A multi-agent system for laboratory automation and research, specializing in automated research paper monitoring, device control, and AI-powered analysis for condensed matter physics labs.

# Clone the repository
git clone https://github.com/caidish/labAgent.git
cd labAgent

# Install dependencies
pip install -r requirements.txt

# Set up environment variables
cp .env.example .env

Edit .env file with your API keys:

# Required for ArXiv Daily Updates and MCP Tools
OPENAI_API_KEY=your_openai_api_key_here
OPENAI_MODEL=gpt-5

# Optional  
GOOGLE_API_KEY=your_google_api_key_here
GEMINI_MODEL=gemini-2.0-flash-exp
DEBUG=false
LOG_LEVEL=INFO

# Web Interface (Recommended)
streamlit run lab_agent/web/app.py

# Command Line Interface
python -m lab_agent.main

• Automated Scraping: Fetches latest papers from arXiv cond-mat/new
• AI-Powered Scoring: GPT-4 rates paper relevance (1-3 priority levels)
• Smart Filtering: Focuses on 2D materials, graphene, TMDs, quantum devices
• Beautiful Reports: Generates HTML reports with priority sections
• Web Interface: Manual trigger, report viewing, and management
• No Database: Simple file-based storage

Usage: Navigate to "📚 ArXiv Daily" tab → Click "Generate Daily Report"

• MCP Server for Microscope Control: Nikon glovebox microscope automation
• Deep Learning Tool Integration: Automated flake analysis and scoring
• Multi-Agent Coordination: OpenAI and Google Gemini integration
• Real-time Communication: WebSocket support for agent coordination

labAgent/
├── lab_agent/                     # Main package
│   ├── agents/                    # Agent implementations
│   │   ├── base_agent.py         # Abstract base class
│   │   └── arxiv_daily_agent.py  # ArXiv monitoring agent
│   ├── tools/                    # Agent capabilities
│   │   ├── arxiv_daily_scraper.py # Web scraping for ArXiv
│   │   ├── paper_scorer.py       # GPT-4 paper scoring
│   │   ├── daily_report_generator.py # HTML/JSON reports
│   │   ├── web_scraper.py        # General web scraping
│   │   └── arxiv_parser.py       # ArXiv API integration
│   ├── utils/                    # Utilities
│   │   ├── config.py             # Environment configuration
│   │   └── logger.py             # Logging setup
│   ├── config/                   # Configuration files
│   │   ├── interestKeywords.txt  # Research interest keywords
│   │   └── promptArxivRecommender.txt # GPT scoring prompts
│   └── web/                      # Streamlit interface
│       └── app.py                # Main web application
├── requirements.txt              # Python dependencies
├── setup.py                     # Package configuration
├── .env.example                 # Environment template
├── CLAUDE.md                    # Project context for Claude Code
├── TESTING_GUIDE.md            # Comprehensive testing protocol
└── README.md                   # This file

Follow the comprehensive testing guide in TESTING_GUIDE.md:

# Quick test - verify system works
streamlit run lab_agent/web/app.py
# → Go to ArXiv Daily tab → Generate Daily Report

# Full testing protocol
cat TESTING_GUIDE.md

Edit lab_agent/config/interestKeywords.txt to customize paper filtering:

# 2D Materials & Graphene
2D materials
graphene
monolayer graphene
transition metal dichalcogenides
van der Waals heterostructures
# ... add your research areas

Modify lab_agent/config/promptArxivRecommender.txt for different evaluation criteria.

from lab_agent.agents.arxiv_daily_agent import ArxivDailyAgent
import asyncio

# Initialize agent
agent = ArxivDailyAgent({'reports_dir': './reports'})
await agent.initialize()

# Generate report
task = {
    'type': 'generate_daily_report',
    'url': 'https://arxiv.org/list/cond-mat/new'
}
result = await agent.process_task(task)
print(f"Generated report with {result['total_papers']} papers")

1. Start Application: streamlit run lab_agent/web/app.py
2. Navigate: Go to "📚 ArXiv Daily" tab
3. Generate: Click "🔄 Generate Daily Report"
4. View: Expand report section to see results
5. Manage: Use "🗑️ Clear All Reports" to cleanup

pip install -e .

# Use console entry points
lab-agent        # CLI version
lab-agent-web    # Web version

• Async-first: All agents built on asyncio
• Modular Design: Separate agents, tools, and utilities
• Configuration-driven: Environment-based settings
• Tool-based: Agents use composable tools for capabilities

"ArXiv agent not available"

• Check OpenAI API key in .env file
• Verify API key is valid and has credits

"No papers found"

• Check internet connection
• Verify ArXiv website is accessible

Slow GPT scoring

• Normal for 20+ papers (includes rate limiting)
• Consider upgrading to higher tier OpenAI plan

Permission errors

• Ensure write permissions in project directory
• Check ./reports/ directory can be created

Enable detailed logging:

# In .env file
DEBUG=true
LOG_LEVEL=DEBUG

• ✅ Milestone 0: ArXiv Daily Update System
• 🚧 Milestone 1: MCP Server for Nikon Microscope
• ✅ Milestone 2: Deep Learning Tool Integration
• 📋 Milestone 3: Multi-Agent SDK Wiring

• General scaling framework and best practices

• Advanced automation capabilities

MIT License - see LICENSE file for details

1. Fork the repository
2. Create feature branch: git checkout -b feature/amazing-feature
3. Commit changes: git commit -m 'Add amazing feature'
4. Push to branch: git push origin feature/amazing-feature
5. Open a Pull Request

• Issues: GitHub Issues
• Documentation: See CLAUDE.md for detailed project context
• Testing: Follow TESTING_GUIDE.md for validation

Built for condensed matter physics research labs 🔬⚛️