LangGraph Router Agent

A semi-autonomous agent using the Router Pattern in LangGraph.

The graph routes user queries to one of two specialist agents:

• RAG Agent — answers questions grounded in your vector store
• Web Search Agent — answers questions requiring fresh / real-time information via Tavily

User Query
    │
    ▼
[Router Node]  ──── "rag"    ──►  [RAG Agent Node]        ──►  [Output]
                └── "web"    ──►  [Web Search Agent Node]  ──►  [Output]

Setup

Prerequisites

• Ollama installed and running with qwen2.5 model
• Tavily API Key (for web search functionality)

Install & Configure

1. Install Ollama: Download from ollama.ai

2. Pull Qwen2.5 model:

   ollama pull qwen2.5

3. Start Ollama server (if not already running):

   ollama serve

   This will start at http://localhost:11434 by default.

4. Set up the project:

   cd /path/to/langgraph-router-agent
   cp .env.example .env          # Edit .env and add your TAVILY_API_KEY

5. Install dependencies using uv:

   uv sync

Run

uv run python -m src.main

Or use the script entry point:

uv run langgraph-router

Project Layout

src/
  config/       – env & model settings
  graph/        – LangGraph state, nodes, edges, compiled graph
  agents/       – RAG agent & Web Search agent logic
  tools/        – Tavily tool wrapper, RAG retriever wrapper
  utils/        – shared helpers (logging, prompt builders)
tests/          – pytest test suite