README.md

MAGI: Manus Artificial General Intelligence for academic research assistance with human-in-the-loop AI agents #

MAGI (Manus Artificial General Intelligence) is a system designed to assist with materials science research by coordinating multiple specialized AI agents under human supervision.

Overview #

MAGI consists of a Master Research Agent that coordinates several specialized sub-agents:

  1. Master Research Agent: Coordinates the workflow between sub-agents and serves as the main interface for users.
  2. Planning Agent: Handles research planning, literature review, and experimental design.
  3. Execution Agent: Manages experiment execution, simulation runs, and data collection.
  4. Analysis Agent: Performs data analysis, visualization, and interpretation.
  5. Knowledge Agent: Manages research artifacts, references, and knowledge retrieval.

Installation #

  1. Clone the repository:

    git clone https://github.com/yourusername/gemini-ai-agent.git
    cd gemini-ai-agent/agents/academic-research
    
  2. Create and activate a virtual environment:

    python -m venv venv
    source venv/bin/activate  # On Windows: venv\Scripts\activate
    
  3. Install dependencies:

    pip install -r requirements.txt
    
  4. Set up environment variables: Create a .env file in the project root with your Google Cloud credentials:

    GOOGLE_APPLICATION_CREDENTIALS="path/to/your/service-account-key.json"
    

Usage #

from MAGI.agent import root_agent

# Initialize the MAGI system
agent = root_agent

# Start a research task
response = agent.process("Help me design an experiment to study the properties of graphene oxide.")
print(response)

Architecture #

QueryDelegatesDelegatesDelegatesQueriesStoresStoresStoresUserMaster Research AgentPlanning AgentExecution AgentAnalysis AgentKnowledge Agent

Features #

  • Human-in-the-loop: Critical decisions require human approval
  • Modular Design: Easy to extend with new agents and tools
  • Knowledge Management: Built-in vector database for research artifacts
  • Reproducibility: Full audit trail of all research activities

License #

This project is licensed under the Apache License 2.0 - see the LICENSE file for details.