🎙️ AI Meeting Lead (Action-Items Extractor)

A sophisticated Full-Stack AI Application that transforms voice recordings and audio files into organized, actionable task lists. No more manual note-taking—let the AI handle your meeting minutes.

🚀 Live Demo

Experience the tool here: [Your-HuggingFace-Link-Here]

✨ Core Features

• Live Voice Recording: Record your meetings directly from the browser using the MediaRecorder API.
• Audio File Processing: Upload existing MP3/WAV files for instant analysis.
• Speech-to-Text (ASR): Powered by OpenAI's Whisper-Tiny for high-accuracy transcription.
• Action-Item Extraction: Uses the BART model to distill long conversations into concise Bullet Points.
• On-Demand Translation: Integrated English-to-Arabic translation for the extracted tasks.

🛠️ Technical Architecture

The project follows a modern asynchronous flow:

1. Frontend: JavaScript captures audio as a Blob and transmits it via Multipart/form-data.
2. API Layer: FastAPI handles the file stream and manages temporary storage.
3. AI Engine:
   • Whisper converts audio waves to raw text.
   • BART-Large-CNN filters the text to extract specific tasks.
   • Helsinki-NLP provides the final translation layer.
4. Deployment: Containerized using Docker with ffmpeg integration for audio decoding.

📦 Project Structure

• main.py: The FastAPI backend logic and AI pipelines.
• index.html: Modern violet-themed UI with recording capabilities.
• Dockerfile: Custom container configuration for Hugging Face Spaces.
• packages.txt: System-level dependencies (ffmpeg).
• requirements.txt: Python library dependencies.

⚙️ How to Run Locally

1. Clone the repository.
2. Install FFmpeg on your system.
3. Install dependencies: pip install -r requirements.txt.
4. Launch the app: uvicorn main:app --reload.

---

Built with 🧠 and 🎙️ by PhilopateerDev