GraphRAG + AutoGen + Ollama + Chainlit UI = Local Multi-Agent RAG Superbot

This application integrates GraphRAG with AutoGen agents, powered by local LLMs from Ollama, for free and offline embedding and inference. Key highlights include:

• Agentic-RAG: - Integrating GraphRAG's knowledge search method with an AutoGen agent via function calling.
• Offline LLM Support: - Configuring GraphRAG (local & global search) to support local models from Ollama for inference and embedding.
• Non-OpenAI Function Calling: - Extending AutoGen to support function calling with non-OpenAI LLMs from Ollama via Lite-LLM proxy server.
• Interactive UI: - Deploying Chainlit UI to handle continuous conversations, multi-threading, and user input settings.

Useful Links 🔗

• Full Guide: Microsoft's GraphRAG + AutoGen + Ollama + Chainlit = Fully Local & Free Multi-Agent RAG Superbot Medium.com 📚

📦 Installation and Setup

Follow these steps to set up and run AutoGen GraphRAG Local with Ollama and Chainlit UI:

1. Install LLMs:

   Visit Ollama's website for installation files.

   ollama pull mistral
   ollama pull nomic-embed-text
   ollama pull llama3
   ollama serve

2. Create conda environment and install packages:

   conda create -n RAG_agents python=3.12
   conda activate RAG_agents
   git clone https://github.com/karthik-codex/autogen_graphRAG.git
   cd autogen_graphRAG
   pip install -r requirements.txt

3. Initiate GraphRAG root folder:

   mkdir -p ./input
   python -m graphrag.index --init  --root .
   mv ./utils/settings.yaml ./

4. Replace 'embedding.py' and 'openai_embeddings_llm.py' in the GraphRAG package folder using files from Utils folder:

   sudo find / -name openai_embeddings_llm.py
   sudo find / -name embedding.py

5. Create embeddings and knowledge graph:

   python -m graphrag.index --root .

6. Start Lite-LLM proxy server:

   litellm --model ollama_chat/llama3

7. Run app:

   chainlit run appUI.py