--- title: Personalized Movie Recommender Assistant emoji: 🎬 colorFrom: blue colorTo: purple sdk: gradio sdk_version: 5.33.1 app_file: app.py pinned: false license: apache-2.0 short_description: Your Al Movie Buddy tailored to your watching habits tags: - agent-demo-track --- # 🎬 Personalized Movie Recommender A chatbot-style movie recommendation system built for the [🤖 Gradio Agents & MCP Hackathon 2025 🚀](https://huggingface.co/Agents-MCP-Hackathon). This project leverages the powerful Google `gemma-2b-it` model (quantized to 4-bit for speed and efficiency), FastAPI for serving, and Gradio for an engaging interactive UI. --- ## 🚀 Features ✅ **Personalized Movie Recommendations** - Understands user’s mood and genre preferences to deliver tailored suggestions. - Fully customizable: choose the agent’s personality, expertise level, and response language. ✅ **Gemma-2B-It Integration** - Efficient 4-bit quantized loading using `bitsandbytes`. - Runs on H100 GPUs for blazing-fast inference. ✅ **Multilingual Support** - Supports English, Korean, and Japanese. --- ## 🏗️ Architecture - **Backend**: FastAPI app deployed on Modal with GPU acceleration. - **Model**: Google’s `gemma-2b-it` with 4-bit quantization. - **Frontend**: Gradio chatbot UI for a natural conversation experience. - **Infrastructure**: Modal’s NetworkFileSystem for model caching and deployment. --- ## 🔧 How to Use ### Modal Deployment 1. **Set up secrets**: Store your `modal-key` secret. 2. **Deploy**: ```bash modal deploy Local Development 1. Install dependencies: pip install torch transformers accelerate einops huggingface_hub bitsandbytes fastapi gradio requests 2. Run locally: python app.py ⸻ 🪄 Live Demo A video overview of the app in action: 👉 Watch here! (replace with your video recording link) ⸻ 🏷️ Hackathon Track This project participates in the Agentic Demo Showcase. ⸻ 🧑‍💻 Team • Jadob ⸻ 📜 License This project is licensed under the MIT License. ⸻ 🌟 Acknowledgments • Thanks to the 🤖 Gradio Agents & MCP Hackathon 2025 🚀 team and sponsors for organizing this event. • Huge shoutout to Hugging Face, Modal, and all the contributors to open-source tools powering this project. ⸻ 💡 This project was built from scratch during the Hackathon week, following all official rules and submission guidelines.