chore: re-lock uv.lock to match the bench optional-dependency extra

[mattmillerai]

ELI-5

pyproject.toml gained an optional bench extra a while back (in #490), but the lockfile was never regenerated to match. So the lock and the manifest disagreed: uv lock --check failed on main, and any bare uv run — like the one in our own test instructions — would silently re-lock and leave a ~287-line uv.lock diff in your working tree. That has already been committed into an unrelated PR by accident once, and reverted to keep that PR's diff clean. This regenerates the lockfile so the two agree again.

What changed

uv.lock only, regenerated with uv lock (uv 0.11.26) from a clean checkout of main:

• Adds the anthropic dependency tree pulled in by the bench extra: anthropic, pydantic, pydantic-core, jiter, docstring-parser, annotated-types, typing-inspection.
• Records bench in provides-extras.

Risk surface

The diff is 285 additions and 2 deletions. Everything else is purely additive. Both deletions:

1. provides-extras = ["dev", "preview"] → ["bench", "dev", "preview"] — this is the manifest/lock disagreement being fixed.
2. exceptiongroup's dependency on typing-extensions lost its python_full_version < '3.13' marker. This is a semantic no-op: exceptiongroup is only ever reachable under python_full_version < '3.11' (via anyio and pytest), so a < '3.13' marker is tautologically true everywhere it is installed, and uv simplified it away.

No existing package version changed — no other lines were removed. requires-python = ">=3.10" and the lockfile revision are unchanged.

Verification

• uv lock --check → exits 0 (fails on main today).
• uv run -p 3.10 --extra dev pytest tests/comfy_cli/command/ -q → 1157 passed, 2 skipped.
• uv run -p 3.10 --extra dev pytest tests/comfy_cli/ -q → 2471 passed, 13 skipped.
• ruff check . reports 14 pre-existing UP038 errors — identical on a pristine origin/main checkout, so unrelated to this change. ruff format --check . → 235 files already formatted.
• uv run --extra bench python -c "import anthropic, pydantic" → installs the 7 new packages and imports cleanly (anthropic 0.117.0 / pydantic 2.13.4), confirming the newly-locked tree actually resolves rather than just looking plausible.
• After a bare uv run, the working tree shows no new lock churn and uv lock --check still exits 0 — the silent re-dirty behavior is gone.

Judgment calls

• Landed without waiting for fix(registry): emit pyproject hint blocks as single-line tomlkit comments #533/fix: build telemetry providers lazily so comfy install can upgrade pydantic_core (BE-3289) #535/fix(registry): emit pyproject hints as standalone comments for tomlkit 0.15 #536. The plan suggested landing after those merge to avoid a lockfile conflict. None of them touch uv.lock, and only fix(registry): emit pyproject hints as standalone comments for tomlkit 0.15 #536 touches pyproject.toml — so there is no conflict to wait for and this can land in any order.
• fix(registry): emit pyproject hints as standalone comments for tomlkit 0.15 #536 will re-introduce the drift. fix(registry): emit pyproject hints as standalone comments for tomlkit 0.15 #536 pins tomlkit>=0.13,<0.16. Applying that change on top of this branch makes uv lock --check fail again, so whichever of the two merges second needs a re-lock. Cheapest fix is for fix(registry): emit pyproject hints as standalone comments for tomlkit 0.15 #536 to run uv lock and include the result.
• The CI uv lock --check gate is deliberately not included here. The hardening is worth doing — fix(registry): emit pyproject hints as standalone comments for tomlkit 0.15 #536 above is concrete proof the drift recurs — but it is a repo-wide CI policy change that would immediately fail an in-flight PR, and it touches nothing this PR touches. It is better as its own PR that can be weighed on its own merits than bundled into a mechanical re-lock. Recommending it as a follow-up.

[coderabbitai]

Important

Review skipped

Review was skipped due to path filters

⛔ Files ignored due to path filters (1)

• uv.lock is excluded by !**/*.lock

CodeRabbit blocks several paths by default. You can override this behavior by explicitly including those paths in the path filters. For example, including **/dist/** will override the default block on the dist directory, by removing the pattern from both the lists.

⚙️ Run configuration

Configuration used: Organization UI

Review profile: ASSERTIVE

Plan: Pro Plus

Run ID: cdf499f7-5daf-4f4b-9a5a-21fa1ac1a07e

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch matt/be-3312-relock-uv-lock

✨ Simplify code

• Create PR with simplified code
• Commit simplified code in branch matt/be-3312-relock-uv-lock

---

_{Comment @coderabbitai help to get the list of available commands.}

[github-actions]

🔍 Cursor Review — Consolidated panel

Triggered by @mattmillerai.

✅ No high-signal findings.

Panel: 6/8 reviewers contributed findings.

Reviewers that did not contribute: kimi-k2.5:adversarial (empty), kimi-k2.5:edge-case (empty)

[mattmillerai]

The failing build check is not caused by this PR — it is an environmental break on main that landed today, and it is already fixed by #536.

Root cause

tomlkit 0.15.1 was released today (2026-07-17 01:48 UTC). It makes Item.comment() reject line breaks, which breaks create_comfynode_config() in comfy_cli/registry/config_parser.py:75 and :250. Bisected locally across versions:

tomlkit	Item.comment(multiline)
0.13.3	OK
0.15.0	OK
0.15.1	ValueError: Comment cannot contain line breaks

Timeline lines up exactly: main last passed pytest.yml on 2026-07-13 (a732f5c); tomlkit 0.15.1 published 01:48 today; this PR's run was 10:32 today.

Why this PR cannot be the cause

pytest.yml installs with pip install pytest pytest-cov jsonschema + pip install -e .. It never reads uv.lock, and uv.lock is the only file this PR touches (285+/2-). pip resolves the unpinned "tomlkit" from pyproject.toml to whatever is newest, so it picked up 0.15.1. Any PR opened against main today fails the same 9 tests.

The fix is #536, not here

#536 (fix(registry): emit pyproject hints as standalone comments for tomlkit 0.15) pins tomlkit>=0.13,<0.16 and emits standalone comments instead — its build check is green. Duplicating that fix into this lockfile-only PR would just conflict with it.

Merge-order note

#536 changes pyproject.toml but not uv.lock, so the two interact:

• This PR merges first → fix(registry): emit pyproject hints as standalone comments for tomlkit 0.15 #536 then needs a re-lock, otherwise uv lock --check drifts again.
• fix(registry): emit pyproject hints as standalone comments for tomlkit 0.15 #536 merges first → this branch needs a rebase + uv lock to pick up the tomlkit pin, which also turns this PR's CI green.

Verified the lockfile is still correct as-is: uv lock --check exits 0 on this branch and fails on a pristine main clone — the drift this PR fixes is still real and still unfixed on main.

Leaving the code untouched. Happy to rebase + re-lock the moment #536 lands.

[mattmillerai]

🤖 Reviews-resolver re-checked this PR — status unchanged and confirmed:

• Lockfile still correct & still needed. uv lock --check exits 0 on this branch; main still has provides-extras = ["dev", "preview"] (missing bench), a 289-line drift. No new base conflict — PR is MERGEABLE.
• The red build check is not caused by this PR. It's the environmental tomlkit-0.15.1 break in config_parser.py (multi-line Item.comment()), and pytest.yml installs via pip install -e . — it never reads uv.lock, the only file this PR touches. Same 9 tests fail on any PR opened against main today.
• Fix belongs in fix(registry): emit pyproject hints as standalone comments for tomlkit 0.15 #536, not here. fix(registry): emit pyproject hints as standalone comments for tomlkit 0.15 #536 emits standalone comments + pins tomlkit. I deliberately did not duplicate that fix into this lockfile-only PR — it would just conflict with fix(registry): emit pyproject hints as standalone comments for tomlkit 0.15 #536 on the same lines.

⚠️ Blocker update: #536 is currently a draft and has a failing test check, so it can't land yet. This PR is therefore blocked on #536 being de-drafted and made green. Recommended order: land #536 first, then rebase + uv lock here (which also turns this PR's build green). Happy to do that re-lock the moment #536 merges.