An intelligent, full-stack recipe generation workspace. Chef AI leverages machine learning to transform user-provided ingredients, nutritional targets, or meal types into structured, actionable cooking instructions.
This application is engineered as a decoupled system to ensure a clean separation of concerns between the user interface and the machine learning inference engine.
- Frontend: React 18 + Vite (Custom responsive glass-morphism design).
- Backend: Python 3.x + FastAPI.
- AI Engine: HuggingFace / Local Llama Models.
- Assets: Unsplash API for dynamic food imagery.
Detailed technical documentation and installation guides are maintained in dedicated files:
- Data Extraction: Develop Python scripts to scrape and clean raw text from PDF/eBook sources.
- Dataset Preparation: Convert cleaned text into structured .jsonl files for LLM fine-tuning.
- Model Training: Fine-tune Llama 3/Mistral on custom culinary datasets using Unsloth/HuggingFace.
- Inference Optimization: Implement GGUF quantization for faster local response times.
- API Logic: Connect the fine-tuned model to the FastAPI POST /generate endpoint.
- CORS & Security: Configure middleware to allow secure communication between services.
- State Management: Implement LocalStorage in React to persist user recipe history.
- Error Handling: Add robust try/except blocks for API timeouts and model inference failures.
- Dockerization: Write Dockerfiles for both Frontend and Backend to ensure environment parity.
- Orchestration: Use docker-compose to launch the entire stack with a single command.
- Environment Control: Standardize .env management across local and production builds.
- Backend (Render): Deploy the FastAPI server with a persistent disk for model weights.
- Frontend (Vercel): Deploy the React/Vite UI to Vercel for global edge-network delivery.
- Automated CI/CD: Set up GitHub Actions to trigger deployments on every successful Pull Request merge.
This project is available under the MIT License.