AlphaAI - Local Multi-Agent RAG System

A powerful multi-agent RAG (Retrieval-Augmented Generation) system combining GraphRAG, AutoGen agents, and Ollama LLMs with an interactive Chainlit UI. Run everything locally and offline for complete privacy and control.

🌟 Features

• 🤖 Agentic RAG: Integrates GraphRAG's knowledge search with AutoGen agents via function calling
• 💻 Offline LLM Support: Uses local models from Ollama for both inference and embedding - completely free and private
• 🔌 Non-OpenAI Function Calling: Extends AutoGen to support function calling with non-OpenAI LLMs via LiteLLM proxy
• 🎨 Interactive UI: Chainlit interface for continuous conversations, multi-threading, and user input settings
• 📊 Knowledge Graphs: Automatically builds and visualizes knowledge graphs from your data
• 🗄️ Supabase Integration: Full-stack database and authentication with PostgreSQL
• 🐳 Docker Ready: Complete Docker Compose setup for easy deployment
• 🔒 Privacy First: All processing happens locally - your data never leaves your machine

🏗️ Architecture

AlphaAI combines several powerful technologies:

• GraphRAG: For knowledge graph generation and advanced RAG capabilities
• AutoGen: Multi-agent orchestration and conversation management
• Ollama: Local LLM inference (Mistral, Llama3, etc.)
• LiteLLM: Proxy server enabling function calling with local models
• Chainlit: Modern web UI for chat interactions
• Supabase: Self-hosted backend with PostgreSQL, Auth, and Real-time
• Docker: Containerized deployment for consistency

📋 Prerequisites

• Python 3.12+
• Docker and Docker Compose
• Ollama installed locally
• 16GB+ RAM recommended
• CUDA-capable GPU (optional, for faster inference)

🚀 Quick Start

1. Install Ollama and Models

Visit Ollama's website for installation instructions.

# Install and pull required models
ollama pull mistral-large
ollama pull llama3
ollama pull nomic-embed-text

# Start Ollama server
ollama serve

2. Clone and Setup Environment

git clone https://github.com/aimanyounises1/AlphaAI.git
cd AlphaAI

# Copy environment file
cp .env.example .env

# Edit .env with your configuration
nano .env

3. Option A: Run with Docker (Recommended)

# Start all services
docker-compose up -d

# View logs
docker-compose logs -f rag_agents

# Access the UI
open http://localhost:8002

4. Option B: Run Locally

# Create conda environment
conda create -n RAG_agents python=3.12
conda activate RAG_agents

# Install dependencies
pip install -r requirements.txt

# Initialize GraphRAG
mkdir -p ./input
python -m graphrag.index --init --root .
cp settings.yaml ./

# Create embeddings and knowledge graph
python -m graphrag.index --root .

# Start LiteLLM proxy (in separate terminal)
litellm --model ollama_chat/llama3

# Run the application
chainlit run appUI.py

📁 Project Structure

AlphaAI/
├── agents/              # AutoGen agent configurations
├── chat/                # Chat management and conversation handlers
├── database/            # Database connection and models
├── file_handlers/       # Document processing utilities
├── graph/               # Knowledge graph management
│   ├── graph_handler.py
│   ├── visualization.py
│   ├── communities.py
│   └── centrality.py
├── input/               # Input documents for processing
├── tasks/               # Task initialization and management
├── appUI.py             # Main Chainlit application
├── agents_creation.py   # Agent factory and configuration
├── config.py            # Application configuration
├── settings.yaml        # GraphRAG settings
├── docker-compose.yml   # Docker orchestration
└── requirements.txt     # Python dependencies

⚙️ Configuration

Environment Variables

Key configuration options in .env:

# GraphRAG
GRAPHRAG_API_KEY=your_api_key_here

# Ollama
OLLAMA_API_BASE=http://localhost:11434

# Supabase
SUPABASE_URL=http://kong:8000
POSTGRES_PASSWORD=your_secure_password

# Ports
STUDIO_PORT=3000
CHAINLIT_PORT=8002
LITELLM_PORT=8001

GraphRAG Settings

Edit settings.yaml to configure:

• LLM model selection
• Embedding model
• Chunk sizes and overlap
• Community detection parameters
• Query types (local vs global)

🎯 Usage

Basic Chat

1. Open http://localhost:8002 in your browser
2. Type your question or query
3. AlphaAI will:
   • Search the knowledge graph
   • Retrieve relevant context
   • Generate informed responses
   • Cite sources

Adding Documents

# Add documents to the input folder
cp your_document.txt ./input/

# Rebuild the knowledge graph
python -m graphrag.index --root .

Query Types

Local Search: For specific, targeted questions

# Finds specific facts and relationships
"What is the relationship between X and Y?"

Global Search: For broad, analytical questions

# Provides comprehensive overviews
"Summarize the main themes in the documents"

🧩 Components

GraphRAG

Creates knowledge graphs from unstructured text:

• Entity extraction
• Relationship mapping
• Community detection
• Hierarchical clustering

AutoGen Agents

Multi-agent system for:

• Task decomposition
• Collaborative problem-solving
• Function calling
• Response synthesis

Chainlit UI

Modern chat interface featuring:

• Real-time streaming responses
• File uploads
• Conversation history
• User settings
• Message threading

🐳 Docker Services

The Docker Compose setup includes:

• rag_agents: Main application container
• studio: Supabase Studio UI
• kong: API gateway
• auth: GoTrue authentication
• rest: PostgREST API
• realtime: Real-time subscriptions
• storage: File storage service
• db: PostgreSQL database
• vector: pgvector for embeddings
• analytics: Logflare analytics

📊 Knowledge Graph Visualization

Access the knowledge graph visualization:

# Generate visualization
open graph.html

Features:

• Interactive node exploration
• Community highlighting
• Relationship filtering
• Centrality metrics
• Export capabilities

🔧 Advanced Configuration

Custom Agents

Create custom agents in agents_creation.py:

from autogen import AssistantAgent

custom_agent = AssistantAgent(
    name="MyAgent",
    system_message="Your agent's role",
    llm_config=llm_config
)

Custom Tools

Add tools for agents to call:

@tool
def my_custom_tool(query: str) -> str:
    """Tool description for the agent"""
    # Implementation
    return result

🔍 Troubleshooting

Ollama Connection Issues

# Check Ollama is running
curl http://localhost:11434/api/tags

# Restart Ollama
pkill ollama && ollama serve

Docker Issues

# Rebuild containers
docker-compose down
docker-compose up --build -d

# Check logs
docker-compose logs -f

GraphRAG Indexing Errors

# Clear cache and reindex
rm -rf cache/
python -m graphrag.index --root . --verbose

Memory Issues

For large documents, adjust settings.yaml:

parallelization:
  num_threads: 4  # Reduce if needed

chunks:
  size: 256  # Reduce chunk size

📈 Performance Optimization

GPU Acceleration

Enable GPU support in Ollama:

# NVIDIA GPUs
ollama serve --gpu

# Check GPU usage
nvidia-smi

Model Selection

Choose models based on your hardware:

# Fast, lower memory (settings.yaml)
llm:
  model: mistral:7b

# Better quality, more memory
llm:
  model: mistral-large:latest

Caching

Enable caching for faster repeated queries:

# In config.py
ENABLE_CACHE = True
CACHE_TTL = 3600

🤝 Contributing

Contributions are welcome! Please:

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Add tests if applicable
5. Submit a pull request

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

🙏 Acknowledgments

• GraphRAG by Microsoft
• AutoGen by Microsoft
• Ollama for local LLM inference
• Chainlit for the amazing UI framework
• Supabase for the backend infrastructure
• LiteLLM for LLM proxy capabilities

📚 Further Reading

• GraphRAG Documentation
• AutoGen Documentation
• Ollama Model Library
• Chainlit Documentation

🆘 Support

If you encounter issues:

1. Check the Troubleshooting section
2. Search existing GitHub issues
3. Create a new issue with:
   • System information
   • Error logs
   • Steps to reproduce

🗺️ Roadmap

• Multi-modal support (images, PDFs)
• Advanced visualization options
• Custom embedding models
• Distributed processing
• API endpoint for external integrations
• Mobile-responsive UI
• Voice interaction support
• Multi-language support

---

Built with ❤️ for the AI community