models.js's ~2700 lines were entirely built around v4's Studio "roles" and
"candidates per role" concept, neither of which exists in v5 -- the
pipeline is a flat, swappable slot list, not a fixed named-role set.
Replaced with a renderer over the live slots (GET /api/pipeline) grouped
the way the pipeline itself groups them (Loaders/Preparation/Sampling/
Render), each slot's real widget inputs read live from GET /api/nodes and
edited through the same /api/pipeline POST. The "not ready to generate
yet" / incomplete-inputs list is now real and visible, not implied.
Node add/remove/swap/wire is NOT built here -- v4's graph-editing machinery
assumed nodes that no longer exist, and a real editor is its own piece of
work; this establishes the actual working file-picker surface first.
Extracted `pipeline_state.js`: engine_settings.js and models.js both edit
the SAME live pipeline, and each independently caching its own `slots`
copy (as Engine Settings' fix earlier this session did) would let one
panel's save silently revert the other's un-saved edits the moment the
user switched between them -- the exact stale-state bug already fixed
once this session, about to recur in a second file. One shared module now
owns `slots`/the manifest/the save queue; both panels are thin renderers
over it. engine_settings.js refactored to consume it (no behavior change,
confirmed live).
Live-verified end to end: picked real model/clip/vae files scanned off
disk, watched "incomplete" clear field by field down to "ready to
generate", and got back a genuine queueable ComfyUI prompt graph. Then
confirmed the cross-panel fix specifically: saved an Engine Settings
toggle afterward and the POST still carried every file pick Models &
Pipeline had made -- neither panel clobbers the other.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Round-2 review: the reopen-persistence fix scanned every string input on
every slot for anything that JSON-decodes to a plain object, with no check
that it's actually the real settings blob -- a coincidental match (any
future node with an unrelated JSON-object string input) could inject a key
under a module id that doesn't exist, which then round-trips forever
through every future save via place()'s blind overwrite. No route exposes
the real sink's (node, input) identity to the client to look it up
properly (core doesn't name an implementation, by design), so the fix
narrows the heuristic instead: a decoded key is only merged when it names
a module this session already knows is installed. Not reachable with any
real module today (verified), but was a live hazard for the first one that
collides.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
engine_settings.js's ~1100 lines were entirely hardcoded around
FunPackStudio/Chain Sampler fields that don't exist in v5. Replaced with a
renderer over the real module manifest (GET /api/modules): groups every
loaded module's settings by category, builds a typed control per setting
(bool/int/float/enum/color/text), honors `when` conditional visibility, and
saves through the real /api/pipeline POST -- the same settings_sink
mechanism core/graph.py already had, just never had a caller. Outer
registration/DOM shell (SettingsWindow section, EngineSettingsModal.open/
close) kept identical so every existing entry point still works. Added
three new real api.js functions (modules, pipeline, editPipeline) -- v5-
native capability, no v4 equivalent existed.
Adversarial review found and fixed two real defects before this landed:
(1) every reopen re-fetched the (stateless) server default and reseeded
from schema defaults, silently discarding every previously-set value
across every module -- fixed by only fetching once per session and
holding `slots`/`values` in memory across mount/unmount, since the server
itself remembers nothing between GETs; (2) a second edit made while the
first's save was still in flight was dropped rather than queued -- fixed
with a trailing-edge retry so `commit()` re-sends after the in-flight
request resolves if anything changed meanwhile. Also fixed api.js's error
reader missing /api/pipeline's one inconsistently-named 400 field
("refused" instead of "problems"). Live-verified: toggled settings on two
different modules across two categories, closed and reopened the panel,
both survived; the real graph's FunPackModifierSettings node's `settings`
input genuinely updated end to end.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
The port plan calls for every module under modules/ to declare a
CATEGORY so Settings/Engine panels can group by it. Only alg, sharpen,
appearance, and full_control had one; the other 12 (loaders, empty
latent, media load, minimax_h3, output decode, sampling modifiers/
sampler, system pipeline) are plain infrastructure with no user-facing
settings dial, so they land in "system" except output_decode (post,
matching its STAGE) and the two sampling modules (sampling). Verified
against core/tests (548 passed) -- category validation and the
registry's field-mapping catch a typo here, per the earlier fixture-gap
fix.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Round-1 adversarial review found the resolve-based unsupported() sentinel
broke every existing try/catch in v4's ~60 caller files (they were written
for a rejecting promise, same as any other failed fetch) -- worst case was
the Models & Pipeline panel hard-crashing on open, the one caller api.js's
own comment claimed was safe. Switched unsupported() to reject; the
existing defensive code across the callers now does exactly what it always
did. Also fixed a self-inflicted double-unwrap in listProjects/listMedia/
shortcuts (api.js unwrapped {field:[...]} to the array, but every real
caller still destructures .field off the result -- onboarding's "resume a
project" list was silently always empty), and switched five stubs that
threw synchronously (escaping before a caller's .then/.catch chain could
even attach) to async so the throw becomes a normal rejection. Live-
verified: Models & Pipeline now shows "Could not load" instead of
crashing, "Continue with existing project" reappears in onboarding,
Export Settings shows its real failure message instead of hanging.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Three buckets per the port plan: (a) same-shape routes just renamed
(projects, media, shortcuts, packs, git/system, temp, log), (b) a small
adapter where v5's shape differs (parsePrompt -> /api/prompt/expand,
resultUrl -> ComfyUI's own /view), (c) no v5 equivalent yet -- refinement
keys, REINS, probes, render/stitch, Models/pipeline-roles, Generate's
actual queue flow.
Every stub resolves to {unsupported, reason} instead of throwing, so a
module nobody has built yet can't take a caller down with it -- the
"feature-independent" backend the port asked for. Live-verified: wizard
through project creation into the real editor, zero console errors, every
real route round-tripping 200 OK.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Two real findings from adversarial review of the previous commit (the v4
UI port's route change), both fixed:
- app/tests-layout/*.spec.js (20 files, 133 tests): every one of these
navigated to /funpack/ and asserted on composer-specific DOM
(.cx-workspace*, window.FunPack) that no longer exists there now that
the ported v4 frontend is served at that URL instead. Left as-is they
would silently time out on every run. Each file now carries an explicit
test.skip(true, "...") with a comment explaining why and pointing at
the plan's pixel-exactness step, which is what will eventually rewrite
these against the new DOM -- declared and tracked, not quietly broken.
- core/tests/test_registry.py's fixture-based tests never exercised the
new category field through the real announcement_of()/scan() path --
only test_schema.py's direct validate() calls did, which bypass
registry.py's FIELDS name-mapping entirely. A typo there (e.g.
"Category" instead of "CATEGORY") would have passed every existing
test. The "good module" fixture gained CATEGORY = "guidance" and
test_the_good_module_loads now asserts on it, closing the gap.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Two pieces of the "port v4's exact UI onto v5's backend" plan:
- core/contract.py: new optional ModuleSpec.category field (closed set:
continuity, guidance, conditioning, sampling, post, system), validated in
core/schema.py, threaded through core/registry.py's FIELDS and
to_manifest(). Presentation grouping, independent of STAGE (execution
order) -- a module can be stage="sampling" and category="guidance" at
once. Only the 4 existing modules that actually render a settings panel
(alg, sharpen, appearance, full_control) got a category assigned; the
other 12 have none to categorize (no SETTINGS block) and stay
uncategorised, which is valid.
- v4's frontend (movie_editor/frontend/*, ~29k lines, 61 files) copied
verbatim into app/legacy/. core/routes.py's "/" route now serves
legacy/index.html instead of the composer entry point; a <base> tag in
that file resolves its own relative script/link tags against
/funpack/app/legacy/ regardless of what URL it's served at, so every
file moved as a unit with zero per-file path edits. Verified live: all
61 assets load 200 OK, the real v4 onboarding screen renders pixel-for-
pixel (it's the actual DOM/CSS, not a reimplementation), and several v4
API calls (/api/media, /api/projects, /api/git/status) already succeed
against v5's real backend by coincidence of shared route naming -- the
rest 404/405 as expected, since api.js's ~90 functions haven't been
rewired to v5's actual routes yet (next piece of this plan).
The old composer entry point (app/index.html, app/composer/**) stays in
the tree, unreferenced, until a later cleanup pass once nothing imports it.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Round 1 + round 2 of adversarial review on the single-track timeline
(2092c8b) found and fixed:
- Clip position now stays absolute (seconds*pxPerSecond) instead of the
MIN_CLIP_PX floor cascading into every later clip -- a floored short
clip no longer pushes everything after it out of sync with the ruler.
- pendingSeek is cleared unconditionally right after project.select(),
closing a gap where a no-op select() (clicking within the already
selected clip) left it dangling for a later, unrelated draw to
wrongly replay.
- draw() now always resyncs the playhead to the current scene's start
on any non-track-click selection change, rather than only when the
old value fell outside the new scene's span -- the old check let a
stale playhead survive a wholesale project switch whenever it
coincidentally still fit the new project's first scene.
- The timeline moved from the centre column's own inner split to an
outer split above/below the whole three-column workspace, matching
v4's actual layout (full-width timeline under resizable columns,
not squeezed inside one of them).
- That restructuring put the Properties column under real height
pressure for the first time, exposing a latent bug: .cx-stack's
base rule set min-block-size:0 on every stack, letting a plain
grouping stack (not meant to scroll) get flex-crushed below its own
content height, so its overflow painted over the next row ("Settings"
over "Loosen the starting image"). Removed from the base rule --
.cx-stack-fill already carries it for the one case that should
actually give up height to scroll.
- Documented, rather than "fixed" by reintroducing the rejected
cascading-position bug: several under-floor clips in a row still
visually occlude each other (later DOM paints over earlier). Click-
seek still resolves correctly since it reads time, not paint order.
- Guarded framesOf() against a negative/NaN scene.length the same way
fps() already guards its own input.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Replaces the scene-index-tick ruler with composer's new track.default
element: clips sized by real duration (frames/fps, not equal-width
boxes), a ruler in genuine elapsed seconds, and a playhead that follows
the selected scene but can be dragged to any exact second within a
clip via click-to-seek.
Also fixes a real crash found while live-testing this: addScene()
built a scene object missing source_image/references, which crashed
the Inspector the moment it tried to draw the new scene.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Groups the pipeline window into a card grid with a "+ Add group/node"
card beside it -- matching v4's card-per-loader layout instead of a
separate toolbar button. Picking it offers a new empty group or a node
outside any group (landed in the existing "Other" bucket).
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
GGUF: port v4's gguf_support.py (backend selection, content-based .gguf
detection, dequantize fallback) and wire it into both the diffusion-model
and CLIP loaders.
Pipeline editor: reopening the Models-and-pipeline settings window was
silently discarding the whole edited pipeline (a loaded preset, every
loader file, every sampler setting) because GET /api/pipeline is
deliberately stateless server-side and boot.js never fed its own
already-saved slots back into a freshly remounted editor -- found live
while walking through the MiniMax-H3 setup flow. Fixed by re-validating
the saved slots through check() on reopen instead of always replaying
the raw stateless default.
Also fixes, found via the extensive_testing adversarial loop over both
of the above: a stale file-listing cache in the pipeline window (no way
to see a newly-added model file without a full page reload), and a
chain of races/leaks in its own refresh/commit error-banner handling
that surfaced across five review rounds -- a rejection from a superseded
request overwriting a fresher success, a failure banner never clearing
on success, banners bleeding across groups on navigation, and a
commit()'s refusal landing on the wrong view after the user backs out
before it answers.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
keep_proportion/extra_padding/invert_crop (ImageTransformKJ) and format
(SaveVideo) were wrapped in a redundant {"field": "value"} dict instead of
the plain string ComfyUI's own DynamicCombo expansion expects at the wire
level -- every one of them silently failed graph validation and fell back
to the node's own default, undermining the "total_pixels" resolution-fit
behavior this preset exists for. Found by loading the preset in a real
browser against real node schemas; the dev server never had comfy_extras
or KJNodes registered before now, so this was invisible locally until
tools/devserver.py also gained real node registration (init_extra_nodes +
KJNodes' load_custom_node, matching what the pytest `registered` fixture
already did).
The two pipeline_preset tests that were supposed to catch this had too
broad a filter -- it excluded ANY "is not one of" message as an assumed
"combo not chosen yet" state, which in this codebase actually shows up as
"needs X and nothing fills it" instead. Tightened to exempt only the three
loader-file combos genuinely left unset on a fresh install.
Verified live: preset loads with zero combo-validation problems, only the
three expected unset-loader-file messages remain, and Generate correctly
refuses with an honest reason rather than a false success.
wireReferences() (new, extracted from boot.js's queueInputs) now reports an
unwired count instead of silently losing a reference when: the pipeline has
fewer reference slots than the scene has references, two references' roles
resolve to the same slot or downstream destination, or a reference's slot
was already claimed by the scene's source-image assignment. boot.js toasts
whenever unwired > 0, worded to cover both the capacity and collision cases.
None of these are reachable through the one shipped preset (H3's R2V, one
role per slot) -- all are latent contract violations a hand-edited or future
preset could trigger. 10 new tests in reference_wiring.test.js.
The last missing piece connecting the backend preset mechanism to something
a person can actually click: pipeline.js gets a presets() fetcher matching
load/describe/check/search's own shape, and pipeline_window.js's group-card
index offers a "Load preset..." button (only when a caller passes one) that
lists what /api/pipeline/presets returns and replaces the pipeline through
the same commit()/check() path every other structural edit already goes
through -- a preset that does not actually build is refused here, same as
a hand-drawn bad wire would be, never shown as if it had loaded.
Writing the test for this caught a real bug before it shipped: open()'s own
destructuring silently dropped `presets` before ever reaching mount(), so
the button would never have appeared no matter what boot.js passed it.
Pieces B and C: Scene gains source_image (aspect-ratio reference only --
resolution itself stays a project.video setting, ImageTransformKJ's
"total_pixels" keep_proportion fits the picture's aspect into that budget)
and references (media ids, in order). Inspector's Scene tab gets pick/clear/
remove controls; picking routes through the existing Media bin via a small
pending-pick handshake in boot.js (one bin, one onPick, so "which field
asked" has to be tracked between the click that asked and the click that
answered).
References wire into MiniMaxH3ReferenceToVideo's ref_images autogrow input
dynamically, per run, only for what a scene actually has -- an unwired slot,
not an empty image, is what "not in use" means to that node. The target is
data on the role itself (wireTo), not a slot id boot.js would otherwise have
to know, keeping the "core never names an implementation" contract intact
one layer up into the app.
Verifying this exposed a second real core gap (after the MatchType one):
core/graph.py had no idea what an Autogrow (V3's io.Autogrow) input even
is -- "ref_images" reported as one opaque COMFY_AUTOGROW_V3 slot, its real
per-instance sockets (ref_image_0, ref_image_1, ...) invisible to
declared()/build(), so a wire naming one was refused as an input the node
does not have. Fixed generically in comfy_types.py (autogrow_instances())
and graph.py, confirmed against ComfyUI's own _io.py (finalize_prefix's
dotted join, build_nested_inputs's runtime un-nesting) -- not guessed, and
proven end to end with a real override()+build() test using the exact edit
boot.js makes.
Piece A of the H3 workflow port: the loader/SigmaShift/ImageTransformKJ/
R2V/sampler-at-CFG-1/decode/video-mux chain the H3 user runs by hand today,
offered as a preset (GET /api/pipeline/presets) loadable into the same
editable pipeline core/graph.py already supports -- not a second "default",
which module-load order would make ambiguous.
References are deliberately NOT pre-wired into R2V's ref_images autogrow;
an unwired optional slot is what "not in use" means to that node (absent
from its ref_images dict), not an empty image degrading its <Picture i>
numbering. boot.js wires them per-run based on what's actually picked --
next commit.
Building this against a real ComfyUI+KJNodes registration (not a fake
schema) surfaced a genuine core bug: core/graph.py's from_comfyui() had no
idea what to do with a V3 MatchType socket (ImageTransformKJ's own image
input) -- its bare type name "COMFY_MATCHTYPE_V3" matched nothing, so a
completely legal IMAGE wire into it was refused. Fixed generically in
comfy_types.py (match_type_union()) and graph.py, for any node using
MatchType, not special-cased to this one.
The bridge piece B and C (resolution source, reference images) both need:
Media bin picks have no way into a ComfyUI graph today. One node, media_id
in, IMAGE out, backed by core/media.py's existing store.
Also fixes a real .gitignore bug found along the way: a bare "media/"
pattern matched modules/media/ (this node's own source) as well as the
intended root-level user-data directory. Rooted both media/ and projects/.
A variable referencing the same other variable more than once is not a
cycle, so memoization alone still leaves the OUTPUT exponential in size
(each level's re.sub concatenates the memoized child result multiple
times). Adds a shared output-size budget, checked before using either
a cached or freshly-computed result.
Also ignore the runtime shortcuts.json data file, and fix the fan-out
test's assertion, which was checking against a vacuous condition.
Round 4 of adversarial review, both reproduced live:
1. removeButton's mousedown-only wiring (round 3's own fix) dropped keyboard
activation entirely -- Tab-to-focus, Enter/Space did nothing, silently.
onClick is wired back alongside the mousedown handler; a real mouse click
still fires both, but the removal it triggers is idempotent (see below),
so the second call is a safe no-op.
2. Deleting a row that ITSELF still has an uncommitted edit could corrupt a
DIFFERENT, already-saved row. removeButton's own synchronous redraw
detaches whatever field still has a pending edit; a browser fires `blur`
on a detached-while-focused element even with preventDefault() called on
the mousedown, and that reentrant commit -- firing INSIDE the very
setVariables() call still removing a row -- was still indexed against
the row's OLD numeric position in an array that had already shrunk by
one underneath it.
Both share one cause round 3's fix narrowed but didn't remove: every commit
and removal in this editor addressed a row by the numeric index a render
closed over, and a synchronous redraw can invalidate that index before a
still-in-flight commit reads it. Fixed by addressing rows by the row OBJECT
itself, found by reference in the CURRENT array at the moment a commit or
removal actually fires -- a row no longer there (already removed, whether by
the user's own click or a reentrant one) is a no-op instead of a slot number
that now means something else. Confirmed with two new Playwright tests:
editing then removing the SAME row leaves the neighbour untouched, and
Enter/Space on a focused ✕ removes it.
Round 3 of adversarial review, reproduced live: type an uncommitted edit
into one Variables row, then click a DIFFERENT row's own delete button
without blurring the first. The click's mousedown blurs the edited field,
which commits and triggers a full non-keyed redraw of the whole list
(inspector.js draw() -> node.replaceChildren()) -- destroying the delete
button mid-gesture, so the click that started on it never arrives. No
error, nothing removed, no sign anything happened.
Fixed with the same mousedown+preventDefault idiom this session's own
autocomplete.js already uses for the same reason: stop the browser's
default focus-move (and the blur/commit/redraw it would trigger) before it
can pull the target out from under its own pending click. Confirmed with a
Playwright test reproducing the exact sequence.
Round 2 of adversarial review found a real bug in the fix the previous
round's own review prompted: mutating the live project.variables array (and
its row objects) in place before calling setVariables() meant projects.js's
remember() -- which deep-clones the project for the undo stack BEFORE
applying the new value -- captured the already-edited state as the "before"
snapshot. Undo after any variable edit or removal was a silent no-op.
Fixed by never mutating what project.variables already points at: every
commit now reads project.variables fresh and hands setVariables a brand new
array, which is also what closes the SAME stale-closure race the in-place
version was built to fix in the first place (redraw-mid-edit still reads
whatever the previous commit already saved, since it reads live rather than
from a snapshot taken at render time). Also fixed core/prompt_build.py's
docstring, which claimed the opposite of what resolve-variables-last
actually does and what its own test already pinned.
Adversarial review found the "picks at random each time" hint in the
Shortcuts editor was false: boot.js never sent a seed to /api/prompt/expand,
so core/shortcuts.py's own no-seed fallback (hash of the literal text) made
the pick a fixed function of the prompt, identical across every Generate on
unedited text. Fixed by drawing a fresh random seed per queue -- confirmed
with a Playwright test asserting two consecutive Generate clicks send
different seeds.
A trigger->replacement library (core/shortcuts.py, a close port of v4's
apply_prompt_shortcuts/resolve_variables), a Settings section to manage it,
autocomplete on the real prompt textarea (composer.autocomplete.default, so
it outranks the Constructor modal it lives inside), and a stateless
/api/prompt/expand that combines anchor + shortcuts + $variables + postfix
at the moment a run is queued -- scene.text itself stays exactly what was
typed. Project gained anchor/postfix/variables editing UI on its own tab.
Along the way: a real layout bug in the Properties column (every composer
stack can flex-shrink below its own content, and once the Project tab grew
past a couple of rows the column started squeezing its own children's
content into each other and, at the outer level, into the sibling Generation
section below) -- fixed with flex-shrink:0 at each level that needs it,
caught by a Playwright regression test.
Alongside the existing negative prompt: anchor (prepended to every scene)
and postfix + postfix_enabled (appended, with its own on/off switch --
v4 proved a postfix worth keeping needs to be turned off for a quick test
without losing it, which clearing the field would do).
Data layer only, deliberately: unlike negative, which maps 1:1 onto an
existing CLIPTextEncode "text" input via a role, anchor and postfix need
to be COMBINED with each scene's own prompt text, and nothing in the
pipeline does that yet -- that is Studio's job, not built. Wiring a UI
field to a mechanism that does not exist yet is the "half-assed port"
this project has already been steered away from once this session.
Also backfilled round-trip and never-trust-a-file tests for the existing
negative field, which had none at the core/projects.py level despite
video already having them.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Every node's own OUTPUTS were invisible on its own panel in the pipeline
editor -- wiring an input already let you pick a source, but opening the
node that PRODUCES something told you nothing about what it makes or
whether anything reads it. Found live, by a real user just looking at a
node, after two full review rounds that only ever checked the consuming
side. Fixed with a read-only "Outputs" section per node, naming each
output, its type, and who (if anyone) currently wires from it.
Also: model family detection. core/probe.py reads a safetensors header
(tensor names only, no weights) and asks every installed model module,
through the same capability-lookup mechanism already used for
empty_latent/decode, whether the file is its own -- core never names a
family itself. A real, adversarially-found bug in the ported signature:
one of MiniMax H3's three candidate prefixes doesn't correspond to any
prefix ComfyUI's own loader actually resolves to, traced through
comfy/model_detection.py and comfy/sd.py -- the probe could have called a
file H3 that ComfyUI's real loader would refuse. Removed, and pinned with
a dedicated test module. A second finding narrowed the folder search to
only the folders FunPack's own loaders actually read from.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Pipeline wiring: core/graph.py gains wire()/unwire(), and the "Models and
pipeline" editor can now actually connect one node's output to another's
input -- sockets and widgets alike, which is what makes "linked inputs"
(several widgets sharing one Primitive node's value) buildable at all,
without a second mechanism. Three adversarial review rounds found and fixed
real bugs along the way: check()'s request builder silently dropped the
wire-specific fields; the widgets loop had no wiring UI at all despite the
backend supporting it, and a widget already holding a link rendered as an
editable box with the wrong (default) value; wiring an input that already
had an unsaved drafted value let the stale value silently win on Save. Also
fixed, found live rather than by review: the Add-node picker stacked two
disagreeing search boxes (an outer server search and filterList's own
redundant client-side one).
Updates window: "Restart ComfyUI" is now a plain, always-available button,
not just something offered when an update already landed. Rolling back or
switching branch on a dirty tree now auto-stashes instead of refusing --
updating still refuses, unchanged, since pulling new commits onto changes
in progress is not something to do without asking. Three review rounds
here too: stashing before reading the rollback target shifted what HEAD@{1}
meant (a real reflog hazard, not just reasoned about); a stash's disclosure
was lost if the step right after it failed; that fix was too narrow and
missed pull()'s own failures inside checkout(pull_after=True) -- the actual
default path -- and rollback()'s own reset failing; and finally, the
broadened fix still only caught GitUpdateError, missing subprocess-level
failures like a hung fetch timing out, which is the realistic failure mode
on a rented GPU box's network.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Files the user brings IN (a reference image, an imported clip) now have a
real store: core/media.py, /api/media routes, and a "Media" tab beside the
bin's "Results" tab in the Assets zone. Upload, list, fetch, delete all
round-trip through the real server; nothing wires into a pipeline input yet,
since no model module or Studio conditioning takes a reference image today.
Two adversarial review rounds found and fixed three real issues along the
way: MAX_BYTES (512MB) was dead code because aiohttp's own client_max_size
capped every upload far lower first in both prod and dev, with a JSON-parse
error leaking to the user on rejection instead of a clean message; the
index.json read-modify-write had no guard against concurrent writers; and
the devserver's bare Application didn't match ComfyUI's own 100MB default,
which also broke an unrelated e2e assertion once the review polluted the
real, unisolated media/ directory testing the fix.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Found by the adversarial pass on the commit that moved it to the Project
tab: it had no persistence and no per-project sync at all. width/height/
length (project.video) get both -- a real dict field on Project, and
syncVideo() re-reading it into the controls on every project switch.
Negative got neither: it lived only in the pipeline-override controls'
own in-memory state, so it silently carried over from whichever project
was open last. Typed for project A, it was sent as project B's negative
prompt the moment someone switched -- with nothing on screen to say so,
and the Project tab's own copy ("Size and length live in the
Constructor") actively implying every field there was project-scoped.
Project.negative is now a real field (core/projects.py, round-tripped
through the existing whole-project PUT with no route changes needed),
and boot.js gained syncNegative(), wired everywhere syncVideo() already
is: onOpen, and after the pipeline changes underneath. Reproduced the
exact leak end-to-end before fixing it, and the new project_video.spec.js
test is that same reproduction, now asserting it doesn't happen.
Three things the unified Settings window was still missing next to v4's
own, all found by looking at it next to v4 rather than in isolation:
- The sidebar is a 52px icon rail now, expanding to a full list on
hover/focus and overlaying the content instead of a splitPane's
permanently-docked two-pane shape -- v4's own sidebar, not an
approximation of it. Fixing this surfaced two real CSS bugs along the
way: filterList's own min-width:220px silently defeated the collapse
(clipping the search box to "Sear" instead of showing an icon), and
.cx-filter-list's unset overflow-x (computed to auto once overflow-y
was set) put a horizontal scrollbar under it. Centering the icon in
the collapsed row needed the label's flex-basis AND the row's own gap
to both go to zero -- an invisible label still claims its share of
both otherwise, and the icon renders off-centre.
- The modal now has a real height (v4's own 660px) rather than sizing to
whichever section is showing, so switching between a short section
(About) and a long one (Node packs) doesn't visibly resize the whole
window. That exposed a THIRD latent bug: .cx-modal-body never had
flex:1 1 auto, invisible until a modal had a real height to fill --
every prior modal shrink-wrapped its content, so head+body+foot always
summed to exactly the card's height by accident. Fixed at the shared
rule, not just for Settings.
- About now shows what v4's does: commit and dirty state, a codename
when the checkout has one, and the machine's real hardware/software
facts (chip, memory, GPU or MPS, disk, OS, ComfyUI/Python/Torch/
attention-backend versions, hostname) -- a new core/sysinfo.py
(ported from v4's own, individually guarded so a missing probe never
takes the panel down) behind a new GET /funpack/api/system, injected
into settings_window.js the same way every other network call here is.
Also: a nav row's tone (the coloured icon chip) is now validated against
its closed set the same way button.js already validates its own tone --
an unknown one used to render silently wrong instead of failing loudly.
Most FunPack pipelines run at CFG 1, where a negative prompt does
nothing -- it was one more box beside the prompt every Generate reads,
for a field only worth touching occasionally. The reference pipeline's
negative role now targets "project.negative" instead of
"generation.prompt"; the inspector's Project tab hosts it, in a STABLE
node outside the part draw() tears down and rebuilds on every scene
selection, so a field mounted there survives normal use.
An outside click on the Settings window closed it unconditionally --
pipeline_window.js's own standalone modal has always refused to close
that way (a group edit's draft is the only copy of it), and hosting it
inside Settings silently dropped that guarantee, with no way left for a
real user to reach the protected version (boot.js has no other caller
of pipeline_window.js any more). Settings now holds the door the same
way, for every section: an accidental backdrop click during any
half-finished edit no longer costs it.
filterList's icon `tone` claimed to be a closed set, the same rule
el()'s style refusal already holds -- but nothing checked it, so a
typo would have rendered a silent, unstyled chip instead of failing
loudly the way button.js's own tone check already does for the same
mistake. Now it throws, matching that sibling pattern.
A bare glyph character had no framing at all; filterList's icon now gets a
coloured chip, drawn from a fixed tone (accent/danger/warn/good/neutral) --
never a raw colour from data, the same rule el()'s style refusal already
holds everywhere else. Each Settings section picks one.
The sidebar was the filterList element's small dropdown-list shape (a
240px-capped box with its own border, floating in a large pane) rather
than a real full-height sidebar; .cx-filter-full opts a host into the
pane version without changing the compact one pipeline_window's own node
picker still uses.
Round one had a per-row hint line under the icon+label, which doesn't
exist in v4's own sidebar and broke alignment the moment a hint wrapped
(icon and label baseline-aligned against a hint floating as an independent
sibling rather than stacked under its own label). Fixed the real bug --
label+hint are a column now, not loose flex siblings -- then removed the
hint entirely to match v4: a section's subtitle belongs in its own head
once picked, not repeated under every row whether picked or not.
About is centred -- mark, name, version/branch -- rather than a plain
settings-row list, the one section built fresh rather than mounting an
existing window.
Picking Node packs, Updates, the log or Temp files used to close the
Settings window and open that feature's own separate modal, with no way
back except reopening Settings from the menu bar -- not what v4 does,
and not what "unified" is supposed to mean. Each of those five files
(plus pipeline_window.js) now exports mount({ setFooter, close }), the
same content its standalone open() already built, taking the modal
chrome's footer and close from whoever hosts it instead of owning a
modal directly. open() is unchanged for anyone opening one alone --
pipeline_window.test.js and every Playwright spec for these five prove
it, unedited, still passing.
Also fixes two findings from the adversarial pass on the earlier
deep-link version: filterList's search only matched a section's visible
label/hint, so a section findable only by a keyword (e.g. "loaders" for
"Models and pipeline") silently vanished from the filtered list; and
createSettingsWindow's "one at a time" state lived at module scope, so
a second independent instance would have redrawn the first one's modal
instead of opening its own. Both now have regression tests.
Models and pipeline, Updates, Node packs, the log, and Temp files were
five separate dropdown entries with no way to search across them. One
button now opens a single searchable window; About is drawn in place,
everything else is a deep link -- each of those five already has real,
tested modal chrome of its own, so this reuses it rather than rebuilding
it inline.
Uses the existing composer.splitPane.h for the two-pane layout rather
than a new element -- an earlier pass added one, and its CSS collided
with splitPane's own .cx-split class, silently breaking the pipeline
window's own list/parameters split (caught by pipeline.spec.js before
this landed).
Found by the adversarial pass on the exclude commit: every other
early-return in generateAll() says something (empty load, cancelled,
N failed) except this one -- excluding every scene made the button do
nothing with no feedback, indistinguishable from a hang.
The Edit menu gains Toggle Exclude, next to Add/Delete scene. An excluded
scene stays on the timeline, dimmed in the strip, and Generate All skips it
-- both in its count and its loop, so "2 of 2 did not generate" never counts
a scene that was never going to run.
Covered by a jsdom render test, a real end-to-end menu-to-strip test, and a
Generate All exclude test.
Player: Save to bin (whatever the viewer shows, if it has a real file
identity) and Save frame (canvas-capture the video on screen, upload
through ComfyUI's own /upload/image, no new route needed). Both add
without navigating the viewer -- an upload is a real round trip, and
the result is added to the bin, not shown as a stale finished pick.
Save/Load Project File: a real download route with Content-Disposition,
and an import route that never trusts an id in the file -- an import
always lands as its own project, never overwrites one already here.
Generate All: every scene in order, waiting for each to finish before
starting the next -- not a montage (v5 has no render/stitch stage),
N separate generations each landing on its own scene.
Both project-switching (open/newProject/importProject) and the earlier
run-reattach path shared a bug class this session kept finding: state
captured at one moment, applied at a later async moment with no check
that nothing had moved on in between. Closed it properly this time --
a generation counter every switch bumps and checks, including start()
(the page's own handlers are live before it resolves) and undo/redo
(also "a different project is open now"), and scheduleSave() captures
the object an edit actually touched instead of re-reading a variable
that can point somewhere else by the time its debounce timer fires.
Four adversarial rounds against this range, each fixed before landing.
Drag a clip onto another to reorder; an S/M/L control changes clip width
(--strip-w), remembered per browser. Trim already existed as the Scene
inspector's length field.
The dragged clip is tracked by its DataTransfer payload, not a closure
variable keyed on DOM position or id -- a redraw mid-drag (any project
change, not just a removal) tears down and rebuilds every cell, and
dragend fires on the original source node regardless of whether a
redraw detached it from the tree. Two adversarial rounds against this
file specifically, both fixed before landing.
Two routes silently shared /api/log; a pack's "updates available" badge
outlived the update; nothing server-side stopped a restart mid-generation,
and the fixes for that had their own dead end, a lock scoped to the wrong
event loop, an unlocked status fetch, and a check that only ran once
instead of at the moment it mattered. A run's result was lost outright if
the project it belonged to wasn't the one open when it finished -- on a
project switch or a page reload. A brand-new project inherited the
previous one's width, then inherited the wrong "default" for it.
Found and confirmed via an adversarial pass over 9dcd94c..HEAD, one round
per fix.
Fifth off the port list. ComfyUI writes previews and other transient outputs to
a temp directory and wipes it on restart, so everything in there is findable
for exactly as long as the server has been up -- which is the window in which
anyone asks where something went.
Read-only on purpose. This is a place to FIND a file, not to manage one;
deleting temp files is ComfyUI's job and it does it on restart.
Rows rather than tiles, which is the opposite of the bin and for the opposite
reason: a row shows the whole name, the size and when it was written, and the
grid draws none of those. Only media is listed -- a temp directory fills with
whatever any node felt like writing, and a browser listing 400 .pt files is not
a media bin. No <video> thumbnails, which wedged this app's API once already.
Three empty states, because they are three different facts: nothing was written
yet, the directory was wiped, and ComfyUI is not here to ask. An empty list
reads as the first one whichever is true.
One test taught me something about this environment again: opening a file puts
it in the Preview, the dev server has no /view, so the picture 404s and the kit
replaces it with its fallback -- correctly, and leaving nothing in the DOM to
assert on. What the viewer FETCHES is the evidence, so the test reads the
request rather than the picture.
Fourth off the port list. Users do not open devtools, and "it did not work"
with nothing to read is where a bug report dies -- especially on a rented box
behind a tunnel, where the terminal that has the answer is not on this desk.
Read from ComfyUI's own log FILE rather than from a buffer in this process.
v4 kept both and its own comment says why the file is the half that matters: a
crash and restart empties an in-memory buffer, and the lines from before the
crash are exactly the ones worth reading. So this is the smaller half and the
more useful one.
Three things the panel refuses to do, each because of what a log is FOR:
* It does not redraw while text is being selected. Copying a stack trace is the
one thing anybody opens a log to do, and a panel that repaints under the drag
makes it impossible.
* It keeps the last lines when the server stops answering. The most interesting
thing a log can say is that ComfyUI just died, and replacing the lines from
just before it went with the error that says it is gone throws away the only
evidence of why.
* No log file is an ANSWER, with the reason and where to look instead. An empty
list reads as a quiet log, and the difference matters precisely when somebody
opened the panel because something else already broke.
Both findings from the adversarial pass, confirmed and fixed.
RESTARTING FOR NOTHING. Every successful git call scheduled a restart, whether
or not the checkout had moved. Pressing Update while already up to date is a
normal thing to do -- there is no way to know until it has been asked -- and it
cost a full boot. The route now restarts only when something actually moved: a
commit changed, or a branch did. Two branches at the same commit still count,
because they stop being the same the moment anything lands on either.
And the worse half, which the pass rated first: nothing checked whether a
generation was in flight. The restart takes the run with it, GPU time already
spent, with no warning anywhere. The window now says a run is going and turns
off the three buttons that would end it.
CHECK FOR UPDATES COULD NOT REPORT ONE. check_updates() answers under `checked`
and the app read `nodes`. The request succeeded, the JSON was valid, and every
pack reported nothing to update however far behind it was -- a fetch per pack
over the network for an answer that could never be shown. Neither side's tests
could see it: the route test asserted the server's shape, the window test never
pressed the button, and nothing compared the two.
So the tests now name each other. The Python one asserts the key the app reads;
the browser one stubs the server's real shape and reads the words off the row.
Third off the port list. Users will want to add the one pack a workflow needs
without being sent to another UI to do it, and a pack that ships experimental
work needs the same for itself.
custom_nodes.py comes across as core/nodes_manager.py, changed in two places:
where FunPack's root is, and one guard that the port made necessary.
THE GUARD. custom_nodes_root() asks ComfyUI first and falls back to "FunPack
lives inside custom_nodes" -- a reliable anchor in an install, and a wrong one
in a development checkout, where the parent is somebody's home directory. Every
rule protecting the delete measures against that answer, so with ComfyUI's
paths unavailable, "remove Foo" resolved against ~/Documents and all four
guards happily agreed. It now checks the assumption it is named after and
refuses with where it thinks it is. Found by running the function, not by
reading it.
Two faults of my own, both in the window:
* Update and Remove were a button GROUP -- radios -- for two actions that are
not a choice between two states. It also left Remove looking selected after a
click, which is an alarming thing to look at.
* A failure was drawn and then wiped by the redraw in the same finally block:
the message appeared and vanished in one frame. What went wrong is state now,
not an argument to one draw call.
Second off the port list, and the one that gets reached for most: a pack whose
value is experimental work shipped continuously lives on switch-branch and
update.
git_update.py comes across from v4 as core/update.py, changed in three places
and no more: where the repo root is, a `remote=False` mode, and a 20-second
timeout on the one call that touches the network. It has been switching
branches on real rentals for months and every refusal in it -- no git on PATH,
not a checkout, a dirty tree, a branch that exists on neither side -- is a
lesson that cost somebody an afternoon. Rewriting it to look more like v5 would
trade proven plumbing for a fresh set of edge cases.
It is core rather than a module because it is about the INSTALL: no module can
announce the thing that updates every module.
Four routes, and every one of them ends by relaunching ComfyUI -- so the answer
is written first and the restart is a callback scheduled after it. `restart` is
looked up on the module rather than captured, or running the suite would
relaunch the test runner.
The window reads the checkout TWICE: once without the network, so there is
something to see instantly, then again with it for ahead and behind. A fetch
takes seconds on a good day and blocks until it times out on a machine that
cannot reach the remote, and neither is a reason to show an empty box. A dirty
tree is said before either button is pressed rather than as a refusal after.
Then it waits: an overlay that polls /api/health until the server answers, and
says so when it does not come back rather than spinning forever. That last part
needed the kit's blocking overlay to be able to change its message -- it could
not, so my give-up branch would have failed silently, which is the exact
pattern this project keeps finding.
First off the port list. Every edit in v5 was final -- v4 has had undo since
store.js, which is why it never showed up in a file-by-file inventory: there
is no undo.js to find.
A project is plain JSON and a few kilobytes, so undo is a snapshot before each
edit rather than an inverse per action. Capped at 40. The SELECTION travels
with the snapshot: undoing a removal and landing on a different scene than the
one that came back is not what anyone means by "take that back".
Two things it deliberately does not do. A new edit ends the redo line, or a
redo walks into a version that never followed from what is on screen. And a
project just opened has no past -- otherwise undo reaches back across a
project switch and replaces the open project with a version of a different one.
Edit menu: undo, redo, add and delete scene, move clip either way. Built on
each press so Undo is offered only when there is something behind it; a menu
that always offers it teaches people it does nothing. ⌘Z and ⇧⌘Z do the same,
except inside a text box, where the browser's own undo is the right one --
taking it over means one keystroke un-typing a paragraph instead of a word.
View gains Reset layout, for the person who closed both panels and cannot
remember what brings them back.
Found by the adversarial pass, reproduced twice.
Switching projects put the new project's numbers in the store and left the
OLD ones in the controls that produce what a run actually sends. The store's
own getter reported the new project correctly and the Inspector agreed, so
the only place the stale value showed was the Constructor -- a window nobody
has a reason to reopen after switching. Generate then ran at the size of the
project you left.
The attacker found it on the File-menu path. It is on three: opening from the
menu, opening programmatically, and applying a pipeline change, which rebuilds
every control from the PIPELINE's defaults and discards the project's. So the
fix is not in the menu handler -- the store now says when a DIFFERENT project
is open, which is not the same event as a field changing, and everything that
follows a project is wired to that instead of to the one route in that
happened to be reported.
Second finding, same pass: setVideo said "nothing on screen draws from this"
and did not report the change. True when written; the timeline started
computing its clip widths and its ruler from that value an hour later. A
comment is not a thing the next change has to keep true.
Also serial-mode for the project spec: two tests making and switching projects
share one real store on the dev server, and the config runs a file in
parallel. A race over a JSON directory, not a bug in the app.
dev went binary on H3 and said why: the steering reads the SIGN of a rating
and never its size, so a ten-point scale was the UI claiming a precision
nothing downstream used -- and at the sample sizes this actually runs at, a
mediocre middle rating only diluted the direction. v5 had invented a
four-way scale of its own; it now stores what dev stores.
Pressing the one that is on takes it back. Without a way out a mis-click is
a permanent opinion, and "no opinion" is a real answer nothing should have
to guess at.
The surfaces v4 has that v5 did not. Each announces itself and each can be
absent without the rest noticing, which is the point of the exercise: the
frontend should read as v4, and nothing in it should be load-bearing for
anything else.
* SCENE AND PROJECT, in the column v4 keeps them in, which now says which
scene it is about. Length on the timeline is the scene's own CROP and says
outright that a regenerate reads the project's -- the rule that mixing the
two is how a per-scene value quietly becomes everyone's.
* RATINGS on a scene, and a clip carries its rating as a colour on its edge.
Stored, not yet read: nothing in v5 learns from them yet.
* A RULER, and clips as wide as they are long, so the strip reads as time
rather than a row of equal boxes. Both ends bounded -- one long clip must
not become the whole strip, and one short one must stay hittable.
* THE PICKER WHEEL, on the middle button anywhere. It shows whatever
announced itself to actions.js, so a part of the app that grew a new action
is in the wheel without the wheel being told, and a part that failed to
load is simply not among them. v4's three pinned buttons had a fixed idea
of what could go in them.
* THE WAY IN, and it closes -- on Escape, on the backdrop, on its own button.
v4's could only be completed: a page refresh was the only way out, which
makes it a trap rather than a wizard. Dismissing is a real answer.
* File > rename, and the listing moves with it rather than going stale until
a reload.
Two things the kit refused, correctly, and now nobody has to know: a wheel of
fewer than two items throws (there is nothing to aim between), so the shell
does not offer one; and a strip cell with no ceiling took the whole strip.
Steps, Sampler and Scheduler are on the main window -- the three v4 buried
several clicks deep in Engine Settings. No new UI: the sampler slot declares
roles the way the prompt and the size already do.
What made that nearly free is that the controls now come from the NODE's own
description, through the settingFor/rendererFor pair the pipeline window has
always used. So a combo is a dropdown of the real choices and a number obeys
the real bounds, and prompt.js lost both of its hand-rolled branches: it no
longer decides what a control looks like, it asks.
File: New project, and every project listed with the open one ticked. The
store has been able to do this since the day it arrived and there was no way
in -- 20 lines of menu.
Three faults found on the way, all in the shared input helper:
* ENTER COULD NOT BE TYPED IN A TEXTAREA. `wire()` bound Enter to commit for
every text control, textareas included, so a multi-line value was
impossible -- the prompt above all, which is the longest multi-line thing
in the app. It has been that way in the pipeline window since it was
written; nobody had typed a second line into one.
* A control committed on blur alone, so autofill, a password manager and any
script setting a value went unheard. It listens to `change` as well now,
and commit() only acts on a real difference so the pair costs nothing.
* A control commits on blur, which means text still being typed is not in
the project yet -- and a tab closing is exactly when that is lost. boot
blurs before it saves.
