create fake test repository to eval suggestions

[dielduarte]

Task: Create Test Repository for Evaluating Documentation Suggestions

Goal: Build a realistic test repository with documentation and code that can be used to systematically evaluate JanusDoc's suggestion quality across different scenarios.

Success Criteria:

• Repository contains realistic documentation (API docs, guides, architecture)
• Multiple test PRs covering different change types
• Clear baseline for what suggestions should be made
• Can be used for regression testing as JanusDoc evolves

Repository Contents Needed:

1. Documentation Structure (/docs):
   - API reference (functions, classes, types)
   - Getting started guide
   - Architecture overview
   - Configuration guide
   - At least 10-15 markdown files
2. Codebase (TypeScript/JavaScript):
   - Simple but realistic project (e.g., task manager, blog API, or CLI tool)
   - Code that actually matches the documentation initially
   - Multiple modules/files that docs reference
3. Test Scenarios via PRs (at least 6-8):
   - Obvious updates needed: Add new API endpoint not documented
   - Subtle changes: Rename function parameter that's referenced in docs
   - Breaking changes: Change function signature documented in API ref
   - New feature: Add feature that needs guide section
   - Deprecation: Mark something as deprecated
   - No update needed: Internal refactoring that doesn't affect public API
   - Configuration change: Add new config option
   - Behavior change: Modify how existing feature works
4. Expected Results Document:
   - For each PR, document what suggestions JanusDoc should make
   - Include specific doc files and sections that should be flagged
   - Note edge cases or tricky scenarios

Deliverables:

• GitHub repository (use janusdoc-evals)
• README explaining the test structure
• Branches and PRs set up for each test scenario
• EXPECTED_RESULTS.md with baseline expectations

Bonus:

• Integration with https://www.evalite.dev/ for automated evaluation
• Evalite test suite to run JanusDoc on all test PRs and score against expected results
• Metrics tracking (precision, recall, false positives)
• Different documentation styles to test (API-first vs guide-first)