chore(ci): DRY release.yml + size-limit PR comment#76
Merged
Conversation
Replace the inlined format/lint/typecheck/test/build steps in release.yml with a single `npm run verify` call so future verify changes auto-propagate. Verify also runs typedoc, gaining the release pipeline a docs-generation safety net. Add an `andresz1/size-limit-action` job to PR CI that posts a sticky comment with the gzip delta for each of the 5 budgets in package.json#size-limit. Job runs only on `pull_request` events and needs `pull-requests: write`. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
size-limit report 📦
|
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 93a03b3207
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
Per Codex review, two fixes: 1. Drop workflow-level `pull-requests: write`. Only the `size-limit-comment` job needs PR write; granting it at workflow scope expanded the GITHUB_TOKEN's blast radius across every parallel job. Job-level permissions already carry the write grant on the one job that posts the comment. 2. Drop `needs: build` from `size-limit-comment`. `build` runs `npx size-limit` and fails when budgets are exceeded — exactly the regression case where the sticky delta comment is most useful. Without `needs`, the comment posts regardless and the reviewer sees the offending bytes. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: ec20f83a77
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
Per Codex review (P1): the `script` override to `andresz1/size-limit-action` was `size-limit --json`, which assumes the binary is on PATH. On GitHub-hosted runners the binary lives only in `node_modules/.bin`, so the action would fail with command-not-found before posting the sticky delta comment. Restore `npx` so the action's documented default behavior holds. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Owner
Author
|
@codex review |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: c9ec83749a
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
Two fixes from Codex's third review: 1. P1 — actionlint job uses `reviewdog/action-actionlint` with `reporter: github-check`, which writes through the Checks API. The previous workflow-level `contents: read` change stripped `checks: write` from every job, including this one. Restore `checks: write` at the actionlint job's permissions block so the check reporter can post normally. 2. P2 — size-limit-comment was decoupled from `build` to ensure the delta comment posts when the size budget fails. Reconcile by restoring `needs: build` for ordering but using `if: always() && needs.build.result != 'cancelled' && != 'skipped'` so the comment still fires when build's `npx size-limit` step fails. Avoids parallel install/build duplication while keeping the regression-feedback property intact. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Owner
Author
|
@codex review |
|
Codex Review: Didn't find any major issues. 🚀 ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
If Codex has suggestions, it will comment; otherwise it will react with 👍. Codex can also answer questions or update the PR. Try commenting "@codex address that feedback". |
3 tasks
Luis85
added a commit
that referenced
this pull request
Apr 25, 2026
…pdate (#85) * feat(llm,docs): MockLlmProvider example + README LLM section + plan update - examples/llm-mock/: minimal Node example (index.ts + package.json + README.md). MockLlmProvider → tiny LlmReasoner → createAgent under SeededRng + ManualClock for 5 ticks; re-runs and asserts byte-identical DecisionTrace[] across runs. Demonstrates the LlmProviderPort end-to-end without network or a real provider key. - README.md: new "LLM provider port (preview)" section between reactive-store binding and Development. Five-line MockLlmProvider snippet, links to the example, calls out Phase B for concrete Anthropic/OpenAI adapters. - docs/specs/vision.md: one-paragraph status note under P3 — port + mock ship in 1.0; concrete adapters Phase B. - docs/plans/2026-04-25-comprehensive-polish-and-harden.md: new "Track B" section under "What's already shipped" listing PRs #76 through #85 with their roadmap row mapping. Workflow section gets a sixth rule pinning plan/doc updates to the same PR that lands the work. - CLAUDE.md (Plans & specs section): same plan-rides-with-PR rule inline, with rationale (stale plans force second review cycles + degrade Codex review quality). Roadmap row 15. No library code change. No changeset (docs + example only). Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * fix(example): drop parameter property in llm-mock for strip-types compat Per Codex review (P1): the LlmReasoner constructor used a TypeScript parameter property (`constructor(private readonly provider: ...)`), which Node's `--experimental-strip-types` mode (the runtime the example uses via `npm run start`) cannot parse. The example failed immediately with `ERR_UNSUPPORTED_TYPESCRIPT_SYNTAX` before any determinism logic ran. Replace with a manual field declaration + body assignment. Visibility modifiers on regular methods (`private async runAsync`) stay — those ARE strippable in place, only parameter properties require an actual transformation. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * fix(example): pre-compute LLM decisions to remove shared-state hazard Per Codex review (P2): the original example wired `provider.complete` inside `selectIntention` and stored the result in a module-scoped `lastDecision` variable. Run B reused the same slot — a stale write from run A's pending completion could land while run B was already ticking, masking real determinism violations. Restructure the example to pre-compute one decision per tick by awaiting `provider.complete` sequentially BEFORE the run starts. A new sync `ScriptedReasoner` then walks the pre-computed list with a private cursor. No shared state, no late writes from prior runs, and the example now actually demonstrates determinism rather than coincidentally passing. `runOnce()` now constructs both the provider AND the reasoner per call — the second run is a clean replay from a fresh provider with identical scripts, exactly as the example claims. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * fix(example): only swallow queue-exhausted errors in precomputeDecisions Per Codex review (P2): the bare `catch` masked every `provider.complete(...)` failure as queue exhaustion, so a real regression — budget enforcement change, malformed-request handling, runtime error in a future adapter — would silently downgrade to `''` and the example would still print success. That defeats the example's job as a smoke test for the LLM port. Narrow the catch via an `isQueueExhausted(err)` predicate that matches on `MockLlmProvider`'s documented error message ("script queue exhausted"). Anything else rethrows. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Luis Mendez <hallo@luis-mendez.de> Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This was referenced Apr 25, 2026
Luis85
added a commit
that referenced
this pull request
Apr 26, 2026
## Summary
Doc-audit pass over `docs/plans` + `docs/specs`. Three things land
together:
- **`docs/archive/{plans,specs}/`** — new home for plans whose roadmap
rows have all shipped (or whose goals were folded into a successor)
and specs whose design is now reflected in code. Includes a
`README.md` explaining the policy; `CLAUDE.md` documents the
convention.
- **`git mv` 23 plans + 3 specs into the archive.** The active live
set is now the comprehensive polish-and-harden plan plus three
specs (post-tfjs improvements, mvp-demo, vision), each with a
refreshed status banner.
- **Refresh the live comprehensive plan** against current `develop`:
- PR column updated for rows 16/19/20/3/4/22 (now shipped via
PRs #91 / #98 / #104 / #110 / #113 / #111).
- New "Post-roadmap follow-ups" section covers PRs #92 → #125
(review-bot infra, tracker findings, demo + tfjs hotfixes,
tooling).
- Stale prose-baked counts dropped (size budgets now reference
`package.json#size-limit` only).
- Coverage-thresholds section gains a pointer to the sticky PR
comment shipped in PR #124.
## Other doc fixes
- `README.md`: drop the unverifiable "Phase A milestones (M0–M15) are
all green" claim — the milestones don't exist as documented IDs
anywhere; replace with a pointer to the live polish plan.
- `vision.md`: refresh cadence note (was pinned to 2026-04-19 + "next
review at 1.0").
- `2026-04-24-post-tfjs-improvements.md`: mark recommended-order items
that have shipped (PRs #61, #76, #77, #83, #84, #91, #94, #96,
#104, #113), link the active roadmap as the heir.
- `mvp-demo.md`: status banner explaining where active polish work is
now tracked.
## Mechanical
- Update inline cross-refs in `CLAUDE.md`, `eslint.config.js`,
`src/agent/{Agent,AgentModule}.ts`, `tests/unit/exports.test.ts`,
and `docs/daily-reviews/2026-04-25.md` to point at the new
`docs/archive/` paths so links keep resolving.
No code change beyond comment-path updates.
## Test plan
- [x] `npm run verify` green (`format:check` + `lint` + `typecheck` +
`test` + `build` + `docs`). 523 tests pass; the 2 lint warnings
are pre-existing (`CognitionPipeline.invokeSkillAction` complexity
+ `scoreFailure` param count) and on the ratchet menu.
- [x] `git ls-files docs/archive/` shows the moved files; renames are
preserved (`git log --follow` works for any moved file).
- [ ] Codex review: clean, no blockers.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
---------
Co-authored-by: Luis Mendez <hallo@luis-mendez.de>
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
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
2 participants
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.
Summary
.github/workflows/release.ymlwith a singlenpm run verifycall. Future verify changes auto-propagate; release also gains thedocs(typedoc) safety net.andresz1/size-limit-actionjob to PR CI. Posts a sticky comment with the gzip delta for each of the 5 budgets inpackage.json#size-limit(main 50 KB, integrations 2 KB each, tfjs 7 KB).permissions: pull-requests: writeso the bot can comment.Roadmap row 1 of
docs/plans/2026-04-25-comprehensive-polish-and-harden.md.Test plan
npm run verifygreen locally.main.Notes for review
size-limit-commentjob runs only onpull_requestevents (uses base ref to compute delta).buildso it shares the artifact ordering with the existingEnforce bundle-size budgetstep rather than racing with it.🤖 Generated with Claude Code