suraj143/ai-multi-agent-fastapi-mcp-db — reverse-engineered prompt

Reverse engineered prompt

Build me a small but solid multi agent AI app for customer data analysis. I want a public FastAPI endpoint where I can send a question, then an orchestrator uses Gemini to decide whether Agent A, Agent B, or both should handle it, pulls data from separate PostgreSQL schemas for each customer domain, and returns one combined plain English insight.

Please make it runnable with Docker Compose, with env based config for the Gemini key and database settings. Include the internal orchestrator route too, so I can test routing directly. Seed the database with simple sample customer records in two schemas and make sure each agent only sees its own schema. If Gemini is unavailable, return a clear error instead of crashing.

I’d like the app organized into a gateway service, an orchestrator service, two agent services, and shared Gemini logic. Add just enough README setup steps so I can run it locally or in Docker and quickly test the API. Look up current docs online if you need to.

Want more depth? Deep Reverse