# MeetVault

Self-hosted meeting recording and intelligence platform. The user machine handles local transcription; the backend API manages auth, metadata, uploads, summarization jobs, tasks, and reminders. Large media files are stored in SeaweedFS via its S3-compatible API.

## Architecture

```text
User machine:
  - Record audio/video
  - Local Whisper STT → transcript
  - Upload to SeaweedFS (presigned URLs)
  - Send transcript to backend

Backend API:
  - Auth, users, subscriptions
  - Meeting metadata
  - Transcript ingestion
  - Summarization jobs
  - Task extraction
  - Reminder scheduling
  - PostgreSQL for structured data
```

## Quick start

### Backend

1. Start PostgreSQL and run the schema SQL from `Overview/README.md` (includes `pgcrypto` extension, all tables, indexes, triggers).
2. Run the backend API server.
3. Spin up SeaweedFS and configure presigned URL endpoints on the backend.

### Desktop client

Run the Tauri + React demo in `Wireframes/Desktop/MeetVault-Desktop-Demo/`. It supports local Whisper transcription, speaker diarization via diart, live overlay window, recording capture, and mock speech-engine mode for UI testing without ML models.

## Stack (MVP)

Frontend: React / Next.js  
Backend: FastAPI or Node.js  
Database: PostgreSQL + `processing_jobs` table as job queue  
Media storage: SeaweedFS  
Local STT: Whisper / whisper.cpp  

No Kafka, Elasticsearch, Kubernetes, vector DBs, or Redis are needed for the MVP.