Commit graph

5 commits

Author SHA1 Message Date
Lasse Rune Hansen
aa3db295ce fix(backend): Fix Mistral story generation with proper chat completion format
- Update MistralChatRequest to remove unused properties (User, Stop, FrequencyPenalty, PresencePenalty) causing 422 errors
- Add explicit JsonPropertyName attributes to ensure correct JSON serialization
- Update MistralService to send minimal requests matching Mistral API expectations
- Pass SegmentCount and CustomPrompt through StoryGenerationService to build proper prompts
- Add request logging in MistralConnector to debug API calls
- Update default model to mistral-small-latest
- Enhance error handling to log raw JSON responses

Generated by Mistral Vibe.
Co-Authored-By: Mistral Vibe <vibe@mistral.ai>
2026-06-18 15:21:04 +02:00
Mistral Vibe
518d1c8f27 feat(backend/story-integration): Phase 5 - Audio Generation
- Fixed audio file path generation in StoryGenerationService.GenerateAudioAsync
- Audio files now stored at wwwroot/audio/story/level{id}-segment{order}.wav
- Added static file serving in Program.cs (app.UseStaticFiles)
- Added wwwroot/audio/story directory creation on startup
- Updated StoryController.GetSegmentAudioAsync to return audio URL
- Static file middleware serves audio files automatically
- All 324 tests still pass

Generated by Mistral Vibe.
Co-Authored-By: Mistral Vibe <vibe@mistral.ai>
2026-06-13 14:47:32 +02:00
Mistral Vibe
536382641d feat(backend/story-integration): Phase 4 - AI Integration with CEFR-level prompts
All checks were successful
ci/woodpecker/push/woodpecker Pipeline was successful
- Enhanced StoryGenerationService with CEFR-level-specific prompt templates
- Added GetStoryRequirements() method with level-specific requirements for A1-C1
- Added GetSegmentRequirements() method with level-specific requirements
- Story prompts now include detailed CEFR-level constraints:
  - A1: Simple present tense, basic SVO structure, max 500 words
  - A2: Present/past/future, subordinate clauses, max 1000 words
  - B1: All tenses, modal verbs, complex sentences, max 2000 words
  - B2: Nuanced tenses, all cases, Konjunktiv I/II, Passiv, max 3000 words
  - C1: All tenses/moods, sophisticated vocabulary, literary devices
- Updated docs/features/story-integration.md for Phase 4 completion
- All 324 tests still pass

Generated by Mistral Vibe.
Co-Authored-By: Mistral Vibe <vibe@mistral.ai>
2026-06-13 14:38:30 +02:00
Lasse Rune Hansen
6693165f83 feat(backend/story-integration): Phase 2 - Backend Services
Implement story integration backend services:
- Create StoryService (Application/Services/StoryService.cs) with full CRUD operations
- Create StoryGenerationService (Application/Services/StoryGenerationService.cs) for AI-powered story generation
  - Uses IMistralService for text generation
  - Uses ITtsService for audio generation
  - Splits stories into segments per lesson
- Create StoryUnlockService (Application/Services/StoryUnlockService.cs) for progress management
  - Handles lesson completion → story segment unlocking
- Create StoryController (Presentation/Controllers/StoryController.cs) with 12 endpoints:
  - GET /api/story/levels - list levels with stories
  - GET /api/story/levels/{levelId}/segments - get segments for level
  - GET /api/story/segments/{id} - get specific segment
  - POST /api/story/segments - create segment (Admin)
  - PUT /api/story/segments/{id} - update segment (Admin)
  - DELETE /api/story/segments/{id} - delete segment (Admin)
  - POST /api/story/levels/{levelId}/generate - generate story with AI (Admin)
  - POST /api/story/segments/{segmentId}/audio - generate audio (Admin)
  - GET /api/story/levels/{levelId}/progress - get user progress
  - POST /api/story/segments/{segmentId}/complete - mark as completed
  - GET /api/story/levels/{levelId}/next - get next segment
  - GET /api/story/segments/{segmentId}/unlocked - check if unlocked
  - GET /api/story/segments/{segmentId}/audio - get audio URL
  - GET /api/story/levels/{levelId}/lessons-with-stories - get lessons with story status
- Register all services in Program.cs DI container
- Update feature document to reflect Phase 2 completion

Next: Phase 3 - AI Integration (unit tests for services)

Generated by Mistral Vibe.
Co-Authored-By: Mistral Vibe <vibe@mistral.ai>
2026-06-13 13:37:16 +02:00
Lasse Rune Hansen
e002868b74 feat(backend/application): implement Phase 5 AI Service Integration
- Create higher-level AI services:
  - StoryGenerationService (uses MistralService)
  - WritingFeedbackService (uses MistralService)
  - SpeechExerciseService (uses VoskService)
  - AudioGenerationService (uses TtsService)
  - AiFallbackService (fallback mechanisms for service failures)
- Register AiFallbackService in Program.cs DI container
- Add comprehensive unit tests for all Phase 5 services:
  - AiFallbackServiceTests (14 tests)
  - AudioGenerationServiceTests (16 tests)
  - MistralServiceTests (16 tests)
  - SpeechExerciseServiceTests (12 tests)
  - StoryGenerationServiceTests (13 tests)
  - WritingFeedbackServiceTests (11 tests)
  - VoskServiceTests (15 tests)
  - TtsServiceTests (21 tests)
- Update feature document (ai-services.md) to mark Phase 5 as complete

Generated by Mistral Vibe.
Co-Authored-By: Mistral Vibe <vibe@mistral.ai>
2026-06-13 12:44:10 +02:00