README.md

---
title: Meeting Minutes Generator
emoji: 📝
colorFrom: blue
colorTo: purple
sdk: gradio
sdk_version: 5.49.1
app_file: app.py
pinned: false
license: mit
---

# 📝 Meeting Minutes Generator

Upload meeting audio and get professionally formatted minutes instantly!

## 🚀 Features

- **Audio Transcription** using Groq Whisper Large v3
- **Smart Formatting** using Groq GPT-OSS-120B
- **Instant Results** in clean Markdown format
- **No Login Required** - privacy-focused
- **Free to Use** - no hidden costs

## 🎯 How to Use

1. **Upload** your meeting audio file (MP3, WAV, M4A, WEBM)
2. **Click** "Generate Minutes"
3. **Wait** ~30-40 seconds for processing
4. **Copy** your formatted minutes

## 📊 Supported Files

- **Formats:** MP3, WAV, M4A, WEBM, FLAC
- **Max Size:** 25 MB (~25 minutes at 128kbps)
- **Languages:** English (optimized)

## 🛠️ Tech Stack

- **Backend:** FastAPI
- **Frontend:** Gradio
- **Transcription:** Groq Whisper Large v3 (temperature: 0.1)
- **Minutes Generation:** Groq GPT-OSS-120B (120B parameters, temperature: 0.1)

## 💻 Local Setup

If you want to run this locally:

1. Clone the repository
2. Install dependencies: `pip install -r requirements.txt`
3. Add your `GROQ_API_KEY` to `.env` file
4. Run: `python app.py`
5. Open: http://localhost:7860

## 🔒 Privacy & Security

- ✅ No data is stored or logged
- ✅ Audio files are processed in memory only
- ✅ Temporary files are deleted after processing
- ✅ Your meetings stay completely private
- ✅ Open source - audit the code yourself

## 📋 Output Format

The generated minutes include:
- **Meeting metadata** (date, attendees)
- **Summary** (2-3 sentence overview)
- **Key discussions** (structured agenda items)
- **Action items** (with owners and deadlines when mentioned)
- **Open issues/concerns**
- **Additional context** (when relevant)

## 🎓 Use Cases

Perfect for:
- Team standups and sprint reviews
- Client meetings and consultations
- Board meetings and executive sessions
- Educational lectures and seminars
- Medical consultations (keep HIPAA in mind)
- Legal depositions (with appropriate consent)

## ⚡ Performance

- **Transcription**: ~30 seconds for 10-minute audio
- **Minutes Generation**: ~5 seconds
- **Total Processing**: ~35-40 seconds end-to-end

## 🔧 Technical Details

### API Endpoints (Backend)
- `POST /transcribe` - Audio to text transcription
- `POST /generate-minutes` - Text to formatted minutes
- `GET /` - Health check

### Models Used
- **Whisper Large v3**: 1550M parameters, multilingual
- **GPT-OSS-120B**: 120B parameters, optimized for structured output

### Processing Pipeline
```
Audio Upload → Validation → Whisper Transcription → 
LLM Formatting → Markdown Output → Copy/Download
```

## 🤝 Contributing

This is an MVP. Future enhancements could include:
- Speaker diarization (who said what)
- Multi-language support (Hindi, Spanish, etc.)
- PDF/DOCX export
- Meeting history with search
- Real-time transcription
- Calendar integration

## 👨‍💻 Built By

Made with ❤️ by Jai

## 📄 License

MIT License - Free to use and modify

---

**⚠️ Note**: This app requires a Groq API key. Add yours in the HF Spaces Settings → Repository Secrets with the name `GROQ_API_KEY`.
```

---

## 📊 Files Summary for HF Upload

**Upload these 6 files to HF Spaces:**

1. ✅ `app.py` (NEW - just gave you)
2. ✅ `backend.py` (your existing file)
3. ✅ `frontend.py` (your existing file)
4. ✅ `requirements.txt` (your existing file)
5. ✅ `README.md` (updated version above)
6. ✅ `.gitignore` (NEW - just gave you)

---

## ✅ Final Checklist

**In Cursor, verify you now have:**
```
hackathon/
├── app.py              ✅ NEW
├── backend.py          ✅ Existing
├── frontend.py         ✅ Existing
├── requirements.txt    ✅ Existing
├── README.md           ✅ Updated
└── .gitignore          ✅ NEW