A value of the right type under an unknown name may mean something
else (set_prompt {style: "anime"} is no prompt), so the correction now
takes only the other names light models were seen to write in the tool
exams (name/workflow_name, file/filename/image_name, text/positive,
model/id, node/id); any other name is answered with what the tool takes.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A call lacking its one needed argument but carrying one unknown
argument of that type is run with it (open_workflow {name} is
{workflow}), and the AI is told the right name. The timeline names
zoom_image steps; a test makes every tool have a name in both
languages. tool-exam.md records the second exam, all eight tasks
including the repaint one.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A question the user began to answer is never answered with the
recommended option on time-out; the AI is told they began an answer
and ends its turn. On the page, what was half typed in a question's
line survives the question being drawn again (another question came,
the AI read an answer), with the cursor; that question stays first in
the strip; a narrow window lights the stage tab instead of taking it
from under the user's typing; Enter that picks an input-method word no
longer sends a stage answer line.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Gemini 3.8 Flash took seven exam tasks through Antigravity (107 calls, 19
failed, every task done). Nearly all failures were a tool's first use:
empty or guessed arguments. So:
- An argument error lists what the tool takes ("set_prompt takes: prompt
(needed), role, node_id, mode"), and the next call is right.
- One argument name per thing: set_prompt's text is prompt (as in
run_workflow), open_workflow's name is workflow (as in run_workflow).
- A wait over 55 s waits 55 s instead of failing; search_models takes a
model type in the singular too.
- An unread reply older than two hours no longer draws a new chat's calls
into its old task (a two-day-old mark took the first exam task).
The exam doc gets the setup, the Antigravity command line and the results.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Antigravity, Claude Code, Cursor and VS Code load tool descriptions on
demand themselves; looking tools up through Anomalous (tool_help, use_tool)
only doubled that: in one Gemini 3.8 Flash run 18 of 29 calls went to it,
while full-set chats had none. tool_help, use_tool, action_catalog.DAILY,
ai_gates.tool_set and the Gates row go.
An app still holding an older list is now told apart by the list it fetched
last (a fingerprint per app in anomalous/ai_tool_lists.json), so the note
under the task header also covers an Anomalous update.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- An AI app calling by an older tool list (tool_help or use_tool not offered,
use_tool for a tool offered directly, a tool the list lacks) is told so, and
the task header shows how to reconnect until the app fetches tools/list again.
- add_lora / set_model compare the base models on the model line
(model_advice.py; the page reports the line) and add base_model_note.
- get_model gives author_examples (sampler, steps, CFG, size, a LoRA's weight
from the Civitai record's example pictures) and the version name.
- run_workflow says ComfyUI's refusal as which node, what is wrong and the tool
that fixes it (run_refusal.py; mcp_run.js passes the refusal as data).
- Gates "When I don't answer": wait for me, or go with the recommendation once
the waiting time runs out (choices with a recommended option only), handed to
the AI in the same call and shown on the stage and under Decisions.
- A tool budget test (count, size, description length, required arguments) and
the exam tasks with their scorecard (docs/architecture/tool-exam.md,
tools/tool_exam.mjs). The tool set's text recommends Full for Antigravity.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The user took an option for something to select and then pressed "skip"
to confirm; boxes shown open read as fields to fill.
- Questions (stage panels, the strip over the message box): options are
buttons with an arrow under "One press sends it to the AI" and show the
pressed one at once; "Let the AI decide" moves to the question's top
right (the panel title for ask_user); a line of one's own is folded as
"None of these, I'll say it…". The Gates card folds its reason the same.
- A waiting task is one page: the sentence for the AI leads and sticks
under the task header while the page scrolls (the workbench measures the
header); "Say what to do" and the gallery are optional folds that say
how many pictures are in. The sentence has the AI do what the task says,
asking only when it says nothing. The Get started / Put ready tabs and
their words are gone.
- Gates window texts fixed (canvas covers building, lines are over the
message box, "Ask me" named as on the button); run_workflow waits up to
55 s by default, so a quick run comes back with its pictures.
- The UI map counts words the backend names (label_key) as used.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Old toolbar, scan-wizard and workspace labels and a few hints, in both languages; each
checked against the web files, tests and the backend (no t() use, no key built in a
template, no error message used as a key). The UI map's scanner read a quoted string with
an escape ('\n') wrong and missed the key after it on the same line (delNote is used).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- The 54 screen words that said 配方 / recipe for a saved workflow say 工作流 / workflow
(the Workflows page, saving and exporting, the activity log, backups, the gallery's keep);
配方包 / recipe package stays the .zip given to others. README and the MCP guide too.
- The Workflows page's empty guide drops its fallback texts: translate() returns the key
when a word is missing, so they were never shown and still said 配方.
- The glossary retires 配方 and recipe; a retired name may list where it is not that name
("配方（配方包、搭配方案除外）"), and {placeholders} do not count.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- tools/ui_map.mjs find <words on screen | key | /anomalous/route | file.js>, read from the code
each time (tools/ui_map_scan.mjs: the words, what each web file shows and calls, its CSS
classes, imports, the backend's routes); build writes the whole map to the untracked
docs/ui-map.md, with the routes no page calls and the words no file uses.
- docs/architecture/ui-flows.md: the hand-written part for the AI workbench (what each part
is for, what a click does, why) and a glossary of one name per thing. The structure check
fails when a word in it is no longer on screen, a file or route it names is gone, or a
retired name is back on screen.
- One name each on the AI page: 素材 (not 材料), 留言 (not 消息), 节点结构 (the / command
too), 连接 for the last AI, 副本 for the task's tab, 撤销; English "Node map".
- AGENTS.md: look a button up with the map before changing it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
In one real task the AI built an inpaint branch with 9 add_node and 16
connect_nodes calls. add_nodes takes the whole group in ComfyUI's API form, the
format models know best:
- links inside the group, and to canvas nodes as "#12", with outputs by slot
or name;
- a node's wire_to feeds canvas nodes the way add_node does.
All of it is checked before anything changes. The page lays the group out as
one block in a free spot (canvas_build.placedBlock) and applies it in one
transaction: one Ctrl+Z step, and a failure takes every new node out again
(applyWires now takes several nodes). The answer maps the AI's keys to node ids
and lists what is spliced and what is still unwired.
A splice now uses only an input that nothing else in the plan feeds. A node
whose latent comes from the group, or from wire_from, replaces the sampler's
empty latent instead (replace:true) rather than being fed twice. It is in the
daily tool set.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
GPT's auto-wiring (4afee83) was checked on the isolated test ComfyUI. A splice behind
a node, a new branch, the branch choice and the loop check each took one call. A
ControlNet still failed both ways, because its positive and negative have the same
type as a sampler's. A LoRA after a checkpoint wired only its MODEL line.
- Same-type sockets of the same name pair up (positive to positive).
- wire_to into an input that is already fed puts the new node in front: the old
source now feeds the new node's matching input. A ControlNet before a KSampler
is 6 links in one call; a LoRA before the sampler and both prompts is 5.
- insert_after splices every type a node passes through.
- add_node errors call the new node #new instead of #-1 and say how to pick sockets.
- frontend.md: the canvas_wiring paragraph moves from Verification to the
mutation rules.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- Each run keeps its shape read from the prompt ComfyUI queued (ai_run_map.py): boxes by role
with their settings, typed links, plumbing wired through, and what changed since the task's
run before. The run step shows a one-line summary, the changes as chips and the boxes (top
down, fitted to the column, each locating its node); show_run_map puts it on the stage.
- By default the AI works in a tab of the task's own (ai_task_canvas.py, mcp_task_tab.js): the
open workflow is copied before its first change, the tab is brought back (a click on
ComfyUI's tab) before later changes, reads and runs, and the user's tab stays untouched.
Per-task switch in the header, default in the Gates window; the AI is told once.
- Output pictures can be handed to the task an AI is on (gallery card menu, the inspector's
✦), with the user's words: the AI gets them attached with what made them (ai_pictures.py).
search_images(with_pictures) attaches the first six.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- A stop button on a running run (timeline, and beside the run-status button), and stop_run for
the AI: only this run's prompts leave the queue or are interrupted; the step ends "cancelled"
and a user's stop reaches the AI as a "stop" message (ai_run_control.py).
- The run's progress carries ComfyUI's console since it began (the bar's note, models loading),
time left and per step, and VRAM; once queued, each sampler's working size is estimated from
the prompt, with warnings for an upscaled picture sampled again, a size past the card and a
huge decode, for the user and in the AI's run news (ai_run_watch.py, ui_ai_run_live.js).
- The materials picker opens as a side panel that leaves the page usable (the full-screen
browser makes room), expands to the large window and back, and remembers the choice.
- Esc closes the run-status card only (data-anomalous-escape).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
find_nodes looks up this ComfyUI's node types (inputs, outputs, choices)
from its own /object_info. new_workflow opens a new tab, blank or built and
laid out from a workflow in API form, never touching the open one;
add_node, connect_nodes, set_values and remove_nodes edit it. Everything is
checked against the nodes before the canvas changes (types, loops, settings,
choices with close matches); each is one Ctrl+Z step, logged as the AI's
and shown under its workbench step. Receipts now show rewiring ("wired":
what feeds an input, before and after). A run ComfyUI refuses reports
ComfyUI's own reason instead of "Prompt execution failed".
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
ui_tts_import.js gives its frame and toolbar to ui_tts_import_screens.js and its window
drop to ui_tts_file_drop.js; material_inspector.js gives the PNG chunk reader to
png_metadata.js. 221 texts no code uses any more leave locales.js (both languages).
No behaviour changes.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A download by name says where the list's file is; an entry of no stated type no longer
fits a node of known type; without the Manager, the page says it would be one more
source. docs/model-check-rules.md gets a section on the Manager.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Keep -> Its combo now saves the nodes upstream of the picture's samplers' model and prompt
inputs (loaders, LoRA chains, CLIP skip, both prompts) as a node-structure combo, values
untouched; models are named by where they are here, missing ones kept by the picture's name.
A workflow without a sampler still gives the simple combo.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The combo editor gets a head with its picture (See the picture), a large name with a
Rename button, chips of what it holds and Put on canvas; models show only what is chosen,
Change / Add open the grid of compatible ones. No Save button: everything saves itself
(a tag changed in the bilingual view now saves too). The gallery's ★ Kept lists only the
pictures kept as a workflow, combo or prompt.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A bar over the canvas finds combos, models and saved prompts as you type
(GET /anomalous/quick_index lists the models in a few words) and puts the chosen one
where the pointer was. Also in the empty canvas's right-click menu, now registered.
Saved model paths in a linked model folder get their real place.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A node's menu offers Open in AMB, change model by picture, insert a LoRA before / after,
saved values, the model's pictures and details, save values and save as combo, each where
it applies; the empty canvas's menu opens AMB's pages. Entries can be switched off, moved,
and put in one AMB submenu or right in the menu. The model picker's Esc closes only the
picker, wherever the focus is.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The card shows the output image with the main model's cover as a badge; the name is the
first LoRA (else the main model) and two telling tags; the toast after keeping shows the
picture and renames the combo on the spot, and the editor's title renames on a click and
opens the picture.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A user found model folders cluttered after scans: four files per model.
model_sidecars.py now says where the plugin's own side files are (.info,
.anomalous.json, .civitai_bak.*): beside the model, or in the folder's
hidden .amb (Settings -> Models -> Files that come with models). Every
reader and writer goes through it and reads both places; covers stay
beside the model for ComfyUI and other plugins.
Scans no longer keep a copy of Civitai's picture identical to the cover:
the cover is noted as Civitai's (SHA-256 in the .info) and a copy is kept
aside only while the user's own cover shows. api/sidecar_layout.py moves
existing files and drops the duplicates to the Recycle Bin. The folder
tree leaves .amb out; the duplicated tree builder and cover owner check
are shared now (model_catalog.py, backup.py back under 600 lines).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A user asked to pick tags from a saved prompt instead of pouring the
whole card in. A clicked card now opens between the cards and the prompt
being edited (ui_prompt_reference.js): a click on one of its tags adds
that tag, Add all the whole card, ticks show what the box has. The
preview's Add and dragging a card work as before; beside the canvas the
drawer widens by the column while it is open.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
With many models, or models on a network drive, every visit to the scan
page read every folder and, after a restart, every new model's header
again (79 s for 1041 models on an AutoDL model disk).
api/model_heads.py keeps what each file's header says by real path, size
and time in user/anomalous/model_heads.json (at most 50,000 files, the
longest unused dropped first, written a few seconds after the last
change, only once setup_routes called keep_on_disk); get_metadata reads
headers through it, so a header is read once per file version.
scan_summary keeps the last summary (memory, and scan_summary.json when
it changed); ?quick=1 hands it out at once and the page shows it with
"showing the last counts while they are updated", then the fresh ones.
Requests meanwhile await one count (_fresh). The models found on Civitai
(components apart, at most 5000 kept) are listed by GET
/anomalous/scan_matched; the page's "Found on Civitai" count opens them,
with Civitai's name, base model, a box to find one by name and Open.
Checked on the isolated ComfyUI reading the real loras and checkpoints
read-only (scans stubbed; no file there changed): the last counts showed
at once also after a restart, the fresh ones replaced them, the found list
had 68 models with names, the box filtered, Open and Back returned to it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Before an online scan (the scan page, a card's radar, after a download or
an import), scan_connection.js asks GET /anomalous/civitai_check, which
tries both of Civitai's addresses. When neither answers, a dialog says
why and offers Scan from the files (the offline scan, so nothing waits on
every model), Set a proxy (Settings → Models, Network) or Cancel. A check
that reached Civitai is trusted for five minutes. anomalousConfirm takes
okClass, so this choice is not drawn as a delete.
A model that is a link with no info beside it shows the info beside the
file it links to (metadata._info_base): A1111's Civitai Helper leaves
.civitai.info there. Only read there; covers, edits and scans stay beside
the link.
Checked on the isolated ComfyUI with the check and the scan request
stubbed in the page: the dialog's three choices (from files posts
offline_only, set a proxy opens the Models tab, cancel posts nothing),
an answering Civitai scans online without asking, and a second scan
skips the check.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This reverts merge 25f1f6d (feature/next's 851a8cb and this branch's
adaptations of the scan choice, the editor's covers, Tidy's move and
move_into). The owner keeps the cloud-server work in 1.61
(feature/remote-access); 1.60 is back to 4bc559e.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
1.60's own places that took a model file by name use the new helpers
too: the scan choice (model_file_within), the editor's covers, Tidy's
move (model_reference_within). Opening a folder keeps 1.60's "no file
name opens the folder". move_into tells a rename from a copy by where the
link is (lstat), so Tidy moving a linked model renames the link instead
of copying the model disk's file onto the server. The open-folder test
stubs through path_utils and also refuses "..".
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
On an AutoDL server every model in the models folder is a link into the
platform's shared, read-only model disk, named like 6b4679... without an
extension. Scans wrote the .info beside the link, but get_metadata
resolved the path and read beside the disk's file: a scanned model
showed as not scanned, under the disk's file name. Delete, edit, open
folder, MCP and saved-workflow lookups resolved the file too, so they
refused it as outside the models folder (or, for a link inside it, would
have acted on the file behind it).
path_utils.model_file_within joins a resolved folder with a bare name
(require_filename, which now refuses "." and ".." too);
model_reference_within resolves the subfolders of a relative path inside
the root, through links as before, and takes the file as named.
get_metadata reads beside the path it is given (abspath). Edits, renames
and deletes act on the link; nothing writes into a model file.
Checked with a link stood in by patching realpath (this computer cannot
make links): the info beside the link is read and the name is the
link's; "..", ".", "a/b" and a relative path leaving the root are refused.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
On an AutoDL server civitai.com timed out while civitai.red, Civitai's
other address with the same API, answered in about a second, so scans
there fell back to guessing from the files.
civitai_client.api_hosts gives both addresses, the one that answered
last first; the scan's lookups, import's and Model Check's
(download_sources._civitai_json) and the network test ask the other one
when the first does not answer at all. An HTTP error is Civitai's own
answer and is not asked again. The test answers which host answered,
and Settings says so when it was civitai.red.
The pictures stay on image.civitai.com, which such networks often cannot
reach either: after COVERS_DOWN_AFTER (2) covers in a row that would not
download, a scan fetches no more covers instead of waiting on each.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
On a cloud server with 1041 models not scanned yet, on the platform's
network model disk, GET /anomalous/scan_summary took 79 s, each time the
page opened, and the page stayed blank meanwhile: it read each new
model's header (for "looks like no drawing model") one after another.
The headers are now read 16 at a time (HEAD_READERS) and through
get_metadata, whose cache keeps each answer until the file changes, so
the page opened again reads none of them and the picker after it finds
them there. Opened afresh, a count slower than 400 ms shows "Counting
your models…" instead of a blank page.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Cloud users had to find the plugin folder and make allow_remote_changes
themselves. GET /anomalous/access now also answers another device with
"command": the line that makes the file (touch, or echo off > on Windows,
which works in cmd and PowerShell), or removes it when it is there, with
this server's real path, quoted. The chip's dialog shows it with a Copy
button; when the browser refuses the clipboard the line is selected for
Ctrl+C. The line only runs where someone pastes it, on the server.
anomalousAlert takes options.copy; copyText moved from ui_share.js to
ui_dialog.js so both use it.
Checked behind the stand-in gateway: without the file the dialog offers
the line that makes it, with the file the line that removes it; both
Windows lines tried in cmd and PowerShell.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
ComfyUI on a rented cloud server (AutoDL, RunPod) is only ever opened
from another device, so since v1.58.1 the plugin was view only there.
An empty allow_remote_changes file in the plugin folder, made on the
server and read once at start, now makes local_guard.may_change true for
every request; the middleware and the routes that check for themselves
(import, placement, scan choice, recipe focus) use it. is_local stays for
import's search of this computer's Downloads and for MCP. The page cannot
make or remove the file, and the file is git-ignored.
GET /anomalous/access answers {local, changes}. The header chip is now a
button: "view only" says what is refused and the two ways for cloud users
(SSH tunnel, the file); with the file it reads "remote changes on", in
yellow, and says that anyone with the address may change things.
docs/guides/cloud-server.md explains both ways and the risk; README,
AGENTS.md rule 4, backend.md and the CHANGELOG (v1.61.0) say so too.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The pattern behind issue #22 came back once already (removed in v1.57.2,
back with the icon rail in v1.58.0). tools/check_structure.mjs now fails
on any :has() in the web stylesheets whose argument matches [style...],
pointing at the line; AGENTS.md, docs/guides/testing.md and lessons.md
say why and what to do instead (a class on an ancestor).
Prepared in a cloud session (checkpoint cbfc782), committed now that the
issue #22 fix is on this branch.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
