agent-orchestration

Tool-calling agent service built on a LangGraph StateGraph orchestration loop (planner, tool execution, conditional routing) over a RAG pipeline with OpenAI embeddings and a Chroma vector store; ships a reusable tool library (web search, email, knowledge-base retrieval) exposed through a FastAPI backend with REST and SSE streaming endpoints and a React dashboard.

Overview

Agent Orchestration is a tool-calling agent service built on a LangGraph StateGraph orchestration loop. The agent plans, executes tools, and conditionally routes between steps over a RAG pipeline powered by OpenAI embeddings and a Chroma vector store. A reusable tool library (web search, email, knowledge-base retrieval) is exposed through a FastAPI backend with REST and SSE streaming endpoints, and a React dashboard for interacting with the agent.

Key Features

  • LangGraph Orchestration: Planner → tool execution → conditional routing loop built on a StateGraph
  • RAG Pipeline: OpenAI embeddings with a Chroma vector store for knowledge-base retrieval
  • Reusable Tool Library: Web search, email, and knowledge-base retrieval tools, easy to extend
  • FastAPI Backend: REST endpoints plus SSE streaming for real-time agent responses
  • React Dashboard: Frontend for chatting with the agent and observing tool calls

Technologies Used

  • Python - Core language for the agent service
  • LangChain & LangGraph - Agent framework and StateGraph orchestration
  • FastAPI - Backend API with REST and SSE streaming endpoints
  • Chroma - Vector store for RAG retrieval
  • OpenAI - Embeddings and LLM
  • React - Dashboard frontend

Repository

github.com/ervikassingh/agent-orchestration