RAGAgent

Agentic PDF RAG with LangGraph & Ollama

This project implements an Agentic Retrieval-Augmented Generation (RAG) system using LangGraph, LangChain, and Ollama. Unlike standard RAG pipelines, this implementation uses a ReAct agent that can "decide" when to search the document to provide more accurate, context-aware answers.

Demo

Demo.mp4

Features

• Local LLM Integration: Uses Ollama for both embeddings (nomic-embed-text) and reasoning (qwen3:8b), ensuring data privacy.
• Agentic Reasoning: Leverages LangGraph's create_react_agent to handle complex queries and tool calling.
• Persistent Vector Store: Uses Chroma to store and retrieve document chunks efficiently.
• Stateful Workflow: Built on a StateGraph architecture for modular and scalable AI logic.