A full-stack web application implementing secure, scalable music streaming with HTTP Live Streaming (HLS) protocol and asynchronous audio processing.
Initial test deployment
- Frontend: React 19 SPA with TypeScript, HLS.js for adaptive streaming, Zustand for state management
- Backend: FastAPI with async request handling, SQLAlchemy ORM for PostgreSQL integration
- Task Queue: Celery workers with Redis broker for asynchronous audio processing
- Storage: AWS S3 for static asset delivery with direct public access and CORS configuration
- Database: PostgreSQL with pgvector extension for audio feature embeddings
Audio files are transcoded to HLS format using FFmpeg, generating multiple quality variants (64kbps, 128kbps, 192kbps) with MPEG-TS segments and M3U8 playlists. The client implements HLS.js with custom loaders for seamless playback and bandwidth adaptation.
- Encryption: AES-128 encryption applied during HLS conversion with track-specific 16-byte keys
- Authentication: JWT-based access control with role-based authorization (user/admin)
- Key Management: Encryption keys stored in PostgreSQL, served through authenticated endpoints
- Rate Limiting: IP-based throttling on key retrieval endpoints
Celery workers execute the following tasks asynchronously:
- Format normalization and validation
- Multi-bitrate HLS transcoding with AES-128 encryption
- Audio feature extraction: MFCCs, spectral centroids, chroma features, zero-crossing rate
- Metadata parsing and database insertion
- S3 upload with public-read ACL and proper MIME types
- Listening session management with heartbeat monitoring
- Per-track play count aggregation
- User listening history with temporal granularity
- Foundation for content-based recommendation system
- Implement Redis caching for track metadata and popular songs
- Build content-based recommendation system using audio embeddings
- Implement collaborative filtering for user recommendations
- Add playlist management system
- Fix UI design