When a user makes an incorrect move in a puzzle, now only that move
is undone (like chess.com) instead of resetting the entire puzzle.
This improves usability by letting users retry from the same position
rather than having to redo all previous correct moves.
- Add 30s timeout to Stockfish evaluation to prevent hanging promises
- Add FEN validation and depth cap (max 30) to Stockfish API route
- Create ErrorBoundary component with specialized fallbacks for chess game and tutor
- Extract useChessSounds hook for better audio management
- Add React.memo to EvaluationBar and CapturedPieces for performance
- Add useMemo to CapturedPieces for sorted pieces calculation
- Improve tacticDetection to return empty array instead of "none" type
- Add filterMeaningfulTactics and hasTactics helper functions
- Translate hardcoded UI strings (stockfishLevel, download, evalChange)
- Update translations for EN, DE, FR, IT, PL
- Add uuid to Jest transformIgnorePatterns for ESM compatibility
- Update tests to use new filterMeaningfulTactics function
- Add tutor message guardrail to prevent rapid-fire messages
- Track last message by move index
- Wait for both player and opponent moves before speaking
- Speak immediately when player deviates from theory
- Add deviation dialog with three options:
- Continue Playing (Start Game) - transitions to game mode
- Undo & Return to Opening - returns to theory
- Explore This Variation - continues off-book practice
- Implement smooth game mode transition:
- Pass opening context from trainer to game
- Tutor welcomes player with context about their study
- Computer makes first move if needed in starting position
- Fix generate-test-fixtures script:
- Add safety check to prevent overwriting real data
- Restore full opening database (12,379 openings)
- Rebuild move index (12,377 sequences)
- Add clear warnings about test fixtures
- Update ChessGame to accept openingContext prop
- Update Tutor to display contextual greeting in game mode
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
This commit implements iOS/Android mobile app support using Capacitor
and adds a comprehensive opening training feature with LLM-powered
explanations.
## Mobile App Infrastructure
- Add Capacitor configuration for iOS/Android builds
- Create mobile build script that excludes API routes
- Update Next.js config for conditional static export
- Add layout components with generateStaticParams for static builds
- Generate 500+ static pages for offline mobile use
## Chess Engine Abstraction
- Create ChessEngine interface for pluggable implementations
- Add LocalEngine (GPL - uses stockfish.js in browser)
- Add RemoteEngine (proprietary - calls API server)
- Factory pattern selects engine based on environment
- Enables GPL compliance for web, proprietary for mobile
## Opening Training Feature
- Interactive opening repertoire training
- Move validation with engine-backed feedback
- LLM explanations using Gemini API
- Wikipedia integration for opening context
- Opening family grouping (e4, d4, c4, etc.)
- Session state management
- Real-time move feedback with evaluation
Components:
- OpeningSelector: Browse and select openings by family
- OpeningTrainer: Main training interface with chessboard
- MoveFeedback: Display move quality and LLM explanations
- WikipediaSummary: Show opening history and context
- ErrorBoundary: Graceful error handling
Services:
- openingLoader: Load and filter opening database
- engineService: Engine evaluation wrapper
- moveValidator: Validate moves against repertoire
- feedbackGenerator: Generate contextual feedback
- wikipediaService: Fetch and cache Wikipedia data
- sessionManager: Track training session state
## Wikipedia Integration
- Automatic Wikipedia article fetching for openings
- Client-side and server-side caching
- Sanitized summaries with proper formatting
- Link opening database to Wikipedia slugs
- API endpoints for on-demand fetching
## Docker Improvements
- Add entrypoint script for automatic data setup
- Fetch Wikipedia data on first container startup
- Generate opening move index automatically
- Remove generated data from git (public/openings/*.json, public/wikipedia/*.json)
- Add READMEs explaining data requirements
- Update .gitignore for generated files
## Dual Licensing Strategy
- Add LICENSING.md explaining dual licensing approach
- GPL-3.0 for web builds (includes Stockfish)
- Proprietary option for mobile builds (no GPL code)
- Single codebase, multiple licensing models
- Legal compliance documented
## API Endpoints
- POST /api/v1/llm/opening-explanation - Get LLM move explanations
- GET /api/v1/wikipedia/summary - Fetch Wikipedia summaries
## Type Updates
- Add openingTraining types
- Update Tutor component to use ChessEngine interface
- Add Gemini error handling types
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
- Fix page.test.tsx: Use stable mock router reference to prevent infinite loop
The useRouter mock was returning a new object on every render, causing
the useEffect with router dependency to re-run infinitely
- Update README to document resign button and analyze game features
- Add 'Clear All Data' button in Danger Zone section
- Implement confirmation dialog before clearing
- Clear all localStorage and redirect to onboarding
- Add translations for clearAllData, clearAllDataConfirm, and clearAllDataDescription in all 5 languages (EN, DE, FR, IT, PL)
- Style with red theme to indicate destructive action
- Change from getByText to getByLabelText for navigation buttons
- Buttons now use icon-only display with aria-labels for accessibility
- Fixes tests after UI simplification in previous commit
- Import ArrowLeft icon and useRouter from Next.js
- Add back button below header that navigates to home (/)
- Simplify Previous/Next buttons to show only chevron icons
- Remove text labels from navigation buttons for cleaner UI
- Add aria-labels for accessibility
- Consistent with play page navigation pattern
- Import lookupPossibleOpenings and buildMoveSequenceFromSteps
- Build move sequence from steps array up to currentIndex
- Lookup up to 5 possible openings per position
- Update prompt to include all possible openings
- Update UI to show single opening name or count of possibilities
- Consistent with play page opening detection
- Import ChatSession from @google/generative-ai
- Add chatSession state to maintain conversation history
- Initialize chat session with personality and language context
- Replace stateless generateContent() with chatSession.sendMessage()
- Add FEN before/after to move analysis prompts
- Chat history is preserved when navigating between moves
- LLM can now provide context-aware commentary across the game
- Fix TypeScript errors in analysis page with non-null assertions
- Fix Chessboard component API usage to use options prop
- Fix Analyze Game button placement to be always visible
- Create chessFormatDetector utility for automatic format detection
- Update StartScreen with textarea supporting both FEN and PGN input
- Add real-time format detection with visual feedback indicators
- Update translations for all 4 languages (EN, DE, FR, IT)
- Add markdown rendering for Tutor chat messages
- Add comprehensive tests for format detection (20 tests)
- Update game initialization to handle both FEN and PGN formats
This completes the fix/stale-analysis-data branch with:
- Fixed stale evaluation data in hint/best move requests
- Clarified AI's dual role (opponent + tutor) to prevent hint rejection
- Stored complete evaluation history (P0, P1, P2) for all moves
- Improved end-game analysis with better mistake detection
- Fixed duplicate analysis runs with useRef flag
- Added markdown rendering for formatted analysis output
- Added unified FEN/PGN import with auto-detection
This commit fixes a bug where the AI tutor and end-game analysis would use stale data from previous evaluations.
The `Tutor` component now receives the live `game` instance and has a new function, `evaluateCurrentPosition`, which is called on-demand when a user requests a hint or the best move. This ensures the LLM receives up-to-date information.
The end-game analysis was also corrected to use the proper evaluation data when constructing the move history, preventing incorrect analysis of mistakes and blunders.
The test suite was improved by restoring deleted tests, adding a new test to verify the fix, and making existing tests more robust.
Claude