How moussa-kalla/datathon works
Main Components
• Frontend Application (located in the “datathon-frontend” directory): Built with Next.js (indicated by “next.config.ts”, “tsconfig.json”, and components in “src/app”).
• Backend API (in “datathon-api”): Contains Python files like “app.py”, “config.py”, “main.py” and a requirements file. This service is responsible for serving chatbot requests, handling API endpoints, and orchestrating the RAG process.
• Data Sources: The “data” folder contains various documents (PDF manuals, brochures, etc.) and JSON files (“faq.json”, “responses.json”) used for retrieving information during chatbot interactions. The backend likely processes this data using RAG.
• External Services: Integration with the OpenAI API (as mentioned in the README) to leverage GPT-4’s language model capabilities.
• Notebooks: A separate “notebooks” directory is present with Jupyter notebooks (e.g., “RAG_OPENAI.ipynb”) for data analysis and possibly prototyping the RAG approach.
Establish Relationships and Interactions
• Optionally, include a note for “Notebooks” as a supportive analytical layer directing insights or improvements, but these are not part of the runtime system.
Architectural Patterns and Design Principles
• Use of separation of concerns: Clearly distinguishing between the frontend (user interface) and backend (business logic and data processing).
• API-Driven Interaction: Simple communication model between the frontend and backend, using well-defined endpoints.
• Modular organization that allows for scalability and easier maintenance.
Relevant Technologies and Frameworks
• Frontend: Next.js (React) with Tailwind CSS (as seen from “tailwind.config.ts”).
• Backend: Python (likely using a lightweight framework – possibly FastAPI or Flask – given “app.py” and recommended “uvicorn main:app --reload” comment).
• Additional tools: Git for version control, Jupyter for notebooks and data analysis, and pip/npm for dependency management.
• Placing the “Frontend Application” component at the top or left, feeding into the “Backend API” at the center.