- Improve cache key generation using JSON.stringify for consistent formatting
- Add promise deduplication in DrawingCard to prevent redundant API calls for full drawing data
- Clear full data state when drawing ID changes to ensure fresh data loading
- Fix async cache invalidation in drawing update and collection delete endpoints
- Move cache invalidation after database operations in SQLite import endpoint
- Add HydratedDrawingData type for better type safety in drawing data management
- Moved BACKEND_URL configuration to the frontend service in the README.
- Added validation for the generated nginx configuration in the entrypoint script.
- Removed fallback nginx configuration copy from the Dockerfile.
- Adjusted nginx template to ensure proper header formatting.
This improves the deployment process and clarifies configuration instructions.
- Add 5s in-memory cache for /drawings responses with automatic cleanup
- Split Drawing/DrawingSummary types for efficient data fetching
- Implement lazy loading of drawing data in DrawingCard component
- Add configurable DRAWINGS_CACHE_TTL_MS and RATE_LIMIT_MAX_REQUESTS env vars
- Prevent memory leaks with periodic cleanup of cache and rate limit maps
- Add loading states and better UX for export operations
- Improve JSON parsing with error handling for malformed stored data
Benchmark results (100 drawings, cached):
- Avg latency: 6.94ms (p50: 4ms, p97.5: 8ms)
- Avg throughput: 668 req/s (peak: 1,023)
- 3k requests in 5s with 0 errors
Update .gitignore to exclude generated files, env files, and build artifacts
- Added BACKEND_URL environment variable to docker-compose for frontend service.
- Introduced a new entrypoint script to configure nginx with the BACKEND_URL at runtime.
- Created a template for nginx configuration to handle API and WebSocket requests dynamically.
- Updated README with instructions for configuring reverse proxy setups.
Fixes#12
Zimeng Xiong