TaskWhisper

TaskWhisper turns spoken notes into cleaned transcripts, extracted tasks, summaries, and optional exports.

Architecture

• backend/: FastAPI API, Firebase auth, local JSON storage, in-process cache
• frontend/: Next.js app for recording, reviewing results, and configuring providers

Runtime data is stored locally under backend/data/, and uploaded audio is stored under backend/data/uploads/ by default.

Quickstart

1. Create a backend virtual environment and install dependencies:

cd backend
python -m venv venv
.\venv\Scripts\activate
pip install -r requirements.txt

2. Copy backend/.env.example to backend/.env.local or backend/.env and fill in the values you need.
3. Start the backend:

cd backend
uvicorn main:app --reload

4. In a second terminal, install frontend dependencies and start Next.js:

cd frontend
npm install
npm run dev

5. Open http://localhost:3000. The backend runs on http://localhost:8000.

Configuration

Backend

• LOCAL_DB_PATH: path to the local JSON data file
• DATABASE_URL: optional PostgreSQL connection string
• POSTGRES_URL: supported alias for DATABASE_URL
• REDIS_URL: Redis connection string used for LLM and pipeline result caching
• FILE_STORAGE_PATH: path for uploaded and intermediate audio files
• BLOB_READ_WRITE_TOKEN or taskwhisper_VERCEL_READ_WRITE_TOKEN: Vercel Blob token used to store uploaded voice files
• FIREBASE_PROJECT_ID: Firebase project used for token verification
• ALLOW_INSECURE_AUTH=true: allows local development without strict Firebase token validation
• SUPABASE_URL: Supabase project URL for TaskWhisper Cloud sync
• SUPABASE_SERVICE_ROLE_KEY: Supabase service role key used by the backend sync route
• SUPABASE_SCHEMA: Supabase PostgREST schema, defaults to public

Frontend

• NEXT_PUBLIC_BACKEND_URL: backend base URL, defaults to http://localhost:8000
• Firebase client variables in frontend/.env.example

Pipeline

1. Voice cleaning
2. Speech-to-text
3. Text cleanup and optional translation
4. Task extraction
5. Optional export connectors

Progress updates stream over WebSocket at /ws/jobs/{job_id}.

Uploaded voice files are stored in Vercel Blob when a blob token is configured. During processing, the backend downloads the blob to a temporary local file so the current cleanup and STT adapters can read it.

TaskWhisper Cloud Sync

Cloud sync is handled by the backend, not directly by the browser. When a signed-in user clicks sync in Settings, the frontend sends its current notes and extracted tasks to the backend, and the backend upserts them into Supabase using the service-role key.

Before using it:

1. Apply supabase/taskwhisper_cloud_schema.sql in your Supabase SQL editor.
2. Set SUPABASE_URL and SUPABASE_SERVICE_ROLE_KEY in the backend environment.
3. Restart the backend.