Production-grade multi-agent AI research system built with FastAPI, LangGraph, Gemini, PostgreSQL, and Docker.
Nexus AI is a backend-focused multi-agent AI platform that simulates an autonomous research workflow. Instead of relying on a single prompt-response interaction, the system coordinates multiple AI agents through a LangGraph execution pipeline where each agent performs a specialized task such as:
The platform supports real-time streaming responses, tool calling, persistent report storage, and production-ready deployment using Docker.
Built using LangGraph to coordinate multiple AI agents through a stateful execution graph. Agents operate independently while sharing structured state across the workflow.
Integrated tool-calling architecture allows agents to dynamically invoke external tools during execution. Current tools include real-time web search and external context retrieval.
Implements asynchronous token streaming for real-time response generation. Supports ChatGPT-style incremental output delivery using FastAPI streaming responses.
Research reports and session outputs are stored in PostgreSQL using SQLAlchemy ORM. The system is designed around cloud-first infrastructure principles, integrating the Gemini API for LLM inference and Neon PostgreSQL for managed database hosting.
┌─────────────────────┐
│ Frontend UI │
└─────────┬───────────┘
│
▼
┌─────────────────────┐
│ FastAPI Server │
└─────────┬───────────┘
│
▼
┌─────────────────────┐
│ LangGraph Engine │
└─────────┬───────────┘
│
┌───────────────────┼───────────────────┐
▼ ▼ ▼
┌──────────────┐ ┌──────────────┐ ┌────────────────┐
│ Planner Agent│ │ Search Agent │ │ Summary Agent │
└──────┬───────┘ └──────┬───────┘ └────────┬───────┘
│ │ │
▼ ▼ ▼
┌───────────┐ ┌────────────┐ ┌─────────────┐
│ Gemini API│ │ Web Search │ │ PostgreSQL │
└───────────┘ └────────────┘ └─────────────┘