Meeting Notes AI

A full-stack app that transcribes meeting audio and generates structured summaries, decisions, and action items.

Features

• Audio upload (wav/mp3)
• Full transcript
• Summary, participants, decisions, action items
• Word (DOCX) export
• Action items export as ICS
• MOCK_MODE for local demos without API keys

Architecture

• Backend: FastAPI, Whisper-compatible transcription API, LLM summarization (Gemini by default)
• Frontend: React (Vite)

System Prompt

The LLM system prompt used for summaries lives at:

• backend/app/prompts/meeting_summary_system.txt

Local Setup

Backend

cd backend
python -m venv .venv
.venv\\Scripts\\activate
pip install -r requirements.txt
copy .env.example .env
notepad .env
uvicorn app.main:app --reload

Frontend

cd frontend
npm install
npm run dev

Open http://localhost:5173 and set API Base URL to http://localhost:8000.

Environment Variables

Create backend/.env based on backend/.env.example.

• GEMINI_API_KEY: Gemini API key for summarization.
• GEMINI_MODEL: Optional, defaults to gemini-1.5-flash.
• WHISPER_API_URL: Whisper-compatible transcription endpoint (multipart file upload).
• WHISPER_API_KEY: API key for the transcription service.
• WHISPER_MODEL: Optional, defaults to whisper-1.
• MOCK_MODE: true to enable deterministic mock responses for demo.

Notes

• If you don’t have API keys, set MOCK_MODE=true to demo the full flow locally.
• WHISPER_API_URL should accept a file upload and return JSON with text or transcript.
• The Word export is generated server-side in backend/app/services/export_docx.py.
• Process documentation is in PROCESS.md.

Deliverables

• GitHub repo: https://github.com/Rivka387/meeting-notes-ai
• Live demo (or local demo instructions): https://meeting-notes-ai.onrender.com/