Skip to content

Latest commit

 

History

History
106 lines (80 loc) · 3.66 KB

README.md

File metadata and controls

106 lines (80 loc) · 3.66 KB

Wikipedia AI agent research assistant.

AI agent research assistant that uses Wikipedia's vast knowledge base to deliver comprehensive, well-researched answers to your questions.

Featuring a ReAct (Reasoning and Action) agent architecture that uses Wikipedia, or RAG (retrieval augmented generation), semantic chunking, embeddings, GraphRAG. Offers flexible access through both terminal and web interfaces.

Built on a modern stack including LangChain's LangGraph's ReAct agent, Wikipedia API, FAISS vector storage, Microsoft's GraphRAG, support for leading LLM providers (OpenAI, Anthropic, Google), and Streamlit frontend.

License: MIT

image

Features

  • 🤖 Multiple LLM Support from OpenAI, Anthropic, and Google, such as:
    • OpenAI: gpt-4o, chatgpt-4o-latest, o1-2024-12-17, gpt-4-turbo
    • Anthropic: claude-3-5-sonnet-latest
    • Google: gemini-exp-1206, gemini-1.5-pro, gemini-2.0-flash-exp, gemini-2.0-flash-thinking-exp-1219
  • 🔍 Wikipedia search capabilities
  • 📚 RAG (Retrieval Augmented Generation) support, soon finished GraphRAG
  • 💾 Conversation memory and thread management
  • 🖥️ Both terminal and web interfaces
  • 🔄 Streaming responses for real-time interaction
  • 📝 Logging system

Architecture

The project uses a modular architecture with:

  • LangGraph's ReAct agent for agent orchestration
  • LangChain for LLM interactions and tool integration
  • Wikipedia API as agent's tool
  • FAISS for vector storage (when using RAG), or soon finished GraphRAG
  • Streamlit for the web interface

Setup

  1. Clone the repository:
git clone https://github.com/yourusername/wikipedia-agent.git
cd wikipedia-agent
  1. Install dependencies:
pip install -r requirements.txt
  1. Configure by setting up environment variables in .env:
OPENAI_API_KEY="your_key_here" # if you want to use OpenAI
ANTHROPIC_API_KEY="your_key_here" # if you want to use Anthropic
GOOGLE_API_KEY="your_key_here" # if you want to use Google
LLM="claude-3-5-sonnet-latest"  # or other supported models
USE_RAG="false"  # set to true to enable RAG, is false by default
SYSTEM_PROMPT="your_prompt_here" # if you want to change the default prompt

Usage

Web Interface

python -m streamlit run app.py

or

streamlit run app.py

Then visit http://localhost:8501 in your browser.

Terminal Interface

python app_terminal.py

RAG Setup

To use RAG functionality instead of Wikipedia search:

  1. Place wikipedia pages text files in the rag_data/ directory (by default includes only artificial_intelligence.txt, physics.txt)
  2. Set USE_RAG=true in your .env file

GraphRAG will work soon too.

Contributing

Contributions are welcome! Please feel free to submit a pull request.

TODO

  • Finish GraphRAG support
  • Implement automatic Wikipedia scraping for RAG data collection
  • Integrate text-to-speech capabilities
  • Add support for more LLM providers (e.g., Mistral, Cohere)
  • Create a Docker container for easy deployment
  • Implement rate limiting and API usage tracking
  • Add support for custom knowledge bases beyond Wikipedia
  • Create a web API endpoint for programmatic access
  • Implement caching to reduce API calls
  • Add support for image and diagram generation
  • Add benchmarking tools for different LLM models
  • Add support for concurrent user sessions
  • Create a configuration UI for easy setup

License

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