fix: use local binary for hooks + parallelize translation workflow

[NiveditJain]

Summary

• Replace npx -y failproofai with bun ./bin/failproofai.mjs in committed .claude/settings.json to fix self-referencing conflict when developing inside the failproofai repo itself
• Add contributor note in README.md and CLAUDE.md explaining this is specific to this repo
• Parallelize translate-docs.yml workflow: 14-language matrix + consolidation job
• Parallelize file translations within each job: upfront cache filtering, concurrent API calls (max 10), batch cache write
• Fix cache merge to only use each fragment's authoritative language entries
• Consolidation job runs on partial failure (if: always()) so successful translations aren't dropped
• Add workflow_dispatch inputs: force (ignore cache) and languages (comma-separated filter) for manual translation runs

Test plan

• bun ./bin/failproofai.mjs --hook PreToolUse runs successfully
• All 929 unit tests pass (bun run test:run)
• Type check passes (bunx tsc --noEmit)
• Hooks fire correctly in this repo
• bun run translate --languages zh --dry-run works with concurrency limiter
• Translation workflow runs 14 parallel jobs and consolidation creates PR
• Manual dispatch with force and languages inputs works

🤖 Generated with Claude Code

[coderabbitai]

📝 Walkthrough

Walkthrough

This PR updates the failproofai hook configuration to use a local Bun-based script instead of npx invocation, modifies the translation CI workflow to run translations in parallel across 14 languages with cache consolidation, and documents the repository-specific hook setup changes.

Changes

Cohort / File(s)	Summary
Hook Configuration \.claude/settings.json, CLAUDE.md, README.md, CHANGELOG.md	Updated all hook commands to invoke bun ./bin/failproofai.mjs --hook <EventType> instead of npx -y failproofai. Added documentation explaining the local binary approach due to self-referencing conflicts within this repository, with guidance for other projects to use the standard failproofai policies --install command.
CI/CD Workflow Parallelization \.github/workflows/translate-docs.yml	Replaced single sequential translation execution with a language-level parallel matrix strategy running bun run translate --languages ${{ matrix.lang }} per language. Added per-language artifact uploads and cache fragments, plus a new consolidate job that merges translation caches via inline Node script before saving the unified cache and proceeding with the PR-creation flow.

Estimated code review effort

🎯 3 (Moderate) | ⏱️ ~22 minutes

Possibly related PRs

• [ef-46] fix: bundle CLI for Node.js — support npm install -g without bun #46: Modifies CLI invocation and packaging to support different runtimes; related to how the local binary is invoked in hooks.
• [luv-54] docs: rewrite README to focus on hooks management #54: Updates README documentation regarding hooks management; related to the contributor-facing hook configuration guidance added here.

Poem

🐰 With Bun so swift and hooks held tight,
Translations fly in parallel flight,
Cache fragments merge where paths align,
No npx conflicts—this repo's just fine!
Local binaries hop, workflow runs true, ✨🥕

🚥 Pre-merge checks | ✅ 2 | ❌ 1

❌ Failed checks (1 inconclusive)

Check name	Status	Explanation	Resolution
Description check	❓ Inconclusive	The pull request description provides a clear summary of changes and test plan, but does not follow the required template structure with Type of Change and Checklist sections.	Restructure the description to match the template: add a Type of Change section (selecting Bug fix, New feature, Refactor, and/or Documentation) and complete the Checklist section by marking which tests were run.

✅ Passed checks (2 passed)

Check name	Status	Explanation
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.
Title check	✅ Passed	The pull request title accurately captures the two main changes: fixing hooks to use a local binary and parallelizing the translation workflow.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch luv-98

---

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.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

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

[coderabbitai]

Actionable comments posted: 1

🧹 Nitpick comments (1)

.github/workflows/translate-docs.yml (1)

18-21: Avoid duplicating the tier-3 language list in CI.

This matrix hardcodes the same selection logic that scripts/translate-docs/cli.ts already owns. A future language add/remove or tier change will make bun run translate --tier 3 and this workflow disagree. Consider generating the matrix from the same source of truth, or add a guard that fails when the lists drift.

🤖 Prompt for AI Agents

Verify each finding against the current code and only fix it if needed.
In @.github/workflows/translate-docs.yml around lines 18 - 21, The CI workflow
hardcodes the tier-3 language matrix (lang) which duplicates the logic in
scripts/translate-docs/cli.ts and can drift from bun run translate --tier 3;
change the workflow to derive or validate the matrix from the single source of
truth: either invoke scripts/translate-docs/cli.ts (or a new exportable function
it uses) at workflow runtime to emit the lang list, or add a step that runs a
script which compares the workflow's lang list against
scripts/translate-docs/cli.ts and fails the job if they differ; update
references to the matrix name (lang) and the script invocation (bun run
translate --tier 3 or the exported function) so maintainers can locate and keep
a single canonical list.

🤖 Prompt for all review comments with AI agents

Verify each finding against the current code and only fix it if needed.
Inline comments:
In @.github/workflows/translate-docs.yml:
- Around line 93-115: The merge currently uses
Object.assign(merged.translations, data.translations) which lets a stale
full-snapshot overwrite other languages; change the logic to merge per-language
using data.lastUpdated to avoid clobbering: create a map (e.g.,
mergedLastUpdatedByLang) and for each file's data iterate
Object.keys(data.translations) and for each lang replace
merged.translations[lang] only if mergedLastUpdatedByLang[lang] is missing or
data.lastUpdated > mergedLastUpdatedByLang[lang], update
mergedLastUpdatedByLang[lang] and set merged.lastUpdated =
max(merged.lastUpdated, data.lastUpdated); remove the Object.assign usage and
ensure the final write still writes merged (and sourceHash handling if needed).
---
Nitpick comments:
In @.github/workflows/translate-docs.yml:
- Around line 18-21: The CI workflow hardcodes the tier-3 language matrix (lang)
which duplicates the logic in scripts/translate-docs/cli.ts and can drift from
bun run translate --tier 3; change the workflow to derive or validate the matrix
from the single source of truth: either invoke scripts/translate-docs/cli.ts (or
a new exportable function it uses) at workflow runtime to emit the lang list, or
add a step that runs a script which compares the workflow's lang list against
scripts/translate-docs/cli.ts and fails the job if they differ; update
references to the matrix name (lang) and the script invocation (bun run
translate --tier 3 or the exported function) so maintainers can locate and keep
a single canonical list.

🪄 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: defaults

Review profile: CHILL

Plan: Pro

Run ID: 0dd438bc-f482-472f-9ffd-900b057ad318

📥 Commits

Reviewing files that changed from the base of the PR and between 5d460f0 and ff33fa1.

📒 Files selected for processing (5)

• .claude/settings.json
• .github/workflows/translate-docs.yml
• CHANGELOG.md
• CLAUDE.md
• README.md

[NiveditJain]

@CodeRabbit review

[coderabbitai]

✅ Actions performed

Review triggered.

Note: CodeRabbit is an incremental review system and does not re-review already reviewed commits. This command is applicable only when automatic reviews are paused.