feat: add automatic report generation on build

[zeke]

This PR implements automatic report generation as requested in issue #9.

Changes

• Added script/report.js to generate comprehensive model statistics report
• Updated build process to include report generation
• Report includes models with most runs this week, most runs ever, new models, and recently updated models
• Generated report is saved as report.md in repository root

Testing

Run npm run build to generate the report and verify the output.

Fixes #9

Generated with Claude Code

[claude]

Claude finished @zeke's task —— View job

---

PR Code Review

I've completed a thorough review of this automatic report generation feature. Here's my feedback:

Todo List:

• Read package.json changes
• Analyze script/report.js implementation
• Check for CLAUDE.md repo guidelines
• Review code quality and best practices
• Check for potential bugs or issues
• Assess performance considerations
• Evaluate security concerns
• Check test coverage
• Provide comprehensive review feedback

✅ Strengths

• Clean integration: The build script addition in package.json:20 is simple and logical
• Good structure: Well-organized code with clear sections and comments
• Proper error handling: Catches errors and exits with non-zero code
• Export support: Function is exported, making it testable
• Readable output: Generated markdown report is well-formatted

⚠️ Issues & Concerns

Critical Bug - New Model Detection (script/report.js:52-54)

const createdDate = new Date(model.url.includes('created_at=') ? 
  decodeURIComponent(model.url.split('created_at=')[1]) : 
  new Date(0)); // fallback for models without creation date in URL

Problem: This logic is fragile and unreliable:

• Assumes creation date is always in URL parameters
• URL parsing can easily break if URL format changes
• Fallback to new Date(0) (1970) could incorrectly flag old models as "new"

Recommendation: Consider using model metadata fields or a more reliable date source.

Performance Considerations

• Memory usage: Loads entire datasets into memory (script/report.js:9-10)
• Multiple iterations: Processes the same data multiple times (lines 21-29, 39-47, 62-80)
• Potential optimization: Could combine some loops or use streaming for large datasets

Code Quality Issues

• Magic numbers: Hardcoded values (7 days, 3 days, 10 results) should be constants
• Date arithmetic: Repeated date calculations could be extracted to helper functions
• No input validation: Doesn't verify that required JSON files exist before reading

Missing Test Coverage

• No tests provided for the new functionality
• Should test:
  • Report generation with sample data
  • Edge cases (empty data, malformed dates)
  • Error handling scenarios

🔒 Security Assessment

• Low risk: Only reads local JSON files and writes markdown
• No external dependencies: Uses only Node.js built-ins
• Proper file handling: Uses async/await for file operations

📋 Recommendations

1. Fix new model detection: Replace URL parsing with a more reliable method
2. Add constants: Extract hardcoded values to named constants
3. Add tests: Create test cases for the report generation logic
4. Optimize performance: Consider combining data processing loops
5. Add validation: Check file existence before reading
6. Extract helpers: Create date utility functions

📊 Impact Assessment

• Low risk change: Adds optional reporting without affecting core functionality
• Build integration: Properly integrated into existing build process
• Backwards compatible: No breaking changes to existing code

The implementation addresses the core requirements well but has some reliability issues that should be addressed before merging.

---