The source-available audio lecture platform & semantic intelligence engine
Shruti Engine powers rich audio lecture applications — listen to audio content, follow along with synchronized transcripts, take notes, and interact with an AI companion grounded in the transcript corpus.
Shruti Engine is structured as a clean, modular microservice platform:
chat(Python / FastAPI / LangGraph): Semantic AI companion grounded in lecture transcripts with exact timestamp attributions.discovery(Go / PgVector): Semantic search, vector embeddings, entity extraction, and content discovery.auth(Go): JWT authentication supporting OAuth (Google, Apple) and device identities.orchestrator&ingest(Go): Audio processing pipeline orchestration and worker jobs.billing&profile(Go): User profile, preferences, listen history sync, and subscription entitlement management.share-audio,share-video,share-transcript(Go / Python): Dynamic snippet rendering (audiograms, subtitles, waveforms, and cards).shruti-corpus-mcp(Go): Model Context Protocol (MCP) server for corpus querying and tools.storage-sync&cleanup-worker(Go): Object storage sync, cache invalidation, and background maintenance.
mobile(Ionic / Capacitor / Vue 3): Offline-first cross-platform mobile player with synced transcripts, embedded SQLite database, and background audio controls.
transcriber-service: Speech recognition (ASR) and transcript alignment service.denoiser-service: Neural audio denoiser.
app/compose/: Docker Compose configurations for local development and production VPS.app/compose/caddy/: Reverse proxy with automatic SSL and metrics routing.observability/: Prometheus, Loki, Grafana, and alerting configurations.
# Setup development environment and generate local secrets
make stack-setup
# Start the full stack with Docker Compose
make stack-up
# Check service health and readiness
make stack-statusShruti Engine is source-available under the PolyForm Noncommercial License 1.0.0.

