How david01px2018/mcp-hackathon works
Project type and purpose
• This is a full-stack analytics application consisting of:
• A React/Next.js frontend (under frontend/)
• A Python FastAPI backend (app.py)
• Vector databases (Pinecone) and LLM services (OpenAI)
• Ingestion script (import_data.py) to populate Pinecone
File structure
• Top level:
• app.py – the FastAPI server exposing / and /chat endpoints
• import_data.py – CLI script to create/populate Pinecone indexes
• .env & requirements.txt – backend config
• frontend/ – Next.js app directory with app/, components/, hooks/, lib/, config files (Tailwind, PostCSS, tsconfig)
• Notice patterns:
• Frontend is organized with UI component library and custom hooks
• Backend separates ingestion (import_data) from runtime API (app.py)
README for architectural insights
• Agents: User Researcher, Market Analyst, Competitor Analyst, Strategic Advisor, Orchestrator
• External services: OpenAI API, Pinecone (with two indexes)
• Data flow: User query → Orchestrator → Agent(s) → Pinecone/OpenAI → Aggregate → JSON response