                "list_collections",
+                "GET",
+                BASE,
+                None,
+                {},
+                "Collections could not be read.",
+            ),
+            (
+                "file_collections",
+                "GET",
+                MEMBERSHIP,
+                None,
+                {"root_index": "0", "root_scope": self.scope(), "path": "a.png"},
+                "Collections could not be read.",
+            ),
+            (
+                "add_to_collection",
+                "POST",
+                ITEMS,
+                {"collection_id": cid, "items": [self.item("a.png")]},
+                {},
+                "The collection could not be changed.",
+            ),
+            (
+                "remove_from_collection",
+                "DELETE",
+                ITEMS,
+                {"collection_id": cid, "items": [self.item("a.png")]},
+                {},
+                "The collection could not be changed.",
+            ),
+            (
+                "rename_collection",
+                "PATCH",
+                f"{BASE}/{cid}",
+                {"name": "x"},
+                {},
+                "The collection could not be changed.",
+            ),
+            (
+                "delete_collection",
+                "DELETE",
+                f"{BASE}/{cid}",
+                None,
+                {},
+                "The collection could not be changed.",
+            ),
+        )
+        for method_name, method, url, body, params, message in cases:
+            with self.subTest(method=method_name):
+                with patch.object(self.api.db, method_name, side_effect=broken):
+                    data = await self.call(method, url, body, 500, **params)
+                self.assertEqual(data, {"success": False, "error": message})
+
+
+class TestCollectionListingFilter(CollectionRoutesTestCase):
+    """#392's "all N matching" filter carries ``collection`` like the listing."""
+
+    async def test_a_filter_with_a_collection_resolves_only_members(self):
+        second = self.two_sources()
+        self.make("a.png", "b.png")
+        self.make("c.png", root=second)
+        made = await self.add(
+            name="Set", items=[self.item("a.png"), self.item("c.png", 1)]
+        )
+        cid = str(made["collection"]["id"])
+        selection = await self.api.resolve_listing_filter({"collection": cid})
+        self.assertEqual(
+            sorted((index, path) for index, _scope, path in selection.targets),
+            [(0, "a.png"), (1, "c.png")],
+        )
+        selection = await self.api.resolve_listing_filter(
+            {"collection": cid, "root": "1", "root_scope": self.scope(1)}
+        )
+        self.assertEqual([t[2] for t in selection.targets], ["c.png"])
+        for bad in ("0", "x", "1234567890"):
+            with self.subTest(bad=bad):
+                with self.assertRaises(ValueError):
+                    from py.api.images import parse_listing_filter
+
+                    parse_listing_filter({"collection": bad})
+
+    async def test_bulk_category_acts_on_collection_members_only(self):
+        self.make("a.png", "b.png")
+        inside = self._save_prompt("Inside", category="Old")
+        outside = self._save_prompt("Outside", category="Old")
+        self.api.db.link_image_to_prompt(inside, str(self.output_dir / "a.png"))
+        self.api.db.link_image_to_prompt(outside, str(self.output_dir / "b.png"))
+        made = await self.add(name="Set", items=[self.item("a.png")])
+        cid = str(made["collection"]["id"])
+        data = await self.call(
+            "POST",
+            "/prompt_manager/bulk/category",
+            {"images_filter": {"collection": cid}, "category": "New"},
+        )
+        self.assertEqual((data["updated_count"], data["prompt_count"]), (1, 1))
+        self.assertEqual(self.api.db.get_prompt_by_id(inside)["category"], "New")
+        self.assertEqual(self.api.db.get_prompt_by_id(outside)["category"], "Old")
+
+    async def test_a_deleted_collection_in_a_bulk_filter_is_a_404(self):
+        self.make("a.png")
+        data = await self.call(
+            "POST",
+            "/prompt_manager/bulk/category",
+            {"images_filter": {"collection": "77"}, "category": "New"},
+            status=404,
+        )
+        self.assertEqual(data["code"], "collection-not-found")
+        with self.assertRaises(Exception) as caught:
+            await self.api.resolve_listing_filter({"collection": "77"})
+        self.assertEqual(type(caught.exception).__name__, "CollectionMissing")
+
+
+class TestReviewFindings(CollectionRoutesTestCase):
+    """PR #400 review: loop work, counts snapshot, id and query rules."""
+
+    async def test_no_realpath_runs_on_the_event_loop(self):
+        import os
+        import threading
+
+        names = [f"img{n}.png" for n in range(20)]
+        self.make(*names)
+        await self.call("GET", BASE)  # warm the listing cache
+        loop_thread = threading.current_thread()
+        on_loop = []
+        real = os.path.realpath
+
+        def watched(*args, **kwargs):
+            if threading.current_thread() is loop_thread:
+                on_loop.append(args[0])
+            return real(*args, **kwargs)
+
+        items = [self.item(name) for name in names]
+        cid = None
+        with patch("os.path.realpath", side_effect=watched):
+            made = await self.add(name="Set", items=items)
+            cid = made["collection"]["id"]
+            await self.call("DELETE", ITEMS, {"collection_id": cid, "items": items[:5]})
+            await self.call(
+                "GET",
+                MEMBERSHIP,
+                root_index="0",
+                root_scope=items[0]["root_scope"],
+                path="img9.png",
+            )
+            await self.call("GET", BASE)
+        self.assertEqual(on_loop, [])
+
+    async def test_missing_is_never_negative(self):
+        self.make("a.png", "b.png")
+        made = await self.add(
+            name="Set", items=[self.item("a.png"), self.item("b.png")]
+        )
+        cid = made["collection"]["id"]
+        real = self.api.db.collection_count_snapshot
+
+        def raced(root_keys):
+            members, stored = real(root_keys)
+            return members, {cid: 1}  # a remove landed between two reads
+
+        with patch.object(self.api.db, "collection_count_snapshot", side_effect=raced):
+            listed = await self.listed()
+        self.assertEqual((listed[0]["count"], listed[0]["missing"]), (2, 0))
+
+    async def test_path_ids_follow_the_listing_rule(self):
+        self.make("a.png")
+        made = await self.add(name="Set", items=[self.item("a.png")])
+        cid = made["collection"]["id"]
+        for raw in ("0", "007", f"0{cid}"):
+            with self.subTest(raw=raw):
+                await self.call("PATCH", f"{BASE}/{raw}", {"name": "x"}, 400)
+                await self.call("DELETE", f"{BASE}/{raw}", None, 400)
+        self.assertEqual(len(await self.listed()), 1)
+
+    async def test_membership_refuses_repeated_query_keys(self):
+        self.make("a.png", "b.png")
+        response = await self.client.get(
+            MEMBERSHIP
+            + f"?root_index=0&root_scope={self.scope()}&path=a.png&path=b.png"
+        )
+        self.assertEqual(response.status, 400)
+        response = await self.client.get(
+            MEMBERSHIP
+            + f"?root_index=0&root_index=0&root_scope={self.scope()}&path=a.png"
+        )
+        self.assertEqual(response.status, 400)
+
+    async def test_deleting_a_collection_leaves_everything_else(self):
+        from py.api.images import _favorite_root_key
+
+        try:
+            from tests.test_annotation_schema import insert_group_rows, table_rows
+        except ImportError:
+            from test_annotation_schema import insert_group_rows, table_rows
+
+        self.make("a.png")
+        db = self.api.db
+        db.set_image_favorite(_favorite_root_key(self.output_dir), "a.png")
+        prompt = self._save_prompt("Linked")
+        db.link_image_to_prompt(prompt, str(self.output_dir / "a.png"))
+        with db.model.get_connection() as conn:
+            insert_group_rows(conn, locator=str(self.output_dir / "a.png"))
+        made = await self.add(name="Set", items=[self.item("a.png")])
+        tables = (
+            "image_favorites",
+            "generated_images",
+            "prompts",
+            "tags",
+            "media_annotation_files",
+            "media_annotation_bytes",
+            "media_image_tags",
+        )
+        before = table_rows(db.model.get_connection(), tables)
+        file_bytes = (self.output_dir / "a.png").read_bytes()
+        await self.call("DELETE", f"{BASE}/{made['collection']['id']}")
+        self.assertEqual(table_rows(db.model.get_connection(), tables), before)
+        self.assertEqual((self.output_dir / "a.png").read_bytes(), file_bytes)
+        self.assertTrue(all(before[t] for t in tables))

tests/test_api_duplicate_files.py

@@ -693,6 +693,39 @@ class TestCleanupJob(DuplicateFilesTestCase):
             "Moved 2 database links to the kept copies", "\n".join(cleanup.log)
         )
 
+    async def test_duplicate_removal_moves_collection_memberships(self):
+        self.trio()
+        db = self.api.db
+        root_key = duplicate_files.favorite_root_key(self.output_dir)
+        grids = db.add_to_collection(name="LoRA grids", items=[(root_key, "b.png")])
+        grids = grids["collection"]["id"]
+        other = db.add_to_collection(
+            name="Other", items=[(root_key, "sub/c.png"), (root_key, "d.png")]
+        )["collection"]["id"]
+        db.add_to_collection(collection_id=grids, items=[(root_key, "sub/c.png")])
+        _plan, cleanup = await self.remove_all_but("a.png")
+        # The union: a.png joins both; c.png's second grids row was a repeat.
+        self.assertEqual(db.collection_members(grids, root_key), {"a.png"})
+        self.assertEqual(db.collection_members(other, root_key), {"a.png", "d.png"})
+        self.assertEqual(sorted(db.file_collections(root_key, "a.png")), [grids, other])
+        self.assertEqual(db.file_collections(root_key, "b.png"), [])
+        self.assertEqual(db.file_collections(root_key, "sub/c.png"), [])
+        self.assertIn(
+            "Moved 2 collection memberships to the kept copies", "\n".join(cleanup.log)
+        )
+
+    def test_one_collection_membership_is_singular(self):
+        job = type(
+            "Job", (), {"log": [], "log_line": lambda s, line: s.log.append(line)}
+        )()
+        totals = dict.fromkeys(
+            ("moved", "merged", "favorites", "tags", "review", "collections"), 0
+        )
+        duplicate_files.DuplicateFilesMixin._log_moved_records(
+            job, {**totals, "collections": 1}
+        )
+        self.assertEqual(job.log, ["Moved 1 collection membership to the kept copies"])
+
     async def test_a_failed_removal_leaves_its_links_on_the_kept_copy(self):
         self.trio()
         prompt = self.save("Stays on c")
@@ -849,8 +882,11 @@ class TestLinkQueries(DuplicateFilesTestCase):
                 with self.subTest(sql=sql):
                     self.assertNotIn("SCAN generated_images", plan)
                     self.assertNotIn("SCAN image_favorites", plan)
+                    self.assertNotIn("SCAN image_collection_items", plan)
         finally:
             conn.close()
+        statements = duplicate_files.query_plan_sql()
+        self.assertTrue(any("image_collection_items" in sql for sql in statements))
 
 
 class TestReviewHelpers(DuplicateFilesTestCase):

tests/test_api_vocabulary.py

@@ -62,12 +62,16 @@ class TestVocabularyAPI(ImageAPITestCase):
             },
         )
         self.assertEqual(
-            categories["categories"], [{"value": "Animal", "prompt_count": 1}]
+            categories["categories"],
+            [{"value": "Animal", "prompt_count": 1, "image_count": 0}],
         )
         _, rest = await self.get(
             "categories", limit=1, cursor=categories["next_cursor"]
         )
-        self.assertEqual(rest["categories"], [{"value": "Pet", "prompt_count": 1}])
+        self.assertEqual(
+            rest["categories"],
+            [{"value": "Pet", "prompt_count": 1, "image_count": 0}],
+        )
         self.assertFalse(rest["has_more"])
 
     async def test_vocabulary_cursor_is_signed_epoch_scoped(self):
@@ -488,6 +492,7 @@ class TestVocabularyAPI(ImageAPITestCase):
             "other_prompts": 0,
             "negative_variants": 1,
             "archived_negative_variants": 0,
+            "images": 0,
         }
         self.assertEqual(
             preview,
@@ -528,6 +533,73 @@ class TestVocabularyAPI(ImageAPITestCase):
         self.assertEqual(response.status, 409)
         self.assertEqual(replay["code"], "vocabulary-conflict")
 
+    async def test_category_rows_report_images_and_image_only_values(self):
+        self._save_prompt("Owner", category="Prompt value")
+        _, options_before = await self.get_options()
+        self.api.db.set_image_categories(
+            [("root", "a.png"), ("root", "b.png")], "Prompt value"
+        )
+        self.api.db.set_image_categories([("root", "c.png")], "Image value")
+        response, first = await self.get("categories", limit=1)
+        self.assertEqual(response.status, 200)
+        self.assertEqual(
+            first["categories"],
+            [{"value": "Prompt value", "prompt_count": 1, "image_count": 2}],
+        )
+        self.assertEqual(first["total"], 2)
+        _, rest = await self.get("categories", limit=1, cursor=first["next_cursor"])
+        self.assertEqual(
+            rest["categories"],
+            [{"value": "Image value", "prompt_count": 0, "image_count": 1}],
+        )
+        self.assertFalse(rest["has_more"])
+        # The Prompt library's category list does not gain image values.
+        _, options_after = await self.get_options()
+        self.assertEqual(options_after, options_before)
+        self.assertEqual(options_after["categories"], ["Prompt value"])
+
+    async def get_options(self):
+        response = await self.client.get("/prompt_manager/library/options")
+        return response, await response.json()
+
+    async def test_category_routes_rename_image_only_value(self):
+        epoch = self.api.db.get_execution_epoch()
+        self._save_prompt("Bystander", category="Kept")
+        self.api.db.set_image_categories(
+            [("root", "a.png"), ("second", "b.png")], "Image value"
+        )
+        prompts_before = self.category_rows()
+        body = {"expected_epoch": epoch, "source": "Image value", "target": "Moved"}
+        response, preview = await self.send("POST", "categories/preview", body)
+        self.assertEqual(response.status, 200, preview)
+        impact = {
+            "positive_prompts": 0,
+            "other_prompts": 0,
+            "negative_variants": 0,
+            "archived_negative_variants": 0,
+            "images": 2,
+        }
+        self.assertEqual(preview["impact"], impact)
+        stale = preview["impact_token"]
+        self.api.db.set_image_categories([("root", "late.png")], "Image value")
+        response, refused = await self.send(
+            "POST", "categories/apply", {**body, "impact_token": stale}
+        )
+        self.assertEqual(response.status, 409)
+        self.assertEqual(refused["code"], "vocabulary-conflict")
+        self.assertEqual(self.api.db.image_category("root", "a.png"), "Image value")
+        _, preview = await self.send("POST", "categories/preview", body)
+        response, applied = await self.send(
+            "POST",
+            "categories/apply",
+            {**body, "impact_token": preview["impact_token"]},
+        )
+        self.assertEqual(response.status, 200, applied)
+        self.assertEqual(applied["impact"], {**impact, "images": 3})
+        for key in (("root", "a.png"), ("second", "b.png"), ("root", "late.png")):
+            self.assertEqual(self.api.db.image_category(*key), "Moved")
+        self.assertEqual(self.category_rows(), prompts_before)
+
     async def test_preview_validation(self):
         epoch = self.api.db.get_execution_epoch()
         self._save_prompt("Owner", category="Scenery")

tests/test_api_workspace_image_categories.py

@@ -0,0 +1,262 @@
+"""Images filters by the effective category (#332 MC4, MC5, DC6).
+
+An image's effective category is its own category when it has one, else the
+category its link supplies (the recipe's when set, else the positive's), else
+none. The category filter, "select all N matching" and saved views match it.
+"""
+
+import os
+import sqlite3
+from unittest.mock import patch
+
+try:
+    import test_api_workspace_images as workspace
+    from test_image_categories import category_checksum
+except ImportError:
+    from tests import test_api_workspace_images as workspace
+    from tests.test_image_categories import category_checksum
+
+from py.api.images import image_category_key, parse_listing_filter
+
+
+class WorkspaceCategoryCase(workspace.ImageAPITestCase):
+    """Listing helpers plus own-category writes keyed as the routes key them."""
+
+    media = workspace.TestWorkspaceImages.media
+    get = workspace.TestWorkspaceImages.get
+    tag = workspace.TestWorkspaceImageTags.tag
+    names = workspace.TestWorkspaceImageTags.names
+
+    async def tearDownAsync(self):
+        self.api.db.close_all()
+        workspace.assert_closed(self, self._temp_db.name, self._output_tmp.name)
+        await super().tearDownAsync()
+
+    def own(self, path, value, sources=None):
+        sources = sources or self.api._get_all_output_dirs()
+        self.api.db.set_image_categories([image_category_key(sources, path)], value)
+
+    async def listed(self, **params):
+        response, body = await self.get(sort="name", **params)
+        self.assertEqual(response.status, 200, body)
+        return self.names(body), body
+
+
+class TestEffectiveCategoryFilter(WorkspaceCategoryCase):
+    def batch(self):
+        """Four images of "castle at dusk" (Landscapes); one is overridden."""
+        castle = self._save_prompt("castle at dusk", category="Landscapes", rating=4)
+        files = [self.media(f"castle_{n}.png", mtime=100 + n) for n in range(4)]
+        for path in files:
+            self.api.db.link_image_to_prompt(castle, str(path))
+        self.own(files[0], "Portraits")
+        return castle, files
+
+    async def test_category_filter_prefers_own_category(self):
+        castle, _files = self.batch()
+        conn = self.api.db.model.get_connection()
+        before = category_checksum(conn)
+        names, body = await self.listed(category="Portraits")
+        self.assertEqual(names, ["castle_0.png"])
+        # The caption still counts the prompt behind the image shown.
+        self.assertEqual(body["positive_count"], 1)
+        names, body = await self.listed(category="Landscapes")
+        self.assertEqual(names, ["castle_1.png", "castle_2.png", "castle_3.png"])
+        self.assertEqual(body["total"], 3)
+        names, _ = await self.listed()
+        self.assertEqual(len(names), 4)
+        self.assertEqual(category_checksum(conn), before)
+        self.assertEqual(self.api.db.get_prompt_by_id(castle)["category"], "Landscapes")
+
+    async def test_category_filter_lists_unlinked_with_own(self):
+        scans = [self.media(f"scan_{n}.png", mtime=100 + n) for n in range(2)]
+        self.media("plain.png", mtime=50)
+        for path in scans:
+            self.own(path, "Scans")
+        names, body = await self.listed(category="Scans")
+        self.assertEqual(names, ["scan_0.png", "scan_1.png"])
+        self.assertEqual(body["positive_count"], 0)
+
+    async def test_recipe_category_is_inherited(self):
+        """MC4: a recipe's own category, not the positive's, is inherited."""
+        db = self.api.db
+        self._save_prompt("Pair positive", category="Night")
+        pair = db.save_prompt_pair("Pair positive", "Noise", category="Recipes")
+        blank = db.save_prompt_pair("Pair positive", "Other noise", category="  ")
+        recipe = self.media("recipe.png", mtime=103)
+        blank_recipe = self.media("blank_recipe.png", mtime=102)
+        positive = db.get_prompt_pair(pair)["positive_prompt_id"]
+        for path, pair_id in ((recipe, pair), (blank_recipe, blank)):
+            image = db.link_image_to_prompt(positive, str(path))
+            db.link_image_to_pair(pair_id, image, {"source": "fixture"})
+        self.assertEqual((await self.listed(category="Recipes"))[0], ["recipe.png"])
+        # A blank recipe category falls back to the positive's.
+        self.assertEqual((await self.listed(category="Night"))[0], ["blank_recipe.png"])
+        # An own category still wins over the recipe's.
+        self.own(recipe, "Mine")
+        self.assertEqual((await self.listed(category="Recipes"))[0], [])
+        self.assertEqual((await self.listed(category="Mine"))[0], ["recipe.png"])
+
+    async def test_ambiguous_files_inherit_nothing_but_keep_their_own(self):
+        first = self._save_prompt("Owner one", category="Night")
+        second = self._save_prompt("Owner two", category="Night")
+        ambiguous = self.media("ambiguous.png")
+        for owner in (first, second):
+            self.api.db.link_image_to_prompt(owner, str(ambiguous))
+        self.assertEqual((await self.listed(category="Night"))[0], [])
+        self.own(ambiguous, "Night")
+        self.assertEqual((await self.listed(category="Night"))[0], ["ambiguous.png"])
+
+    async def test_combines_with_tags_search_and_rating(self):
+        stone = self._save_prompt(
+            "castle at dusk", category="Landscapes", tags=["stone"], rating=4
+        )
+        files = [self.media(f"castle_{n}.png", mtime=100 + n) for n in range(2)]
+        for path in files:
+            self.api.db.link_image_to_prompt(stone, str(path))
+        self.own(files[0], "Portraits")
+        other = self._save_prompt("harbor at noon", category="Portraits", rating=2)
+        harbor = self.media("harbor.png", mtime=90)
+        self.api.db.link_image_to_prompt(other, str(harbor))
+        names, _ = await self.listed(category="Portraits", tags="stone")
+        self.assertEqual(names, ["castle_0.png"])
+        names, _ = await self.listed(category="Portraits", q="harbor")
+        self.assertEqual(names, ["harbor.png"])
+        names, _ = await self.listed(category="Portraits", q="castle_0")
+        self.assertEqual(names, ["castle_0.png"])
+        response, body = await self.get(category="Portraits", sort="rating")
+        self.assertEqual(response.status, 200, body)
+        self.assertEqual(self.names(body), ["castle_0.png", "harbor.png"])
+        self.assertEqual(body["positive_count"], 2)
+
+    async def test_image_tag_admitted_file_keeps_its_inherited_category(self):
+        night = self._save_prompt("Night owl", category="Night", tags=["a"])
+        tagged = self.media("tagged.png", mtime=103)
+        unlinked = self.media("unlinked.png", mtime=102)
+        self.api.db.link_image_to_prompt(night, str(tagged))
+        self.tag(tagged, ["b"])
+        self.tag(unlinked, ["a", "b"])
+        names, _ = await self.listed(tags="a,b", category="Night")
+        self.assertEqual(names, ["tagged.png"])
+        self.own(unlinked, "Night")
+        names, _ = await self.listed(tags="a,b", category="Night")
+        self.assertEqual(names, ["tagged.png", "unlinked.png"])
+
+    async def test_cursor_pages_walk_only_the_category(self):
+        files = [self.media(f"img_{n:02}.png", mtime=100 + n) for n in range(7)]
+        for path in files[::2]:
+            self.own(path, "Even")
+        seen, cursor = [], None
+        while True:
+            params = {"category": "Even", "limit": 2, "sort": "name"}
+            if cursor:
+                params["cursor"] = cursor
+            response, body = await self.get(**params)
+            self.assertEqual(response.status, 200, body)
+            seen += self.names(body)
+            self.assertEqual(body["total"], 4)
+            cursor = body["next_cursor"]
+            if not cursor:
+                break
+        self.assertEqual(seen, ["img_00.png", "img_02.png", "img_04.png", "img_06.png"])
+
+    async def test_an_outer_source_lists_a_nested_sources_own_category(self):
+        inner = self.output_dir / "nested"
+        nested = self.media("nested/a.png")
+        self.media("b.png")
+        self.api._get_all_output_dirs = lambda: [self.output_dir, inner]
+        self.own(nested, "Scans")
+        names, _ = await self.listed(category="Scans", root="0")
+        self.assertEqual(names, ["nested/a.png"])
+        names, _ = await self.listed(category="Scans", root="1")
+        self.assertEqual(names, ["a.png"])
+        # The outer source's own key for that path is not the file's key.
+        from py.api.images import _favorite_root_key
+
+        self.api.db.clear_image_categories([image_category_key([inner], nested)])
+        self.api.db.set_image_categories(
+            [(_favorite_root_key(self.output_dir), "nested/a.png")], "Stale"
+        )
+        self.assertEqual((await self.listed(category="Stale"))[0], [])
+
+    async def test_a_gone_files_category_is_not_listed(self):
+        gone = self.media("gone.png")
+        self.own(gone, "Scans")
+        os.unlink(gone)
+        self.api.invalidate_gallery_cache()
+        self.assertEqual((await self.listed(category="Scans"))[0], [])
+
+    async def test_a_failed_own_category_read_is_listing_failed(self):
+        self.media("a.png")
+
+        def broken(*_args, **_kwargs):
+            raise sqlite3.OperationalError("secret detail")
+
+        with patch.object(self.api.db, "own_categories_of", side_effect=broken):
+            with self.assertLogs("prompt_manager", level="ERROR"):
+                response, body = await self.get(category="Scans")
+            self.assertEqual(response.status, 500)
+            self.assertNotIn("secret", str(body))
+            # Without a category nothing new is read.
+            response, body = await self.get()
+            self.assertEqual(response.status, 200, body)
+
+    async def test_select_all_matching_uses_the_effective_category(self):
+        _castle, files = self.batch()
+        resolved = await self.api.resolve_listing_filter(
+            parse_listing_filter({"category": "Portraits"}), 100
+        )
+        self.assertEqual([t[2] for t in resolved.targets], ["castle_0.png"])
+        resolved = await self.api.resolve_listing_filter(
+            parse_listing_filter({"category": "Landscapes"}), 100
+        )
+        self.assertEqual(len(resolved.targets), 3)
+        self.assertEqual(len(files), 4)
+
+    async def test_saved_view_category_means_the_effective_category(self):
+        """CUJ 8: a saved view's category is the listing's query value."""
+        _castle, files = self.batch()
+        self.own(files[1], "Portraits")
+        names, _ = await self.listed(category="Portraits")
+        self.assertEqual(names, ["castle_0.png", "castle_1.png"])
+
+
+class TestFilterAndChipAgreeOnBlanks(WorkspaceCategoryCase):
+    """The filter and the inspector use one inherited-category rule."""
+
+    async def chip(self, path):
+        response = await self.client.get(
+            f"/prompt_manager/images/inspect/{path}", params={"root": "0"}
+        )
+        body = await response.json()
+        self.assertEqual(response.status, 200, body)
+        return body["library"]["category"]
+
+    async def test_blank_recipe_values_agree(self):
+        db = self.api.db
+        positive = self._save_prompt("Pair positive", category="Night")
+        blanks = {"spaces": "   ", "tab": "\t", "nbsp": " ", "set": "Recipes"}
+        for index, (label, value) in enumerate(blanks.items()):
+            pair = db.save_prompt_pair("Pair positive", f"noise {label}")
+            path = self.media(f"{label}.png", mtime=100 + index)
+            image = db.link_image_to_prompt(positive, str(path))
+            db.link_image_to_pair(pair, image, {"source": "fixture"})
+            with db.model.get_connection() as conn:
+                conn.execute(
+                    "UPDATE prompt_pairs SET category=? WHERE id=?", (value, pair)
+                )
+        blob = db.save_prompt_pair("Pair positive", "noise blob")
+        path = self.media("blob.png", mtime=90)
+        image = db.link_image_to_prompt(positive, str(path))
+        db.link_image_to_pair(blob, image, {"source": "fixture"})
+        with db.model.get_connection() as conn:
+            conn.execute(
+                "UPDATE prompt_pairs SET category=CAST(' ' AS BLOB) WHERE id=?",
+                (blob,),
+            )
+        for name in ("spaces", "tab", "nbsp", "set", "blob"):
+            with self.subTest(name=name):
+                shown = await self.chip(f"{name}.png")
+                for value in ("Night", "Recipes"):
+                    listed = f"{name}.png" in (await self.listed(category=value))[0]
+                    self.assertEqual(listed, shown == value, (name, value, shown))

tests/test_api_workspace_images.py

@@ -1347,7 +1347,7 @@ class TestWorkspaceImages(ImageAPITestCase):
         _, data = await self.get(q="harbor")
         self.assertEqual(self._names(data), ["night_a.png"])
         self.assertEqual(
-            self.api.db.get_image_paths_matching("zebra", None, []),
+            self.api.db.get_image_paths_matching("zebra", []),
             {
                 normalize_image_path(str(files[name]))
                 for name in ("night_ab", "day_ab", "titled")
@@ -1407,8 +1407,10 @@ class TestWorkspaceImages(ImageAPITestCase):
             for query in statements
             if query.lstrip().upper().startswith(("SELECT", "WITH"))
         ]
-        # Plus the one stream of tagged image records q and tags read (B3).
-        self.assertEqual(len(selects), 4, selects)
+        # Plus the one stream of tagged image records q and tags read (B3),
+        # and the category's own-category read of the one source (#332).
+        self.assertEqual(len(selects), 5, selects)
+        self.assertEqual(sum("image_categories" in q for q in selects), 1, selects)
         self.assertEqual(sum("image_favorites" in q for q in selects), 1, selects)
         self.assertEqual(sum("media_image_tags" in q for q in selects), 1, selects)
 
@@ -1528,15 +1530,16 @@ class TestWorkspaceImages(ImageAPITestCase):
         elsewhere = self.api.db.get_prompt_id_for_image("/elsewhere/7.png")
         self.assertEqual(
             self.api.db.get_image_link_matches(
-                "", "Night", [], every_link=True, wanted=wanted
+                "", [], every_link=True, wanted=wanted, category="Night"
             ),
             {
                 normalize_image_path(str(files["night_ab"])): (
                     True,
                     5,
                     self.api.db.get_prompt_id_for_image(str(files["night_ab"])),
+                    "Night",
                 ),
-                "/elsewhere/7.png": (True, 5, elsewhere),
+                "/elsewhere/7.png": (True, 5, elsewhere, "Night"),
             },
         )
 
@@ -1588,6 +1591,24 @@ class TestWorkspaceImages(ImageAPITestCase):
                     for i in range(links)
                 ],
             )
+        # 2,000 images (one in 25 at 50,000 files) carry their own category
+        # (#332), so a category filter also reads and matches those.
+        from py.api.images import _favorite_root_key
+
+        root_key = _favorite_root_key(self.output_dir)
+        with self.api.db.model.get_connection() as conn:
+            conn.executemany(
+                "INSERT INTO image_categories VALUES(?,?,?,?)",
+                [
+                    (
+                        root_key,
+                        listed[i][0].relative_to(self.output_dir).as_posix(),
+                        f"cat-{i % 10}",
+                        "2026-10-27T00:00:00+00:00",
+                    )
+                    for i in range(1, files, 25)[:2_000]
+                ],
+            )
         return listed
 
     _BUDGET_REQUESTS = (
@@ -1615,7 +1636,7 @@ class TestWorkspaceImages(ImageAPITestCase):
         import math
 
         from py.api import images as images_module
-        from database.operations import link_filter_query
+        from database.operations import link_filter_query, link_matches_json_query
 
         listed = await asyncio.get_running_loop().run_in_executor(
             None, self._budget_fixture, 5_000, 2_000
@@ -1678,7 +1699,13 @@ class TestWorkspaceImages(ImageAPITestCase):
                     # q and tags also stream the tagged image records once
                     # (#249 B3); none are tagged here, so nothing is stat'ed.
                     streamed = "q" in params or "tags" in params
-                    self.assertEqual(len(found), 2 if streamed else 1, found)
+                    # A category also reads the own categories, one primary
+                    # key range per source (#332).
+                    own = "category" in params
+                    self.assertEqual(len(found), 1 + streamed + own, found)
+                    self.assertEqual(
+                        sum("image_categories" in query for query in found), int(own)
+                    )
                     self.assertEqual(
                         sum("media_image_tags" in query for query in found),
                         int(streamed),
@@ -1738,13 +1765,42 @@ class TestWorkspaceImages(ImageAPITestCase):
             self.assertEqual(self.api._workspace_listed_cache, {})
             self.assertEqual(self.api._workspace_order_cache, {})
         # The per-row ambiguity probe is an index search, never a scan.
-        sql, params = link_filter_query("lantern", "cat-3", ["tag-1"], every_link=True)
-        with self.api.db.model.get_connection() as conn:
-            plan = [row[3] for row in conn.execute("EXPLAIN QUERY PLAN " + sql, params)]
-        self.assertIn(
-            "SEARCH other USING COVERING INDEX idx_image_file_path (file_path=?)", plan
+        # The read runs it wrapped in one JSON aggregate (#354): same plan.
+        sql, params = link_filter_query("lantern", ["tag-1"], every_link=True)
+        for query in (sql, link_matches_json_query(sql)):
+            with self.subTest(wrapped=query is not sql):
+                with self.api.db.model.get_connection() as conn:
+                    plan = [
+                        row[3]
+                        for row in conn.execute("EXPLAIN QUERY PLAN " + query, params)
+                    ]
+                self.assertIn(
+                    "SEARCH other USING COVERING INDEX idx_image_file_path "
+                    "(file_path=?)",
+                    plan,
+                )
+                self.assertFalse(
+                    [step for step in plan if step.startswith("SCAN other")]
+                )
+        # The category's narrowing clause (#332) keeps every join an index
+        # search, plain and wrapped with the raw categories.
+        sql, params = link_filter_query(
+            "", ["tag-1"], every_link=True, category="cat-3"
         )
-        self.assertFalse([step for step in plan if step.startswith("SCAN other")])
+        for query in (sql, link_matches_json_query(sql, categories=True)):
+            with self.subTest(category_wrapped=query is not sql):
+                with self.api.db.model.get_connection() as conn:
+                    plan = [
+                        row[3]
+                        for row in conn.execute("EXPLAIN QUERY PLAN " + query, params)
+                    ]
+                self.assertIn(
+                    "SEARCH other USING COVERING INDEX idx_image_file_path "
+                    "(file_path=?)",
+                    plan,
+                )
+                scans = [step for step in plan if step.startswith("SCAN")]
+                self.assertTrue(all(step.startswith("SCAN gi") for step in scans), plan)
 
     @unittest.skipUnless(
         PERF_BUDGET and not TRACED,
@@ -2494,3 +2550,207 @@ class TestListingFilterSelection(ImageAPITestCase):
 
         self.assertEqual(response.status, 409)
         self.assertEqual((await response.json())["code"], "source-changed")
+
+
+class TestWorkspaceCollectionFilter(ImageAPITestCase):
+    """#331: ``collection=<id>`` keeps only that collection's listed members."""
+
+    media = TestWorkspaceImages.media
+    get = TestWorkspaceImages.get
+    _names = TestWorkspaceImages._names
+
+    async def tearDownAsync(self):
+        self.api.db.close_all()
+        assert_closed(self, self._temp_db.name, self._output_tmp.name)
+        await super().tearDownAsync()
+
+    def keys(self, root, relative):
+        from py.api.images import _favorite_relative_key, _favorite_root_key
+
+        return (_favorite_root_key(root), _favorite_relative_key(relative))
+
+    def collect(self, name, *members):
+        """A collection holding ``(root, relative)`` members; its id."""
+        result = self.api.db.add_to_collection(
+            name=name, items=[self.keys(root, rel) for root, rel in members]
+        )
+        return result["collection"]["id"]
+
+    def two_sources(self):
+        second = self.comfy_dir / "second"
+        second.mkdir()
+        self.api._get_all_output_dirs = lambda: [self.output_dir, second]
+        return second
+
+    async def test_listing_keeps_only_collection_members(self):
+        second = self.two_sources()
+        self.media("a.png", 300)
+        self.media("b.png", 200)
+        self.media("sub/c.png", 100, root=second)
+        self.media("d.png", 50, root=second)
+        grids = self.collect(
+            "Grids",
+            (self.output_dir, "a.png"),
+            (second, "sub/c.png"),
+            (self.output_dir, "gone.png"),  # stored, never listed
+        )
+        self.collect("Other", (self.output_dir, "b.png"))
+        response, data = await self.get(collection=str(grids))
+        self.assertEqual(response.status, 200, data)
+        self.assertEqual(self._names(data), ["a.png", "c.png"])
+        self.assertEqual(data["total"], 2)
+        _, data = await self.get(collection=str(grids), root="1")
+        self.assertEqual(self._names(data), ["c.png"])
+        # Without the filter the listing is unchanged.
+        _, data = await self.get()
+        self.assertEqual(data["total"], 4)
+
+    async def test_collection_filter_combines_with_the_other_filters(self):
+        night = self._save_prompt("Neon zebra", category="Night", tags=["a"])
+        day = self._save_prompt("Morning field", category="Day", tags=["a"])
+        files = {
+            name: self.media(name, 100 + index)
+            for index, name in enumerate(
+                ("night.png", "day.png", "zebra_plain.png", "clip.mp4", "out.png")
+            )
+        }
+        self.api.db.link_image_to_prompt(night, str(files["night.png"]))
+        self.api.db.link_image_to_prompt(day, str(files["day.png"]))
+        members = [(self.output_dir, name) for name in files if name != "out.png"]
+        grids = str(self.collect("Grids", *members))
+        cases = (
+            ({}, ["clip.mp4", "day.png", "night.png", "zebra_plain.png"]),
+            ({"q": "zebra"}, ["night.png", "zebra_plain.png"]),
+            ({"category": "Night"}, ["night.png"]),
+            ({"tags": "a"}, ["day.png", "night.png"]),
+            ({"media": "video"}, ["clip.mp4"]),
+            ({"media": "image", "q": "day"}, ["day.png"]),
+        )
+        for params, expected in cases:
+            with self.subTest(**params):
+                response, data = await self.get(collection=grids, sort="name", **params)
+                self.assertEqual(response.status, 200, data)
+                self.assertEqual(self._names(data), expected)
+                self.assertEqual(data["total"], len(expected))
+
+    async def test_collection_filter_combines_with_an_images_own_category(self):
+        """#331 with #332: a member is admitted by its effective category."""
+        from py.api.images import image_category_key
+
+        night = self._save_prompt("Neon zebra", category="Night")
+        files = {
+            name: self.media(name, 100 + index)
+            for index, name in enumerate(
+                ("linked.png", "overridden.png", "own.png", "outside.png")
+            )
+        }
+        for name in ("linked.png", "overridden.png"):
+            self.api.db.link_image_to_prompt(night, str(files[name]))
+        sources = self.api._get_all_output_dirs()
+
+        def own(name, value):
+            key = image_category_key(sources, files[name])
+            self.api.db.set_image_categories([key], value)
+
+        own("overridden.png", "Portraits")
+        own("own.png", "Night")  # unlinked: only its own category
+        own("outside.png", "Night")  # not a member
+        members = [(self.output_dir, name) for name in files if name != "outside.png"]
+        grids = str(self.collect("Grids", *members))
+        for category, expected in (
+            ("Night", ["linked.png", "own.png"]),
+            ("Portraits", ["overridden.png"]),
+        ):
+            with self.subTest(category=category):
+                response, data = await self.get(
+                    collection=grids, category=category, sort="name"
+                )
+                self.assertEqual(response.status, 200, data)
+                self.assertEqual(self._names(data), expected)
+                self.assertEqual(data["total"], len(expected))
+
+    async def test_invalid_collection_is_a_400(self):
+        for value in ("x", "0", "-1", "1.5", "1234567890", " 1"):
+            with self.subTest(value=value):
+                response, data = await self.get(collection=value)
+                self.assertEqual(response.status, 400, data)
+
+    async def test_unknown_collection_is_a_404(self):
+        self.media("a.png")
+        response, data = await self.get(collection="77")
+        self.assertEqual(response.status, 404)
+        self.assertEqual(data["code"], "collection-not-found")
+        self.assertNotIn("images", data)
+        self.api._get_all_output_dirs = lambda: []
+        response, data = await self.get(collection="77")
+        self.assertEqual((response.status, data["code"]), (404, "collection-not-found"))
+
+    async def test_a_collection_deleted_during_the_read_is_a_404(self):
+        self.media("a.png")
+        grids = self.collect("Grids", (self.output_dir, "a.png"))
+        real = self.api.db.collection_members
+
+        def deleted_meanwhile(collection_id, root_key):
+            self.api.db.delete_collection(collection_id)
+            return real(collection_id, root_key)
+
+        with patch.object(
+            self.api.db, "collection_members", side_effect=deleted_meanwhile
+        ):
+            response, data = await self.get(collection=str(grids))
+        self.assertEqual((response.status, data["code"]), (404, "collection-not-found"))
+
+    async def test_a_failed_members_read_is_a_500_never_an_unfiltered_page(self):
+        self.media("a.png")
+        self.media("b.png")
+        grids = self.collect("Grids", (self.output_dir, "a.png"))
+
+        def broken(*_args, **_kwargs):
+            raise sqlite3.OperationalError("database is locked")
+
+        with patch.object(self.api.db, "collection_members", side_effect=broken):
+            response, data = await self.get(collection=str(grids))
+        self.assertEqual(response.status, 500)
+        self.assertNotIn("images", data)
+        self.assertNotIn("locked", json.dumps(data))
+
+    async def test_reordering_sources_keeps_members(self):
+        second = self.two_sources()
+        self.media("a.png", root=second)
+        self.media("b.png")
+        grids = str(self.collect("Grids", (second, "a.png")))
+        _, data = await self.get(collection=grids)
+        self.assertEqual(
+            [(i["root_index"], i["path"]) for i in data["images"]], [(1, "a.png")]
+        )
+        self.api._get_all_output_dirs = lambda: [second, self.output_dir]
+        _, data = await self.get(collection=grids)
+        self.assertEqual(
+            [(i["root_index"], i["path"]) for i in data["images"]], [(0, "a.png")]
+        )
+
+    async def test_cursor_from_another_collection_is_refused(self):
+        for index, name in enumerate(("a.png", "b.png", "c.png")):
+            self.media(name, 100 + index)
+        everything = [(self.output_dir, name) for name in ("a.png", "b.png", "c.png")]
+        first = str(self.collect("First", *everything))
+        second = str(self.collect("Second", *everything))
+        _, page = await self.get(collection=first, limit=1)
+        token = page["next_cursor"]
+        self.assertIsInstance(token, str)
+        response, _ = await self.get(collection=first, limit=1, cursor=token)
+        self.assertEqual(response.status, 200)
+        for params in ({"collection": second}, {}, {"favorites": "1"}):
+            with self.subTest(**params):
+                response, _ = await self.get(cursor=token, limit=1, **params)
+                self.assertEqual(response.status, 400)
+
+    async def test_collection_wins_over_favorites(self):
+        self.media("a.png")
+        self.media("b.png")
+        grids = str(self.collect("Grids", (self.output_dir, "a.png")))
+        from py.api.images import _favorite_root_key
+
+        self.api.db.set_image_favorite(_favorite_root_key(self.output_dir), "b.png")
+        _, data = await self.get(collection=grids, favorites="1")
+        self.assertEqual(self._names(data), ["a.png"])

tests/test_autotag_batch.py

@@ -553,6 +553,35 @@ class TestStartRefusals(BatchTestCase):
         self.assertEqual(self.api._jobs.list(), [])
         self.assertIsNone(model_gate.holder())
 
+    async def test_a_missing_backend_package_refuses_the_start(self):
+        self.tagger.missing["gguf"] = ["llama-cpp-python"]
+        status, data = await self.start_batch(model_type="gguf")
+        self.assertEqual((status, data["code"]), (409, "model-unavailable"))
+        self.assertEqual(data["success"], False)
+        self.assertTrue(
+            data["error"].startswith("GGUF JoyCaption needs llama-cpp-python."), data
+        )
+        self.assertIn("pip install llama-cpp-python", data["error"])
+        self.assertEqual(self.api._jobs.list(), [])
+        self.assertEqual(self.tagger.load_calls, [])
+        self.assertIsNone(model_gate.holder())
+
+    async def test_a_package_lost_after_the_start_fails_the_job_without_a_traceback(
+        self,
+    ):
+        from py.autotag import MissingBackendPackages
+
+        self.image("a.png", (1, 1, 1))
+        self.tagger.load_error = MissingBackendPackages("gguf", ["llama-cpp-python"])
+        with self.assertNoLogs(self.api.logger, level="ERROR"):
+            job = await self.run_batch(model_type="gguf")
+        self.assertEqual(job.state, "failed")
+        self.assertTrue(
+            job.status_line.startswith("GGUF JoyCaption needs"), job.status_line
+        )
+        self.assertTrue(any("pip install llama-cpp-python" in line for line in job.log))
+        self.assertIsNone(model_gate.holder())
+
     async def test_model_busy_when_the_gate_is_held(self):
         owner = model_gate.new_owner("image-annotations")
         self.assertTrue(model_gate.try_acquire(owner))
@@ -658,6 +687,20 @@ class TestPreview(BatchTestCase):
             self.assertIsNone(llm["estimate_device"])
         self.assertEqual([s for s in _strings(skip) if _absolute(s)], [])
 
+    async def test_preview_names_missing_backend_packages(self):
+        self.image("a.png", (1, 1, 1))
+        data = (await self.preview(model_type="gguf"))[1]
+        self.assertEqual(data["model"]["missing_packages"], [])
+        self.assertIsNone(data["model"]["install_command"])
+        self.tagger.missing["gguf"] = ["llama-cpp-python"]
+        data = (await self.preview(model_type="gguf"))[1]
+        self.assertTrue(data["model"]["downloaded"])
+        self.assertEqual(data["model"]["missing_packages"], ["llama-cpp-python"])
+        self.assertEqual(
+            data["model"]["install_command"], "pip install llama-cpp-python"
+        )
+        self.assertIn("DGGML_CUDA", data["model"]["install_note"])
+
     async def test_preview_uses_the_graphics_card_rate_when_wd14_runs_on_cuda(self):
         self.image("a.png", (1, 1, 1))
         self.tagger.wd14_device = lambda: "gpu"

tests/test_autotag_engine.py

@@ -800,8 +800,12 @@ class TestDefaultLoaders(EngineTestCase):
         transformers, model, processor, inputs = self._fake_transformers()
         pixel = inputs["pixel_values"]
         service = AutoTagService(models_dir=self.models_dir)
+        # The 8-bit load also needs these two (#424); stand-ins satisfy the check.
+        stubs = {
+            name: types.ModuleType(name) for name in ("accelerate", "bitsandbytes")
+        }
 
-        with _patch_modules({"torch": torch, "transformers": transformers}):
+        with _patch_modules({"torch": torch, "transformers": transformers, **stubs}):
             service.load_model("hf")
             tags = service.generate_tags(str(self.image_path))
 

tests/test_autotag_legacy.py

@@ -161,6 +161,39 @@ class TestLegacyBatch(LegacyTestCase):
         self.assertEqual(self.tagger.calls[-1]["prompt"], AutotagConfig.TEMPLATE)
         self.assertEqual(self.tagger.custom_prompt, "legacy prompt")
 
+    async def test_a_missing_backend_package_is_one_error_frame(self):
+        self.tagger.missing["gguf"] = ["llama-cpp-python"]
+        _, frames = await self.legacy("model_type=gguf")
+        self.assertEqual(len(frames), 1)
+        self.assertEqual(frames[0]["type"], "error")
+        self.assertTrue(
+            frames[0]["message"].startswith("GGUF JoyCaption needs llama-cpp-python.")
+        )
+        self.assertEqual(self.tagger.load_calls, [])
+
+    async def test_legacy_single_refuses_a_missing_backend_package(self):
+        from py.autotag import MissingBackendPackages
+
+        self.image("a.png", (1, 1, 1))
+        await self.relist()
+        self.tagger.missing["gguf"] = ["llama-cpp-python"]
+        status, data = await self.single(
+            {"path": "a.png", "root_index": 0, "model_type": "gguf"}
+        )
+        self.assertEqual((status, data["code"]), (409, "model-unavailable"))
+        self.assertIn("pip install llama-cpp-python", data["error"])
+        self.assertEqual(self.tagger.load_calls, [])
+        # Lost after the check: the same 409, no ERROR record, nothing held.
+        self.tagger.missing["gguf"] = []
+        self.tagger.load_error = MissingBackendPackages("gguf", ["llama-cpp-python"])
+        with self.assertNoLogs(self.api.logger, level="ERROR"):
+            status, data = await self.single(
+                {"path": "a.png", "root_index": 0, "model_type": "gguf"}
+            )
+        self.assertEqual((status, data["code"]), (409, "model-unavailable"))
+        self.assertIn("pip install llama-cpp-python", data["error"])
+        self.assertIsNone(model_gate.holder())
+
     async def test_each_refusal_is_one_error_frame(self):
         self.tagger.downloaded["wd14-vit"] = False
         _, frames = await self.legacy()

tests/test_autotag_packages.py

@@ -0,0 +1,248 @@
+"""#424: each tagging backend reports the Python packages it cannot import.
+
+``find_spec`` is patched throughout: no ML backend is installed or imported.
+"""
+
+import os
+import sys
+import types
+import unittest
+from unittest.mock import patch
+
+sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
+
+try:
+    from test_autotag_engine import (
+        EngineTestCase,
+        FakeGgufModel,
+        _patch_modules,
+        _touch_model_files,
+    )
+except ImportError:  # pragma: no cover - run as a package
+    from tests.test_autotag_engine import (
+        EngineTestCase,
+        FakeGgufModel,
+        _patch_modules,
+        _touch_model_files,
+    )
+
+from py import autotag  # noqa: E402
+from py.autotag import (  # noqa: E402
+    BACKEND_PACKAGES,
+    MODELS,
+    AutoTagService,
+    MissingBackendPackages,
+    missing_packages,
+    package_fields,
+)
+
+GGUF_CUDA = 'For a CUDA build: CMAKE_ARGS="-DGGML_CUDA=on" pip install llama-cpp-python'
+
+
+def _find_spec_without(*absent):
+    """A ``find_spec`` that knows every module except *absent*."""
+
+    def find_spec(name, *args):
+        return None if name in absent else object()
+
+    return find_spec
+
+
+def _missing_from(*absent):
+    return patch.object(
+        autotag.importlib.util, "find_spec", side_effect=_find_spec_without(*absent)
+    )
+
+
+class TestMissingPackages(unittest.TestCase):
+
+    def test_every_backend_names_its_modules_and_pip_names(self):
+        self.assertEqual(BACKEND_PACKAGES["gguf"], (("llama_cpp", "llama-cpp-python"),))
+        self.assertEqual(
+            [pip for _module, pip in BACKEND_PACKAGES["hf"]],
+            ["torch", "transformers", "accelerate", "bitsandbytes"],
+        )
+        for model in ("wd14-swinv2", "wd14-vit"):
+            self.assertEqual(BACKEND_PACKAGES[model], (("onnxruntime", "onnxruntime"),))
+        self.assertEqual(set(BACKEND_PACKAGES), set(MODELS))
+
+    def test_reports_pip_names_of_modules_find_spec_cannot_find(self):
+        with _missing_from("llama_cpp", "accelerate", "bitsandbytes"):
+            self.assertEqual(missing_packages("gguf"), ["llama-cpp-python"])
+            self.assertEqual(missing_packages("hf"), ["accelerate", "bitsandbytes"])
+            self.assertEqual(missing_packages("wd14-vit"), [])
+
+    def test_find_spec_errors_count_as_missing(self):
+        for error in (ValueError("no spec"), ImportError("broken"), OSError("io")):
+            with self.subTest(error=type(error).__name__):
+                with patch.object(
+                    autotag.importlib.util, "find_spec", side_effect=error
+                ):
+                    self.assertEqual(missing_packages("wd14-vit"), ["onnxruntime"])
+
+    def test_a_module_already_imported_is_present_without_a_lookup(self):
+        with _patch_modules({"llama_cpp": types.ModuleType("llama_cpp")}):
+            with patch.object(
+                autotag.importlib.util, "find_spec", side_effect=AssertionError
+            ):
+                self.assertEqual(missing_packages("gguf"), [])
+
+    def test_the_check_imports_nothing(self):
+        before = set(sys.modules)
+        missing_packages("gguf")
+        missing_packages("hf")
+        missing_packages("wd14-vit")
+        imported = set(sys.modules) - before
+        heavy = {"llama_cpp", "torch", "transformers", "onnxruntime"}
+        self.assertFalse(imported & heavy, imported)
+
+    def test_unknown_model_has_no_packages(self):
+        self.assertEqual(missing_packages("nope"), [])
+
+    def test_package_fields_and_the_one_sentence(self):
+        self.assertEqual(
+            package_fields("gguf", []),
+            {"missing_packages": [], "install_command": None, "install_note": None},
+        )
+        fields = package_fields("gguf", ["llama-cpp-python"])
+        self.assertEqual(fields["missing_packages"], ["llama-cpp-python"])
+        self.assertEqual(fields["install_command"], "pip install llama-cpp-python")
+        self.assertEqual(fields["install_note"], GGUF_CUDA)
+        self.assertEqual(
+            package_fields("hf", ["torch", "bitsandbytes"])["install_command"],
+            "pip install torch bitsandbytes",
+        )
+        self.assertIsNone(package_fields("hf", ["torch"])["install_note"])
+        self.assertIn(
+            "onnxruntime-gpu",
+            package_fields("wd14-vit", ["onnxruntime"])["install_note"],
+        )
+        error = MissingBackendPackages("gguf", ["llama-cpp-python"])
+        self.assertIsInstance(error, RuntimeError)
+        self.assertEqual(
+            str(error),
+            "GGUF JoyCaption needs llama-cpp-python. Install it into ComfyUI's "
+            "Python: pip install llama-cpp-python. " + GGUF_CUDA,
+        )
+        self.assertEqual(
+            str(MissingBackendPackages("hf", ["accelerate", "bitsandbytes"])),
+            "HuggingFace JoyCaption needs accelerate and bitsandbytes. Install them "
+            "into ComfyUI's Python: pip install accelerate bitsandbytes.",
+        )
+
+
+class TestStatusReportsPackages(EngineTestCase):
+
+    def test_downloaded_keeps_its_files_only_meaning(self):
+        _touch_model_files(self.models_dir, "gguf")
+        _touch_model_files(self.models_dir, "wd14-vit")
+        service = AutoTagService(models_dir=self.models_dir)
+
+        with _missing_from("llama_cpp"):
+            status = service.get_models_status()
+
+        self.assertTrue(status["gguf"]["downloaded"])
+        self.assertEqual(status["gguf"]["missing_packages"], ["llama-cpp-python"])
+        self.assertEqual(
+            status["gguf"]["install_command"], "pip install llama-cpp-python"
+        )
+        self.assertEqual(status["gguf"]["install_note"], GGUF_CUDA)
+        self.assertEqual(status["wd14-vit"]["missing_packages"], [])
+        self.assertIsNone(status["wd14-vit"]["install_command"])
+        # A model not on disk still says which packages it will need.
+        self.assertFalse(status["wd14-swinv2"]["downloaded"])
+        self.assertIn("missing_packages", status["hf"])
+
+    def test_an_injected_loader_brings_its_own_backend(self):
+        service = self._service(("gguf", FakeGgufModel("x")))
+        with _missing_from("llama_cpp"):
+            self.assertEqual(
+                service.get_models_status()["gguf"]["missing_packages"], []
+            )
+
+    def test_an_injected_probe_is_used_for_status_and_load(self):
+        _touch_model_files(self.models_dir, "gguf")
+        service = AutoTagService(
+            models_dir=self.models_dir,
+            model_loader=lambda *_: ("gguf", FakeGgufModel("x")),
+            package_probe=lambda model_type: ["llama-cpp-python"],
+        )
+        self.assertEqual(
+            service.get_models_status()["gguf"]["missing_packages"],
+            ["llama-cpp-python"],
+        )
+        with self.assertRaises(MissingBackendPackages):
+            service.load_model("gguf")
+
+
+class TestLoadFailsFast(EngineTestCase):
+
+    def test_missing_package_refuses_before_the_loader_without_an_error_log(self):
+        _touch_model_files(self.models_dir, "gguf")
+        service = AutoTagService(models_dir=self.models_dir)
+        calls = []
+        service._model_loader = lambda *args: calls.append(args)
+
+        with (
+            _missing_from("llama_cpp"),
+            self.assertLogs(service.logger, level="DEBUG") as logs,
+        ):
+            with self.assertRaises(MissingBackendPackages) as ctx:
+                service.load_model("gguf")
+
+        self.assertEqual(calls, [])
+        self.assertEqual(ctx.exception.model_type, "gguf")
+        self.assertEqual(ctx.exception.missing, ["llama-cpp-python"])
+        self.assertIn("pip install llama-cpp-python", str(ctx.exception))
+        self.assertFalse(service.is_model_loaded())
+        levels = {record.levelname for record in logs.records}
+        self.assertNotIn("ERROR", levels)
+        self.assertIn("WARNING", levels)
+        self.assertTrue(all(record.exc_info is None for record in logs.records))
+
+    def test_not_downloaded_is_reported_before_packages(self):
+        service = AutoTagService(models_dir=self.models_dir)
+        with _missing_from("llama_cpp"):
+            with self.assertRaises(RuntimeError) as ctx:
+                service.load_model("gguf")
+        self.assertNotIsInstance(ctx.exception, MissingBackendPackages)
+
+    def test_a_module_not_found_from_the_loader_becomes_the_same_refusal(self):
+        _touch_model_files(self.models_dir, "gguf")
+
+        def loader(model_type, use_gpu):
+            raise ModuleNotFoundError("No module named 'llama_cpp'", name="llama_cpp")
+
+        service = AutoTagService(
+            models_dir=self.models_dir,
+            model_loader=loader,
+            package_probe=lambda _model: [],
+        )
+        with self.assertLogs(service.logger, level="DEBUG") as logs:
+            with self.assertRaises(MissingBackendPackages) as ctx:
+                service.load_model("gguf")
+        self.assertEqual(ctx.exception.missing, ["llama-cpp-python"])
+        self.assertNotIn("ERROR", {record.levelname for record in logs.records})
+        self.assertFalse(service.is_model_loaded())
+
+    def test_an_unknown_module_not_found_names_the_module(self):
+        def loader(model_type, use_gpu):
+            raise ModuleNotFoundError(
+                "No module named 'sentencepiece'", name="sentencepiece"
+            )
+
+        hf_dir = self.models_dir / MODELS["hf"]["subdir"]
+        hf_dir.mkdir(parents=True, exist_ok=True)
+        (hf_dir / "config.json").write_text("{}")
+        service = AutoTagService(
+            models_dir=self.models_dir,
+            model_loader=loader,
+            package_probe=lambda _model: [],
+        )
+        with self.assertRaises(MissingBackendPackages) as ctx:
+            service.load_model("hf")
+        self.assertEqual(ctx.exception.missing, ["sentencepiece"])
+
+
+if __name__ == "__main__":
+    unittest.main()

tests/test_cached_runs.py

@@ -0,0 +1,863 @@
+"""Cached re-runs of runtime-linked prompt text count once per job (#347).
+
+Jobs are simulated end to end: the queue hook (submit + dequeue), the
+``execution_cached`` replay, the real node executions and ComfyUI's output
+cache (fakes of ``execution`` and ``comfy_execution.caching``, as in
+tests/test_execution_context_host.py).
+"""
+
+import logging
+import os
+import sys
+import tempfile
+import threading
+import types
+import unittest
+from collections import namedtuple
+from unittest import mock
+
+sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
+
+try:
+    from tests.open_handles import assert_closed
+except ImportError:  # discovered with tests/ as the top-level directory
+    from open_handles import assert_closed
+
+from database.operations import PromptDatabase
+from utils import cached_runs as module
+from utils.cached_runs import (
+    CachedRuns,
+    Facts,
+    Role,
+    Seen,
+    graph_role,
+    graph_seen,
+    own_counts,
+    register_cached_runs,
+    stored_facts,
+)
+from utils.execution_context_host import cached_string_outputs
+from utils.hashing import generate_prompt_hash
+from utils.run_keys import prompt_node_key, upstream_closure, upstream_nodes
+from utils.usage_tracking import (
+    PendingFirstUse,
+    QueuedRuns,
+    find_pair_id,
+    find_pair_id_by_hashes,
+    handle_queued_prompt,
+    on_dequeue,
+)
+
+Entry = namedtuple("CacheEntry", "ui outputs")
+LLM = "a lighthouse the LLM wrote"
+
+
+class _Cache:
+    def __init__(self):
+        self.entries = {}
+
+    def get_local(self, node_id):
+        return self.entries.get(node_id)
+
+
+class HierarchicalCache(_Cache):
+    pass
+
+
+class LRUCache(_Cache):
+    pass
+
+
+class RAMPressureCache(LRUCache):
+    pass
+
+
+class NullCache(_Cache):
+    pass
+
+
+def host_modules():
+    """Fake ``execution`` and ``comfy_execution.caching`` for the cache reader."""
+    execution = types.ModuleType("execution")
+    execution.CacheEntry = Entry
+
+    class PromptExecutor:
+        def add_message(self, event, data, broadcast):
+            return None
+
+    execution.PromptExecutor = PromptExecutor
+    caching = types.ModuleType("comfy_execution.caching")
+    for cls in (HierarchicalCache, LRUCache, RAMPressureCache, NullCache):
+        setattr(caching, cls.__name__, cls)
+    package = types.ModuleType("comfy_execution")
+    package.caching = caching
+    return {
+        "execution": execution,
+        "comfy_execution": package,
+        "comfy_execution.caching": caching,
+    }
+
+
+def graph(
+    kind="PromptManager",
+    text=None,
+    role="positive",
+    sampler=True,
+    negative_text="",
+    seed=1,
+):
+    """PrimitiveString -> PreviewAny -> ``kind`` -> KSampler (runtime text).
+
+    ``text`` set: typed text instead of the runtime link. ``role`` "negative":
+    the node feeds the sampler's negative input. Pair: its negative editor holds
+    ``negative_text`` and it feeds both sampler inputs.
+    """
+    g = {
+        "str": {"class_type": "PrimitiveString", "inputs": {"value": "seed text"}},
+        "llm": {"class_type": "PreviewAny", "inputs": {"source": ["str", 0]}},
+        "ckpt": {"class_type": "CheckpointLoaderSimple", "inputs": {"ckpt": "m"}},
+        "pm": {
+            "class_type": kind,
+            "inputs": {"text": ["llm", 0] if text is None else text},
+        },
+        "other": {
+            "class_type": "CLIPTextEncode",
+            "inputs": {"text": "", "clip": ["ckpt", 1]},
+        },
+        "save": {"class_type": "SaveImage", "inputs": {"images": ["ks", 0]}},
+    }
+    if kind != "PromptManagerText":
+        g["pm"]["inputs"]["clip"] = ["ckpt", 1]
+    if kind == "PromptManagerPair":
+        g["pm"]["inputs"]["negative_text"] = negative_text
+    source = ["pm", 0]
+    if kind == "PromptManagerText":
+        g["enc"] = {
+            "class_type": "CLIPTextEncode",
+            "inputs": {"text": ["pm", 0], "clip": ["ckpt", 1]},
+        }
+        source = ["enc", 0]
+    positive, negative = (source, ["other", 0])
+    if role == "negative":
+        positive, negative = negative, positive
+    if kind == "PromptManagerPair":
+        positive, negative = ["pm", 0], ["pm", 1]
+    if sampler:
+        g["ks"] = {
+            "class_type": "KSampler",
+            "inputs": {
+                "model": ["ckpt", 0],
+                "positive": positive,
+                "negative": negative,
+                "seed": seed,
+            },
+        }
+    else:
+        g["save"]["inputs"]["images"] = positive
+    return g
+
+
+class JobCase(unittest.TestCase):
+    """A database, the queue hook stashes and one ``CachedRuns`` per test."""
+
+    def setUp(self):
+        self.tmp = tempfile.TemporaryDirectory()
+        self.addCleanup(self.tmp.cleanup)
+        self.addCleanup(assert_closed, self, self.tmp.name)
+        self.db = PromptDatabase(os.path.join(self.tmp.name, "runs.db"))
+        self.addCleanup(self.db.close_all)
+        patcher = mock.patch.dict(sys.modules, host_modules())
+        patcher.start()
+        self.addCleanup(patcher.stop)
+        self.queued, self.pair_queued = QueuedRuns(), QueuedRuns()
+        self.pending, self.pair_pending = PendingFirstUse(), PendingFirstUse()
+        self.context = (None, None)
+        self.runs = CachedRuns(context=lambda: self.context)
+        self.queue = object()
+        self.executor = types.SimpleNamespace(
+            server=types.SimpleNamespace(prompt_queue=self.queue),
+            caches=types.SimpleNamespace(outputs=HierarchicalCache()),
+        )
+        self.item_id = 0
+        self.trackers = []
+
+    # -- nodes ----------------------------------------------------------------
+
+    def node(self, kind):
+        if kind == "PromptManagerPair":
+            from prompt_manager_pair import PromptManagerPair as cls
+        elif kind == "PromptManagerText":
+            from prompt_manager_text import PromptManagerText as cls
+        else:
+            import prompt_manager
+
+            cls = getattr(prompt_manager, kind)
+        node = cls.__new__(cls)
+        node.db = self.db
+        node.logger = logging.getLogger("test.cached_runs")
+        node.prompt_tracker = mock.Mock()
+        self.trackers.append(node.prompt_tracker)
+        node.comfyui_integration = mock.Mock()
+        node._inject_lora_trigger_words = lambda text: text
+        node.pending_first_use = self.pending
+        node.pending_pair_first_use = self.pair_pending
+        node.cached_runs = self.runs
+        return node
+
+    def execute(self, g, node_id="pm", text=LLM, negative="", index=None):
+        """Run the node as ComfyUI would; returns its STRING outputs."""
+        kind = g[node_id]["class_type"]
+        node = self.node(kind)
+        self.context = (self.context[0], index)
+        clip = mock.Mock()
+        # The node receives its switch as the queued graph holds it (#90)
+        save = g[node_id]["inputs"].get("save_to_library", True)
+        if kind == "PromptManagerText":
+            return node.process_text(
+                text, prompt=g, unique_id=node_id, save_to_library=save
+            )
+        if kind == "PromptManagerPair":
+            return node.encode_prompt(
+                clip,
+                text,
+                negative_text=negative,
+                prompt=g,
+                unique_id=node_id,
+                save_to_library=save,
+            )
+        return node.encode_prompt(
+            clip, text, prompt=g, unique_id=node_id, save_to_library=save
+        )
+
+    def cache(self, node_id, *results):
+        """Store executed results (one per list index) as ComfyUI's cache entry."""
+        width = len(results[0])
+        outputs = [[result[slot] for result in results] for slot in range(width)]
+        self.executor.caches.outputs.entries[node_id] = Entry({}, outputs)
+
+    # -- jobs -----------------------------------------------------------------
+
+    def job(self, g, cached=(), run=None, outputs=("save",), prompt_id=None):
+        """One queued job: hook, dequeue, ``execution_cached``, executions, done."""
+        self.item_id += 1
+        prompt_id = prompt_id or f"job-{self.item_id}"
+        request = handle_queued_prompt(
+            {"prompt": g, "prompt_id": prompt_id},
+            self.queued,
+            pair_queued=self.pair_queued,
+        )
+        on_dequeue(
+            request["prompt_id"],
+            self.queued,
+            lambda: self.db,
+            self.pending,
+            pair_queued=self.pair_queued,
+            pair_pending=self.pair_pending,
+        )
+        self.runs.begin_attempt(
+            ((self.item_id, prompt_id, g, {}, list(outputs), {}), self.item_id)
+        )
+        self.context = (prompt_id, None)
+        self.runs.replay(self.executor, prompt_id, list(cached), lambda: self.db)
+        result = run() if run else None
+        self.runs.end_attempt(self.item_id)
+        self.context = (None, None)
+        return result
+
+    def first(self, g, text=LLM, negative="", node_id="pm"):
+        """A job that executes the node and fills the cache."""
+        result = self.job(
+            g, run=lambda: self.execute(g, node_id, text=text, negative=negative)
+        )
+        self.cache(node_id, result)
+        return result
+
+    def count(self, text):
+        row = self.db.get_prompt_by_hash(generate_prompt_hash(text))
+        return row["run_count"] if row else None
+
+    def pair_count(self, positive, negative):
+        pair_id = find_pair_id(self.db, positive, negative)
+        return self.db.get_prompt_pair(pair_id)["run_count"] if pair_id else None
+
+
+class TestCachedJobsCount(JobCase):
+    def test_runtime_text_counts_every_cached_job_for_each_node_kind(self):
+        for kind in (
+            "PromptManager",
+            "PromptManagerFocused",
+            "PromptManagerSplit",
+            "PromptManagerText",
+        ):
+            with self.subTest(kind=kind):
+                text = f"{LLM} for {kind}"
+                g = graph(kind)
+                self.first(g, text=text)
+                self.job(graph(kind, seed=2), cached=["pm"])
+                self.job(graph(kind, seed=3), cached=["pm"])
+                self.assertEqual(self.count(text), 3)
+
+    def test_typed_text_still_counts_once_per_job(self):
+        g = graph(text="typed castle")
+        self.first(g, text="typed castle")
+        self.job(g, cached=["pm"])
+        self.job(g, cached=["pm"])
+        self.assertEqual(self.count("typed castle"), 3)
+
+    def test_rewired_to_negative_while_cached_counts_nothing(self):
+        self.first(graph())
+        self.job(graph(role="negative"), cached=["pm"])
+        self.assertEqual(self.count(LLM), 1)
+        # Rewired back to positive: the receipt still proves the text
+        self.job(graph(), cached=["pm"])
+        self.assertEqual(self.count(LLM), 2)
+
+    def test_no_sampler_counts_every_node_on_replay(self):
+        g = graph(sampler=False)
+        self.first(g)
+        self.job(g, cached=["pm"])
+        self.assertEqual(self.count(LLM), 2)
+
+    def test_empty_editor_replays_nothing(self):
+        g = graph()
+        self.first(g, text="")
+        self.job(g, cached=["pm"])
+        self.assertEqual(self.db.search_prompts(), [])
+
+    def test_disabled_node_then_cached_jobs_counts_0(self):
+        # #90 x #347: a node with Save to library off records no receipt and is
+        # no run prompt, so its cached jobs replay nothing. The prompt exists
+        # (an enabled run saved it), so any replayed count would show.
+        for kind in (
+            "PromptManager",
+            "PromptManagerFocused",
+            "PromptManagerSplit",
+            "PromptManagerText",
+            "PromptManagerPair",
+        ):
+            with self.subTest(kind=kind):
+                text = f"{LLM} not saved by {kind}"
+                negative = "blurry" if kind == "PromptManagerPair" else ""
+                self.first(graph(kind, negative_text=negative), text, negative)
+                before = (self.count(text), self.pair_count(text, negative))
+
+                def off(seed):
+                    g = graph(kind, negative_text=negative, seed=seed)
+                    g["pm"]["inputs"]["save_to_library"] = False
+                    return g
+
+                self.first(off(1), text, negative)
+                self.job(off(2), cached=["pm"])
+                self.job(off(3), cached=["pm"])
+                after = (self.count(text), self.pair_count(text, negative))
+                self.assertEqual(before[0], 1)
+                self.assertEqual(after, before)
+
+    def test_alternating_upstream_values_each_count_their_own_prompt(self):
+        g = graph()
+        self.first(g, text="value A")
+        self.first(g, text="value B")
+        self.cache("pm", self.execute_outputs("value A"))
+        self.job(g, cached=["pm"])
+        self.assertEqual((self.count("value A"), self.count("value B")), (2, 1))
+
+    def execute_outputs(self, text):
+        return (mock.ANY, text)
+
+    def test_list_items_replay_each_item_in_order(self):
+        g = graph()
+        results = []
+
+        def run():
+            for index, text in enumerate(("one", "two", "three")):
+                results.append(self.execute(g, text=text, index=index))
+
+        self.job(g, run=run)
+        self.cache("pm", *results)
+        self.job(g, cached=["pm"])
+        self.assertEqual([self.count(t) for t in ("one", "two", "three")], [2, 2, 2])
+
+    def test_a_repeated_id_or_message_never_counts_a_node_twice(self):
+        g = graph()
+        self.first(g)
+        self.job(g, cached=["pm", "pm"])
+        self.assertEqual(self.count(LLM), 2)
+        self.item_id += 1
+        self.runs.begin_attempt(((1, "again", g, {}, ["save"], {}), self.item_id))
+        for _ in range(2):
+            self.runs.replay(self.executor, "again", ["pm"], lambda: self.db)
+        self.assertEqual(self.count(LLM), 3)
+
+    def test_a_shorter_list_keeps_its_receipt_after_a_longer_one(self):
+        g = graph()
+        self.first(g, text="alpha")
+        results = []
+
+        def run():
+            for index, text in enumerate(("alpha", "beta")):
+                results.append(self.execute(g, text=text, index=index))
+
+        self.job(g, run=run)
+        self.cache("pm", (mock.ANY, "alpha"))
+        self.job(g, cached=["pm"])
+        self.assertEqual((self.count("alpha"), self.count("beta")), (3, 1))
+
+    def test_a_job_reusing_a_prompt_id_counts_and_executes_normally(self):
+        g = graph()
+        self.first(g)
+        self.job(g, cached=["pm"], prompt_id="same")
+        self.job(g, run=lambda: self.execute(g), prompt_id="same")
+        self.job(g, cached=["pm"], prompt_id="same")
+        self.assertEqual(self.count(LLM), 4)
+
+
+class TestFailClosed(JobCase):
+    def setUp(self):
+        super().setUp()
+        self.g = graph()
+        self.first(self.g)
+
+    def assertNotCounted(self):
+        self.assertEqual(self.count(LLM), 1)
+
+    def test_changed_cached_text_is_not_counted(self):
+        self.cache("pm", (mock.ANY, LLM + " edited"))
+        self.job(self.g, cached=["pm"])
+        self.assertNotCounted()
+
+    def test_no_receipt_in_this_process_is_not_counted(self):
+        self.runs = CachedRuns(context=lambda: self.context)
+        self.job(self.g, cached=["pm"])
+        self.assertNotCounted()
+
+    def test_missing_cache_entry_is_not_counted(self):
+        self.executor.caches.outputs.entries.clear()
+        self.job(self.g, cached=["pm"])
+        self.assertNotCounted()
+
+    def test_unsupported_cache_is_not_counted(self):
+        entries = self.executor.caches.outputs.entries
+        self.executor.caches.outputs = NullCache()
+        self.executor.caches.outputs.entries = entries
+        self.job(self.g, cached=["pm"])
+        self.assertNotCounted()
+
+    def test_more_list_values_than_the_bound_are_not_counted(self):
+        self.cache("pm", *[(mock.ANY, LLM)] * 65)
+        self.job(self.g, cached=["pm"])
+        self.assertNotCounted()
+
+    def test_missing_list_index_is_not_counted(self):
+        self.cache("pm", (mock.ANY, LLM), (mock.ANY, LLM))
+        self.job(self.g, cached=["pm"])
+        self.assertNotCounted()
+
+    def test_node_the_job_does_not_run_is_not_counted(self):
+        self.job(self.g, cached=["pm"], outputs=("other",))
+        self.assertNotCounted()
+
+    def test_message_for_another_job_is_not_counted(self):
+        self.item_id += 1
+        self.runs.begin_attempt(((1, "this", self.g, {}, ["save"], {}), self.item_id))
+        self.runs.replay(self.executor, "another", ["pm"], lambda: self.db)
+        self.assertNotCounted()
+
+    def test_unkeyable_node_is_not_counted(self):
+        g = graph()
+        g["pm"]["inputs"]["extra"] = ["missing", 0]
+        self.job(g, cached=["pm"])
+        self.assertIsNone(prompt_node_key(g, "pm"))
+        self.assertNotCounted()
+
+    def test_deleted_prompt_is_skipped(self):
+        self.db.delete_prompt(
+            self.db.get_prompt_by_hash(generate_prompt_hash(LLM))["id"]
+        )
+        self.job(self.g, cached=["pm"])
+        self.assertIsNone(self.count(LLM))
+
+    def test_database_errors_never_raise(self):
+        def broken():
+            raise RuntimeError("database unavailable")
+
+        self.item_id += 1
+        self.runs.begin_attempt(((1, "j", self.g, {}, ["save"], {}), self.item_id))
+        self.runs.replay(self.executor, "j", ["pm"], broken)
+        self.assertNotCounted()
+
+    def test_malformed_messages_and_results_are_ignored(self):
+        self.runs.begin_attempt(None)
+        self.runs.begin_attempt((("x",), "not an int"))
+        self.runs.replay(self.executor, None, ["pm"], lambda: self.db)
+        self.item_id += 1
+        self.runs.begin_attempt(((1, "j", self.g, {}, ["save"], {}), self.item_id))
+        self.runs.replay(self.executor, "j", "pm", lambda: self.db)
+        self.runs.replay(self.executor, "j", ["pm"] * 501, lambda: self.db)
+        self.runs.replay(self.executor, "j", [7, None], lambda: self.db)
+        self.assertNotCounted()
+
+
+class TestReplayThenExecution(JobCase):
+    def test_execution_after_replay_in_the_same_job_is_not_counted_again(self):
+        # RAM-pressure frees the entry after execution_cached: the node executes
+        g = graph()
+        self.first(g)
+        self.job(g, cached=["pm"], run=lambda: self.execute(g))
+        self.assertEqual(self.count(LLM), 2)
+        # The suppressed execution still leaves a receipt: the next job counts
+        self.job(g, cached=["pm"])
+        self.assertEqual(self.count(LLM), 3)
+
+    def test_suppression_keeps_the_pending_catch_up_and_the_tracker_queue(self):
+        g = graph()
+        self.first(g)
+        prompt_id = self.db.get_prompt_by_hash(generate_prompt_hash(LLM))["id"]
+
+        def run():
+            # The replay counted, then the prompt was deleted and re-saved
+            self.db.delete_prompt(prompt_id)
+            self.pending.add(generate_prompt_hash(LLM))
+            return self.execute(g)
+
+        self.job(g, cached=["pm"], run=run)
+        self.assertEqual(self.count(LLM), 1)  # the catch-up, not the own count
+        push = self.trackers[-1].set_current_prompt.call_args.kwargs["push_to_queue"]
+        self.assertTrue(push)
+
+    def test_without_host_context_the_node_counts_as_before(self):
+        g = graph()
+        self.runs = CachedRuns(context=lambda: (None, None))
+        self.first(g)
+        self.job(g, cached=["pm"])
+        self.job(g, run=lambda: self.execute(g))
+        self.assertEqual(self.count(LLM), 2)
+
+
+class TestPairReplay(JobCase):
+    def test_runtime_positive_counts_positive_and_pair_every_job(self):
+        g = graph("PromptManagerPair", negative_text="blurry")
+        self.first(g, negative="blurry")
+        self.job(graph("PromptManagerPair", negative_text="blurry", seed=2), ["pm"])
+        self.job(graph("PromptManagerPair", negative_text="blurry", seed=3), ["pm"])
+        self.assertEqual(self.count(LLM), 3)
+        self.assertEqual(self.pair_count(LLM, "blurry"), 3)
+        self.assertEqual(self.count("blurry"), 0)
+
+    def test_empty_negative_pair_replays_its_pair_count(self):
+        g = graph("PromptManagerPair")
+        self.first(g)
+        self.job(g, cached=["pm"])
+        self.assertEqual(self.pair_count(LLM, ""), 2)
+        self.assertEqual(self.count(LLM), 2)
+
+    def test_pair_execution_after_replay_is_not_counted_again(self):
+        g = graph("PromptManagerPair", negative_text="blurry")
+        self.first(g, negative="blurry")
+        self.job(g, ["pm"], run=lambda: self.execute(g, negative="blurry"))
+        self.assertEqual(self.count(LLM), 2)
+        self.assertEqual(self.pair_count(LLM, "blurry"), 2)
+
+    def test_pair_role_follows_the_current_graph(self):
+        g = graph("PromptManagerPair", negative_text="blurry")
+        self.first(g, negative="blurry")
+        rewired = graph("PromptManagerPair", negative_text="blurry")
+        rewired["ks"]["inputs"]["positive"] = ["other", 0]
+        self.job(rewired, cached=["pm"])
+        self.assertEqual(self.count(LLM), 1)
+        self.assertEqual(self.pair_count(LLM, "blurry"), 1)
+
+
+class TestOwnCounts(unittest.TestCase):
+    """The node path and replay decide through the same function."""
+
+    def test_single_node_table(self):
+        facts = Facts("p")
+        for used, seen, expected in (
+            ({"text"}, False, [("prompt", "p")]),
+            ({"text"}, True, []),
+            (set(), False, []),
+        ):
+            with self.subTest(used=used, seen=seen):
+                counts = own_counts(
+                    "PromptManager", facts, Role(frozenset(used)), Seen(text=seen)
+                )
+                self.assertEqual(counts, expected)
+        self.assertEqual(
+            own_counts("PromptManager", Facts(None), Role(frozenset({"text"})), Seen()),
+            [],
+        )
+
+    def test_pair_table(self):
+        both = Role(frozenset({"text", "negative_text"}), True)
+        self.assertEqual(
+            own_counts("PromptManagerPair", Facts("p", "n"), both, Seen()),
+            [("prompt", "p"), ("prompt", "n"), ("pair", "p", "n")],
+        )
+        self.assertEqual(
+            own_counts("PromptManagerPair", Facts("p", "p"), both, Seen()),
+            [("prompt", "p"), ("pair", "p", "p")],
+        )
+        self.assertEqual(
+            own_counts("PromptManagerPair", Facts("p"), both, Seen(text=True)),
+            [("pair", "p", None)],
+        )
+        self.assertEqual(
+            own_counts(
+                "PromptManagerPair", Facts("p", "n"), both, Seen(True, True, True)
+            ),
+            [],
+        )
+
+    def test_node_and_replay_agree_on_role_and_seen(self):
+        for g, node_id in (
+            (graph(), "pm"),
+            (graph(text="typed"), "pm"),
+            (graph(role="negative"), "pm"),
+            (graph(sampler=False), "pm"),
+            (graph("PromptManagerPair", negative_text="n"), "pm"),
+            (graph("PromptManagerPair", text="typed", negative_text="n"), "pm"),
+        ):
+            kind = g[node_id]["class_type"]
+            text = "typed" if isinstance(g[node_id]["inputs"]["text"], str) else LLM
+            facts = stored_facts(text, g[node_id]["inputs"].get("negative_text"))
+            with self.subTest(kind=kind, inputs=g[node_id]["inputs"]):
+                replay = own_counts(
+                    kind,
+                    facts,
+                    graph_role(g, node_id, kind),
+                    graph_seen(g, node_id, kind, facts),
+                )
+                self.assertEqual(replay, self.node_counts(g, node_id, kind, text))
+
+    @staticmethod
+    def node_counts(g, node_id, kind, text):
+        """What the executing node counted itself, from a recording database."""
+        db = mock.Mock()
+        db.get_prompt_by_hash.side_effect = lambda h: {"id": h}
+        db.save_prompt_pair.return_value = 1
+        negative = g[node_id]["inputs"].get("negative_text", "")
+        db.get_prompt_pair.return_value = {
+            "positive_prompt_id": generate_prompt_hash(text),
+            "negative_prompt_id": (
+                generate_prompt_hash(negative) if negative.strip() else None
+            ),
+        }
+        case = JobCase()
+        case.db = db
+        case.pending, case.pair_pending = PendingFirstUse(), PendingFirstUse()
+        case.runs = CachedRuns(context=lambda: (None, None))
+        case.trackers = []
+        case.context = (None, None)
+        if kind == "PromptManagerPair":
+            node = case.node(kind)
+            node.encode_prompt(
+                mock.Mock(), text, negative_text=negative, prompt=g, unique_id=node_id
+            )
+        else:
+            node = case.node(kind)
+            node._track_prompt_execution(text, text, None, None, {}, g, node_id)
+        counts = [("prompt", c.args[0]) for c in db.record_prompt_use.call_args_list]
+        if kind == "PromptManagerPair":
+            counts += [
+                ("pair", generate_prompt_hash(text), db.get_prompt_pair()[k])
+                for k in ["negative_prompt_id"]
+                if db.record_prompt_pair_use.called
+            ]
+        return counts
+
+
+class TestReceipts(unittest.TestCase):
+    def setUp(self):
+        self.context = ("j1", None)
+        self.runs = CachedRuns(
+            context=lambda: self.context, max_keys=2, max_receipts=2, max_items=3
+        )
+        self.g = {
+            name: {"class_type": "PromptManager", "inputs": {"text": name}}
+            for name in ("a", "b", "c")
+        }
+
+    def begin(self, item_id, prompt_id="j1"):
+        self.context = (prompt_id, None)
+        self.runs.begin_attempt(
+            ((item_id, prompt_id, self.g, {}, ["a", "b", "c"], {}), item_id)
+        )
+
+    def keys(self):
+        return list(self.runs._receipts)
+
+    def test_least_recently_used_run_keys_are_dropped(self):
+        self.begin(1)
+        for name in ("a", "b", "c"):
+            self.runs.record_item(name, (name,), Facts("h"))
+        self.assertEqual(self.keys(), [prompt_node_key(self.g, n) for n in ("b", "c")])
+
+    def test_receipts_per_key_are_bounded_and_equal_outputs_replace(self):
+        key = prompt_node_key(self.g, "a")
+        for item_id, text in ((1, "x"), (2, "y"), (3, "z"), (4, "y")):
+            self.begin(item_id)
+            self.runs.record_item("a", (text,), Facts(text))
+            self.runs.end_attempt(item_id)
+        receipts = self.runs._receipts[key]
+        self.assertEqual([r.item_id for r in receipts], [4, 3])
+
+    def test_list_indices_past_the_bound_are_not_recorded(self):
+        self.begin(1)
+        self.context = ("j1", 3)
+        self.runs.record_item("a", ("x",), Facts("x"))
+        self.assertEqual(self.keys(), [])
+
+    def test_non_string_outputs_and_unknown_nodes_are_not_recorded(self):
+        self.begin(1)
+        self.runs.record_item("a", (None,), Facts("x"))
+        self.runs.record_item("zzz", ("x",), Facts("x"))
+        self.assertEqual(self.keys(), [])
+
+    def test_task_done_clears_only_its_own_attempt(self):
+        self.begin(1)
+        self.runs.end_attempt(2)
+        self.runs.record_item("a", ("x",), Facts("x"))
+        self.runs.end_attempt(1)
+        self.runs.record_item("b", ("x",), Facts("x"))
+        self.assertEqual(self.keys(), [prompt_node_key(self.g, "a")])
+
+    def test_job_nodes_follow_the_outputs_to_execute(self):
+        self.runs.begin_attempt(((1, "j1", self.g, {}, ["b"], {}), 1))
+        self.assertEqual(list(self.runs._attempt.nodes), ["b"])
+
+
+class TestCacheReader(unittest.TestCase):
+    def setUp(self):
+        patcher = mock.patch.dict(sys.modules, host_modules())
+        patcher.start()
+        self.addCleanup(patcher.stop)
+        self.cache = LRUCache()
+        self.executor = types.SimpleNamespace(
+            caches=types.SimpleNamespace(outputs=self.cache)
+        )
+
+    def read(self, outputs, kind="PromptManager"):
+        self.cache.entries["n"] = Entry({}, outputs)
+        return cached_string_outputs(self.executor, "n", kind)
+
+    def test_reads_each_string_slot_in_order(self):
+        self.assertEqual(self.read([["c"], ["a", "b"]]), (("a", "b"),))
+        self.assertEqual(self.read([["t"]], "PromptManagerText"), (("t",),))
+        self.assertEqual(
+            self.read([[1], [2], ["p"], ["n"]], "PromptManagerPair"), (("p",), ("n",))
+        )
+
+    def test_fails_closed(self):
+        for outputs, kind in (
+            ([["c"], ["a", 1]], "PromptManager"),
+            ([["c"]], "PromptManager"),
+            ([["c"], ["a"] * 65], "PromptManager"),
+            ([[1], [2], ["p", "q"], ["n"]], "PromptManagerPair"),
+            ([["c"], "text"], "PromptManager"),
+            ([["c"], ["a"]], "CLIPTextEncode"),
+            ([[]] * 5, "PromptManager"),
+        ):
+            with self.subTest(outputs=outputs, kind=kind):
+                self.assertIsNone(self.read(outputs, kind))
+        self.cache.entries["m"] = ({}, [["c"], ["a"]])  # not a CacheEntry
+        self.assertIsNone(cached_string_outputs(self.executor, "m", "PromptManager"))
+        self.executor.caches.outputs = NullCache()
+        self.assertIsNone(self.read([["c"], ["a"]]))
+
+
+class TestGraphHelpers(unittest.TestCase):
+    def test_upstream_nodes_and_closure(self):
+        g = graph()
+        self.assertEqual(
+            set(upstream_nodes(g, ["save"])),
+            {"save", "ks", "pm", "llm", "str", "ckpt", "other"},
+        )
+        self.assertEqual(upstream_nodes(g, ["missing"]), [])
+        self.assertEqual(upstream_nodes(None, ["save"]), [])
+        self.assertIsNone(upstream_closure(g, ["save"], limit=2))
+        self.assertIsNone(upstream_closure({"a": {"inputs": {"x": ["b", 0]}}}, ["a"]))
+
+    def test_find_pair_by_hashes_matches_the_text_lookup(self):
+        with tempfile.TemporaryDirectory() as tmp:
+            db = PromptDatabase(os.path.join(tmp, "p.db"))
+            try:
+                pair_id = db.save_prompt_pair("cat", "")
+                self.assertEqual(
+                    find_pair_id_by_hashes(db, generate_prompt_hash("cat"), None),
+                    pair_id,
+                )
+                self.assertIsNone(
+                    find_pair_id_by_hashes(
+                        db, generate_prompt_hash("cat"), generate_prompt_hash("x")
+                    )
+                )
+            finally:
+                db.close_all()
+            assert_closed(self, tmp)
+
+
+class TestRegistration(unittest.TestCase):
+    def setUp(self):
+        self.modules = host_modules()
+        patcher = mock.patch.dict(sys.modules, self.modules)
+        patcher.start()
+        self.addCleanup(patcher.stop)
+        flag = mock.patch.object(module, "_registered", threading.Event())
+        flag.start()
+        self.addCleanup(flag.stop)
+
+    class Queue:
+        def __init__(self):
+            self.done = []
+
+        def get(self, timeout=None):
+            return self.result
+
+        def task_done(self, item_id, *args, **kwargs):
+            self.done.append(item_id)
+
+    def test_registers_once_and_follows_the_job(self):
+        queue = self.Queue()
+        server = types.SimpleNamespace(prompt_queue=queue)
+        runs = mock.Mock()
+        self.assertTrue(register_cached_runs(server, lambda: "db", runs))
+        self.assertFalse(register_cached_runs(server, lambda: "db", runs))
+        queue.result = (("item",), 5)
+        queue.get()
+        runs.begin_attempt.assert_called_once_with((("item",), 5))
+        executor = self.modules["execution"].PromptExecutor()
+        executor.server = server
+        executor.add_message("execution_start", {"prompt_id": "j"}, broadcast=False)
+        executor.add_message(
+            "execution_cached", {"prompt_id": "j", "nodes": ["n"]}, broadcast=False
+        )
+        runs.replay.assert_called_once()
+        self.assertEqual(runs.replay.call_args.args[1:3], ("j", ["n"]))
+        other = self.modules["execution"].PromptExecutor()
+        other.server = types.SimpleNamespace(prompt_queue=object())
+        other.add_message("execution_cached", {"prompt_id": "j", "nodes": []}, False)
+        self.assertEqual(runs.replay.call_count, 1)
+        queue.task_done(5, None, status=None)
+        runs.end_attempt.assert_called_once_with(5)
+
+    def test_hosts_without_the_interfaces_register_nothing(self):
+        class NoTaskDone:
+            def get(self):
+                return None
+
+        self.assertFalse(
+            register_cached_runs(
+                types.SimpleNamespace(prompt_queue=NoTaskDone()), lambda: None
+            )
+        )
+        self.assertFalse(register_cached_runs(types.SimpleNamespace(), lambda: None))
+        del self.modules["execution"].PromptExecutor.add_message
+        self.assertFalse(
+            register_cached_runs(
+                types.SimpleNamespace(prompt_queue=self.Queue()), lambda: None
+            )
+        )
+
+
+if __name__ == "__main__":
+    unittest.main()

tests/test_database_backup.py

@@ -614,6 +614,50 @@ class TestRestoreFromFile(BackupTestCase):
             )
         assert_closed(self, str(backups[0]))
 
+    def test_backup_and_restore_keep_collections(self):
+        with self.db.model.get_connection() as conn:
+            conn.execute(
+                "INSERT INTO image_collections (id, name, name_key, created_at) "
+                "VALUES (4, 'LoRa testing', 'lora testing', '2026-10-27T00:00:00Z')"
+            )
+            conn.execute(
+                "INSERT INTO image_collection_items VALUES "
+                "(4, 'root', 'grid.png', '2026-10-27T00:00:00Z')"
+            )
+        with tempfile.TemporaryDirectory() as directory:
+            # The rows still sit in the WAL; a backup must carry them.
+            backup = os.path.join(directory, "backup.db")
+            self.assertTrue(self.db.model.backup_database(backup))
+            self.assertEqual(
+                PromptModel.verify_database_file(backup, require_library_state=True),
+                (True, "ok"),
+            )
+            with self.db.model.get_connection() as conn:
+                conn.execute("DELETE FROM image_collections")
+
+            self.db.model.restore_from_file(backup)
+
+            conn = self.db.model.get_connection()
+            self.assertEqual(
+                [
+                    tuple(r)
+                    for r in conn.execute("SELECT id, name FROM image_collections")
+                ],
+                [(4, "LoRa testing")],
+            )
+            self.assertEqual(
+                [
+                    tuple(r)
+                    for r in conn.execute(
+                        "SELECT collection_id, root_key, relative_path "
+                        "FROM image_collection_items"
+                    )
+                ],
+                [(4, "root", "grid.png")],
+            )
+            PromptModel.close_all_instances()
+            assert_closed(self, directory)
+
     def test_restore_without_a_live_database_returns_empty_backup_path(self):
         src = self._make_source_db(["fresh"])
         self.db.close_all()

tests/test_database_link_matches.py

@@ -0,0 +1,456 @@
+"""The prompt-link read of the workspace Images page (#354).
+
+``PromptDatabase.get_image_link_matches`` returns every row of its filter
+SELECT in one SQLite step (one JSON value) instead of one step per row, so the
+read does not have to win the GIL back once per linked image while ComfyUI
+generates. Its result must equal the streamed read it replaces, which stays as
+the fallback, on every filter and on odd values.
+"""
+
+import itertools
+import os
+import sqlite3
+import sys
+import tempfile
+import unittest
+from pathlib import Path
+from unittest.mock import patch
+
+sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
+
+from database import operations  # noqa: E402
+from database.operations import PromptDatabase, link_filter_query  # noqa: E402
+
+try:
+    from tests.open_handles import assert_closed  # noqa: E402
+except ImportError:  # discovered with tests/ as the top-level directory
+    from open_handles import assert_closed  # noqa: E402
+
+LOGGER = "prompt_manager.database"
+MISSING_FUNCTION_SQL = (
+    "SELECT pm_missing_function(json_group_array(json_array("
+    "file_path, rating, matched, positive))) FROM ({select})"
+)
+QUERIES = ("", "lantern", "LANTERN title", "no-such-text")
+CATEGORIES = (None, "cat-a", "cat-b", "cat-none")
+TAG_SETS = ([], ["t1"], ["t1", "t2"], ["t-none"])
+
+PROMPTS = (
+    # id, text, category, rating, tags, library title
+    (1, "red lantern at dusk", "cat-a", 5, ("t1", "t2"), None),
+    (2, "blue harbor", "cat-a", None, ("t1",), None),
+    (3, "green field", "cat-b", 3, (), "Lantern title"),
+    (4, "pair positive", "cat-b", 4, ("t2",), None),
+    (5, "pair negative lantern", "cat-a", 1, ("t1",), None),
+)
+QUOTED = '/out/it\'s "quoted".png'
+BACKSLASHED = "C:\\out\\back\\slash.png"
+UNICODE = "/out/ünïcødé 画像.png"
+NEWLINE = "/out/new\nline.png"
+IMAGES = (
+    # prompt id, file path
+    (1, "/out/a.png"),
+    (2, QUOTED),
+    (2, BACKSLASHED),
+    (3, UNICODE),
+    (1, NEWLINE),
+    (1, "/out/ambiguous.png"),
+    (2, "/out/ambiguous.png"),
+    (5, "/out/pair.png"),  # linked through pair 1: its positive is prompt 4
+    (3, "/out/two-pairs.png"),  # two pairs: no single pair, so prompt 3
+    (1, None),  # no file path: never listed
+)
+
+
+class LinkMatchesTestCase(unittest.TestCase):
+    def setUp(self):
+        self.temp = tempfile.TemporaryDirectory()
+        self.root = Path(os.path.realpath(self.temp.name))
+        self.db = PromptDatabase(str(self.root / "prompts.db"))
+        for name in ("_one_step_unsupported", "_one_step_warned"):
+            patcher = patch.object(operations, name, False)
+            patcher.start()
+            self.addCleanup(patcher.stop)
+        self.statements = []
+        original = self.db.model.get_connection
+
+        def traced_connection():
+            connection = original()
+            connection.set_trace_callback(self.statements.append)
+            return connection
+
+        patcher = patch.object(
+            self.db.model, "get_connection", side_effect=traced_connection
+        )
+        patcher.start()
+        self.addCleanup(patcher.stop)
+
+    def tearDown(self):
+        self.db.close_all()
+        assert_closed(self, self.root)
+        self.temp.cleanup()
+
+    def seed(self):
+        with self.db.model.get_connection() as conn:
+            for prompt_id, text, category, rating, tags, title in PROMPTS:
+                conn.execute(
+                    "INSERT INTO prompts(id,text,category,rating,hash) "
+                    "VALUES(?,?,?,?,?)",
+                    (prompt_id, text, category, rating, f"hash-{prompt_id}"),
+                )
+                for tag in tags:
+                    conn.execute("INSERT OR IGNORE INTO tags(name) VALUES(?)", (tag,))
+                    conn.execute(
+                        "INSERT INTO prompt_tags(prompt_id,tag_id) "
+                        "SELECT ?, id FROM tags WHERE name=?",
+                        (prompt_id, tag),
+                    )
+                if title:
+                    conn.execute(
+                        "INSERT INTO prompt_library_state(prompt_id,title) "
+                        "VALUES(?,?)",
+                        (prompt_id, title),
+                    )
+            ids = {}
+            for prompt_id, path in IMAGES:
+                cursor = conn.execute(
+                    "INSERT INTO generated_images"
+                    "(prompt_id,image_path,filename,file_path) VALUES(?,?,?,?)",
+                    (prompt_id, path or "/elsewhere/x.png", "x.png", path),
+                )
+                ids.setdefault(path, cursor.lastrowid)
+            conn.execute(
+                "INSERT INTO prompt_pairs(id,positive_prompt_id,negative_prompt_id) "
+                "VALUES(1,4,5),(2,3,5),(3,3,NULL)"
+            )
+            conn.executemany(
+                "INSERT INTO image_prompt_pairs(image_id,pair_id,provenance) "
+                "VALUES(?,?,'{}')",
+                [
+                    (ids["/out/pair.png"], 1),
+                    (ids["/out/two-pairs.png"], 2),
+                    (ids["/out/two-pairs.png"], 3),
+                ],
+            )
+
+    def add_image(self, prompt_id, path):
+        with self.db.model.get_connection() as conn:
+            conn.execute(
+                "INSERT INTO generated_images(prompt_id,image_path,filename,file_path)"
+                " VALUES(?,?,?,?)",
+                (prompt_id, "/blob/x.png", "x.png", path),
+            )
+
+    def streamed(self, q, category, tags, every_link=False, wanted=None):
+        """Today's read, called directly (the fallback)."""
+        sql, params = link_filter_query(
+            q, tags, every_link=every_link, category=category
+        )
+        with self.db.model.get_connection() as conn:
+            return operations._streamed_link_matches(
+                conn, sql, params, wanted, category is not None
+            )
+
+    def selects(self):
+        found = [s for s in self.statements if s.lstrip().upper().startswith("SELECT")]
+        self.statements.clear()
+        return found
+
+
+class TestOneStepLinkRead(LinkMatchesTestCase):
+    def test_plain_read_returns_every_unambiguous_link(self):
+        self.seed()
+        matches = self.db.get_image_link_matches("", [], every_link=True)
+        self.assertEqual(
+            matches,
+            {
+                "/out/a.png": (True, 5, 1, None),
+                QUOTED: (True, None, 2, None),
+                BACKSLASHED: (True, None, 2, None),
+                UNICODE: (True, 3, 3, None),
+                NEWLINE: (True, 5, 1, None),
+                "/out/pair.png": (True, 4, 4, None),
+                "/out/two-pairs.png": (True, 3, 3, None),
+            },
+        )
+        self.assertIsNone(matches[QUOTED][1])
+        self.assertIsInstance(matches["/out/a.png"][1], int)
+        self.assertIsInstance(matches["/out/a.png"][2], int)
+
+    def test_search_hits_follow_text_title_and_tags(self):
+        self.seed()
+        self.assertEqual(
+            self.db.get_image_link_matches("lantern", []),
+            {
+                "/out/a.png": (True, 5, 1, None),
+                NEWLINE: (True, 5, 1, None),
+                UNICODE: (True, 3, 3, None),
+                "/out/two-pairs.png": (True, 3, 3, None),
+            },
+        )
+        self.assertEqual(
+            self.db.get_image_link_matches("lantern", [], category="cat-a"),
+            {
+                "/out/a.png": (True, 5, 1, "cat-a"),
+                NEWLINE: (True, 5, 1, "cat-a"),
+                QUOTED: (False, None, 2, "cat-a"),
+                BACKSLASHED: (False, None, 2, "cat-a"),
+            },
+        )
+
+    def test_every_filter_equals_the_streamed_read(self):
+        self.seed()
+        wanted_sets = (None, {"/out/a.png", UNICODE, "/out/pair.png", "/absent.png"})
+        for q, category, tags, every_link, wanted in itertools.product(
+            QUERIES, CATEGORIES, TAG_SETS, (False, True), wanted_sets
+        ):
+            with self.subTest(
+                q=q, category=category, tags=tags, every=every_link, wanted=wanted
+            ):
+                expected = self.streamed(q, category, tags, every_link, wanted)
+                self.statements.clear()
+                actual = self.db.get_image_link_matches(
+                    q,
+                    list(tags),
+                    every_link=every_link,
+                    wanted=wanted,
+                    category=category,
+                )
+                self.assertEqual(actual, expected)
+                self.assertEqual(
+                    {path: tuple(map(type, row)) for path, row in actual.items()},
+                    {path: tuple(map(type, row)) for path, row in expected.items()},
+                )
+                found = self.selects()
+                self.assertEqual(len(found), 1, found)
+                self.assertIn("json_group_array", found[0])
+
+    def test_empty_database_returns_nothing_in_one_statement(self):
+        self.assertEqual(self.db.get_image_link_matches("", [], every_link=True), {})
+        self.assertEqual(self.db.get_image_link_matches("x", ["t"], category="c"), {})
+        found = self.selects()
+        self.assertEqual(len(found), 2, found)
+        self.assertTrue(all("json_group_array" in query for query in found))
+
+    def test_wanted_limits_the_result_to_the_listing(self):
+        self.seed()
+        self.assertEqual(
+            self.db.get_image_link_matches(
+                "", [], every_link=True, wanted={"/out/pair.png", "/x.png"}
+            ),
+            {"/out/pair.png": (True, 4, 4, None)},
+        )
+        self.assertEqual(
+            self.db.get_image_link_matches("", [], every_link=True, wanted=set()),
+            {},
+        )
+
+    def test_category_read_carries_the_inherited_category_in_one_step(self):
+        # A recipe's own category wins over its positive's (#332 MC4); the
+        # read only narrows to links holding the value, the caller compares.
+        self.seed()
+        with self.db.model.get_connection() as conn:
+            conn.execute("UPDATE prompt_pairs SET category='cat-a' WHERE id=1")
+        self.statements.clear()
+        for category, expected in (
+            ("cat-a", {"/out/pair.png": (True, 4, 4, "cat-a")}),
+            ("cat-b", {"/out/pair.png": (True, 4, 4, "cat-a")}),
+        ):
+            with self.subTest(category=category):
+                actual = self.db.get_image_link_matches(
+                    "", [], every_link=True, category=category, wanted={"/out/pair.png"}
+                )
+                self.assertEqual(actual, expected)
+                self.assertEqual(
+                    actual, self.streamed("", category, [], True, {"/out/pair.png"})
+                )
+        found = [query for query in self.selects() if "json_group_array" in query]
+        self.assertEqual(len(found), 2, found)
+        self.assertTrue(all("pair_category" in query for query in found))
+
+    def test_plain_read_leaves_the_categories_out_of_the_json(self):
+        self.seed()
+        self.statements.clear()
+        self.db.get_image_link_matches("", [], every_link=True)
+        (query,) = self.selects()
+        aggregate = query[query.index("json_array(") :].split(")", 1)[0]
+        self.assertEqual(aggregate, "json_array(file_path, rating, matched, positive")
+
+    def test_read_releases_its_snapshot(self):
+        # A statement left open after the read would pin its WAL snapshot, and
+        # a TRUNCATE checkpoint from another connection would report busy.
+        self.seed()
+        self.assertTrue(self.db.get_image_link_matches("", [], every_link=True))
+        other = sqlite3.connect(str(self.root / "prompts.db"))
+        try:
+            other.execute("UPDATE prompts SET rating=2 WHERE id=1")
+            other.commit()
+            busy, _log, _done = other.execute(
+                "PRAGMA wal_checkpoint(TRUNCATE)"
+            ).fetchone()
+        finally:
+            other.close()
+        self.assertEqual(busy, 0)
+        self.assertEqual(os.path.getsize(self.root / "prompts.db-wal"), 0)
+
+    def test_wrapped_query_keeps_the_ambiguity_probe_an_index_search(self):
+        self.seed()
+        sql, params = link_filter_query(
+            "lantern", ["t1"], every_link=True, category="cat-a"
+        )
+        with self.db.model.get_connection() as conn:
+            plan = [
+                row[3]
+                for row in conn.execute(
+                    "EXPLAIN QUERY PLAN "
+                    + operations.link_matches_json_query(sql, categories=True),
+                    params,
+                )
+            ]
+        self.assertIn(
+            "SEARCH other USING COVERING INDEX idx_image_file_path (file_path=?)", plan
+        )
+        self.assertFalse([step for step in plan if step.startswith("SCAN other")])
+
+
+class TestStreamedFallback(LinkMatchesTestCase):
+    def test_missing_json_functions_stream_and_stop_trying(self):
+        self.seed()
+        expected = self.streamed("", None, [], True)
+        self.statements.clear()
+        wrapper = operations.link_matches_json_query
+        with (
+            patch.object(operations, "_LINK_MATCHES_JSON_SQL", MISSING_FUNCTION_SQL),
+            patch.object(
+                operations, "link_matches_json_query", side_effect=wrapper
+            ) as attempts,
+            self.assertLogs(LOGGER, "WARNING") as logged,
+        ):
+            first = self.db.get_image_link_matches("", [], every_link=True)
+            found = self.selects()
+            self.assertEqual(attempts.call_count, 1)
+            second = self.db.get_image_link_matches("", [], every_link=True)
+            again = self.selects()
+        self.assertEqual(first, expected)
+        self.assertEqual(second, expected)
+        self.assertEqual(len(logged.records), 1, logged.output)
+        self.assertIn("no such function: pm_missing_function", logged.output[0])
+        self.assertTrue(operations._one_step_unsupported)
+        # SQLite refuses the wrapped statement before it runs (not traced).
+        self.assertEqual(len(found), 1, found)
+        self.assertNotIn("json_group_array", found[0])
+        # The next read skips the attempt for the rest of the process.
+        self.assertEqual(attempts.call_count, 1)
+        self.assertEqual(len(again), 1, again)
+        self.assertNotIn("json_group_array", again[0])
+
+    def test_blob_path_streams_that_read_and_keeps_trying(self):
+        self.seed()
+        self.add_image(2, b"/out/blob.png")
+        expected = self.streamed("", None, [], True)
+        self.assertEqual(expected[b"/out/blob.png"], (True, None, 2, None))
+        self.statements.clear()
+        with self.assertLogs(LOGGER, "WARNING") as logged:
+            first = self.db.get_image_link_matches("", [], every_link=True)
+            second = self.db.get_image_link_matches("", [], every_link=True)
+        self.assertEqual(first, expected)
+        self.assertEqual(second, expected)
+        self.assertEqual(len(logged.records), 1, logged.output)
+        self.assertFalse(operations._one_step_unsupported)
+        found = self.selects()
+        self.assertEqual(sum("json_group_array" in query for query in found), 2)
+        self.assertEqual(len(found), 4, found)
+
+    def test_blob_that_reads_as_json_still_streams(self):
+        # SQLite 3.45+ reads a BLOB that is valid JSONB as JSON (b"\x00" is
+        # null) instead of refusing it; the streamed read keeps the bytes.
+        self.seed()
+        self.add_image(2, b"\x00")
+        expected = self.streamed("", None, [], True)
+        self.assertIn(b"\x00", expected)
+        with self.assertLogs(LOGGER, "WARNING"):
+            self.assertEqual(
+                self.db.get_image_link_matches("", [], every_link=True), expected
+            )
+        self.assertFalse(operations._one_step_unsupported)
+
+    def test_blob_rating_streams(self):
+        self.seed()
+        with self.db.model.get_connection() as conn:
+            # Only a database written around the CHECK can hold one.
+            conn.execute("PRAGMA ignore_check_constraints=ON")
+            conn.execute("UPDATE prompts SET rating=? WHERE id=3", (b"\x00",))
+            conn.execute("PRAGMA ignore_check_constraints=OFF")
+        expected = self.streamed("", None, [], True)
+        self.assertEqual(expected[UNICODE], (True, b"\x00", 3, None))
+        with self.assertLogs(LOGGER, "WARNING"):
+            self.assertEqual(
+                self.db.get_image_link_matches("", [], every_link=True), expected
+            )
+
+    def test_blob_category_streams_a_category_read(self):
+        self.seed()
+        with self.db.model.get_connection() as conn:
+            # Admitted by its recipe's text category; the positive's is a BLOB.
+            conn.execute("UPDATE prompts SET category=? WHERE id=4", (b"cat-b",))
+            conn.execute("UPDATE prompt_pairs SET category='cat-b' WHERE id=1")
+        expected = self.streamed("", "cat-b", [], True)
+        self.assertEqual(expected["/out/pair.png"], (True, 4, 4, "cat-b"))
+        with self.assertLogs(LOGGER, "WARNING") as logged:
+            actual = self.db.get_image_link_matches(
+                "", [], every_link=True, category="cat-b"
+            )
+        self.assertEqual(actual, expected)
+        self.assertIn("BLOB", logged.output[0])
+        self.assertFalse(operations._one_step_unsupported)
+
+    def test_json_text_over_the_length_limit_streams(self):
+        self.seed()
+        expected = self.streamed("", None, [], True)
+        conn = self.db.model.get_connection()
+        limit = conn.setlimit(sqlite3.SQLITE_LIMIT_LENGTH, 200)
+        try:
+            with self.assertLogs(LOGGER, "WARNING") as logged:
+                actual = self.db.get_image_link_matches("", [], every_link=True)
+        finally:
+            conn.setlimit(sqlite3.SQLITE_LIMIT_LENGTH, limit)
+        self.assertEqual(actual, expected)
+        self.assertIn("too big", logged.output[0])
+        self.assertFalse(operations._one_step_unsupported)
+
+    def test_an_interrupted_read_raises_without_a_fallback(self):
+        # Move library interrupts a retired connection: the read fails as
+        # before, and the one warning is kept for a real fallback.
+        self.seed()
+        conn = self.db.model.get_connection()
+        conn.set_progress_handler(lambda: 1, 1)
+        self.statements.clear()
+        try:
+            with (
+                self.assertNoLogs(LOGGER, "WARNING"),
+                self.assertRaisesRegex(sqlite3.OperationalError, "^interrupted$"),
+            ):
+                self.db.get_image_link_matches("", [], every_link=True)
+        finally:
+            conn.set_progress_handler(None, 1)
+        self.assertFalse(operations._one_step_warned)
+        self.assertFalse(operations._one_step_unsupported)
+        self.assertFalse([s for s in self.statements if "json_group_array" not in s])
+
+    def test_the_maintenance_link_rows_reader_is_not_shadowed(self):
+        from database.maintenance import MaintenanceOperations
+
+        self.assertIs(PromptDatabase._link_rows, MaintenanceOperations._link_rows)
+
+    def test_a_failed_streamed_read_still_raises(self):
+        with (
+            patch.object(operations, "_LINK_MATCHES_JSON_SQL", MISSING_FUNCTION_SQL),
+            patch.object(operations, "_LINK_FILTER_SQL", "SELECT broken FROM nowhere"),
+            self.assertLogs(LOGGER, "WARNING"),
+        ):
+            with self.assertRaises(sqlite3.OperationalError):
+                self.db.get_image_link_matches("", [], every_link=True)
+
+
+if __name__ == "__main__":
+    unittest.main()

tests/test_duplicate_image_categories.py

@@ -149,3 +149,36 @@ class TestDuplicateCleanupKeepsCategories(DuplicateFilesTestCase):
             kept = image_category_key([link], link / "a.png")
         self.assertFalse((self.output_dir / "b.png").exists())
         self.assertEqual(self.api.db.image_category(*kept), "Portraits")
+
+    async def test_a_copy_with_a_collection_and_a_category_moves_both(self):
+        # #331 + #332: one removed copy carries both records; one cleanup
+        # moves its membership and copies its category onto the kept copy.
+        self.trio()
+        db = self.api.db
+        root_key = duplicate_files.favorite_root_key(self.output_dir)
+        b = self.output_dir / "b.png"
+        grids = db.add_to_collection(name="LoRA grids", items=[(root_key, "b.png")])
+        grids = grids["collection"]["id"]
+        self.set_category(b, "Portraits")
+        seen = {}
+        real_log = duplicate_files.DuplicateFilesMixin._log_moved_records
+
+        def log_moved_records(job, totals):
+            seen.update(totals)
+            real_log(job, totals)
+
+        with patch.object(
+            duplicate_files.DuplicateFilesMixin,
+            "_log_moved_records",
+            staticmethod(log_moved_records),
+        ):
+            cleanup = await self.run_cleanup()
+        self.assertFalse(b.exists())
+        self.assertEqual(db.collection_members(grids, root_key), {"a.png"})
+        self.assertEqual(db.file_collections(root_key, "b.png"), [])
+        self.assertEqual(self.category(self.output_dir / "a.png"), "Portraits")
+        self.assertEqual(seen["collections"], 1)
+        self.assertEqual(seen["categories"], 1)
+        log = "\n".join(cleanup.log)
+        self.assertIn("Moved 1 collection membership to the kept copies", log)
+        self.assertIn("Kept 1 image category on the kept copies", log)

tests/test_image_collections.py

@@ -0,0 +1,268 @@
+"""Collections (#331): the database operations."""
+
+import sqlite3
+import tempfile
+import unittest
+from pathlib import Path
+from unittest.mock import patch
+
+try:
+    from tests.open_handles import assert_closed
+except ImportError:  # discovered with tests/ as the top-level directory
+    from open_handles import assert_closed
+
+from database.models import PromptModel
+from database.operations import PromptDatabase
+
+try:  # guarded so a missing module fails assertions, not test loading
+    from database.image_collections import (
+        CollectionFilesMissing,
+        CollectionInvalid,
+        CollectionMissing,
+        CollectionNameTaken,
+        READ_SQL,
+    )
+except ImportError:
+    CollectionFilesMissing = CollectionMissing = CollectionNameTaken = None
+    CollectionInvalid = ValueError
+    READ_SQL = None
+
+ROOT = "/real/output"
+OTHER = "/real/other"
+
+
+class ImageCollectionOperationsTests(unittest.TestCase):
+    def setUp(self):
+        self.temp = tempfile.TemporaryDirectory()
+        self.addCleanup(self.temp.cleanup)
+        self.addCleanup(assert_closed, self, self.temp.name)
+        self.addCleanup(PromptModel.reset_schema_cache)
+        self.db = PromptDatabase(str(Path(self.temp.name) / "prompts.db"))
+        self.addCleanup(self.db.close_all)
+
+    def require(self):
+        self.assertTrue(
+            hasattr(self.db, "add_to_collection"),
+            "ImageCollectionOperations is not implemented",
+        )
+
+    def add(self, items, **target):
+        return self.db.add_to_collection(items=items, **target)
+
+    def test_add_creates_a_collection_by_name(self):
+        self.require()
+        result = self.add(
+            [(ROOT, "a.png"), (ROOT, "b.png"), (OTHER, "a.png")], name="LoRa testing"
+        )
+        collection = result["collection"]
+        self.assertEqual(collection["name"], "LoRa testing")
+        self.assertEqual(
+            result,
+            {"collection": collection, "created": True, "added": 3, "already": 0},
+        )
+        self.assertEqual(self.db.list_collections(), [collection])
+        self.assertTrue(self.db.collection_exists(collection["id"]))
+        self.assertEqual(self.db.get_collection(collection["id"]), collection)
+        self.assertEqual(
+            self.db.collection_members(collection["id"], ROOT), {"a.png", "b.png"}
+        )
+        self.assertEqual(
+            self.db.root_collection_members(OTHER), {collection["id"]: {"a.png"}}
+        )
+        self.assertEqual(self.db.collection_counts(), {collection["id"]: 3})
+        self.assertEqual(self.db.file_collections(ROOT, "a.png"), [collection["id"]])
+        self.assertEqual(self.db.file_collections(ROOT, "zzz.png"), [])
+
+    def test_add_by_name_ignores_case_and_unicode_form(self):
+        self.require()
+        first = self.add([(ROOT, "a.png")], name="Café grids")["collection"]
+        # NFD spelling (macOS) in another case joins the same collection.
+        second = self.add([(ROOT, "b.png")], name="CAFÉ  GRIDS")
+        self.assertEqual(second["collection"], first)
+        self.assertFalse(second["created"])
+        self.assertEqual(len(self.db.list_collections()), 1)
+        self.assertEqual(
+            self.db.collection_members(first["id"], ROOT), {"a.png", "b.png"}
+        )
+
+    def test_add_counts_already_present(self):
+        self.require()
+        made = self.add([(ROOT, "a.png"), (ROOT, "b.png")], name="Set")
+        result = self.add(
+            [(ROOT, "b.png"), (ROOT, "c.png"), (ROOT, "c.png")],
+            collection_id=made["collection"]["id"],
+        )
+        self.assertEqual(
+            (result["created"], result["added"], result["already"]), (False, 1, 1)
+        )
+        self.assertEqual(self.db.collection_counts(), {made["collection"]["id"]: 3})
+
+    def test_add_with_a_new_name_and_nothing_inserted_creates_nothing(self):
+        self.require()
+        with self.assertRaises(CollectionFilesMissing):
+            self.add([], name="Empty")
+        self.assertEqual(self.db.list_collections(), [])
+        # An existing name with nothing to add is not an error.
+        made = self.add([(ROOT, "a.png")], name="Kept")["collection"]
+        self.assertEqual(
+            self.add([], name="kept"),
+            {"collection": made, "created": False, "added": 0, "already": 0},
+        )
+
+    def test_add_to_a_deleted_collection_raises_missing(self):
+        self.require()
+        made = self.add([(ROOT, "a.png")], name="Gone")["collection"]
+        self.db.delete_collection(made["id"])
+        with self.assertRaises(CollectionMissing):
+            self.add([(ROOT, "b.png")], collection_id=made["id"])
+        with self.assertRaises(CollectionMissing):
+            self.db.remove_from_collection(made["id"], [(ROOT, "a.png")])
+        with self.assertRaises(CollectionMissing):
+            self.db.rename_collection(made["id"], "Back")
+        self.assertFalse(self.db.collection_exists(made["id"]))
+        self.assertIsNone(self.db.get_collection(made["id"]))
+
+    def test_remove_returns_the_count(self):
+        self.require()
+        made = self.add([(ROOT, "a.png"), (ROOT, "b.png")], name="Set")["collection"]
+        result = self.db.remove_from_collection(
+            made["id"], [(ROOT, "a.png"), (ROOT, "a.png"), (ROOT, "never.png")]
+        )
+        self.assertEqual(result, {"collection": made, "removed": 1})
+        self.assertEqual(self.db.collection_members(made["id"], ROOT), {"b.png"})
+        # Removing the last member keeps the (now empty) collection.
+        self.db.remove_from_collection(made["id"], [(ROOT, "b.png")])
+        self.assertEqual(self.db.list_collections(), [made])
+        self.assertEqual(self.db.collection_counts(), {made["id"]: 0})
+
+    def test_rename_refuses_a_taken_name(self):
+        self.require()
+        first = self.add([(ROOT, "a.png")], name="First")["collection"]
+        second = self.add([(ROOT, "a.png")], name="Second")["collection"]
+        with self.assertRaises(CollectionNameTaken) as caught:
+            self.db.rename_collection(second["id"], " first ")
+        self.assertEqual(caught.exception.collection_id, first["id"])
+        self.assertEqual(self.db.get_collection(second["id"]), second)
+
+    def test_rename_accepts_a_case_variant_of_its_own_name(self):
+        self.require()
+        made = self.add([(ROOT, "a.png")], name="lora grids")["collection"]
+        renamed = self.db.rename_collection(made["id"], "LoRA   grids ")
+        self.assertEqual(renamed, {"id": made["id"], "name": "LoRA grids"})
+        self.assertEqual(self.db.list_collections(), [renamed])
+        other = self.db.rename_collection(made["id"], "Portraits")
+        self.assertEqual(other["name"], "Portraits")
+
+    def test_delete_unknown_raises_missing(self):
+        self.require()
+        made = self.add([(ROOT, "a.png")], name="Doomed")["collection"]
+        self.assertTrue(self.db.delete_collection(made["id"]))
+        self.assertEqual(self.db.root_collection_members(ROOT), {})
+        with self.assertRaises(CollectionMissing):
+            self.db.delete_collection(made["id"])
+        with self.assertRaises(CollectionMissing):
+            self.db.delete_collection(999)
+
+    def test_list_is_in_name_key_order(self):
+        self.require()
+        for name in ("beta", "Alpha", "été", "Zeta", "alpha two"):
+            self.add([(ROOT, "a.png")], name=name)
+        self.assertEqual(
+            [row["name"] for row in self.db.list_collections()],
+            ["Alpha", "alpha two", "beta", "Zeta", "été"],
+        )
+
+    def test_names_are_trimmed_and_collapsed(self):
+        self.require()
+        made = self.add([(ROOT, "a.png")], name="  LoRa \t testing\n ")["collection"]
+        self.assertEqual(made["name"], "LoRa testing")
+        for bad in ("", "   ", "x" * 101, "bell\x07", "c1\x81", "del\x7f", 5, None):
+            with self.subTest(name=bad):
+                with self.assertRaises(CollectionInvalid):
+                    self.add([(ROOT, "a.png")], name=bad)
+        self.assertEqual(len(self.db.list_collections()), 1)
+
+    def test_targets_and_items_are_validated(self):
+        self.require()
+        made = self.add([(ROOT, "a.png")], name="Set")["collection"]
+        for target in (
+            {},
+            {"name": "x", "collection_id": made["id"]},
+            {"collection_id": True},
+            {"collection_id": 0},
+            {"collection_id": "1"},
+        ):
+            with self.subTest(target=target):
+                with self.assertRaises(CollectionInvalid):
+                    self.add([(ROOT, "a.png")], **target)
+        for items in ([("root",)], [(ROOT, 5)], "a.png", [(ROOT, "x" * 4097)]):
+            with self.subTest(items=items):
+                with self.assertRaises(CollectionInvalid):
+                    self.add(items, collection_id=made["id"])
+
+    def test_a_failure_after_the_membership_move_rolls_it_back(self):
+        self.require()
+        made = self.add([(ROOT, "b.png")], name="Grids")["collection"]
+        move = {
+            "removed": [],
+            "kept": [],
+            "targets": {},
+            "kept_path": "/real/output/a.png",
+            "favorites": [(ROOT, "b.png", ROOT, "a.png")],
+            "removed_locator": "/real/output/b.png",
+            "kept_locator": "/real/output/a.png",
+            "sha256": "0" * 64,
+            "kept_size": 1,
+            "kept_mtime_ns": 1,
+        }
+        with patch.object(
+            type(self.db), "_move_copy_tags", side_effect=RuntimeError("boom")
+        ):
+            with self.assertRaises(RuntimeError):
+                self.db.move_copy_records(move)
+        self.assertEqual(self.db.file_collections(ROOT, "b.png"), [made["id"]])
+        self.assertEqual(self.db.file_collections(ROOT, "a.png"), [])
+        # Without the failure the same move succeeds.
+        self.assertEqual(self.db.move_copy_records(move)["collections"], 1)
+        self.assertEqual(self.db.file_collections(ROOT, "a.png"), [made["id"]])
+
+    def test_count_snapshot_reads_in_one_transaction(self):
+        self.require()
+        made = self.add([(ROOT, "a.png"), (OTHER, "b.png")], name="Set")["collection"]
+        statements = []
+        conn = self.db.model.get_connection()
+        conn.set_trace_callback(statements.append)
+        try:
+            members, stored = self.db.collection_count_snapshot([ROOT, OTHER, ROOT])
+        finally:
+            conn.set_trace_callback(None)
+        self.assertEqual(
+            members, {ROOT: {made["id"]: {"a.png"}}, OTHER: {made["id"]: {"b.png"}}}
+        )
+        self.assertEqual(stored, {made["id"]: 2})
+        self.assertEqual(statements[0], "BEGIN")
+        self.assertEqual(statements[-1], "ROLLBACK")
+        self.assertFalse(conn.in_transaction)
+
+    def test_reads_use_an_index(self):
+        self.require()
+        self.add([(ROOT, "a.png")], name="Set")
+        conn = sqlite3.connect(self.db.model.db_path)
+        try:
+            for sql in READ_SQL:
+                plan = " ".join(
+                    row[3]
+                    for row in conn.execute(
+                        "EXPLAIN QUERY PLAN " + sql, [None] * sql.count("?")
+                    )
+                )
+                with self.subTest(sql=sql):
+                    self.assertNotIn("SCAN image_collection_items", plan)
+                    if "?" in sql:  # a lookup, not the whole list
+                        self.assertNotIn("SCAN image_collections", plan)
+        finally:
+            conn.close()
+
+
+if __name__ == "__main__":
+    unittest.main()

tests/test_image_collections_schema.py

@@ -0,0 +1,521 @@
+"""Collections schema (#331): install, validation and restore safety."""
+
+import re
+import sqlite3
+import tempfile
+import unittest
+import unicodedata
+from contextlib import closing
+from pathlib import Path
+from unittest.mock import patch
+
+try:
+    from tests.open_handles import assert_closed
+    from tests.test_annotation_schema import LEGACY_TABLES, table_rows
+except ImportError:  # discovered with tests/ as the top-level directory
+    from open_handles import assert_closed
+    from test_annotation_schema import LEGACY_TABLES, table_rows
+
+from database.models import PromptModel
+
+try:  # guarded so a missing module fails assertions, not test loading
+    from database import image_collections
+    from database.image_collections import (
+        image_collections_schema_error,
+        migrate_image_collections,
+    )
+except ImportError:
+    image_collections = None
+    image_collections_schema_error = None
+    migrate_image_collections = None
+
+COLLECTIONS = "image_collections"
+ITEMS = "image_collection_items"
+INDEX = "idx_image_collection_items_file"
+TABLES = (COLLECTIONS, ITEMS)
+DDL = {
+    COLLECTIONS: (
+        f"CREATE TABLE {COLLECTIONS} (id INTEGER PRIMARY KEY AUTOINCREMENT, "
+        "name TEXT NOT NULL, name_key TEXT NOT NULL UNIQUE, "
+        "created_at TEXT NOT NULL)"
+    ),
+    ITEMS: (
+        f"CREATE TABLE {ITEMS} (collection_id INTEGER NOT NULL REFERENCES "
+        f"{COLLECTIONS}(id) ON DELETE CASCADE, root_key TEXT NOT NULL, "
+        "relative_path TEXT NOT NULL, added_at TEXT NOT NULL, "
+        "PRIMARY KEY (collection_id, root_key, relative_path))"
+    ),
+}
+INDEX_DDL = f"CREATE INDEX {INDEX} ON {ITEMS}(root_key, relative_path)"
+NOW = "2026-10-27T00:00:00.000000+00:00"
+IMAGES = 23_000
+
+
+def _compact(sql):
+    return re.sub(r"\s+", "", sql or "").lower()
+
+
+def expected_catalog():
+    return {
+        ("table", COLLECTIONS, _compact(DDL[COLLECTIONS])),
+        ("table", ITEMS, _compact(DDL[ITEMS])),
+        ("index", INDEX, _compact(INDEX_DDL)),
+        ("index", f"sqlite_autoindex_{COLLECTIONS}_1", ""),
+        ("index", f"sqlite_autoindex_{ITEMS}_1", ""),
+    }
+
+
+def catalog(conn):
+    return {
+        (kind, name, _compact(sql))
+        for kind, name, sql in conn.execute(
+            "SELECT type,name,sql FROM sqlite_master "
+            "WHERE name IN (?,?,?) OR tbl_name IN (?,?)",
+            (*TABLES, INDEX, *TABLES),
+        )
+    }
+
+
+def key(name):
+    return unicodedata.normalize("NFC", unicodedata.normalize("NFC", name).casefold())
+
+
+def schema(conn, exclude=()):
+    return sorted(
+        tuple(row)
+        for row in conn.execute(
+            "SELECT type,name,tbl_name,sql FROM sqlite_master ORDER BY name"
+        )
+        if row[1] not in exclude and row[2] not in exclude
+    )
+
+
+def row_counts(conn):
+    tables = [
+        name
+        for (name,) in conn.execute(
+            "SELECT name FROM sqlite_master WHERE type='table' ORDER BY name"
+        )
+    ]
+    return {
+        table: conn.execute(f"SELECT COUNT(*) FROM {table}").fetchone()[0]  # nosec
+        for table in tables
+    }
+
+
+class ImageCollectionsSchemaTests(unittest.TestCase):
+    def setUp(self):
+        self.temp = tempfile.TemporaryDirectory()
+        self.addCleanup(self.temp.cleanup)
+        self.addCleanup(assert_closed, self, self.temp.name)
+        self.addCleanup(PromptModel.reset_schema_cache)
+        self.models = []
+        self.addCleanup(self.close_models)
+        self.path = str(Path(self.temp.name) / "live.db")
+
+    def close_models(self):
+        for model in self.models:
+            model.close_all()
+
+    def model(self, path=None):
+        model = PromptModel(path or self.path)
+        self.models.append(model)
+        return model
+
+    def require_module(self):
+        self.assertIsNotNone(
+            image_collections_schema_error,
+            "database.image_collections is not implemented",
+        )
+
+    def closed_current(self, name="current.db", *statements):
+        """A closed current database, altered afterwards by *statements*."""
+        path = str(Path(self.temp.name) / name)
+        model = self.model(path)
+        model.close_all()
+        PromptModel.reset_schema_cache()
+        with closing(sqlite3.connect(path)) as raw, raw:
+            raw.execute("PRAGMA foreign_keys=OFF")
+            for statement in statements:
+                if isinstance(statement, tuple):
+                    raw.execute(*statement)
+                else:
+                    raw.execute(statement)
+        return path
+
+    @staticmethod
+    def drop_group(path):
+        with closing(sqlite3.connect(path)) as raw, raw:
+            raw.execute(f"DROP TABLE {ITEMS}")
+            raw.execute(f"DROP TABLE {COLLECTIONS}")
+
+    def test_migration_creates_tables_and_index(self):
+        self.require_module()
+        model = self.model()
+        with model.get_connection() as conn:
+            self.assertEqual(catalog(conn), expected_catalog())
+            self.assertIsNone(image_collections_schema_error(conn, allow_missing=False))
+            self.assertEqual(
+                conn.execute(f"SELECT COUNT(*) FROM {COLLECTIONS}").fetchone()[0], 0
+            )
+            self.assertEqual(
+                conn.execute(f"SELECT COUNT(*) FROM {ITEMS}").fetchone()[0], 0
+            )
+        self.assertEqual(
+            PromptModel.verify_database_file(self.path, require_library_state=True),
+            (True, "ok"),
+        )
+
+    def test_migration_is_idempotent(self):
+        self.require_module()
+        model = self.model()
+        with model.get_connection() as conn:
+            conn.execute(
+                f"INSERT INTO {COLLECTIONS} (name, name_key, created_at) "
+                "VALUES ('Kept', 'kept', ?)",
+                (NOW,),
+            )
+            conn.execute(f"INSERT INTO {ITEMS} VALUES (1, 'root', 'a.png', ?)", (NOW,))
+            before = schema(conn)
+        model.close_all()
+        for _ in range(2):
+            PromptModel.reset_schema_cache()
+            model = self.model()
+            with model.get_connection() as conn:
+                migrate_image_collections(conn)
+                self.assertEqual(schema(conn), before)
+                self.assertEqual(
+                    [tuple(r) for r in conn.execute(f"SELECT * FROM {ITEMS}")],
+                    [(1, "root", "a.png", NOW)],
+                )
+            model.close_all()
+
+    def test_migration_reads_only_its_own_tables(self):
+        self.require_module()
+        model = self.model()
+        with model.get_connection() as conn:
+            conn.execute("INSERT INTO prompts(id,text,hash) VALUES(1,'positive','p')")
+            conn.executemany(
+                "INSERT INTO generated_images(prompt_id,image_path,filename,"
+                "file_path) VALUES(1,?,?,?)",
+                ((f"/out/{n}.png", f"{n}.png", f"/out/{n}.png") for n in range(IMAGES)),
+            )
+            conn.execute("INSERT INTO image_favorites VALUES('root','a.png',?)", (NOW,))
+        model.close_all()
+        self.drop_group(self.path)
+        with closing(sqlite3.connect(self.path)) as raw:
+            raw.execute("PRAGMA foreign_keys=ON")
+            counts_before = row_counts(raw)
+            schema_before = schema(raw)
+            legacy_before = table_rows(raw, LEGACY_TABLES)
+            self.assertEqual(counts_before["generated_images"], IMAGES)
+            touched = []
+
+            def record(action, arg1, arg2, _db, _trigger):
+                touched.append((action, arg1, arg2))
+                return sqlite3.SQLITE_OK
+
+            raw.set_authorizer(record)
+            try:
+                migrate_image_collections(raw)
+            finally:
+                raw.set_authorizer(None)
+            raw.commit()
+            reads = {
+                arg1 for action, arg1, _arg2 in touched if action == sqlite3.SQLITE_READ
+            }
+            self.assertTrue(reads)
+            self.assertLessEqual(reads, {"sqlite_master", COLLECTIONS, ITEMS})
+            writes = {
+                arg1
+                for action, arg1, _arg2 in touched
+                if action
+                in (sqlite3.SQLITE_INSERT, sqlite3.SQLITE_UPDATE, sqlite3.SQLITE_DELETE)
+            }
+            self.assertLessEqual(writes, {"sqlite_master"})
+            self.assertEqual(catalog(raw), expected_catalog())
+            counts_after = row_counts(raw)
+            self.assertEqual(counts_after.pop(COLLECTIONS), 0)
+            self.assertEqual(counts_after.pop(ITEMS), 0)
+            self.assertEqual(counts_after, counts_before)
+            self.assertEqual(schema(raw, exclude=TABLES), schema_before)
+            self.assertEqual(table_rows(raw, LEGACY_TABLES), legacy_before)
+            # A second run is a no-op.
+            after_first = schema(raw)
+            changes = raw.total_changes
+            migrate_image_collections(raw)
+            raw.commit()
+            self.assertEqual(schema(raw), after_first)
+            self.assertEqual(raw.total_changes, changes)
+
+    def test_migration_refuses_a_foreign_object_of_the_same_name(self):
+        self.require_module()
+        for statements in (
+            (f"CREATE TABLE {COLLECTIONS} (x)",),
+            (f"CREATE VIEW {COLLECTIONS} AS SELECT 1",),
+            (f"CREATE INDEX {INDEX} ON prompts(text)",),
+        ):
+            with self.subTest(statements=statements):
+                path = str(Path(self.temp.name) / f"foreign{len(self.models)}.db")
+                self.model(path).close_all()
+                PromptModel.reset_schema_cache()
+                self.drop_group(path)
+                with closing(sqlite3.connect(path)) as raw, raw:
+                    for statement in statements:
+                        raw.execute(statement)
+                with closing(sqlite3.connect(path)) as raw:
+                    before = schema(raw)
+                    self.assertIsNotNone(image_collections_schema_error(raw))
+                with self.assertRaises(sqlite3.DatabaseError):
+                    self.model(path).get_connection()
+                PromptModel.close_all_instances()
+                PromptModel.reset_schema_cache()
+                with closing(sqlite3.connect(path)) as raw:
+                    self.assertEqual(schema(raw), before)
+                    self.assertIsNone(
+                        raw.execute(
+                            "SELECT 1 FROM sqlite_master WHERE name=?", (ITEMS,)
+                        ).fetchone()
+                    )
+                self.assertFalse(PromptModel.verify_database_file(path)[0])
+
+    def test_migration_rolls_back_on_error(self):
+        self.require_module()
+        self.model().close_all()
+        PromptModel.reset_schema_cache()
+        self.drop_group(self.path)
+        with closing(sqlite3.connect(self.path)) as raw:
+            with patch.object(
+                image_collections,
+                "image_collections_schema_error",
+                side_effect=[None, "Image collections went wrong"],
+            ):
+                with self.assertRaisesRegex(sqlite3.DatabaseError, "went wrong"):
+                    migrate_image_collections(raw)
+            raw.commit()
+            self.assertEqual(catalog(raw), set())
+
+    def test_startup_accepts_a_value_saved_under_older_rules(self):
+        self.require_module()
+        path = self.closed_current(
+            "older.db",
+            (
+                f"INSERT INTO {COLLECTIONS} (name, name_key, created_at) "
+                "VALUES (?, 'not the key', ?)",
+                (" " + "x" * 150 + " ", NOW),
+            ),
+        )
+        model = self.model(path)
+        with model.get_connection() as conn:
+            self.assertIsNone(image_collections_schema_error(conn, check_values=False))
+            self.assertIsNotNone(image_collections_schema_error(conn))
+            self.assertEqual(
+                conn.execute(f"SELECT COUNT(*) FROM {COLLECTIONS}").fetchone()[0], 1
+            )
+        model.close_all()
+        self.assertFalse(
+            PromptModel.verify_database_file(path, require_library_state=True)[0]
+        )
+
+    def test_restore_accepts_an_old_backup_without_collections(self):
+        self.require_module()
+        path = self.closed_current("old.db")
+        self.drop_group(path)
+        with closing(sqlite3.connect(path)) as raw:
+            self.assertIsNone(image_collections_schema_error(raw))
+            self.assertEqual(
+                image_collections_schema_error(raw, allow_missing=False),
+                "Missing image collections schema",
+            )
+        self.assertEqual(PromptModel.verify_database_file(path), (True, "ok"))
+        model = self.model(path)
+        with model.get_connection() as conn:
+            self.assertEqual(catalog(conn), expected_catalog())
+        model.close_all()
+        self.assertEqual(
+            PromptModel.verify_database_file(path, require_library_state=True),
+            (True, "ok"),
+        )
+
+    def test_restore_requires_collections_on_a_current_stage(self):
+        self.require_module()
+        path = self.closed_current("stage.db")
+        self.drop_group(path)
+        self.assertEqual(
+            PromptModel.verify_database_file(path, require_library_state=True),
+            (False, "Missing image collections schema"),
+        )
+        # One table without the other is never acceptable.
+        partial = self.closed_current("partial.db", f"DROP TABLE {ITEMS}")
+        self.assertFalse(PromptModel.verify_database_file(partial)[0])
+
+    def assert_refused(self, name, *statements):
+        self.require_module()
+        path = self.closed_current(
+            f"{name}.db",
+            (
+                f"INSERT INTO {COLLECTIONS} (id, name, name_key, created_at) "
+                "VALUES (1, 'Good', 'good', ?)",
+                (NOW,),
+            ),
+            (f"INSERT INTO {ITEMS} VALUES (1, 'root', 'a.png', ?)", (NOW,)),
+            *statements,
+        )
+        with closing(sqlite3.connect(path)) as raw:
+            self.assertIsNone(image_collections_schema_error(raw, check_values=False))
+            error = image_collections_schema_error(raw)
+        self.assertIsNotNone(error)
+        self.assertIn("collections", error.lower())
+        ok, reason = PromptModel.verify_database_file(path)
+        self.assertFalse(ok)
+        self.assertEqual(reason, error)
+        self.assertFalse(
+            PromptModel.verify_database_file(path, require_library_state=True)[0]
+        )
+
+    def test_a_valid_current_database_is_accepted(self):
+        self.require_module()
+        path = self.closed_current(
+            "valid.db",
+            (
+                f"INSERT INTO {COLLECTIONS} (id, name, name_key, created_at) "
+                "VALUES (1, ?, ?, ?)",
+                ("LoRa testing", key("LoRa testing"), NOW),
+            ),
+            (f"INSERT INTO {ITEMS} VALUES (1, 'root', ?, ?)", ("é" * 4096, NOW)),
+        )
+        self.assertEqual(
+            PromptModel.verify_database_file(path, require_library_state=True),
+            (True, "ok"),
+        )
+
+    def test_restore_refuses_a_non_text_name(self):
+        # Column affinity turns a number into text; a blob stays a blob.
+        self.assert_refused("blob-name", f"UPDATE {COLLECTIONS} SET name = x'61'")
+
+    def test_restore_refuses_an_empty_name(self):
+        self.assert_refused("empty", f"UPDATE {COLLECTIONS} SET name = '', name_key=''")
+
+    def test_restore_refuses_a_name_over_100_characters(self):
+        name = "é" * 101
+        self.assert_refused(
+            "long",
+            (f"UPDATE {COLLECTIONS} SET name = ?, name_key = ?", (name, key(name))),
+        )
+
+    def test_restore_refuses_a_control_character(self):
+        name = "a\x07b"
+        self.assert_refused(
+            "control",
+            (f"UPDATE {COLLECTIONS} SET name = ?, name_key = ?", (name, key(name))),
+        )
+
+    def test_restore_refuses_a_leading_or_trailing_space(self):
+        for name in (" Good", "Good "):
+            with self.subTest(name=name):
+                self.assert_refused(
+                    f"padded{len(name)}{name[0] == ' '}",
+                    (
+                        f"UPDATE {COLLECTIONS} SET name = ?, name_key = ?",
+                        (name, key(name)),
+                    ),
+                )
+
+    def test_restore_accepts_a_name_key_from_other_case_tables(self):
+        # Python versions fold some characters differently; a key written
+        # by another version must never make a backup unrestorable.
+        self.require_module()
+        path = self.closed_current(
+            "other-key.db",
+            (
+                f"INSERT INTO {COLLECTIONS} (id, name, name_key, created_at) "
+                "VALUES (1, 'Good', 'Good', ?)",
+                (NOW,),
+            ),
+        )
+        self.assertEqual(
+            PromptModel.verify_database_file(path, require_library_state=True),
+            (True, "ok"),
+        )
+
+    def test_restore_refuses_an_empty_or_non_text_name_key(self):
+        self.assert_refused("empty-key", f"UPDATE {COLLECTIONS} SET name_key = ''")
+        self.assert_refused(
+            "blob-name-key", f"UPDATE {COLLECTIONS} SET name_key = x'61'"
+        )
+        self.assert_refused(
+            "long-name-key",
+            (f"UPDATE {COLLECTIONS} SET name_key = ?", ("k" * 1201,)),
+        )
+
+    def test_restore_refuses_a_c1_control_character(self):
+        for name in ("a\x81b", "a\x7fb"):
+            with self.subTest(name=repr(name)):
+                self.assert_refused(
+                    f"c1-{ord(name[1])}",
+                    (
+                        f"UPDATE {COLLECTIONS} SET name = ?, name_key = ?",
+                        (name, key(name)),
+                    ),
+                )
+
+    def test_restore_refuses_a_non_text_key(self):
+        # Column affinity turns a number into text; a blob stays a blob.
+        self.assert_refused("blob-key", f"UPDATE {ITEMS} SET relative_path = x'61'")
+
+    def test_restore_refuses_a_key_over_4096_characters(self):
+        self.assert_refused(
+            "long-key", (f"UPDATE {ITEMS} SET root_key = ?", ("é" * 4097,))
+        )
+
+    def test_restore_refuses_a_timestamp_over_64_bytes(self):
+        self.assert_refused(
+            "stamp", (f"UPDATE {COLLECTIONS} SET created_at = ?", ("t" * 65,))
+        )
+        self.assert_refused(
+            "item-stamp", (f"UPDATE {ITEMS} SET added_at = ?", ("t" * 65,))
+        )
+
+    def test_restore_refuses_an_item_of_a_missing_collection(self):
+        self.assert_refused("orphan", f"INSERT INTO {ITEMS} VALUES (9, 'r', 'b', 'n')")
+
+    def test_restore_refuses_a_non_integer_collection_id(self):
+        self.assert_refused("text-id", f"UPDATE {ITEMS} SET collection_id = 'one'")
+
+    def test_restore_refuses_a_trigger_or_extra_index(self):
+        self.require_module()
+        for name, statement in (
+            (
+                "trigger",
+                f"CREATE TRIGGER t AFTER INSERT ON {ITEMS} BEGIN SELECT 1; END",
+            ),
+            ("extra", f"CREATE INDEX extra ON {COLLECTIONS}(name)"),
+        ):
+            with self.subTest(name=name):
+                path = self.closed_current(f"{name}.db", statement)
+                with closing(sqlite3.connect(path)) as raw:
+                    self.assertIsNotNone(image_collections_schema_error(raw))
+                self.assertFalse(PromptModel.verify_database_file(path)[0])
+
+    def test_deleting_a_collection_deletes_its_members(self):
+        self.require_module()
+        model = self.model()
+        with model.get_connection() as conn:
+            conn.execute(
+                f"INSERT INTO {COLLECTIONS} (id, name, name_key, created_at) "
+                "VALUES (1, 'A', 'a', ?), (2, 'B', 'b', ?)",
+                (NOW, NOW),
+            )
+            conn.executemany(
+                f"INSERT INTO {ITEMS} VALUES (?, 'root', ?, ?)",
+                ((1, "a.png", NOW), (1, "b.png", NOW), (2, "a.png", NOW)),
+            )
+            conn.execute(f"DELETE FROM {COLLECTIONS} WHERE id = 1")
+            self.assertEqual(
+                [tuple(r) for r in conn.execute(f"SELECT collection_id FROM {ITEMS}")],
+                [(2,)],
+            )
+
+
+if __name__ == "__main__":
+    unittest.main()

tests/test_image_tag_jobs.py

@@ -299,6 +299,19 @@ class TestAutoTag(ImageTagJobTestCase):
         self.tagger.download_model.assert_not_called()
         self.assertIsNone(model_gate.holder())
 
+    async def test_a_missing_backend_package_fails_the_job_before_the_first_file(self):
+        self.png("a.png")
+        self.tagger.missing["wd14-vit"] = ["onnxruntime"]
+
+        with self.assertNoLogs(self.api.logger, level="ERROR"):
+            job = await self.run_job(self.auto_body(["a.png"]))
+
+        self.assertEqual(job.state, "failed")
+        self.assertEqual(job.processed, 0)
+        self.assertTrue(job.status_line.startswith("WD14 ViT needs onnxruntime."))
+        self.assertIn("pip install onnxruntime", job.status_line)
+        self.assertEqual(self.tagger.load_calls, [])
+
     async def test_model_unavailable_fails_the_job_before_the_first_file(self):
         self.png("a.png")
         self.tagger.downloaded["wd14-vit"] = False
@@ -884,6 +897,16 @@ class TestFilterStart(ImageTagJobTestCase):
         self.assertEqual((status, data["code"]), (400, "no-images"))
         self.assertEqual(self.api._jobs.list(), [])
 
+    async def test_filter_with_a_deleted_collection_is_a_404_without_a_job(self):
+        self.png("a.png")
+
+        status, data = await self.start(
+            {"mode": "add", "filter": {"collection": "77"}, "tags": ["x"]}
+        )
+
+        self.assertEqual((status, data["code"]), (404, "collection-not-found"))
+        self.assertEqual(self.api._jobs.list(), [])
+
     async def test_filter_over_the_cap_is_refused_without_a_job(self):
         from py.api import image_tag_jobs
 

tests/test_inspect_image_categories.py

@@ -0,0 +1,277 @@
+"""The inspector and History show an image's own category (#332 §5, DC4, DC5)."""
+
+import os
+import sqlite3
+import unittest
+from unittest.mock import patch
+
+from tests import test_api_history_outputs as history
+from tests.test_api_images import make_png
+from py.api.images import _favorite_root_key, image_category_key
+from utils.inspection_response import (
+    OWN_CATEGORY_UNREAD,
+    _library_projection,
+    apply_own_category,
+)
+
+
+class InspectCase(history.ImageAPITestCase):
+    async def tearDownAsync(self):
+        self.api.db.close_all()
+        await super().tearDownAsync()
+
+    def own(self, path, value, sources=None):
+        sources = sources or self.api._get_all_output_dirs()
+        self.api.db.set_image_categories([image_category_key(sources, path)], value)
+
+    async def library(self, path, root="0"):
+        response = await self.client.get(
+            f"/prompt_manager/images/inspect/{path}", params={"root": root}
+        )
+        body = await response.json()
+        self.assertEqual(response.status, 200, body)
+        return body["library"]
+
+
+class TestInspectorOwnCategory(InspectCase):
+    async def test_inspect_reports_image_scope(self):
+        path = make_png(self.output_dir / "image.png")
+        self.own(path, "Portraits")
+        library = await self.library("image.png")
+        self.assertEqual(library["status"], "none")
+        self.assertEqual(
+            (library["category"], library["category_scope"]), ("Portraits", "image")
+        )
+
+    async def test_own_overrides_linked_recipe_and_ambiguous(self):
+        db = self.api.db
+        linked = make_png(self.output_dir / "linked.png")
+        self._link_image(linked, "castle at dusk")
+        prompt = db.get_prompt_id_for_image(str(linked))
+        with db.model.get_connection() as conn:
+            conn.execute(
+                "UPDATE prompts SET category='Landscapes' WHERE id=?", (prompt,)
+            )
+        library = await self.library("linked.png")
+        self.assertEqual(
+            (library["category"], library["category_scope"]),
+            ("Landscapes", "positive"),
+        )
+        self.own(linked, "Portraits")
+        library = await self.library("linked.png")
+        self.assertEqual(library["status"], "linked")
+        self.assertEqual(
+            (library["category"], library["category_scope"]), ("Portraits", "image")
+        )
+        recipe = make_png(self.output_dir / "recipe.png")
+        self._save_prompt("Pair positive", category="Night")
+        pair = db.save_prompt_pair("Pair positive", "Noise", category="Recipes")
+        positive = db.get_prompt_pair(pair)["positive_prompt_id"]
+        image = db.link_image_to_prompt(positive, str(recipe))
+        db.link_image_to_pair(pair, image, {"source": "fixture"})
+        library = await self.library("recipe.png")
+        self.assertEqual(
+            (library["category"], library["category_scope"]), ("Recipes", "pair")
+        )
+        self.own(recipe, "Mine")
+        library = await self.library("recipe.png")
+        self.assertEqual(
+            (library["category"], library["category_scope"]), ("Mine", "image")
+        )
+        ambiguous = make_png(self.output_dir / "ambiguous.png")
+        self._link_image(ambiguous, "owner one")
+        self._link_image(ambiguous, "owner two")
+        self.own(ambiguous, "Scans")
+        library = await self.library("ambiguous.png")
+        self.assertEqual(library["status"], "ambiguous")
+        self.assertEqual(
+            (library["category"], library["category_scope"]), ("Scans", "image")
+        )
+
+    async def test_a_failed_read_hides_the_category(self):
+        linked = make_png(self.output_dir / "linked.png")
+        self._link_image(linked, "castle at dusk")
+
+        def broken(*_args, **_kwargs):
+            raise sqlite3.OperationalError("secret")
+
+        with patch.object(self.api.db, "image_category", side_effect=broken):
+            with self.assertLogs("prompt_manager", level="WARNING"):
+                library = await self.library("linked.png")
+        self.assertEqual(library["status"], "linked")
+        self.assertIsNone(library["category"])
+        self.assertIsNone(library["category_scope"])
+
+    async def test_a_lora_root_reads_no_own_category(self):
+        lora = self.comfy_dir / "loras"
+        make_png(lora / "preview.png")
+        roots = [self.output_dir, lora]
+        reads = []
+        real = self.api.db.image_category
+
+        def recorded(*args):
+            reads.append(args)
+            return real(*args)
+
+        with (
+            patch.object(self.api, "_media_roots", lambda outputs: roots),
+            patch.object(self.api, "_linked_only_error", self._no_error),
+            patch.object(self.api.db, "image_category", side_effect=recorded),
+        ):
+            response = await self.client.get(
+                "/prompt_manager/images/inspect/preview.png", params={"root": "1"}
+            )
+            body = await response.json()
+        self.assertEqual(response.status, 200, body)
+        self.assertEqual(reads, [])
+        self.assertIsNone(body["library"]["category"])
+
+    @staticmethod
+    async def _no_error(*_args):
+        return None
+
+
+class TestProjection(unittest.TestCase):
+    def test_scope_image_is_kept_for_category_only(self):
+        base = {"status": "none", "scope": "saved-library", "tags": []}
+        projected = _library_projection(
+            {**base, "category": "Scans", "category_scope": "image"}
+        )
+        self.assertEqual(projected["category_scope"], "image")
+        projected = _library_projection({**base, "tags_scope": "image"})
+        self.assertNotIn("tags_scope", projected)
+
+    def test_apply_own_category_three_cases(self):
+        library = {"status": "linked", "category": "L", "category_scope": "positive"}
+        self.assertEqual(
+            apply_own_category(library, "Mine"),
+            {"status": "linked", "category": "Mine", "category_scope": "image"},
+        )
+        self.assertEqual(apply_own_category(library, None), library)
+        self.assertEqual(
+            apply_own_category(library, OWN_CATEGORY_UNREAD),
+            {"status": "linked", "category": None, "category_scope": None},
+        )
+        self.assertEqual(library["category"], "L")  # never mutated
+
+
+class TestHistoryOwnCategory(history.ImageAPITestCase):
+    """Only the inspection request reads the own category, keyed as Images."""
+
+    # The History fixture's helpers, without inheriting its tests.
+    _base = history.HistoryOutputAPITests
+    ref = _base.ref
+    capture = _base.capture
+    outputs = _base.outputs
+    current = _base.current
+    media = _base.media
+
+    async def get_application(self):
+        app = await super().get_application()
+        self.temp_dir = self.comfy_dir / "temp"
+        self.temp_dir.mkdir()
+        self.host_output = str(self.output_dir)
+        self.host_temp = str(self.temp_dir)
+        self.patches = history.ExitStack()
+        self.addCleanup(self.patches.close)
+        for name, value in (
+            ("folder_paths.get_output_directory", lambda: self.host_output),
+            ("folder_paths.get_temp_directory", lambda: self.host_temp),
+        ):
+            self.patches.enter_context(patch(name, value, create=True))
+        self.patches.enter_context(
+            patch.object(
+                history.GalleryConfig,
+                "allowed_gallery_parents",
+                lambda: [str(self.comfy_dir)],
+            )
+        )
+        return app
+
+    async def tearDownAsync(self):
+        self.api.close_history_reads()
+        self.patches.close()
+        await super().tearDownAsync()
+
+    async def inspect(self, start):
+        descriptor = await self.media(start)
+        response = await self.current(descriptor["inspection_url"])
+        body = await response.json()
+        self.assertEqual(response.status, 200, body)
+        return body["library"]
+
+    async def test_history_inspection_shows_own_category(self):
+        path = make_png(self.output_dir / "image.png")
+        self.api.db.set_image_categories(
+            [image_category_key([self.output_dir], path)], "Portraits"
+        )
+        start = self.capture([self.ref()])
+        library = await self.inspect(start)
+        self.assertEqual(
+            (library["category"], library["category_scope"]), ("Portraits", "image")
+        )
+
+    async def test_nested_sources_give_the_images_key(self):
+        inner = self.output_dir / "nested"
+        path = make_png(inner / "image.png")
+        self.api._get_all_output_dirs = lambda: [self.output_dir, inner]
+        self.api.db.set_image_categories(
+            [(_favorite_root_key(inner), "image.png")], "S"
+        )
+        start = self.capture([self.ref(subfolder="nested")])
+        library = await self.inspect(start)
+        self.assertEqual(library["category"], "S")
+        self.assertEqual(
+            image_category_key([self.output_dir, inner], path),
+            (_favorite_root_key(inner), "image.png"),
+        )
+
+    async def test_a_symlinked_output_folder_still_shows_it(self):
+        real = self.comfy_dir / "real-output"
+        real.mkdir()
+        link = self.comfy_dir / "linked-output"
+        self._make_symlink(real, link, target_is_directory=True)
+        make_png(real / "image.png")
+        self.host_output = str(link)
+        self.api._get_all_output_dirs = lambda: [real]
+        self.api.db.set_image_categories([(_favorite_root_key(real), "image.png")], "S")
+        start = self.capture([self.ref()])
+        library = await self.inspect(start)
+        self.assertEqual(library["category"], "S")
+
+    async def test_outputs_and_media_read_no_own_category(self):
+        make_png(self.output_dir / "image.png")
+        start = self.capture([self.ref()])
+        reads = []
+        real = self.api.db.image_category
+
+        def recorded(*args):
+            reads.append(args)
+            return real(*args)
+
+        with patch.object(self.api.db, "image_category", side_effect=recorded):
+            descriptor = await self.media(start)
+            for field in ("url", "thumbnail_url"):
+                response = await self.current(descriptor[field])
+                await response.read()
+                self.assertEqual(response.status, 200)
+            self.assertEqual(reads, [])
+            await self.current(descriptor["inspection_url"])
+            self.assertEqual(len(reads), 1)
+
+    async def test_a_temp_output_has_none(self):
+        reference = {"folder_type": "temp", "subfolder": "", "filename": "a.png"}
+        self.assertIsNone(self.api._history_own_category(reference, None))
+
+    async def test_a_failed_read_hides_it(self):
+        make_png(self.output_dir / "image.png")
+        start = self.capture([self.ref()])
+
+        def broken(*_args):
+            raise sqlite3.OperationalError("secret")
+
+        with patch.object(self.api.db, "image_category", side_effect=broken):
+            with self.assertLogs("prompt_manager", level="WARNING"):
+                library = await self.inspect(start)
+        self.assertIsNone(library["category"])
+        self.assertTrue(os.path.exists(self.output_dir / "image.png"))

tests/test_inspection_response.py

@@ -1426,3 +1426,32 @@ class TestPerRecordBudgets(unittest.TestCase):
         self.assertEqual(metadata["record_formats"], {"workflow": "limited"})
         self.assertEqual(metadata["workflow_graph"]["status"], "limited")
         self.assertEqual(metadata["prompt_graph"]["status"], "absent")
+
+
+class TestInheritedCategory(unittest.TestCase):
+    """#332 MC4: one inherited-category rule for the chip, filter and History."""
+
+    def test_recipe_set_blank_and_absent(self):
+        from utils.inspection_response import inherited_category, library_metadata
+
+        cases = [
+            (("Night", "Recipes", True), ("Recipes", "pair")),
+            (("Night", "   ", True), ("Night", "positive")),
+            (("Night", " \t", True), ("Night", "positive")),
+            (("Night", b" ", True), ("Night", "positive")),
+            (("Night", None, True), ("Night", "positive")),
+            (("Night", "Recipes", False), ("Night", "positive")),
+            (("  ", None, False), ("  ", "positive")),
+            ((None, None, False), (None, "positive")),
+        ]
+        for args, expected in cases:
+            with self.subTest(args=args):
+                self.assertEqual(inherited_category(*args), expected)
+                positive, recipe, has_recipe = args
+                metadata = library_metadata(
+                    positive, ["t"], (recipe, []) if has_recipe else None
+                )
+                self.assertEqual(
+                    (metadata["category"], metadata["category_scope"]), expected
+                )
+                self.assertEqual(metadata["tags_scope"], "positive")

tests/test_pair_usage_tracking.py

@@ -85,6 +85,23 @@ class TestPairUsageTracking(unittest.TestCase):
             self.db.get_prompt_by_id(pair["negative_prompt_id"])["run_count"], 0
         )
 
+    def test_execution_records_a_cached_run_receipt_for_both_editors(self):
+        from utils.cached_runs import Facts
+
+        node = self.node()
+        node.cached_runs = Mock()
+        node.cached_runs.replayed.return_value = False
+        node.encode_prompt(
+            Mock(),
+            "cat",
+            negative_text="",
+            prompt=pair_graph(negative=""),
+            unique_id="1",
+        )
+        node.cached_runs.record_item.assert_called_once_with(
+            "1", ("cat", ""), Facts(usage_tracking.generate_prompt_hash("cat"), None)
+        )
+
     def test_validation_rejection_never_counts_or_saves(self):
         self.submit(pair_graph())
         self.assertEqual(self.db.list_prompt_pairs()["total"], 0)

tests/test_prompt_usage.py

@@ -278,5 +278,49 @@ class TestNodeRecordsUse(UsageTestCase):
         self.assertEqual(self.db.get_prompt_by_id(pid)["run_count"], 2)
 
 
+class TestNodeRecordsCachedRunReceipt(UsageTestCase):
+    """Each execution leaves a receipt for cached jobs (#347)."""
+
+    GRAPH = {
+        "1": {"class_type": "PromptManagerText", "inputs": {"text": "typed text"}},
+    }
+
+    def _node(self):
+        from unittest import mock
+
+        from prompt_manager_text import PromptManagerText
+
+        node = PromptManagerText.__new__(PromptManagerText)
+        node.db = self.db
+        node.logger = __import__("logging").getLogger("test.prompt_usage")
+        node.prompt_tracker = mock.Mock()
+        node.comfyui_integration = mock.Mock()
+        node._inject_lora_trigger_words = lambda text: text
+        node.cached_runs = mock.Mock()
+        node.cached_runs.replayed.return_value = False
+        return node
+
+    def test_receipt_holds_the_returned_text_and_the_stored_hash(self):
+        from utils.cached_runs import Facts
+
+        node = self._node()
+        node.process_text(
+            "typed text", prepend_text="pre", prompt=self.GRAPH, unique_id="1"
+        )
+        node.cached_runs.record_item.assert_called_once_with(
+            "1", ("pre typed text",), Facts(generate_prompt_hash("typed text"))
+        )
+        # Hook-visible text: the node counts nothing itself
+        row = self.db.get_prompt_by_hash(generate_prompt_hash("typed text"))
+        self.assertEqual(row["run_count"], 0)
+
+    def test_empty_editor_records_empty_facts(self):
+        from utils.cached_runs import Facts
+
+        node = self._node()
+        node.process_text("  ", prompt=self.GRAPH, unique_id="1")
+        node.cached_runs.record_item.assert_called_once_with("1", ("  ",), Facts(None))
+
+
 if __name__ == "__main__":
     unittest.main()

tests/test_restore_plans.py

@@ -49,6 +49,7 @@ SUMMARY_KEYS = {
     "image_links",
     "runs",
     "saved_views",
+    "collections",
 }
 
 

tests/test_restore_stage.py

@@ -226,6 +226,7 @@ class TestStageAndPublish(StageTestCase):
             for key in ("pairs", "archived_pairs", "image_links", "runs"):
                 self.assertEqual(staged.summary[key], 0, key)
             self.assertEqual(staged.summary["saved_views"], 0)
+            self.assertEqual(staged.summary["collections"], 0)
             self.assertEqual(staged.schema, {"migrated": False, "library_state": True})
             self.assertEqual(
                 staged.digest,
@@ -239,6 +240,48 @@ class TestStageAndPublish(StageTestCase):
             staged.discard()
 
 
+class TestCollectionsSummary(StageTestCase):
+    """#331: collections join the review's summary and the live fingerprint."""
+
+    def add_collection(self, db, name):
+        with db.model.get_connection() as conn:
+            conn.execute(
+                "INSERT INTO image_collections (name, name_key, created_at) "
+                "VALUES (?, ?, '2026-10-27T00:00:00+00:00')",
+                (name, name.casefold()),
+            )
+
+    def test_summary_counts_collections(self):
+        src_path = os.path.join(self.uploads, "with-collections.db")
+        src = PromptDatabase(src_path)
+        try:
+            self.save(src, "kept")
+            self.add_collection(src, "First")
+            self.add_collection(src, "Second")
+        finally:
+            src.close_all()
+        staged = self.db.model.stage_restore(src_path)
+        try:
+            self.assertEqual(staged.summary["collections"], 2)
+        finally:
+            staged.discard()
+        self.assertIn("collections", restore_stage.SUMMARY_KEYS)
+
+    def test_adding_a_collection_changes_the_fingerprint(self):
+        conn = self.db.model.get_connection()
+        summary, before = restore_stage.live_fingerprint(conn)
+        self.assertEqual(summary["collections"], 0)
+        self.add_collection(self.db, "After the review")
+        summary, after = restore_stage.live_fingerprint(conn)
+        self.assertEqual(summary["collections"], 1)
+        self.assertNotEqual(after, before)
+        # A delete and a re-add with the same count still changes it (rowid).
+        with self.db.model.get_connection() as write:
+            write.execute("DELETE FROM image_collections")
+        self.add_collection(self.db, "Replacement")
+        self.assertNotEqual(restore_stage.live_fingerprint(conn)[1], after)
+
+
 class TestOlderBackups(StageTestCase):
     """Rollback-journal and older-schema backups restore exactly as before."""
 

tests/test_save_to_library.py

@@ -0,0 +1,508 @@
+"""#90 item 3: a node's ``save_to_library`` switch.
+
+Off: PromptManager ignores the run completely (no prompt saved, no run count,
+no image link), while the node still returns its normal outputs. On (the
+default, and every workflow or image made before the switch existed): as before.
+"""
+
+import logging
+import os
+import sys
+import unittest
+from pathlib import Path
+from unittest import mock
+
+sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
+
+try:
+    from tests.test_prompt_attribution import MonitorCase, NodeCase
+    from tests.test_execution_contexts import load_module
+except ImportError:  # discovered with tests/ as the top-level directory
+    from test_prompt_attribution import MonitorCase, NodeCase
+    from test_execution_contexts import load_module
+
+from prompt_manager import PromptManager, PromptManagerFocused, PromptManagerSplit
+from prompt_manager_pair import PromptManagerPair
+from prompt_manager_text import PromptManagerText
+from prompt_search_list import PromptSearchList
+from utils.hashing import generate_prompt_hash
+from utils.prompt_graph import (
+    intrinsic_pair_nodes,
+    positive_opt_outs,
+    run_prompt_components,
+    run_prompt_nodes,
+    save_setting,
+)
+from utils.run_keys import prompt_node_key
+from utils.usage_tracking import (
+    PendingFirstUse,
+    QueuedRuns,
+    handle_queued_prompt,
+    on_dequeue,
+    queued_pair_runs,
+    queued_prompt_runs,
+)
+
+POSITIVE = "a lighthouse at dusk"
+NEGATIVE = "blurry"
+NODE_CLASSES = (
+    PromptManager,
+    PromptManagerText,
+    PromptManagerFocused,
+    PromptManagerSplit,
+    PromptManagerPair,
+)
+
+
+def run_graph(save=None, kind="PromptManager", negative_save=None):
+    """A PromptManager positive feeding a KSampler; ``save`` None leaves it out."""
+    inputs = {"text": POSITIVE, "clip": ["4", 1]}
+    if save is not None:
+        inputs["save_to_library"] = save
+    negative_inputs = {"text": NEGATIVE, "clip": ["4", 1]}
+    if negative_save is not None:
+        negative_inputs["save_to_library"] = negative_save
+    return {
+        "1": {"class_type": kind, "inputs": inputs},
+        "2": {"class_type": "PromptManager", "inputs": negative_inputs},
+        "4": {"class_type": "CheckpointLoaderSimple", "inputs": {"ckpt_name": "m"}},
+        "3": {
+            "class_type": "KSampler",
+            "inputs": {
+                "model": ["4", 0],
+                "positive": ["1", 0],
+                "negative": ["2", 0],
+                "latent_image": ["6", 0],
+            },
+        },
+        "6": {"class_type": "EmptyLatentImage", "inputs": {"batch_size": 1}},
+    }
+
+
+def pair_graph(save=None):
+    inputs = {"text": POSITIVE, "negative_text": NEGATIVE, "clip": ["4", 1]}
+    if save is not None:
+        inputs["save_to_library"] = save
+    return {
+        "1": {"class_type": "PromptManagerPair", "inputs": inputs},
+        "4": {"class_type": "CheckpointLoaderSimple", "inputs": {"ckpt_name": "m"}},
+        "3": {
+            "class_type": "KSampler",
+            "inputs": {
+                "model": ["4", 0],
+                "positive": ["1", 0],
+                "negative": ["1", 1],
+                "latent_image": ["6", 0],
+            },
+        },
+        "6": {"class_type": "EmptyLatentImage", "inputs": {"batch_size": 1}},
+    }
+
+
+class TestInputDeclaration(unittest.TestCase):
+    def test_every_prompt_manager_node_declares_a_boolean_defaulting_on(self):
+        for cls in NODE_CLASSES:
+            optional = cls.INPUT_TYPES()["optional"]
+            kind, options = optional["save_to_library"]
+            self.assertEqual(kind, "BOOLEAN", cls.__name__)
+            self.assertIs(options["default"], True, cls.__name__)
+            # Declared last: no saved widget value of an older workflow shifts
+            self.assertEqual(list(optional)[-1], "save_to_library", cls.__name__)
+
+    def test_search_list_has_no_switch(self):
+        inputs = PromptSearchList.INPUT_TYPES()
+        self.assertNotIn("save_to_library", inputs.get("optional", {}))
+        self.assertNotIn("save_to_library", inputs.get("required", {}))
+
+
+class TestSaveSetting(unittest.TestCase):
+    def test_missing_input_is_on(self):
+        self.assertIs(save_setting(run_graph(), "1"), True)
+
+    def test_literals_follow_comfyui_boolean_coercion(self):
+        for value, expected in (
+            (True, True),
+            (False, False),
+            ("", False),
+            (0, False),
+            ({"__value__": False}, False),
+            ({"__value__": True}, True),
+        ):
+            with self.subTest(value=value):
+                self.assertIs(save_setting(run_graph(save=value), "1"), expected)
+
+    def test_link_to_primitive_boolean_is_read(self):
+        for value in (True, False):
+            graph = run_graph(save=["9", 0])
+            graph["9"] = {"class_type": "PrimitiveBoolean", "inputs": {"value": value}}
+            self.assertIs(save_setting(graph, "1"), value)
+
+    def test_link_to_an_unknown_node_is_not_knowable(self):
+        graph = run_graph(save=["9", 0])
+        graph["9"] = {"class_type": "SomeCustomSwitch", "inputs": {}}
+        self.assertIsNone(save_setting(graph, "1"))
+        self.assertIsNone(save_setting(run_graph(save=["missing", 0]), "1"))
+
+
+class TestGraphRoles(unittest.TestCase):
+    def test_opted_out_positive_is_not_a_run_prompt(self):
+        graph = run_graph(save=False)
+        self.assertEqual(run_prompt_nodes(graph), [])
+        self.assertEqual(run_prompt_components(graph), [])
+        self.assertEqual(positive_opt_outs(graph), ["1"])
+
+    def test_enabled_positive_is_unchanged(self):
+        for save in (None, True):
+            graph = run_graph(save=save)
+            self.assertEqual(run_prompt_nodes(graph), ["1"])
+            self.assertEqual(positive_opt_outs(graph), [])
+
+    def test_opted_out_negative_is_no_positive_opt_out(self):
+        graph = run_graph(negative_save=False)
+        self.assertEqual(run_prompt_nodes(graph), ["1"])
+        self.assertEqual(positive_opt_outs(graph), [])
+
+    def test_walk_stops_at_an_opted_out_node(self):
+        graph = run_graph(save=False)
+        graph["1"]["inputs"]["text"] = ["7", 0]
+        graph["7"] = {"class_type": "PromptManagerText", "inputs": {"text": "x"}}
+        self.assertEqual(run_prompt_nodes(graph), [])
+
+    def test_sampler_less_fallback_skips_opted_out_nodes(self):
+        graph = {
+            "1": {
+                "class_type": "PromptManagerText",
+                "inputs": {"text": "a", "save_to_library": False},
+            },
+            "2": {"class_type": "PromptManagerText", "inputs": {"text": "b"}},
+        }
+        self.assertEqual(run_prompt_nodes(graph), ["2"])
+        self.assertEqual(run_prompt_components(graph), [("2", "text")])
+        self.assertEqual(positive_opt_outs(graph), ["1"])
+
+    def test_opted_out_pair_is_not_an_intrinsic_pair(self):
+        self.assertEqual(intrinsic_pair_nodes(pair_graph()), ["1"])
+        self.assertEqual(intrinsic_pair_nodes(pair_graph(save=False)), [])
+
+
+class TestQueueHook(unittest.TestCase):
+    def test_opted_out_node_is_never_stashed(self):
+        self.assertEqual(queued_prompt_runs(run_graph(save=False)), [])
+        self.assertEqual(queued_pair_runs(pair_graph(save=False)), [])
+
+    def test_unknown_setting_is_left_to_the_node(self):
+        graph = run_graph(save=["9", 0])
+        graph["9"] = {"class_type": "SomeCustomSwitch", "inputs": {}}
+        self.assertEqual(queued_prompt_runs(graph), [])
+        pair = pair_graph(save=["9", 0])
+        pair["9"] = {"class_type": "SomeCustomSwitch", "inputs": {}}
+        self.assertEqual(queued_pair_runs(pair), [])
+
+    def test_primitive_boolean_on_still_counts(self):
+        graph = run_graph(save=["9", 0])
+        graph["9"] = {"class_type": "PrimitiveBoolean", "inputs": {"value": True}}
+        self.assertEqual(
+            queued_prompt_runs(graph), [(generate_prompt_hash(POSITIVE), POSITIVE)]
+        )
+
+    def test_dequeue_counts_nothing_for_a_cached_opted_out_node(self):
+        """A cached node never executes: only the queued graph can say it is off."""
+        db = mock.Mock()
+        db.get_prompt_by_hash.return_value = {"id": 5}
+        queued, pending = QueuedRuns(), PendingFirstUse()
+        request = {"prompt": run_graph(save=False), "prompt_id": "job-1"}
+        handle_queued_prompt(request, queued, pair_queued=QueuedRuns())
+        on_dequeue("job-1", queued, lambda: db, pending, pair_queued=QueuedRuns())
+        db.record_prompt_use.assert_not_called()
+        self.assertEqual(pending.consume(generate_prompt_hash(POSITIVE)), 0)
+
+
+class TestNodesSaveNothing(NodeCase):
+    def run_node(self, cls, graph, save):
+        node = self.node(cls)
+        node.record_pair = mock.Mock()
+        kwargs = {"prompt": graph, "unique_id": "1", "save_to_library": save}
+        if cls is PromptManagerText:
+            return node, node.process_text(POSITIVE, **kwargs)
+        if cls is PromptManagerPair:
+            return node, node.encode_prompt(
+                self.clip(), POSITIVE, negative_text=NEGATIVE, **kwargs
+            )
+        return node, node.encode_prompt(self.clip(), POSITIVE, **kwargs)
+
+    def test_off_saves_counts_and_tracks_nothing_but_returns_outputs(self):
+        for cls in NODE_CLASSES:
+            with self.subTest(node=cls.__name__):
+                graph = (
+                    pair_graph(save=False)
+                    if cls is PromptManagerPair
+                    else run_graph(save=False, kind=cls.__name__)
+                )
+                node, result = self.run_node(cls, graph, False)
+                self.assertEqual(self.count_rows("prompts"), 0)
+                self.assertEqual(self.count_rows("prompt_pairs"), 0)
+                self.assertEqual(self.tracker.keyed_snapshot(), {})
+                self.assertEqual(self.tracker._prompt_queue, [])
+                node.comfyui_integration.register_prompt.assert_not_called()
+                if cls is PromptManagerText:
+                    self.assertEqual(result, (POSITIVE,))
+                elif cls is PromptManagerPair:
+                    self.assertEqual(result[0], ("cond", POSITIVE))
+                    self.assertEqual(result[1], ("cond", NEGATIVE))
+                    self.assertEqual(result[2:], (POSITIVE, NEGATIVE))
+                else:
+                    self.assertEqual(result, (("cond", POSITIVE), POSITIVE))
+
+    def test_off_without_graph_context_still_saves_nothing(self):
+        node = self.node(PromptManager)
+        node.encode_prompt(self.clip(), POSITIVE, save_to_library=False)
+        self.assertEqual(self.count_rows("prompts"), 0)
+        self.assertEqual(self.tracker._prompt_queue, [])
+
+    def test_off_leaves_an_existing_prompt_uncounted(self):
+        prompt_id = self.save(POSITIVE)
+        self.run_node(PromptManager, run_graph(save=False), False)
+        row = self.db.get_prompt_by_id(prompt_id)
+        self.assertFalse(row.get("run_count") or 0)
+
+    def test_on_is_unchanged(self):
+        self.run_node(PromptManager, run_graph(save=True), True)
+        self.assertIsNotNone(self.db.get_prompt_by_hash(generate_prompt_hash(POSITIVE)))
+        self.assertIn(
+            prompt_node_key(run_graph(save=True), "1"), self.tracker.keyed_snapshot()
+        )
+
+    def test_unknown_setting_on_at_run_time_counts_at_execution(self):
+        """The hook could not read a wired switch, so the node counts its run."""
+        graph = run_graph(save=["9", 0])
+        graph["9"] = {"class_type": "SomeCustomSwitch", "inputs": {}}
+        prompt_id = self.save(POSITIVE)
+        self.run_node(PromptManager, graph, True)
+        self.assertEqual(self.db.get_prompt_by_id(prompt_id)["run_count"], 1)
+
+
+class TestWiredSwitchOffMarksItsRun(NodeCase):
+    """Review finding: a wired switch off at run time keeps the node a run prompt."""
+
+    def wired_graph(self):
+        graph = run_graph(save=["9", 0])
+        graph["9"] = {"class_type": "SomeCustomSwitch", "inputs": {}}
+        return graph
+
+    def test_an_earlier_entry_of_the_key_never_stands_in(self):
+        graph = self.wired_graph()
+        key = prompt_node_key(graph, "1")
+        node = self.node(PromptManager)
+        node.encode_prompt(
+            self.clip(), POSITIVE, prompt=graph, unique_id="1", save_to_library=True
+        )
+        node.encode_prompt(
+            self.clip(), POSITIVE, prompt=graph, unique_id="1", save_to_library=False
+        )
+        latest, _run, _count = self.tracker.keyed_snapshot()[key]
+        self.assertTrue(latest.get("placeholder"))
+
+    def test_literal_off_marks_nothing(self):
+        node = self.node(PromptManager)
+        node.encode_prompt(
+            self.clip(),
+            POSITIVE,
+            prompt=run_graph(save=False),
+            unique_id="1",
+            save_to_library=False,
+        )
+        self.assertEqual(self.tracker.keyed_snapshot(), {})
+
+
+class TestMonitorLeavesOptedOutFilesUnlinked(MonitorCase):
+    def test_existing_prompt_is_not_linked_from_metadata(self):
+        self.save(POSITIVE)
+        self.assertIsNone(self.process({"prompt": run_graph(save=False)}))
+
+    def test_find_prompt_reports_opted_out(self):
+        self.save(POSITIVE)
+        state, prompt = self.handler._find_prompt_from_metadata(
+            {"prompt": run_graph(save=False)}
+        )
+        self.assertEqual((state, prompt), ("opted_out", None))
+
+    def test_mixed_positive_settings_leave_the_file_unlinked(self):
+        self.save(POSITIVE)
+        graph = run_graph(save=True)
+        graph["8"] = {
+            "class_type": "PromptManager",
+            "inputs": {"text": "other", "save_to_library": False},
+        }
+        graph["5"] = {
+            "class_type": "KSampler",
+            "inputs": {
+                "model": ["4", 0],
+                "positive": ["8", 0],
+                "negative": ["2", 0],
+                "latent_image": ["6", 0],
+            },
+        }
+        self.assertIsNone(self.process({"prompt": graph}))
+
+    def test_workflow_only_metadata_with_switch_off_is_opted_out(self):
+        self.save(POSITIVE)
+        metadata = {
+            "text_encoder_nodes": [
+                {
+                    "type": "PromptManager",
+                    "inputs": [],
+                    "widgets_values": [POSITIVE, "", "", "", "", "", False],
+                }
+            ]
+        }
+        state, _prompt = self.handler._find_prompt_from_metadata(metadata)
+        self.assertEqual(state, "opted_out")
+
+    def test_workflow_only_ignores_muted_nodes_and_reads_subgraphs(self):
+        from utils.prompt_graph import metadata_opted_out
+
+        node = {
+            "type": "PromptManager",
+            "widgets_values": [POSITIVE, "", "", "", "", "", False],
+        }
+        self.assertFalse(
+            metadata_opted_out({"workflow": {"nodes": [{**node, "mode": 4}]}})
+        )
+        self.assertTrue(
+            metadata_opted_out(
+                {
+                    "workflow": {
+                        "nodes": [],
+                        "definitions": {"subgraphs": [{"nodes": [node]}]},
+                    }
+                }
+            )
+        )
+
+    def test_unknown_setting_is_left_to_the_run(self):
+        self.save(POSITIVE)
+        graph = run_graph(save=["9", 0])
+        graph["9"] = {"class_type": "SomeCustomSwitch", "inputs": {}}
+        state, _prompt = self.handler._find_prompt_from_metadata({"prompt": graph})
+        self.assertEqual(state, "unknown")
+
+    def test_enabled_positive_still_links(self):
+        prompt_id = self.save(POSITIVE)
+        self.assertEqual(self.process({"prompt": run_graph(save=True)}), prompt_id)
+
+
+class TestLinkKeepsOptedOutNegativeOut(MonitorCase):
+    def setUp(self):
+        super().setUp()
+        # The real link, not the recording stub of MonitorCase
+        del self.handler.link_image_to_prompt
+
+    def test_opted_out_negative_is_never_stored_as_a_pair(self):
+        prompt_id = self.save(POSITIVE)
+        metadata = {"prompt": run_graph(save=True, negative_save=False)}
+        self.handler.link_image_to_prompt(self.image, {"id": prompt_id}, metadata)
+        self.assertIsNone(self.db.get_prompt_by_hash(generate_prompt_hash(NEGATIVE)))
+        self.assertEqual(self.count_pairs(), 0)
+        linked = (
+            self.db.model.get_connection()
+            .execute("SELECT prompt_id FROM generated_images")
+            .fetchall()
+        )
+        self.assertEqual([row[0] for row in linked], [prompt_id])
+
+    def test_opted_out_positive_links_nothing(self):
+        prompt_id = self.save(POSITIVE)
+        metadata = {"prompt": run_graph(save=False)}
+        self.handler.link_image_to_prompt(self.image, {"id": prompt_id}, metadata)
+        count = (
+            self.db.model.get_connection()
+            .execute("SELECT COUNT(*) FROM generated_images")
+            .fetchone()[0]
+        )
+        self.assertEqual(count, 0)
+
+    def count_pairs(self):
+        conn = self.db.model.get_connection()
+        return conn.execute("SELECT COUNT(*) FROM prompt_pairs").fetchone()[0]
+
+
+class TestScanSkipsOptedOutFiles(NodeCase):
+    def setUp(self):
+        super().setUp()
+        from py.api import admin
+
+        self.mixin = admin.AdminRoutesMixin.__new__(admin.AdminRoutesMixin)
+        self.mixin.db = self.db
+        self.mixin.logger = logging.getLogger("test.save_to_library.scan")
+        self.file = Path(self.tmp.name) / "out.png"
+
+    def test_opted_out_file_stores_nothing(self):
+        status = self.mixin._ingest_scanned_file(
+            self.file, {"prompt": run_graph(save=False)}
+        )
+        self.assertEqual(status, "not_saved")
+        self.assertEqual(self.count_rows("prompts"), 0)
+        self.assertEqual(self.count_rows("generated_images"), 0)
+
+    def test_opted_out_file_is_never_queued_for_review(self):
+        ambiguous = []
+        graph = run_graph(save=False)
+        graph["5"] = dict(graph["3"])  # a second sampler: ambiguous otherwise
+        graph["5"]["inputs"] = {**graph["3"]["inputs"], "positive": ["2", 0]}
+        status = self.mixin._ingest_scanned_file(
+            self.file, {"prompt": graph}, ambiguous=ambiguous
+        )
+        self.assertEqual(status, "not_saved")
+        self.assertEqual(ambiguous, [])
+
+    def test_opted_out_negative_stores_the_positive_only(self):
+        status = self.mixin._ingest_scanned_file(
+            self.file, {"prompt": run_graph(save=True, negative_save=False)}
+        )
+        self.assertEqual(status, "added")
+        self.assertIsNotNone(self.db.get_prompt_by_hash(generate_prompt_hash(POSITIVE)))
+        self.assertIsNone(self.db.get_prompt_by_hash(generate_prompt_hash(NEGATIVE)))
+        self.assertEqual(self.count_rows("prompt_pairs"), 0)
+
+    def test_wired_switch_is_left_to_the_live_run(self):
+        graph = run_graph(save=["9", 0])
+        graph["9"] = {"class_type": "SomeCustomSwitch", "inputs": {}}
+        status = self.mixin._ingest_scanned_file(self.file, {"prompt": graph})
+        self.assertEqual(status, "not_saved_wired")
+        self.assertEqual(self.count_rows("prompts"), 0)
+
+    def test_status_has_a_scan_detail(self):
+        from py.api import admin
+
+        group, row, _line = admin._scan_outcome(self.file, [], "not_saved")
+        self.assertEqual(group, "no_metadata")
+        self.assertIn("does not save", row["detail"])
+
+
+class TestHistoryRecordsNoText(unittest.TestCase):
+    def setUp(self):
+        self.m = load_module("execution_context_graph")
+
+    def test_opted_out_role_has_no_sources_and_no_text(self):
+        recipe = self.m.project_contexts(run_graph(save=False), None).recipes[0]
+        self.assertEqual(recipe.positive.sources, ())
+        self.assertIsNone(recipe.positive.inferred)
+        self.assertEqual(recipe.positive.reason, "not_saved")
+        # The enabled negative is recorded as before
+        self.assertEqual(recipe.negative.inferred.raw.value, NEGATIVE)
+
+    def test_unknown_setting_records_no_text_either(self):
+        graph = run_graph(save=["9", 0])
+        graph["9"] = {"class_type": "SomeCustomSwitch", "inputs": {}}
+        recipe = self.m.project_contexts(graph, None).recipes[0]
+        self.assertIsNone(recipe.positive.inferred)
+        self.assertEqual(recipe.positive.reason, "not_saved")
+
+    def test_enabled_role_is_unchanged(self):
+        recipe = self.m.project_contexts(run_graph(save=True), None).recipes[0]
+        self.assertEqual(recipe.positive.inferred.raw.value, POSITIVE)
+        self.assertEqual(len(recipe.positive.sources), 1)
+
+
+if __name__ == "__main__":
+    unittest.main()

tests/test_vocabulary.py

@@ -13,6 +13,11 @@ from database.library_commands import LibraryCommandError
 from database.operations import PromptDatabase
 from utils.hashing import generate_prompt_hash
 
+try:
+    from test_image_categories import category_checksum
+except ImportError:
+    from tests.test_image_categories import category_checksum
+
 
 class TestVocabularyReads(unittest.TestCase):
     def setUp(self):
@@ -168,7 +173,8 @@ class TestVocabularyReads(unittest.TestCase):
         self.assertEqual(page["total"], 4)
         self.assertFalse(page["has_more"])
         for row in page["categories"]:
-            self.assertEqual(set(row), {"value", "prompt_count"})
+            self.assertEqual(set(row), {"value", "prompt_count", "image_count"})
+            self.assertEqual(row["image_count"], 0)
             self.assertEqual(
                 row["prompt_count"],
                 self.db.list_library(category=row["value"])["total"],
@@ -238,6 +244,111 @@ class TestVocabularyReads(unittest.TestCase):
             with self.assertRaises(ValueError):
                 self.read("list_vocabulary_tags", **options)
 
+    def own(self, value, *paths, root="root"):
+        """Give each relative path under ``root`` the own category ``value``."""
+        self.db.set_image_categories([(root, path) for path in paths], value)
+
+    def all_categories(self, limit=200):
+        rows, pages, after = [], [], None
+        while True:
+            page = self.read("list_vocabulary_categories", limit=limit, after=after)
+            pages.append(page)
+            rows.extend(page["categories"])
+            if not page["has_more"]:
+                return rows, pages
+            after = page["next_key"]
+
+    def test_category_list_includes_image_only_values_last(self):
+        self.save("Oldest", category="Zeta")
+        self.save("Newer", category="Alpha")
+        self.own("Zeta", "a.png", "b.png")
+        self.own("Only images B", "c.png")
+        self.own("Only images A", "d.png", "e.png", "gone.png")
+        rows, _pages = self.all_categories()
+        self.assertEqual(
+            rows,
+            [
+                {"value": "Zeta", "prompt_count": 1, "image_count": 2},
+                {"value": "Alpha", "prompt_count": 1, "image_count": 0},
+                {"value": "Only images A", "prompt_count": 0, "image_count": 3},
+                {"value": "Only images B", "prompt_count": 0, "image_count": 1},
+            ],
+        )
+        page = self.read("list_vocabulary_categories")
+        self.assertEqual(page["total"], 4)
+
+    def test_category_cursor_crosses_into_image_only_values(self):
+        for index in range(3):
+            self.save(f"Owner {index}", category=f"P{index}")
+        self.own("P1", "shared.png")
+        self.own("I-b", "x.png")
+        self.own("I-a", "y.png")
+        self.own("I-c", "z.png", root="other")
+        for limit in (1, 2, 3, 4):
+            with self.subTest(limit=limit):
+                rows, pages = self.all_categories(limit=limit)
+                self.assertEqual(
+                    [row["value"] for row in rows],
+                    ["P0", "P1", "P2", "I-a", "I-b", "I-c"],
+                )
+                self.assertEqual({page["total"] for page in pages}, {6})
+                self.assertEqual(
+                    [row["image_count"] for row in rows], [0, 1, 0, 1, 1, 1]
+                )
+        first = self.read("list_vocabulary_categories", limit=3)
+        self.assertEqual(first["next_key"], [3, "P2"])
+        second = self.read("list_vocabulary_categories", limit=1, after=[3, "P2"])
+        self.assertEqual(second["next_key"], [2**63 - 1, "I-a"])
+
+    def test_image_rows_on_a_prompt_value_count_once(self):
+        self.save("Owner", category="Shared")
+        self.db.save_prompt_pair("Owner", "Noise", category="Shared")
+        self.own("Shared", *[f"{n}.png" for n in range(5)])
+        self.own("Shared", "elsewhere.png", root="second")
+        rows, _pages = self.all_categories()
+        self.assertEqual(
+            rows, [{"value": "Shared", "prompt_count": 1, "image_count": 6}]
+        )
+
+    def test_library_options_ignore_image_categories(self):
+        self.save("Owner", category="Prompt value")
+        before = self.db.get_library_options()
+        self.own("Image value", "a.png")
+        self.own("Prompt value", "b.png")
+        self.assertEqual(self.db.get_library_options(), before)
+        self.assertEqual(before["categories"], ["Prompt value"])
+
+    def test_category_reads_use_the_image_category_index(self):
+        from database import vocabulary
+
+        self.own("Value", "a.png")
+        statements = [
+            vocabulary._CATEGORY_ROWS + "SELECT COUNT(*) FROM counted",
+            vocabulary._CATEGORY_ROWS
+            + "SELECT value,first_pid,prompt_count,image_count FROM counted "
+            "WHERE first_pid>? OR (first_pid=? AND value>?) "
+            "ORDER BY first_pid,value LIMIT ?",
+        ]
+        plans = []
+        with self.db.model.get_connection() as conn:
+            for sql in statements:
+                params = [1] * sql.count("?")
+                plans.extend(
+                    row[-1] for row in conn.execute("EXPLAIN QUERY PLAN " + sql, params)
+                )
+            for sql in (
+                vocabulary._IMAGE_CATEGORY_ROWIDS,
+                *vocabulary._CATEGORY_UPDATES,
+            ):
+                params = [1] * sql.count("?")
+                plans.extend(
+                    row[-1] for row in conn.execute("EXPLAIN QUERY PLAN " + sql, params)
+                )
+        image_steps = [step for step in plans if "image_categories" in step]
+        self.assertTrue(image_steps)
+        for step in image_steps:
+            self.assertIn("idx_image_categories_category", step, step)
+
     def test_reads_share_one_snapshot_and_report_epoch(self):
         self.save("Cat", category="Animal", tags=["cat"])
         epoch = self.db.get_execution_epoch()
@@ -741,6 +852,7 @@ class TestVocabularyCategoryCommands(unittest.TestCase):
                 "other_prompts": 1,
                 "negative_variants": 2,
                 "archived_negative_variants": 1,
+                "images": 0,
             },
         )
         self.assertFalse(result["target_exists"])
@@ -810,6 +922,7 @@ class TestVocabularyCategoryCommands(unittest.TestCase):
                 "other_prompts": 0,
                 "negative_variants": 0,
                 "archived_negative_variants": 0,
+                "images": 0,
             },
         )
         self.assertEqual(self.rows("prompts")[variant]["category"], "Other")
@@ -902,6 +1015,171 @@ class TestVocabularyCategoryCommands(unittest.TestCase):
                 self.apply(bad)
         self.assertEqual(self.snapshot(), before)
 
+    def own(self, value, *paths, root="root"):
+        self.db.set_image_categories([(root, path) for path in paths], value)
+
+    def image_rows(self):
+        with self.db.model.get_connection() as conn:
+            return {
+                (row[0], row[1]): (row[2], row[3], row[4])
+                for row in conn.execute(
+                    "SELECT root_key,relative_path,category,updated_at,rowid "
+                    "FROM image_categories ORDER BY rowid"
+                )
+            }
+
+    def checksum(self):
+        with self.db.model.get_connection() as conn:
+            return category_checksum(conn)
+
+    def test_preview_counts_image_rows(self):
+        self.owners()
+        self.own("Scenery", "a.png", "b.png", "missing.png")
+        self.own("Scenery", "c.png", root="second")
+        self.own("Other", "d.png")
+        before = self.image_rows()
+        result = self.preview()
+        self.assertEqual(result["impact"]["images"], 4)
+        self.assertEqual(self.image_rows(), before)
+        self.assertTrue(self.preview(target="Other")["target_exists"])
+
+    def test_apply_renames_image_categories(self):
+        self.owners()
+        bystander = self.save("Bystander", category="Other")
+        self.own("Scenery", "a.png", "b.png")
+        self.own("Scenery", "c.png", root="second")
+        self.own(" Scenery ", "padded.png")
+        self.own("Other", "d.png")
+        self.age_rows()
+        with self.db.model.get_connection() as conn:
+            conn.execute("UPDATE image_categories SET updated_at=?", (_OLD_STAMP,))
+        images_before = self.image_rows()
+        prompts_before, pairs_before = self.snapshot()
+        preview = self.preview()
+        receipt = self.apply(preview["impact_token"])
+        self.assertEqual(receipt["impact"]["images"], 3)
+        images_after = self.image_rows()
+        self.assertEqual(set(images_after), set(images_before))
+        renamed = {("root", "a.png"), ("root", "b.png"), ("second", "c.png")}
+        for key, (value, stamp, rowid) in images_after.items():
+            old_value, old_stamp, old_rowid = images_before[key]
+            self.assertEqual(rowid, old_rowid, key)
+            if key in renamed:
+                self.assertEqual(value, "Landscape")
+                self.assertNotEqual(stamp, _OLD_STAMP)
+            else:
+                self.assertEqual((value, stamp), (old_value, old_stamp))
+        prompts_after, pairs_after = self.snapshot()
+        self.assertEqual(prompts_after[bystander], prompts_before[bystander])
+        for before, after in (
+            (prompts_before, prompts_after),
+            (pairs_before, pairs_after),
+        ):
+            for row_id, row in after.items():
+                if before[row_id]["category"] != "Scenery":
+                    self.assertEqual(row, before[row_id])
+        # Merging into a value images already hold keeps every file key.
+        merge = self.preview(source="Landscape", target="Other")
+        self.assertTrue(merge["target_exists"])
+        self.assertEqual(merge["impact"]["images"], 3)
+        receipt = self.apply(merge["impact_token"], source="Landscape", target="Other")
+        self.assertTrue(receipt["consolidated"])
+        merged = self.image_rows()
+        self.assertEqual(set(merged), set(images_before))
+        self.assertEqual(
+            sorted(key for key, row in merged.items() if row[0] == "Other"),
+            sorted([*renamed, ("root", "d.png")]),
+        )
+
+    def test_image_only_value_can_be_renamed(self):
+        self.save("Bystander", category="Kept")
+        self.db.save_prompt_pair("Pair", "Noise", category="Kept")
+        self.own("Image only", "a.png", "b.png")
+        self.age_rows()
+        before = self.checksum()
+        preview = self.preview(source="Image only", target="Renamed")
+        self.assertEqual(
+            preview["impact"],
+            {
+                "positive_prompts": 0,
+                "other_prompts": 0,
+                "negative_variants": 0,
+                "archived_negative_variants": 0,
+                "images": 2,
+            },
+        )
+        self.assertFalse(preview["target_exists"])
+        receipt = self.apply(
+            preview["impact_token"], source="Image only", target="Renamed"
+        )
+        self.assertFalse(receipt["consolidated"])
+        self.assertEqual({row[0] for row in self.image_rows().values()}, {"Renamed"})
+        self.assertEqual(self.checksum(), before)
+        # An image-only target counts as existing.
+        self.own("Image target", "c.png")
+        self.assertTrue(
+            self.preview(source="Renamed", target="Image target")["target_exists"]
+        )
+        # Nothing holds the value any more: refused.
+        with self.assertRaises(LibraryCommandError) as caught:
+            self.preview(source="Image only", target="Again")
+        self.assertEqual(caught.exception.code, "vocabulary-conflict")
+
+    def test_token_refused_after_an_image_row_gains_or_loses_the_value(self):
+        self.owners()
+        self.own("Scenery", "a.png", "b.png")
+
+        def gain():
+            self.own("Scenery", "late.png")
+
+        def lose():
+            self.db.clear_image_categories([("root", "a.png")])
+
+        def move_away():
+            self.own("Elsewhere", "b.png")
+
+        def target_on_image():
+            self.own("Landscape", "target.png")
+
+        for mutate in (gain, lose, move_away, target_on_image):
+            with self.subTest(mutate.__name__):
+                token = self.preview()["impact_token"]
+                mutate()
+                before = self.snapshot(), self.image_rows()
+                with self.assertRaises(LibraryCommandError) as caught:
+                    self.apply(token)
+                self.assertEqual(caught.exception.code, "vocabulary-conflict")
+                self.assertEqual((self.snapshot(), self.image_rows()), before)
+
+    def test_image_category_failure_rolls_back_prompts(self):
+        self.owners()
+        self.own("Scenery", "a.png")
+        token = self.preview()["impact_token"]
+        with self.db.model.get_connection() as conn:
+            conn.execute(
+                "CREATE TRIGGER abort_images BEFORE UPDATE ON image_categories "
+                "BEGIN SELECT RAISE(ABORT, 'boom'); END"
+            )
+        before = self.snapshot(), self.image_rows()
+        with self.assertRaises(sqlite3.Error):
+            self.apply(token)
+        self.assertEqual((self.snapshot(), self.image_rows()), before)
+
+    def test_category_exists_reads_the_image_index(self):
+        from database import vocabulary
+
+        with self.db.model.get_connection() as conn:
+            plan = [
+                row[-1]
+                for row in conn.execute(
+                    "EXPLAIN QUERY PLAN " + vocabulary._CATEGORY_EXISTS, {"v": "x"}
+                )
+            ]
+        steps = [step for step in plan if "image_categories" in step]
+        self.assertTrue(steps, plan)
+        for step in steps:
+            self.assertIn("idx_image_categories_category", step)
+
     def test_category_apply_failure_rolls_back(self):
         self.owners()
         token = self.preview()["impact_token"]
@@ -942,6 +1220,18 @@ class TestSharedTokenHashing(unittest.TestCase):
             ),
             "cefe7f6a5544b7f6a203f5527810899c6af568aa7abc1b75adffbfb5ae6251ae",
         )
+        with_images = category_impact_token(
+            "e", "S", "T", True, [(1, True)], [], images=[9, 4]
+        )
+        self.assertRegex(with_images, r"\A[0-9a-f]{64}\Z")
+        self.assertEqual(
+            with_images,
+            category_impact_token("e", "S", "T", True, [(1, True)], [], images=[4, 9]),
+        )
+        self.assertNotEqual(
+            with_images,
+            category_impact_token("e", "S", "T", True, [(1, True)], [], images=[4]),
+        )
 
 
 if __name__ == "__main__":

tests/vendor/browser-coverage.test.mjs

@@ -27,3 +27,23 @@ test('browser coverage maps nested JS/MJS and excludes unrelated and traversal p
     const root = fileURLToPath(new URL('../../web/js/', import.meta.url));
     assert.deepEqual(report.result.map(entry => fileURLToPath(entry.url)), [join(root, 'gallery.js'), join(root, 'shared/controls/button.mjs'), join(root, 'shared/panel.js')]);
 });
+
+// #432: every raw file repeated each script's full source (about 84% of its bytes) and the coverage merge ran out of
+// memory. c8 reads the source from the mapped file on disk, so the entries keep only the URL, id and functions.
+test('browser coverage writes no script source, only what c8 maps', async t => {
+    const directory = mkdtempSync(join(tmpdir(), 'pm-browser-coverage-'));
+    const previous = process.env.PM_JS_COVERAGE;
+    process.env.PM_JS_COVERAGE = directory;
+    t.after(() => {
+        if (previous === undefined) delete process.env.PM_JS_COVERAGE;
+        else process.env.PM_JS_COVERAGE = previous;
+        rmSync(directory, { recursive: true, force: true });
+    });
+    const functions = [{ functionName: 'f', isBlockCoverage: true, ranges: [{ startOffset: 0, endOffset: 9, count: 1 }] }];
+    await saveCoverage({ coverage: { stopJSCoverage: async () => [{ url: 'http://127.0.0.1:8765/prompt_manager/js/gallery.js',
+        scriptId: '7', source: 'x'.repeat(100_000), functions }] } }, 'chromium');
+    const report = JSON.parse(readFileSync(join(directory, readdirSync(directory)[0]), 'utf8'));
+    assert.equal(report.result.length, 1);
+    assert.deepEqual(Object.keys(report.result[0]).sort(), ['functions', 'scriptId', 'url']);
+    assert.deepEqual(report.result[0].functions, functions);
+});

tests/vendor/browser-server-fixture.test.mjs

@@ -2,12 +2,14 @@ import test from 'node:test';
 import assert from 'node:assert/strict';
 import { createServer } from 'node:net';
 import { spawn } from 'node:child_process';
-import { once } from 'node:events';
+import { EventEmitter, once } from 'node:events';
 import { readFileSync } from 'node:fs';
+import { isAbsolute } from 'node:path';
 import { fileURLToPath, pathToFileURL } from 'node:url';
 import { setTimeout as delay } from 'node:timers/promises';
 import { startBrowserServer, createServerSupervisor, attachServerStderr, forwardPrefixedLines, SLOW_RESET_PREFIX } from '../../tools/vendor/start-browser-server.mjs';
-import { workflowWorkers } from '../../tools/vendor/test-workers.mjs';
+import { reporters, workflowWorkers } from '../../tools/vendor/test-workers.mjs';
+import RunningTestsReporter from '../../tools/vendor/running-tests-reporter.mjs';
 
 const fakeServer = fileURLToPath(new URL('./fake-browser-server.mjs', import.meta.url));
 const fake = (...extra) => ({ command: process.execPath, args: port => [fakeServer, '--port', String(port), ...extra] });
@@ -255,3 +257,57 @@ test('worker count follows PM_TEST_WORKERS, then one in CI, then a quarter of th
     assert.throws(() => workflowWorkers({ env: { PM_TEST_WORKERS: 'many' }, cpus: 8 }), /PM_TEST_WORKERS must be a positive integer/);
     assert.throws(() => workflowWorkers({ env: { PM_TEST_WORKERS: '0' }, cpus: 8 }), /PM_TEST_WORKERS must be a positive integer/);
 });
+
+test('CI adds the running-tests reporter and a blob report under test-results/blob/, which the trace upload covers; local runs keep list', () => {
+    assert.equal(reporters({ env: {}, name: 'workflows' }), 'list');
+    const [list, running, blob] = reporters({ env: { CI: '1' }, name: 'workflows' });
+    assert.deepEqual(list, ['list']);
+    // Absolute: a config imported from another directory (workflow-artifacts.test.mjs) still finds it.
+    assert.ok(isAbsolute(running[0]) && running[0].endsWith('running-tests-reporter.mjs'), running[0]);
+    assert.deepEqual(blob, ['blob', { outputDir: './test-results/blob/workflows' }]);
+});
+
+function fakeHost() {
+    const host = new EventEmitter();
+    Object.assign(host, { pid: 4242, signals: [], cwd: () => '/repo/tools/vendor' });
+    host.kill = (pid, signal) => host.signals.push([pid, signal]);
+    return host;
+}
+
+const spec = (id, title, line) => ({ id, title, location: { file: '/repo/tests/browser/a.spec.mjs', line },
+    parent: { project: () => ({ name: 'firefox' }) } });
+
+test('the running-tests reporter names only the tests that never ended, relative to rootDir', () => {
+    let clock = 1_000;
+    const output = [];
+    const host = fakeHost();
+    const reporter = new RunningTestsReporter({ now: () => clock, write: text => output.push(text), host });
+    reporter.onBegin({ rootDir: '/repo/tests/browser' });
+    const done = spec('1', 'finishes', 10);
+    const stuck = spec('2', 'hangs on a fetch', 20);
+    reporter.onTestBegin(done, { retry: 0 });
+    reporter.onTestBegin(stuck, { retry: 0 });
+    clock += 500;
+    reporter.onTestEnd(done, { retry: 0 });
+    clock += 41_500;
+    reporter.onEnd({ status: 'interrupted' });
+    assert.equal(output.join(''), '\nStill running when the run ended (interrupted): 1\n'
+        + '  [firefox] › a.spec.mjs:20 › hangs on a fetch (running 42.0s)\n');
+    assert.equal(host.listenerCount('SIGTERM'), 0, 'the run is over: SIGTERM gets its default again');
+    output.length = 0;
+    reporter.onTestEnd(stuck, { retry: 0 });
+    reporter.onEnd({ status: 'passed' });
+    assert.deepEqual(output, [], 'a run where every test ended prints nothing');
+});
+
+test('on the first SIGTERM the reporter prints the running tests and hands over to Playwright\'s SIGINT', () => {
+    const output = [];
+    const host = fakeHost();
+    const reporter = new RunningTestsReporter({ now: () => 0, write: text => output.push(text), host });
+    reporter.onBegin({ rootDir: '/repo/tests/browser' });
+    reporter.onTestBegin(spec('2', 'hangs on a fetch', 20), { retry: 0 });
+    host.emit('SIGTERM');
+    assert.match(output.join(''), /Still running when the run ended \(SIGTERM\): 1\n {2}\[firefox\] › a\.spec\.mjs:20 › hangs on a fetch/);
+    assert.deepEqual(host.signals, [[4242, 'SIGINT']], 'Playwright\'s own interrupt ends the run and writes the blob');
+    assert.equal(host.listenerCount('SIGTERM'), 0, 'a second SIGTERM kills the process as usual');
+});

tests/vendor/ci-workflow.test.mjs

@@ -247,6 +247,26 @@ test('browser job timeouts keep a real margin over measured runs', () => {
     }
 });
 
+// A hang (#425) must end inside the test step, before the job limit, so the log and the always()
+// uploads survive: the step's timeout-minutes < the job's.
+const BROWSER_TEST_STEP = /npm run --prefix tools\/vendor test:(workflows|coverage)/;
+
+test('each browser test step stops itself before its job limit and its traces upload anyway', () => {
+    for (const id of browserJobIds) {
+        const body = allJobs.get(id);
+        const jobMinutes = Number(/^ {4}timeout-minutes: (\d+)$/m.exec(body)[1]);
+        const jobSteps = steps(body);
+        const testAt = jobSteps.findIndex(step => BROWSER_TEST_STEP.test(step));
+        const testStep = jobSteps[testAt];
+        const stepMinutes = Number(/^ {8}timeout-minutes: (\d+)$/m.exec(testStep)?.[1]);
+        assert.ok(stepMinutes > 0, `${id}: the test step sets its own timeout-minutes`);
+        assert.ok(jobMinutes - stepMinutes >= 2, `${id}: the step limit leaves time for setup and the uploads`);
+        const upload = jobSteps.slice(testAt + 1).find(step => /^ {10}path: tools\/vendor\/test-results\/$/m.test(step));
+        assert.ok(upload, `${id} uploads tools/vendor/test-results/ (traces and blob report) after the tests`);
+        assert.match(upload, /^ {8}if: always\(\)$/m, `${id} uploads them when the step failed or timed out`);
+    }
+});
+
 // Every job is bounded (#350): a hang ends at a measured limit instead of GitHub's 6-hour default.
 const GATING_WORKFLOWS = ['test.yml', 'code-quality.yml', 'viewerjs-check.yml'];
 const BOUNDED_WORKFLOWS = [...GATING_WORKFLOWS, 'portability.yml'];

tests/vendor/comfyui-config.test.mjs

@@ -34,15 +34,17 @@ test('execution script discovers only the new matrix and preserves all 78 origin
     assert.match(executions, /\[classic-start\]/);
     assert.match(executions, /\[nodes2-start\]/);
     // + the fresh-host library location case, + #342 image attribution, + #343 linked text,
-    // + #332 image category routes on a generated output (image-attribution.spec.mjs, 2 projects)
-    assert.match(executions, /Total: 92 tests in 5 files/);
+    // + #332 image category routes on a generated output (image-attribution.spec.mjs, 2 projects),
+    // + #347 cached runtime-linked runs: five node kinds and eight one-renderer cases (execution-text-link.spec.mjs, 2 projects),
+    // + #90 Save to library off (five node types)
+    assert.match(executions, /Total: 128 tests in 5 files/);
     for (const project of ['classic-start', 'nodes2-start']) {
         const cases = executions.split('\n').filter(line => line.includes(`[${project}]`));
         assert.equal(cases.filter(line => line.includes('execution-history.spec.mjs')).length, 15);
         assert.equal(cases.filter(line => line.includes('execution-outputs.spec.mjs')).length, 8);
         assert.equal(cases.filter(line => line.includes('library-location.spec.mjs')).length, 1);
         assert.equal(cases.filter(line => line.includes('image-attribution.spec.mjs')).length, 3);
-        assert.equal(cases.filter(line => line.includes('execution-text-link.spec.mjs')).length, 19);
+        assert.equal(cases.filter(line => line.includes('execution-text-link.spec.mjs')).length, 37);
     }
     assert.doesNotMatch(executions, /prompt-manager.spec/);
 });

tests/vendor/frontend-coverage-cli.test.mjs

@@ -1,9 +1,11 @@
 import test from 'node:test';
 import assert from 'node:assert/strict';
-import { existsSync, mkdtempSync, mkdirSync, rmSync, writeFileSync } from 'node:fs';
+import { existsSync, mkdtempSync, mkdirSync, readFileSync, rmSync, writeFileSync } from 'node:fs';
+import { EventEmitter } from 'node:events';
+import { setTimeout as delay } from 'node:timers/promises';
 import { tmpdir } from 'node:os';
 import { join } from 'node:path';
-import { plan, runCoverage } from '../../tools/vendor/frontend-coverage.mjs';
+import { plan, runCoverage, runNode } from '../../tools/vendor/frontend-coverage.mjs';
 
 const steps = argv => plan(argv).map(({ step, ...options }) => (Object.keys(options).length ? { step, ...options } : step));
 
@@ -91,3 +93,46 @@ test('shard 1 runs the node units into the shared raw dir, then the viewer suite
     assert.ok(calls[1].args.includes(join(paths.tools, 'playwright.config.mjs')));
     assert.deepEqual(calls[2].args.slice(-2), ['--shard', '1/3']);
 });
+
+// #425: npm and a CI step timeout signal only frontend-coverage.mjs; its running child must get it.
+function hostWithSignals() {
+    return new EventEmitter();
+}
+
+test('runNode resolves on exit 0, rejects with the exit code otherwise and leaves no signal listener', async () => {
+    const host = hostWithSignals();
+    await runNode(['-e', 'process.exit(0)'], {}, { host });
+    await assert.rejects(runNode(['-e', 'process.exit(3)'], {}, { host }), /Command failed \(3\): -e process\.exit\(3\)/);
+    assert.equal(host.listenerCount('SIGTERM') + host.listenerCount('SIGINT'), 0);
+});
+
+test('runNode forwards SIGTERM to the running child, which can report before it exits', async t => {
+    const root = mkdtempSync(join(tmpdir(), 'pm-coverage-signal-'));
+    t.after(() => rmSync(root, { recursive: true, force: true }));
+    const ready = join(root, 'ready');
+    const seen = join(root, 'seen');
+    const child = `const fs = require('node:fs');
+process.on('SIGTERM', () => { fs.writeFileSync(${JSON.stringify(seen)}, 'SIGTERM'); process.exit(7); });
+fs.writeFileSync(${JSON.stringify(ready)}, ''); setInterval(() => {}, 1000);`;
+    const host = hostWithSignals();
+    const running = runNode(['-e', child], {}, { host });
+    while (!existsSync(ready)) await delay(20);
+    host.emit('SIGTERM');
+    await assert.rejects(running, /Command failed \(7\)/);
+    assert.equal(readFileSync(seen, 'utf8'), 'SIGTERM', 'the child handled the forwarded signal');
+    assert.equal(host.listenerCount('SIGTERM'), 0);
+});
+
+test('runNode forwards SIGINT and starts the child in its own process group', async () => {
+    const host = hostWithSignals();
+    const kills = [];
+    let options;
+    const fake = new EventEmitter();
+    fake.kill = signal => { kills.push(signal); queueMicrotask(() => fake.emit('exit', null, signal)); };
+    const running = runNode(['x.mjs'], {}, { host, spawnChild: (command, args, spawnOptions) => { options = spawnOptions; return fake; } });
+    assert.equal(options.detached, true, 'a group-wide signal reaches the child only through forwarding, once');
+    host.emit('SIGINT');
+    await assert.rejects(running, /Command failed \(null\): x\.mjs/);
+    assert.deepEqual(kills, ['SIGINT']);
+    assert.equal(host.listenerCount('SIGINT'), 0);
+});

tools/vendor/browser-coverage.mjs

@@ -12,7 +12,10 @@ export async function startCoverage(page, browserName) {
 export async function saveCoverage(page, browserName) {
     if (!process.env.PM_JS_COVERAGE || browserName !== 'chromium') return;
     const entries = await page.coverage.stopJSCoverage();
-    const result = mapBrowserCoverage(entries, { urlPrefix: '/prompt_manager/js/', sourceRoot: join(repositoryRoot, 'web/js') });
+    // The source is left out (#432): c8 reads it from the mapped file on disk, and repeating every script's text in every
+    // raw file made the coverage merge run out of memory.
+    const result = mapBrowserCoverage(entries, { urlPrefix: '/prompt_manager/js/', sourceRoot: join(repositoryRoot, 'web/js') })
+        .map(({ source: _source, ...entry }) => entry);
     mkdirSync(process.env.PM_JS_COVERAGE, { recursive: true });
     writeFileSync(join(process.env.PM_JS_COVERAGE, `browser-${randomUUID()}.json`), JSON.stringify({ result }));
 }

tools/vendor/frontend-coverage-baseline.json

@@ -107,6 +107,12 @@
     "functions": 98,
     "statements": 98
   },
+  "save-to-library.js": {
+    "lines": 98,
+    "branches": 98,
+    "functions": 98,
+    "statements": 98
+  },
   "scan-progress.js": {
     "lines": 98,
     "branches": 98,

tools/vendor/frontend-coverage.mjs

@@ -1,7 +1,7 @@
 import { existsSync, readFileSync, readdirSync, mkdirSync, rmSync, writeFileSync } from 'node:fs';
 import { dirname, join, relative } from 'node:path';
 import { fileURLToPath, pathToFileURL } from 'node:url';
-import { spawnSync } from 'node:child_process';
+import { spawn } from 'node:child_process';
 import { repositoryRoot } from './vendor-viewerjs.mjs';
 
 import { discoverSources, checkCoverage } from '@comfyassets/toolkit/testing/coverage';
@@ -10,10 +10,27 @@ const metrics = ['lines', 'branches', 'functions', 'statements'];
 const SHARD = /^([1-9]\d*)\/([1-9]\d*)$/;
 const USAGE = 'usage: frontend-coverage.mjs [--update-baseline] | --shard I/N | --merge-only [--update-baseline]';
 
-function runNode(args, env = {}) {
-    const result = spawnSync(process.execPath, args, { cwd: repositoryRoot, stdio: 'inherit',
-        env: { ...process.env, ...env } });
-    if (result.status !== 0) throw new Error(`Command failed (${result.status}): ${args.join(' ')}`);
+const FORWARDED_SIGNALS = ['SIGTERM', 'SIGINT'];
+
+// Runs one step's node child and waits for it. npm and a CI step timeout signal only this process,
+// so SIGTERM and SIGINT are forwarded to the running child (#425): Playwright then reports the tests
+// still running and writes its blob. The child gets its own process group, so a signal sent to the
+// whole group (Ctrl-C, `timeout`) still reaches it exactly once, through this forwarding.
+export function runNode(args, env = {}, { spawnChild = spawn, host = process } = {}) {
+    return new Promise((resolve, reject) => {
+        const child = spawnChild(process.execPath, args, { cwd: repositoryRoot, stdio: 'inherit',
+            env: { ...process.env, ...env }, detached: true });
+        const forward = signal => () => child.kill(signal);
+        const handlers = FORWARDED_SIGNALS.map(signal => [signal, forward(signal)]);
+        for (const [signal, handler] of handlers) host.on(signal, handler);
+        const done = () => { for (const [signal, handler] of handlers) host.removeListener(signal, handler); };
+        child.once('error', error => { done(); reject(error); });
+        child.once('exit', code => {
+            done();
+            if (code === 0) resolve();
+            else reject(new Error(`Command failed (${code}): ${args.join(' ')}`));
+        });
+    });
 }
 
 function defaultPaths() {
@@ -112,17 +129,17 @@ export async function runCoverage({ argv = process.argv.slice(2), run = runNode,
         'node-units'() {
             const tests = readdirSync(join(root, 'tests/js')).filter(file => file.endsWith('.test.js'))
                 .map(file => `tests/js/${file}`);
-            run(['--test', ...tests], { NODE_V8_COVERAGE: raw });
+            return run(['--test', ...tests], { NODE_V8_COVERAGE: raw });
         },
         viewer() {
-            run([playwright, 'test', '--config', join(tools, 'playwright.config.mjs')]);
+            return run([playwright, 'test', '--config', join(tools, 'playwright.config.mjs')]);
         },
         workflows({ shard }) {
-            run([playwright, 'test', '--config', join(tools, 'playwright.workflows.config.mjs'),
+            return run([playwright, 'test', '--config', join(tools, 'playwright.workflows.config.mjs'),
                 '--project', 'chromium', '--project', 'touch', ...(shard ? ['--shard', shard] : [])], { PM_JS_COVERAGE: raw });
         },
         report() {
-            run([join(tools, 'node_modules/c8/bin/c8.js'), 'report', '--all', '--src', sourceRoot,
+            return run([join(tools, 'node_modules/c8/bin/c8.js'), 'report', '--all', '--src', sourceRoot,
                 '--include', 'web/js/**/*.js', '--include', 'web/js/**/*.mjs', '--temp-directory', raw, '--reports-dir', reportRoot,
                 '--reporter', 'text', '--reporter', 'json-summary', '--reporter', 'html']);
         },

tools/vendor/playwright.config.mjs

@@ -1,4 +1,5 @@
 import { defineConfig } from '@playwright/test';
+import { reporters } from './test-workers.mjs';
 
 export default defineConfig({
     testDir: './tests',
@@ -8,7 +9,7 @@ export default defineConfig({
     workers: 2,
     forbidOnly: Boolean(process.env.CI),
     retries: process.env.CI ? 1 : 0,
-    reporter: 'list',
+    reporter: reporters({ name: 'viewer' }),
     use: { browserName: 'chromium', headless: true, trace: 'retain-on-failure' },
     projects: [
         { name: 'desktop', use: { viewport: { width: 1280, height: 800 } } },

tools/vendor/playwright.workflows.config.mjs

@@ -1,5 +1,5 @@
 import { defineConfig, devices } from '@playwright/test';
-import { workflowWorkers } from './test-workers.mjs';
+import { reporters, workflowWorkers } from './test-workers.mjs';
 
 // No global webServer: tests/browser/fixtures.mjs starts one server per worker
 // on an OS-chosen port. Files run in parallel; tests inside a file keep their order.
@@ -11,7 +11,7 @@ export default defineConfig({
     fullyParallel: false,
     forbidOnly: Boolean(process.env.CI),
     retries: 0,
-    reporter: 'list',
+    reporter: reporters({ name: 'workflows' }),
     use: { trace: 'retain-on-failure', screenshot: 'only-on-failure' },
     projects: [
         { name: 'chromium', outputDir: './test-results/workflows/chromium', use: { browserName: 'chromium' } },

tools/vendor/running-tests-reporter.mjs

@@ -0,0 +1,58 @@
+import { writeSync } from 'node:fs';
+import { relative } from 'node:path';
+
+// Names the tests still running when a run is cut off (#425). A browser test step that reaches its
+// timeout-minutes gets SIGTERM, which ends Playwright at once: no reporter's onEnd runs, so neither
+// `list` nor the blob report says what was running. On the first SIGTERM this reporter prints every
+// started test that has not ended, then turns the signal into SIGINT, Playwright's own interrupt,
+// which stops the workers, ends the run as interrupted (non-zero exit) and writes the blob report.
+// A second SIGTERM gets the default behaviour: the process dies.
+export default class RunningTestsReporter {
+    constructor({ now = () => Date.now(), write = text => writeSync(1, text), host = process } = {}) {
+        this.now = now;
+        this.write = write;
+        this.host = host;
+        this.rootDir = host.cwd();
+        this.running = new Map();
+        this.onSigterm = () => {
+            this.host.removeListener('SIGTERM', this.onSigterm);
+            this.report('SIGTERM');
+            this.host.kill(this.host.pid, 'SIGINT');
+        };
+    }
+
+    printsToStdio() {
+        return false;
+    }
+
+    onBegin(config) {
+        this.rootDir = config.rootDir;
+        this.host.on('SIGTERM', this.onSigterm);
+    }
+
+    onTestBegin(test, result) {
+        this.running.set(`${test.id}#${result.retry}`, { test, startedAt: this.now() });
+    }
+
+    onTestEnd(test, result) {
+        this.running.delete(`${test.id}#${result.retry}`);
+    }
+
+    describe({ test, startedAt }) {
+        const project = test.parent?.project()?.name;
+        const where = `${relative(this.rootDir, test.location.file)}:${test.location.line}`;
+        const seconds = ((this.now() - startedAt) / 1000).toFixed(1);
+        return `${project ? `[${project}] › ` : ''}${where} › ${test.title} (running ${seconds}s)`;
+    }
+
+    report(reason) {
+        if (!this.running.size) return;
+        const lines = [...this.running.values()].map(entry => `  ${this.describe(entry)}`);
+        this.write(`\nStill running when the run ended (${reason}): ${this.running.size}\n${lines.join('\n')}\n`);
+    }
+
+    onEnd(result) {
+        this.host.removeListener('SIGTERM', this.onSigterm);
+        this.report(result.status);
+    }
+}

tools/vendor/test-workers.mjs

@@ -1,4 +1,8 @@
 import { availableParallelism } from 'node:os';
+import { fileURLToPath } from 'node:url';
+
+// Absolute, so a config that imports this one from another directory still finds it.
+const RUNNING_TESTS_REPORTER = fileURLToPath(new URL('./running-tests-reporter.mjs', import.meta.url));
 
 const MAX_LOCAL_WORKERS = 8;
 const CPUS_PER_WORKER = 4;
@@ -16,3 +20,10 @@ export function workflowWorkers({ env = process.env, cpus = availableParallelism
     if (env.CI) return 1;
     return Math.max(1, Math.min(MAX_LOCAL_WORKERS, Math.floor(cpus / CPUS_PER_WORKER)));
 }
+
+// Under CI, running-tests-reporter.mjs names the tests a step timeout's signal cut off (`list`
+// cannot: they never end), and a blob report beside the traces is uploaded with them.
+export function reporters({ env = process.env, name } = {}) {
+    if (!env.CI) return 'list';
+    return [['list'], [RUNNING_TESTS_REPORTER], ['blob', { outputDir: `./test-results/blob/${name}` }]];
+}

utils/cached_runs.py

@@ -0,0 +1,532 @@
+"""Count a cached PromptManager node like the execution it reuses (#347).
+
+The queue hook (``usage_tracking``) counts text it can read from the queued
+graph. Text only known at run time (a PreviewAny or LLM link, a list node) is
+counted by the node itself when it executes. When ComfyUI later serves the node
+from its output cache the node does not execute, so without this module that
+job is not counted.
+
+When a PromptManager-family node executes it records a *receipt item*: the
+role-independent facts its own count is made from (the stored text hashes)
+and digests of the STRING outputs it returned. When ComfyUI reports the node
+as cached for a later job (``execution_cached``), the receipt whose digests
+equal the text in the cache entry is replayed once for that job. The node's
+role (positive, negative, which Pair outputs are used) and what the hook
+already saw are recomputed from that job's own graph, and both paths count
+through the same ``own_counts`` so they cannot drift.
+
+Fail closed: when the cached text cannot be matched to a receipt recorded in
+this process, nothing is counted (the behavior before #347). In memory only and
+bounded; starts no threads.
+"""
+
+import hashlib
+import threading
+from collections import OrderedDict
+from typing import Any, Callable, Dict, FrozenSet, List, NamedTuple, Optional, Tuple
+
+try:
+    from .hashing import generate_prompt_hash
+    from .logging_config import get_logger
+    from .prompt_graph import (
+        PROMPT_MANAGER_TYPES,
+        intrinsic_pair_nodes,
+        resolve_text,
+        run_prompt_components,
+        run_prompt_nodes,
+    )
+    from .prompt_pair_resolver import resolve_text_input
+    from .queue_observers import subscribe_dequeue, subscribe_task_done
+    from .run_keys import prompt_node_key, upstream_nodes
+    from .usage_tracking import find_pair_id_by_hashes, survive_switch
+except ImportError:
+    from utils.hashing import generate_prompt_hash
+    from utils.logging_config import get_logger
+    from utils.prompt_graph import (
+        PROMPT_MANAGER_TYPES,
+        intrinsic_pair_nodes,
+        resolve_text,
+        run_prompt_components,
+        run_prompt_nodes,
+    )
+    from utils.prompt_pair_resolver import resolve_text_input
+    from utils.queue_observers import subscribe_dequeue, subscribe_task_done
+    from utils.run_keys import prompt_node_key, upstream_nodes
+    from utils.usage_tracking import find_pair_id_by_hashes, survive_switch
+
+logger = get_logger("prompt_manager.cached_runs")
+
+PAIR = "PromptManagerPair"
+MAX_KEYS = 256  # run keys remembered (the same bound as MAX_KEYED_RUNS)
+MAX_RECEIPTS = 4  # receipts per run key: a recurring upstream value (A, B, A)
+MAX_ITEMS = 64  # list items per receipt (the cache reader's bound)
+MAX_CACHED_NODES = 500  # the execution_cached guard the history capture uses
+_EMPTY_HASH = generate_prompt_hash("")
+
+Count = Tuple[Any, ...]  # ("prompt", hash) or ("pair", positive, negative or None)
+
+
+class Facts(NamedTuple):
+    """What one execution stored: hashes of the stored text, None when nothing.
+
+    ``positive_hash`` is None for an empty editor (or saving disabled);
+    ``negative_hash`` (Pair only) is None for an empty negative, which the pair
+    stores as ``negative_prompt_id IS NULL``.
+    """
+
+    positive_hash: Optional[str]
+    negative_hash: Optional[str] = None
+
+
+class Role(NamedTuple):
+    """How a job's graph uses the node: which editors reach a sampler's positive."""
+
+    used_fields: FrozenSet[str]
+    pair_used: bool = False
+
+
+class Seen(NamedTuple):
+    """Which of the node's texts the queue hook resolved, and so already counted."""
+
+    text: bool = False
+    negative_text: bool = False
+    pair: bool = False
+
+
+class _Item(NamedTuple):
+    outputs: Tuple[str, ...]
+    facts: Facts
+
+
+class _Receipt:
+    __slots__ = ("item_id", "items")
+
+    def __init__(self, item_id: int) -> None:
+        self.item_id = item_id
+        self.items: Dict[int, _Item] = {}
+
+    def outputs(self) -> Dict[int, Tuple[str, ...]]:
+        return {index: item.outputs for index, item in self.items.items()}
+
+
+class _Attempt(NamedTuple):
+    item_id: int
+    prompt_id: str
+    graph: Dict[str, Any]
+    nodes: Dict[str, Tuple[str, str]]  # node id -> (class, run key)
+    replayed: set
+
+
+def text_digest(value: str) -> str:
+    """Digest of one returned STRING value, exactly as returned."""
+    return hashlib.sha256(value.encode("utf-8", "surrogatepass")).hexdigest()
+
+
+def stored_facts(text: Any, negative_text: Any = None) -> Facts:
+    """The facts of an execution that stores ``text`` (and a Pair's negative)."""
+    positive = text if isinstance(text, str) and text.strip() else None
+    negative = (
+        negative_text
+        if isinstance(negative_text, str) and negative_text.strip()
+        else None
+    )
+    return Facts(
+        generate_prompt_hash(positive) if positive is not None else None,
+        generate_prompt_hash(negative) if negative is not None else None,
+    )
+
+
+def pair_components(facts: Facts, role: Role) -> List[Tuple[str, str]]:
+    """``(field, hash)`` of each used, stored Pair component; each hash once."""
+    found, counted = [], set()
+    for field, prompt_hash in (
+        ("text", facts.positive_hash),
+        ("negative_text", facts.negative_hash),
+    ):
+        if field not in role.used_fields or prompt_hash is None:
+            continue
+        if prompt_hash in counted:
+            continue
+        counted.add(prompt_hash)
+        found.append((field, prompt_hash))
+    return found
+
+
+def own_counts(node_type: str, facts: Facts, role: Role, seen: Seen) -> List[Count]:
+    """What one execution of a node counts itself: text the hook could not see.
+
+    Shared by the executing node and by replay. Excludes the pending first-use
+    catch-up, which only an execution that saves the prompt can consume.
+    """
+    if facts.positive_hash is None:
+        return []
+    if node_type != PAIR:
+        if "text" in role.used_fields and not seen.text:
+            return [("prompt", facts.positive_hash)]
+        return []
+    counts: List[Count] = [
+        ("prompt", prompt_hash)
+        for field, prompt_hash in pair_components(facts, role)
+        if not getattr(seen, field)
+    ]
+    if role.pair_used and not seen.pair:
+        counts.append(("pair", facts.positive_hash, facts.negative_hash))
+    return counts
+
+
+def graph_role(graph: Any, node_id: str, node_type: str) -> Role:
+    """The node's role in a queued graph, by the functions the nodes use."""
+    if node_type != PAIR:
+        used = {"text"} if node_id in run_prompt_nodes(graph) else set()
+        return Role(frozenset(used))
+    used = {
+        field for source, field in run_prompt_components(graph) if source == node_id
+    }
+    return Role(frozenset(used), node_id in intrinsic_pair_nodes(graph))
+
+
+def _same(resolved: Optional[str], prompt_hash: Optional[str]) -> bool:
+    return resolved is not None and generate_prompt_hash(resolved) == prompt_hash
+
+
+def graph_seen(graph: Any, node_id: str, node_type: str, facts: Facts) -> Seen:
+    """Which texts the queue hook resolved from ``graph`` (as the node decides)."""
+    if node_type != PAIR:
+        return Seen(text=_same(resolve_text(graph, node_id), facts.positive_hash))
+    positive = _same(resolve_text_input(graph, node_id, "text"), facts.positive_hash)
+    negative = _same(
+        resolve_text_input(graph, node_id, "negative_text"),
+        facts.negative_hash or _EMPTY_HASH,
+    )
+    return Seen(positive, negative, positive and negative)
+
+
+def _host_context() -> Tuple[Optional[str], Optional[int]]:
+    """(prompt id, list index) of the node executing on this thread, if known."""
+    try:
+        from comfy_execution.utils import get_executing_context
+    except Exception:
+        return None, None
+    try:
+        context = get_executing_context()
+    except Exception:
+        return None, None
+    prompt_id = getattr(context, "prompt_id", None)
+    index = getattr(context, "list_index", None)
+    return (
+        prompt_id if isinstance(prompt_id, str) and prompt_id else None,
+        index if type(index) is int else None,
+    )
+
+
+def write_count(db: Any, count: Count) -> bool:
+    """Record one counted run; False when the prompt or pair no longer exists."""
+
+    def write() -> bool:
+        if count[0] == "prompt":
+            row = db.get_prompt_by_hash(count[1])
+            if row:
+                db.record_prompt_use(row["id"])
+            return bool(row)
+        pair_id = find_pair_id_by_hashes(db, count[1], count[2])
+        if pair_id:
+            db.record_prompt_pair_use(pair_id)
+        return bool(pair_id)
+
+    return bool(survive_switch(db, write))
+
+
+class CachedRuns:
+    """Receipts of executed nodes and the one job currently executing."""
+
+    def __init__(
+        self,
+        *,
+        max_keys: int = MAX_KEYS,
+        max_receipts: int = MAX_RECEIPTS,
+        max_items: int = MAX_ITEMS,
+        context: Callable[[], Tuple[Optional[str], Optional[int]]] = _host_context,
+    ) -> None:
+        self._max_keys = max_keys
+        self._max_receipts = max_receipts
+        self._max_items = max_items
+        self._context = context
+        self._receipts: "OrderedDict[str, List[_Receipt]]" = OrderedDict()
+        self._attempt: Optional[_Attempt] = None
+        self._lock = threading.Lock()
+
+    # -- the job ------------------------------------------------------------
+
+    def begin_attempt(self, result: Any) -> None:
+        """A job was dequeued: ``result`` is ``PromptQueue.get``'s return value."""
+        attempt = None
+        try:
+            item, item_id = result[0], result[1]
+            prompt_id, graph, outputs = item[1], item[2], item[4]
+            if type(item_id) is int and isinstance(prompt_id, str):
+                if isinstance(graph, dict):
+                    attempt = _Attempt(
+                        item_id, prompt_id, graph, _job_nodes(graph, outputs), set()
+                    )
+        except Exception as error:
+            logger.debug(f"Dequeued job not tracked for cached runs: {error}")
+        with self._lock:
+            self._attempt = attempt
+
+    def end_attempt(self, item_id: Any) -> None:
+        """The job finished: its receipts are complete and replace equal older ones.
+
+        Deduplicating only complete receipts keeps a list execution that is
+        still recording ([A] so far of [A, B]) from dropping an older [A].
+        """
+        with self._lock:
+            if self._attempt is not None and self._attempt.item_id == item_id:
+                self._attempt = None
+            for run_key, receipts in self._receipts.items():
+                if receipts and receipts[0].item_id == item_id:
+                    self._receipts[run_key] = _deduplicated(receipts)
+
+    def _current(self) -> Tuple[Optional[_Attempt], Optional[int]]:
+        """The attempt the executing node belongs to, and its list index."""
+        prompt_id, index = self._context()
+        with self._lock:
+            attempt = self._attempt
+        if attempt is None or prompt_id is None or prompt_id != attempt.prompt_id:
+            return None, None
+        return attempt, index
+
+    # -- node execution -----------------------------------------------------
+
+    def replayed(self, node_id: Any) -> bool:
+        """Whether this job already counted the executing node by replay."""
+        attempt, _ = self._current()
+        if attempt is None:
+            return False
+        with self._lock:
+            return str(node_id) in attempt.replayed
+
+    def record_item(self, node_id: Any, outputs: Tuple[Any, ...], facts: Facts) -> None:
+        """Remember what this execution returned and was entitled to count."""
+        try:
+            attempt, index = self._current()
+            node = attempt.nodes.get(str(node_id)) if attempt else None
+            if node is None:
+                return
+            index = 0 if index is None else index
+            if not 0 <= index < self._max_items:
+                return
+            if any(not isinstance(value, str) for value in outputs):
+                return
+            item = _Item(tuple(text_digest(value) for value in outputs), facts)
+            self._store(node[1], attempt.item_id, index, item)
+        except Exception as error:
+            logger.debug(f"Could not record a cached-run receipt: {error}")
+
+    def _store(self, run_key: str, item_id: int, index: int, item: _Item) -> None:
+        with self._lock:
+            receipts = self._receipts.pop(run_key, [])
+            if receipts and receipts[0].item_id == item_id:
+                receipt = receipts[0]
+            else:
+                # The previous job's receipt is complete (its task_done may
+                # never have fired): it replaces older receipts it equals.
+                receipts = _deduplicated(receipts)
+                receipt = _Receipt(item_id)
+                receipts.insert(0, receipt)
+            receipt.items[index] = item
+            self._receipts[run_key] = receipts[: self._max_receipts]
+            while len(self._receipts) > self._max_keys:
+                self._receipts.popitem(last=False)
+
+    # -- cached nodes -------------------------------------------------------
+
+    def replay(
+        self,
+        executor: Any,
+        prompt_id: Any,
+        nodes: Any,
+        db_factory: Callable[[], Any],
+        reader: Optional[Callable[[Any, str, str], Any]] = None,
+    ) -> None:
+        """Count the job's cached nodes whose cached text a receipt proves."""
+        if reader is None:
+            reader = _host().cached_string_outputs
+        with self._lock:
+            attempt = self._attempt
+        if attempt is None or prompt_id != attempt.prompt_id:
+            return
+        if type(nodes) not in (list, tuple) or len(nodes) > MAX_CACHED_NODES:
+            return
+        db = None
+        for node_id in nodes:
+            node = attempt.nodes.get(node_id) if type(node_id) is str else None
+            if node is None:
+                continue
+            node_type, run_key = node
+            items = self._match(run_key, reader(executor, node_id, node_type))
+            if items is None:
+                logger.debug(
+                    f"Cached node {node_id} not counted: no matching execution"
+                )
+                continue
+            with self._lock:
+                if node_id in attempt.replayed:
+                    continue  # a repeated message or id never counts twice
+                attempt.replayed.add(node_id)
+            counts: List[Count] = []
+            role = graph_role(attempt.graph, node_id, node_type)
+            for item in items:
+                seen = graph_seen(attempt.graph, node_id, node_type, item.facts)
+                counts.extend(own_counts(node_type, item.facts, role, seen))
+            if not counts:
+                continue
+            try:
+                db = db if db is not None else db_factory()
+                for count in counts:
+                    if not write_count(db, count):
+                        logger.info(
+                            f"Cached node {node_id}: its {count[0]} no longer exists, "
+                            "not counted"
+                        )
+            except Exception as error:
+                logger.warning(f"Could not count cached node {node_id}: {error}")
+
+    def _match(self, run_key: str, values: Any) -> Optional[List[_Item]]:
+        """The items of the one receipt whose outputs equal ``values``, in order."""
+        if not values:
+            return None
+        length = len(values[0])
+        if length == 0 or length > self._max_items:
+            return None
+        if any(len(slot) != length for slot in values):
+            return None
+        wanted = {
+            index: tuple(text_digest(slot[index]) for slot in values)
+            for index in range(length)
+        }
+        with self._lock:
+            receipts = self._receipts.get(run_key)
+            if not receipts:
+                return None
+            matches = [receipt for receipt in receipts if receipt.outputs() == wanted]
+            if not matches:
+                return None
+            self._receipts.move_to_end(run_key)
+            facts = {
+                tuple(receipt.items[index].facts for index in range(length))
+                for receipt in matches
+            }
+            if len(facts) != 1:
+                return None
+            return [matches[0].items[index] for index in range(length)]
+
+
+def _deduplicated(receipts: List[_Receipt]) -> List[_Receipt]:
+    """The newest receipt replaces older ones with the same outputs."""
+    if not receipts:
+        return receipts
+    outputs = receipts[0].outputs()
+    return [receipts[0]] + [r for r in receipts[1:] if r.outputs() != outputs]
+
+
+def _job_nodes(graph: Dict[str, Any], outputs: Any) -> Dict[str, Tuple[str, str]]:
+    """PromptManager-family nodes the job's outputs need, with their run keys."""
+    starts = (
+        [str(node) for node in outputs] if isinstance(outputs, (list, tuple)) else []
+    )
+    needed = upstream_nodes(graph, starts)
+    nodes = {}
+    for node_id in needed:
+        node = graph.get(node_id)
+        node_type = node.get("class_type") if isinstance(node, dict) else None
+        if node_type not in PROMPT_MANAGER_TYPES:
+            continue
+        run_key = prompt_node_key(graph, node_id)
+        if run_key is not None:
+            nodes[node_id] = (node_type, run_key)
+    return nodes
+
+
+def _host() -> Any:
+    """The host adapter module (imported late: it needs ComfyUI's modules)."""
+    try:
+        from . import execution_context_host
+    except ImportError:
+        from utils import execution_context_host
+    return execution_context_host
+
+
+# Shared between the nodes and the host observers
+CACHED_RUNS = CachedRuns()
+
+_registered = threading.Event()
+_register_lock = threading.Lock()
+
+
+def register_cached_runs(
+    server: Any, db_factory: Callable[[], Any], runs: Optional[CachedRuns] = None
+) -> bool:
+    """Observe dequeue, ``execution_cached`` and ``task_done``; once per process.
+
+    Registers nothing when the host lacks any of them: counting is then as
+    before #347.
+    """
+    runs = CACHED_RUNS if runs is None else runs
+    with _register_lock:
+        if _registered.is_set():
+            return False
+        queue = getattr(server, "prompt_queue", None)
+        if not callable(getattr(queue, "get", None)) or not callable(
+            getattr(queue, "task_done", None)
+        ):
+            logger.info("Cached runs not counted: the prompt queue is not observable")
+            return False
+        try:
+            from execution import PromptExecutor
+        except Exception:
+            logger.info("Cached runs not counted: no ComfyUI executor")
+            return False
+        if not callable(getattr(PromptExecutor, "add_message", None)):
+            logger.info("Cached runs not counted: executor messages unavailable")
+            return False
+        subscribe_executor_messages = _host().subscribe_executor_messages
+
+        def on_message(executor: Any, args: tuple, kwargs: dict) -> None:
+            try:
+                event = args[0] if args else kwargs.get("event")
+                if event != "execution_cached":
+                    return
+                server_of = getattr(executor, "server", None)
+                if getattr(server_of, "prompt_queue", None) is not queue:
+                    return
+                data = args[1] if len(args) > 1 else kwargs.get("data")
+                if type(data) is not dict:
+                    return
+                runs.replay(
+                    executor, data.get("prompt_id"), data.get("nodes"), db_factory
+                )
+            except Exception as error:
+                logger.warning(f"Could not count cached nodes: {error}")
+
+        unsubscribers = []
+        try:
+            unsubscribers.append(
+                subscribe_dequeue(queue, "cached_runs", runs.begin_attempt)
+            )
+            unsubscribers.append(
+                subscribe_task_done(
+                    queue,
+                    "cached_runs",
+                    lambda args, kwargs: runs.end_attempt(
+                        args[0] if args else kwargs.get("item_id")
+                    ),
+                )
+            )
+            unsubscribers.append(subscribe_executor_messages("cached_runs", on_message))
+        except Exception:
+            for unsubscribe in unsubscribers:
+                unsubscribe()
+            raise
+        _registered.set()
+        logger.info("Counting cached runs of runtime-linked prompt text")
+        return True

utils/execution_context_capture.py

@@ -11,6 +11,7 @@ import weakref
 
 from .execution_context_budget import ContextBudget, PayloadLease, SealedContext
 from .execution_context_graph import (
+    NOT_SAVED,
     GraphProjection,
     GraphRole,
     SourceRef,
@@ -325,6 +326,8 @@ class ContextCapture:
                     nonlocal ambiguous
                     if role.reason == "source-transformed":
                         return _unknown("unsupported")
+                    if role.reason == NOT_SAVED:
+                        return _unknown("unavailable")
                     if not role.sources:
                         return role.inferred or _unknown("unsupported")
                     values = [