.coderabbit.yaml

# CodeRabbit configuration: drydock
language: en-US
early_access: true
enable_free_tier: true

tone_instructions: >-
  Reviewing AI-agent TypeScript code. Terse and direct. State problem and fix
  with code. Skip praise and rhetoric. Flag provider-pattern divergence,
  missed conventions, existing utilities, DRY violations. Only actionable
  feedback.

reviews:
  profile: chill
  request_changes_workflow: false
  high_level_summary: true
  high_level_summary_instructions: >-
    This summary will be read by an AI coding agent (Claude Code), not a human.
    Be direct and structured. Use a changelog with emojis: ✨ Added, 🔧 Changed,
    🐛 Fixed, 🗑️ Removed, ⚠️ Breaking, 🔒 Security. After the changelog, list
    any concerns or issues the agent should address as a flat bullet list — no
    prose, no praise, no context the agent already knows from the diff.
  high_level_summary_placeholder: "@coderabbitai summary"
  high_level_summary_in_walkthrough: false
  collapse_walkthrough: true
  changed_files_summary: false
  sequence_diagrams: false
  estimate_code_review_effort: false
  review_details: false
  review_status: true
  commit_status: true
  fail_commit_status: false
  poem: false
  in_progress_fortune: false
  assess_linked_issues: true
  related_issues: true
  related_prs: true
  suggested_labels: true
  auto_apply_labels: false
  labeling_instructions:
    - label: "second-opinion"
      instructions: |
        Suggest this label only for genuinely high-stakes security-sensitive changes
        (auth, authorization, cryptography, release trust, proxy/socket boundaries,
        workflow permissions, or cross-repository contracts), after a completed
        CodeRabbit source review. Applying it requires explicit maintainer or agent
        opt-in; never apply it automatically. Routine UI changes and generated charts
        do not qualify. The label triggers a one-time Greptile second-opinion summon
        via .github/workflows/greptile.yml.
  suggested_reviewers: false
  auto_assign_reviewers: false
  enable_prompt_for_ai_agents: true
  abort_on_close: true

  auto_review:
    enabled: true
    auto_incremental_review: true
    drafts: false
    # Default is default-branch-only; v1.6+ feature work lands via PRs into
    # dev/vX.Y integration branches and needs the same automatic reviews.
    base_branches:
      - "dev/*"
    ignore_title_keywords:
      - "WIP"
      - "DO NOT MERGE"
      - "[skip review]"

  path_filters:
    - "!**/node_modules/**"
    - "!**/dist/**"
    - "!**/build/**"
    - "!**/.next/**"
    - "!**/coverage/**"
    - "!**/*.min.js"
    - "!**/*.min.css"
    - "!**/package-lock.json"
    - "!**/pnpm-lock.yaml"
    - "!**/*.lock"
    - "!**/*.generated.*"
    - "!CHANGELOG.md"
    # Immutable release archives are byte-bound to their published source by
    # scripts/release-docs-archive.test.mjs. Review the manifest and contract.
    - "!content/docs/v1.7/**"
    - "!content/docs/v1.6/**"

  path_instructions:
    - path: "app/registries/**"
      instructions: |
        Registry providers follow a shared provider pattern. New registries should mirror the structure and error handling of existing ones — flag divergence.
    - path: "app/triggers/**"
      instructions: |
        Notification trigger providers follow a shared provider pattern. Flag divergence from existing triggers and any secret handling in logs.
    - path: "app/api/**"
      instructions: |
        API layer. Auth is enforced in middleware — don't flag missing auth checks in individual handlers.
    - path: "ui/tests/**"
      instructions: |
        UI test suite. Don't suggest error handling or refactors in test code.
    - path: "**/*.test.*"
      instructions: |
        Test files. Don't suggest error handling or null checks in test code.
    - path: "content/docs/**"
      instructions: |
        Documentation content. Review for accuracy against the code, not prose style.

  finishing_touches:
    docstrings:
      enabled: true
    unit_tests:
      enabled: true

  pre_merge_checks:
    docstrings:
      mode: "off"
    title:
      mode: "off"
    description:
      mode: "off"
    issue_assessment:
      mode: "warning"

  tools:
    biome:
      enabled: true
    shellcheck:
      enabled: true
    hadolint:
      enabled: true
    markdownlint:
      enabled: true
    yamllint:
      enabled: true
    gitleaks:
      enabled: true
    actionlint:
      enabled: true
    ast-grep:
      essential_rules: true
    github-checks:
      enabled: true
      timeout_ms: 90000
    languagetool:
      enabled: false

knowledge_base:
  opt_out: false
  learnings:
    scope: local
  issues:
    scope: local
  # Default filePatterns already cover CLAUDE.md / AGENTS.md / .cursorrules
  code_guidelines:
    enabled: true
  web_search:
    enabled: true
  pull_requests:
    scope: local

chat:
  auto_reply: true
  art: false