Track custom node startup errors and expose via API endpoint#13184
Open
Kosinkadink wants to merge 12 commits into
Open
Track custom node startup errors and expose via API endpoint#13184Kosinkadink wants to merge 12 commits into
Kosinkadink wants to merge 12 commits into
Conversation
Store import and prestartup errors in NODE_STARTUP_ERRORS dict (nodes.py, main.py) and add GET /custom_node_startup_errors endpoint (server.py) so the frontend/Manager can distinguish failed imports from missing nodes. Ref: ComfyUI-Launcher#303 Amp-Thread-ID: https://ampcode.com/threads/T-019d2346-6e6f-75e0-a97f-cdb6e26859f7 Co-authored-by: Amp <amp@ampcode.com>
…node) Expand custom-node startup error tracking to differentiate between user-installed custom_nodes, built-in comfy_extras, and partner comfy_api_nodes. Each NODE_STARTUP_ERRORS entry now carries a 'source' field and is keyed by '<source>:<module_name>' so colliding module names across the three locations don't overwrite each other. The /custom_node_startup_errors endpoint returns errors grouped by source so the frontend/Manager can render distinct sections. Also captures previously-missed failures from comfy_entrypoint() (phase='entrypoint'). Introduces nodes.record_node_startup_error() helper used by load_custom_node and main.execute_prestartup_script. Adds tests-unit/node_startup_errors_test.py (6 tests) covering field shape, source mapping for each module_parent, cross-source collisions, and default fallback. Ref: ComfyUI-Launcher#303 Amp-Thread-ID: https://ampcode.com/threads/T-019e23a1-2acc-7619-bd0e-f783d1368ef3 Co-authored-by: Amp <amp@ampcode.com>
When a failing module has a pyproject.toml, parse it via comfy_config.config_parser and attach a 'pyproject' field with the Comfy Registry-style identity (pack_id, display_name, publisher_id, version, repository). This gives the frontend/Manager a stable, user-recognizable handle for the failed pack beyond the on-disk folder name. The lookup is best-effort and never raises: missing toml, missing pydantic-settings dependency, or any parse error simply omits the 'pyproject' key. Ref: ComfyUI-Launcher#303 Amp-Thread-ID: https://ampcode.com/threads/T-019e23a1-2acc-7619-bd0e-f783d1368ef3 Co-authored-by: Amp <amp@ampcode.com>
The absolute on-disk path is internal detail the frontend/Manager has no use for. Keep it in the in-memory NODE_STARTUP_ERRORS dict for server-side debugging, but exclude it from the public API payload. The user-facing identifier remains module_name (and pyproject.pack_id when available). Ref: ComfyUI-Launcher#303 Amp-Thread-ID: https://ampcode.com/threads/T-019e23a1-2acc-7619-bd0e-f783d1368ef3 Co-authored-by: Amp <amp@ampcode.com>
The endpoint covers comfy_extras and comfy_api_nodes failures too, not just user-installed custom nodes, so the path should not pretend otherwise. Ref: ComfyUI-Launcher#303 Amp-Thread-ID: https://ampcode.com/threads/T-019e23a1-2acc-7619-bd0e-f783d1368ef3 Co-authored-by: Amp <amp@ampcode.com>
The public 'source' field on each NODE_STARTUP_ERRORS entry is now the same string as the internal module_parent passed to load_custom_node ('custom_nodes', 'comfy_extras', 'comfy_api_nodes'), rather than being translated to a separate fixed enum. Treating it as a free-form string keeps the contract durable in case the node-source layout evolves (e.g. comfy_api_nodes eventually moving out of core).
The API endpoint now also dynamically groups by whatever sources are present rather than hardcoding the three known top-level keys; consumers should not assume any particular set of keys is always present.
Drops the _NODE_SOURCE_BY_PARENT map, _node_source_from_parent helper, and the related test. Adds a test covering an arbitrary unknown module_parent value passing through unchanged.
Ref: ComfyUI-Launcher#303
Amp-Thread-ID: https://ampcode.com/threads/T-019e23a1-2acc-7619-bd0e-f783d1368ef3
Co-authored-by: Amp <amp@ampcode.com>
Minor cleanup from code review: traceback is stdlib so there's no circular-import concern keeping it inline. The 'from nodes import record_node_startup_error' stays inline because nodes.py imports from contexts that would create a cycle at module load time. Ref: ComfyUI-Launcher#303 Amp-Thread-ID: https://ampcode.com/threads/T-019e23a1-2acc-7619-bd0e-f783d1368ef3 Co-authored-by: Amp <amp@ampcode.com>
Kosinkadink
requested review from
alexisrolland,
comfyanonymous,
guill,
kijai and
rattus128
as code owners
|
Note Reviews pausedIt looks like this branch is under active development. To avoid overwhelming you with review comments due to an influx of new commits, CodeRabbit has automatically paused this review. You can configure this behavior by changing the Use the following commands to manage reviews:
Use the checkboxes below for quick actions:
No actionable comments were generated in the recent review. 🎉 ℹ️ Recent review info⚙️ Run configurationConfiguration used: Path: .coderabbit.yaml Review profile: CHILL Plan: Pro Run ID: 📒 Files selected for processing (2)
🚧 Files skipped from review as they are similar to previous changes (2)
📝 WalkthroughWalkthroughThis PR implements custom node startup error tracking and reporting. It adds a global error registry and helpers in nodes.py to capture failures during import, entrypoint, and prestartup phases (including tracebacks and optional pyproject metadata), integrates recording into load_custom_node() and main's prestartup execution, exposes grouped failures via GET /node_startup_errors, and adds unit tests validating composite-keyed storage, source propagation, collision handling, and conditional pyproject metadata inclusion. 🚥 Pre-merge checks | ✅ 4 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (4 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@main.py`:
- Around line 152-159: Don't import the nodes module inside the prestartup
exception handler; instead capture the failure details locally and defer calling
record_node_startup_error until after the normal import nodes/bootstrap point.
Replace the inline import and record_node_startup_error call in the prestartup
error path by appending a dict with keys/module_path/source/phase/error/tb (use
script_path for module_path, source="custom_nodes", phase="prestartup") to a
local list (e.g., prestartup_errors), and after the standard import nodes stage
(where nodes is normally imported) iterate that list and call
record_node_startup_error for each buffered entry so torch and nodes are only
imported at their intended bootstrap time.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
Run ID: 0d7aab0b-1cc9-4c7c-b77e-7d5aba2f9e2a
📒 Files selected for processing (4)
main.pynodes.pyserver.pytests-unit/node_startup_errors_test.py
Buffer prestartup failures into a module-level list inside main.py instead of importing 'nodes' (and therefore 'torch') from within the exception handler. After the normal 'import nodes' line, drain the buffer via nodes.record_node_startup_error so bootstrap order stays deterministic regardless of whether a prestartup script succeeded. Also convert the explanatory '#' comment on the new /node_startup_errors endpoint into a proper docstring and add a docstring to execute_prestartup_script, addressing CodeRabbit's docstring-coverage warning on this PR. Addresses review feedback on PR #13184. Amp-Thread-ID: https://ampcode.com/threads/T-019e2f90-26fe-7048-9855-5ff39d08a3e0 Co-authored-by: Amp <amp@ampcode.com>
…urce query filters
Two reviewer-requested improvements to GET /node_startup_errors:
1. Emit the pyproject metadata in the same {project: {...}, tool_comfy: {...}}
shape that comfy_config.config_parser.extract_node_configuration already
returns, instead of inventing a flat {pack_id, display_name, ...} bag.
API consumers can now parse the pyproject block straight through the
shared PyProjectConfig pydantic model. Empty / default-valued leaves
are pruned by a small recursive _prune_empty helper so the payload
stays compact, but nesting and field names match the source-of-truth.
2. Add optional source, module_name, and pack_id query parameters
(combined with AND) so a frontend / Manager can ask ?pack_id=foo
instead of grep'ing through the whole grouped response. pack_id
resolves against pyproject.project.name; entries without a parsed
pyproject are naturally excluded from a pack_id query.
The grouping + filtering + module_path stripping moves into
odes.filter_node_startup_errors so the route handler is a one-liner and
the helper is unit-testable without spinning up an aiohttp app.
Tests: 5 new unit tests covering each filter branch, AND-combination, and
empty-result behaviour, plus an updated pyproject-metadata assertion that
checks the nested PyProjectConfig shape, plus a focused test for the
_prune_empty helper.
?source= or ?module_name= or ?pack_id= (param present but blank) would have returned {} because the helper treated the empty string as an exact-match filter. Coalesce to None at the route boundary so a present-but-blank query param behaves the same as the param being absent. The helper's own behaviour is unchanged and locked in by a new assertion.
Amp-Thread-ID: https://ampcode.com/threads/T-019e86fd-b68f-74de-8c91-d2662377424a
Co-authored-by: Amp <amp@ampcode.com>
There was a problem hiding this comment.
Actionable comments posted: 2
🧹 Nitpick comments (2)
tests-unit/node_startup_errors_test.py (2)
223-253: ⚡ Quick winExercise
module_namein an AND-combined filter case.
module_nameis only tested in isolation right now. Since the new contract sayssource,module_name, andpack_idcombine with AND, a bug that ignoresmodule_nameonce another filter is present would still pass. Add one combined case that includesmodule_namealongside another filter.🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@tests-unit/node_startup_errors_test.py` around lines 223 - 253, Add a new unit test that exercises module_name together with another filter to verify AND semantics: update or add a test (e.g., alongside test_filter_node_startup_errors_filters_combine_with_and) that seeds entries using _seed(...) with the same module_name in multiple sources but only one entry matching an additional filter (source or pack_id), then call nodes.filter_node_startup_errors(module_name="A", source="comfy_extras") (or module_name="A", pack_id="comfyui-foo") and assert the result only contains the expected source and module_name; this ensures filter_node_startup_errors respects module_name when combined with other filters.
54-74: ⚡ Quick winAdd a dedicated
entrypointfailure test.These
load_custom_node()tests only exercise import-time exceptions. This PR also records a distinct"entrypoint"phase, so a regression in thecomfy_entrypoint()error path would still leave this suite green. Add one module that imports successfully and raises fromcomfy_entrypoint(), then assert the recordedphase,source, and traceback fields.Also applies to: 98-117, 162-178
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@tests-unit/node_startup_errors_test.py` around lines 54 - 74, Add a new unit test (similar to test_load_custom_node_records_source) that creates a module which imports fine but defines comfy_entrypoint() that raises a RuntimeError; call nodes.load_custom_node(module_path, module_parent=module_parent), assert it returns False, then lookup nodes.NODE_STARTUP_ERRORS[f"{module_parent}:broken_entrypoint"] (or appropriate module name) and assert entry["phase"] == "entrypoint", entry["source"] == module_parent, entry["module_name"] matches the module, and that entry["error"] / entry["traceback"] contain the raised message and exception type; reuse the existing _write_broken_module helper pattern but make it emit the exception from comfy_entrypoint instead of at import time.
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@nodes.py`:
- Around line 2257-2275: The current reshaping of NODE_STARTUP_ERRORS collapses
multiple failures per module by using module_name as a single object value;
change the response so each module_name maps to a collection of failures (e.g.,
a list of failure entries or a dict keyed by phase) instead of a single entry,
preserving all original fields except stripping module_path; update the code
that builds the source->module_name mapping (the logic that currently overwrites
entries) to append or merge entries for the same module_name, and ensure
existing filters (source, module_name, pack_id) still apply and that module_path
remains removed from each returned failure entry.
- Around line 2259-2263: The public API is leaking internal paths by returning
raw tracebacks (via traceback.format_exc()) in public_entry and related places
(also around the code at 2288-2289); update the serializer that builds
public_entry to omit or redact the traceback field (e.g., replace with a generic
message like "internal error" or strip filenames) and keep the full traceback
only in server-side logs/attributes (retain module_path stripping as-is but
ensure any use of traceback.format_exc() is not sent to callers); locate the
code that constructs public_entry and the nearby traceback usage and modify it
to remove/replace the traceback before returning the public-facing structure.
---
Nitpick comments:
In `@tests-unit/node_startup_errors_test.py`:
- Around line 223-253: Add a new unit test that exercises module_name together
with another filter to verify AND semantics: update or add a test (e.g.,
alongside test_filter_node_startup_errors_filters_combine_with_and) that seeds
entries using _seed(...) with the same module_name in multiple sources but only
one entry matching an additional filter (source or pack_id), then call
nodes.filter_node_startup_errors(module_name="A", source="comfy_extras") (or
module_name="A", pack_id="comfyui-foo") and assert the result only contains the
expected source and module_name; this ensures filter_node_startup_errors
respects module_name when combined with other filters.
- Around line 54-74: Add a new unit test (similar to
test_load_custom_node_records_source) that creates a module which imports fine
but defines comfy_entrypoint() that raises a RuntimeError; call
nodes.load_custom_node(module_path, module_parent=module_parent), assert it
returns False, then lookup
nodes.NODE_STARTUP_ERRORS[f"{module_parent}:broken_entrypoint"] (or appropriate
module name) and assert entry["phase"] == "entrypoint", entry["source"] ==
module_parent, entry["module_name"] matches the module, and that entry["error"]
/ entry["traceback"] contain the raised message and exception type; reuse the
existing _write_broken_module helper pattern but make it emit the exception from
comfy_entrypoint instead of at import time.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
Run ID: 524fbbf3-e7be-403d-80b8-895d522e2cc5
📒 Files selected for processing (3)
nodes.pyserver.pytests-unit/node_startup_errors_test.py
Comment on lines
+2257
to
+2275
| """Return `NODE_STARTUP_ERRORS` reshaped for the public HTTP endpoint. | ||
|
|
||
| Entries are grouped by their ``source`` bucket (the same string as the | ||
| internal ``module_parent`` used at load time). The on-disk | ||
| ``module_path`` is stripped from each entry — it's an internal detail | ||
| useful only for server-side logging and would leak absolute filesystem | ||
| layout otherwise. | ||
|
|
||
| Optional filters narrow the response and combine with AND: | ||
|
|
||
| * ``source`` — only entries from this source bucket. | ||
| * ``module_name`` — only entries whose module name matches exactly. | ||
| * ``pack_id`` — only entries whose ``pyproject.project.name`` | ||
| matches exactly. Entries without a parsed | ||
| pyproject.toml can never match this filter. | ||
|
|
||
| A non-matching filter returns an empty dict, not an error — absence of | ||
| a failure is a valid answer for this query. | ||
| """ |
There was a problem hiding this comment.
Preserve multiple phase failures per module.
The new response shape keys each source bucket by module_name, so a pack that fails in more than one startup phase will only expose the last one written. That collapses the new phase dimension the endpoint is supposed to report. Please return a per-module collection (for example, a list or phase-keyed map) instead of a single entry.
Also applies to: 2276-2290
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@nodes.py` around lines 2257 - 2275, The current reshaping of
NODE_STARTUP_ERRORS collapses multiple failures per module by using module_name
as a single object value; change the response so each module_name maps to a
collection of failures (e.g., a list of failure entries or a dict keyed by
phase) instead of a single entry, preserving all original fields except
stripping module_path; update the code that builds the source->module_name
mapping (the logic that currently overwrites entries) to append or merge entries
for the same module_name, and ensure existing filters (source, module_name,
pack_id) still apply and that module_path remains removed from each returned
failure entry.
Comment on lines
+2259
to
+2263
| Entries are grouped by their ``source`` bucket (the same string as the | ||
| internal ``module_parent`` used at load time). The on-disk | ||
| ``module_path`` is stripped from each entry — it's an internal detail | ||
| useful only for server-side logging and would leak absolute filesystem | ||
| layout otherwise. |
There was a problem hiding this comment.
Don’t expose raw tracebacks from the public endpoint.
Stripping module_path is not enough here: traceback.format_exc() already includes absolute filenames and other internal details. Returning it unchanged from the public API still leaks the server’s filesystem layout. Consider omitting or redacting traceback in public_entry and keeping the full traceback server-side only.
Also applies to: 2288-2289
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@nodes.py` around lines 2259 - 2263, The public API is leaking internal paths
by returning raw tracebacks (via traceback.format_exc()) in public_entry and
related places (also around the code at 2288-2289); update the serializer that
builds public_entry to omit or redact the traceback field (e.g., replace with a
generic message like "internal error" or strip filenames) and keep the full
traceback only in server-side logs/attributes (retain module_path stripping
as-is but ensure any use of traceback.format_exc() is not sent to callers);
locate the code that constructs public_entry and the nearby traceback usage and
modify it to remove/replace the traceback before returning the public-facing
structure.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Closes Comfy-Org/Comfy-Desktop#303.
Problem
Today, when a node pack fails at startup, ComfyUI logs a warning and moves on. The frontend / Manager has no way to tell the difference between "this node is missing because the pack isn't installed" and "this node is missing because the pack failed to import" - both look identical to the user.
Change
Track every custom node / built-in extra / API node startup failure in an in-memory dict and expose it via a new HTTP endpoint, so the frontend and Manager can render an explicit "these packs failed to load" UI.
Captured failure points
prestartupmain.execute_prestartup_scriptprestartup_script.pyraisedimportnodes.load_custom_nodeouter exceptimportlib.exec_module()raised (most common: missing dep)entrypointnodes.load_custom_nodeinner exceptcomfy_entrypoint()raised after a successful importCategorized by source
load_custom_nodeis invoked from three call sites with differentmodule_parentvalues, all of which are propagated as-is into the publicsourcefield on each entry:module_parentsourcevaluecustom_nodes"custom_nodes"comfy_extras"comfy_extras"comfy_api_nodes"comfy_api_nodes"sourceis intentionally a free-form string, not a fixed enum. The current values match where things live in the tree today; if the layout ever evolves (e.g.comfy_api_nodesmoving out of core into its own thing) thesourcevalue moves with it and consumers don't have to be re-versioned. Frontends should treat any newsourcevalue as a new bucket rather than rejecting it.Entries are keyed internally by
f"{source}:{module_name}"so a name collision (e.g. a custom pack folder happens to be namednodes_audio) doesn't overwrite the built-in entry of the same name.pyproject.toml identity
When a failing module is a directory containing
pyproject.toml, we attach apyprojectblock parsed via the existingcomfy_config.config_parser(added in #8357) and emit it in the same{project, tool_comfy}shape thatextract_node_configurationalready returns (i.e. thePyProjectConfigpydantic model). Consumers that already speak that schema - Manager, Registry tooling, the existing config_parser users - can parse this block back through the same model directly, without an intermediate translation layer.Empty / default-valued leaves are pruned (so the payload stays compact and you don't get noise from defaulted
requires_python, emptydependencies, etc.) but the nesting and field names are kept 1:1 withPyProjectConfig. Parsing is best-effort: missing toml, missingpydantic-settings, or any parse error simply omits thepyprojectfield.New API endpoint
GET /node_startup_errors- returns failures grouped bysource. The response only contains source buckets that actually had a failure; an entirely healthy install returns{}. Consumers should not assume any particular set of top-level keys is always present.The internal
module_path(absolute on-disk path) is not sent over the wire - it's a server-side detail used for logging only. Tracebacks are sent verbatim (they may contain absolute paths in frame headers; this is intentional for debug value).Query parameters (all optional, all combine with AND)
sourcesourcebucket?source=custom_nodesmodule_namemodule_name(exact)sourcebuckets - same name from different sources both come backpack_idpyproject.project.name(exact)pyproject.toml(comfy_extras,comfy_api_nodes, packs missing the toml) are naturally excludedA non-matching filter returns
{}with HTTP 200 - absence of a failure is a valid answer.Response schema
{ "custom_nodes": { // present only if at least one custom_nodes failure exists "<module_name>": { "source": "custom_nodes", // same string as module_parent - free-form, not a fixed enum "module_name": "ComfyUI-AnimateDiff-Evolved", "phase": "import", // "prestartup" | "import" | "entrypoint" "error": "No module named 'torch'", "traceback": "Traceback (most recent call last):\n ...\nModuleNotFoundError: ...\n", "pyproject": { // OPTIONAL - only present when pyproject.toml exists & parses // Mirrors comfy_config.types.PyProjectConfig 1:1. // Empty / default-valued leaves are pruned. "project": { "name": "comfyui-animatediff-evolved", "version": "1.5.0", "urls": { "repository": "https://github.com/Kosinkadink/ComfyUI-AnimateDiff-Evolved" } }, "tool_comfy": { "publisher_id": "kosinkadink", "display_name": "ComfyUI AnimateDiff Evolved" } } } }, "comfy_extras": { // present only if at least one comfy_extras failure exists "<module_name>": { "source": "comfy_extras", "module_name": "nodes_audio", // file stem of comfy_extras/nodes_audio.py "phase": "import", // effectively always "import" for this source "error": "...", "traceback": "..." // pyproject is never present (single-file modules) } }, "comfy_api_nodes": { // present only if at least one comfy_api_nodes failure exists "<module_name>": { "source": "comfy_api_nodes", "module_name": "nodes_runway", // file stem of comfy_api_nodes/nodes_*.py "phase": "import", "error": "...", "traceback": "..." // pyproject is never present } } }Field semantics
sourcemodule_parent("custom_nodes"/"comfy_extras"/"comfy_api_nodes"today). Free-form - treat unfamiliar values as a new bucket.module_name.py) for comfy_extras / comfy_api_nodes. The fallback display label and the only universal identifier.phase"prestartup"|"import"|"entrypoint"errorstr(exception)- the one-line message, no exception type, no stack. May be empty for argless exceptions.tracebacktraceback.format_exc(). Last line includes the exception type.pyprojectpyproject.toml. Mirrorscomfy_config.types.PyProjectConfig. All inner keys are individually optional (empty values are pruned).pyproject.project.name[project] namefrom pyproject.toml - the canonical Comfy Registry / pip name. Prefer this for cross-referencing with Manager / Registry; this is what thepack_idquery filter matches against.pyproject.project.version[project] version.pyproject.project.urls.repository[project.urls] Repository.pyproject.tool_comfy.display_name[tool.comfy] DisplayName- human-readable label for UI.pyproject.tool_comfy.publisher_id[tool.comfy] PublisherId- registry publisher namespace.Other
PyProjectConfigfields (project.description,project.classifiers,tool_comfy.icon, ...) are emitted verbatim when present and non-empty, but only the fields above are relied on by the typical Manager/Registry cross-referencing flow.Suggested display label fallback chain
Files changed
nodes.py-NODE_STARTUP_ERRORSdict,record_node_startup_error,_read_pyproject_metadata,_prune_empty,filter_node_startup_errors; record errors inload_custom_node(both import and entrypoint paths) usingmodule_parentassource.main.py- buffer prestartup failures and flush them throughrecord_node_startup_errorafter the normalimport nodesline (so a failing prestartup script never drags torch in early).server.py-GET /node_startup_errorsroute, delegates tonodes.filter_node_startup_errorsafter reading thesource/module_name/pack_idquery params.tests-unit/node_startup_errors_test.py- 14 tests covering field shape, source pass-through, cross-source key collisions, pyproject metadata extraction (now asserts thePyProjectConfignesting), no-pyproject case, arbitrary-module_parentpass-through, the_prune_emptyhelper, and every branch of the new filter helper (no filter, each filter individually, AND combination, empty-result behaviour).Verification
Unit tests
14 tests, all passing locally (
pytest tests-unit/node_startup_errors_test.py -v):test_record_node_startup_error_fields- helper writes the expected fields.test_load_custom_node_records_source[custom_nodes|comfy_extras|comfy_api_nodes]- eachmodule_parentis recorded as the entry'ssource.test_load_custom_node_collision_across_sources- samemodule_nameunder different sources keeps both entries.test_load_custom_node_attaches_pyproject_metadata- asserts the nestedpyproject.project.{name,version,urls.repository}andpyproject.tool_comfy.{publisher_id,display_name}shape.test_load_custom_node_no_pyproject_skips_metadata- single-file modules without a toml omit thepyprojectkey entirely.test_load_custom_node_arbitrary_module_parent_passes_through- unknownmodule_parentrecorded as-is.test_prune_empty_drops_empty_leaves_only- recursive pruner drops"",[],{},None; keeps0andFalse.test_filter_node_startup_errors_strips_module_path_and_groups_by_source-module_pathnever leaks; default grouping works.test_filter_node_startup_errors_source_filter-?source=narrows to one bucket; non-match returns{}.test_filter_node_startup_errors_module_name_filter-?module_name=matches across multiple source buckets.test_filter_node_startup_errors_pack_id_filter_matches_only_pyproject_entries-?pack_id=matchespyproject.project.name; entries without pyproject can never match.test_filter_node_startup_errors_filters_combine_with_and- filters AND together.End-to-end test against a running ComfyUI
Deployed an earlier version of this branch onto a real ComfyUI install (Windows, Python 3.13.12, standalone-v0.1.28) via
comfy-runner, planted 5 deliberately broken modules covering all three sources and all three phases, and validated the liveGET /node_startup_errorsresponse. (The same checks apply to the current shape - only thesourcestrings changed from a translated enum to the rawmodule_parentvalue, and thepyprojectblock moved from a flat bag to the nestedPyProjectConfigshape.)custom_nodes/BrokenPackWithToml/custom_nodes__init__.pyraisesRuntimeError; fullpyproject.tomlalongsidecustom_nodes/BrokenPackNoToml/custom_nodes__init__.pydoesimport nonexistent_module_for_e2e_test; no tomlcustom_nodes/BrokenPrestartup/custom_nodesprestartup_script.pyraisesRuntimeError(valid__init__.py)comfy_extras/nodes_apg.pycomfy_extrasraise ImportError(built-in extra; comfy_extras uses a hardcoded file list, so an existing entry had to be temporarily clobbered)comfy_api_nodes/nodes_e2e_broken.pycomfy_api_nodesraise ValueError(api_nodes usesglob.glob("nodes_*.py"))200 OKapplication/json; charset=utf-8source,module_name,phase,error,traceback)sourcefield matches the entry's parent groupmodule_pathstripped from the response payloadpyprojectblock present only onBrokenPackWithToml, with all 5 fields populatedphasevalues correct:prestartupfor the prestartup pack,importfor everything elseerrorstrings carry the right exception messagetracebackstrings end with the correct exception type (RuntimeError,ModuleNotFoundError,ImportError,ValueError)?foo=bar&baz=1)POST /node_startup_errorsrejected with405 Method Not AllowedAfter the test the broken
nodes_apg.pywas restored from backup and all planted directories / files /.pyccache entries were removed.Out of scope (follow-ups)
NODE_STARTUP_ERRORS(and its siblingsNODE_CLASS_MAPPINGS,NODE_DISPLAY_NAME_MAPPINGS,LOADED_MODULE_DIRS,EXTENSION_WEB_DIRS) offnodes.py's module scope onto the server instance. Per-server-instance scoping is the right long-term home for all node-load state, but doing it just for this one dict would create an inconsistency with the four pre-existing module-level globals it sits next to innodes.py. Belongs in a single sweep of all five, separate from this PR.