DeepResearcher: AI-Powered Advanced Research and Writing System

Note: This project is currently under active development. Features and documentation may change frequently.

DeepResearcher is a sophisticated multi-agent system built with LangGraph that automates the research and content creation process. By orchestrating specialized AI agents working together through a defined workflow, it transforms queries into comprehensive, high-quality content with minimal human intervention.

Key Features

🧠 Multi-Agent Architecture

• Research Agent: Finds accurate information from multiple sources including Tavily, ArXiv, and Wikipedia
• Math Agent: Performs calculations and mathematical analysis when needed
• Writing Agent: Transforms research findings into well-written, coherent content
• Publishing Agent: Polishes and formats content to meet publication standards

🔄 Intelligent Workflow

• Dynamically Routed Workflow: Automatically determines which agents to engage based on query analysis
• Safety Mechanisms: Content screening for harmful material
• Quality Verification: Multiple verification and revision steps ensure high-quality output
• Self-Improvement: Procedural memory optimization based on feedback

🛠️ Advanced Features

• Semantic Queries: Generates optimal search queries based on user input
• Fact Verification: Ensures output accuracy through multiple verification passes
• Content Revision: Automatically improves content based on quality assessments
• Tool Usage Verification: Ensures proper processing of all tool outputs

📊 Comprehensive Testing

• Component Testing: Verify each agent and workflow component functions correctly
• End-to-End Evaluation: Test complete user experiences
• LLM-Based Quality Assessment: Automatic evaluation of content quality

Architecture

DeepResearcher uses a state-based workflow architecture with specialized agents:

User Query → Safety Check → Research Team (Research + Math) → Tool Verification → Writing Team → Quality Reflection → Final Output

Each agent has:

• Specialized prompts stored in procedural memory
• Access to relevant tools (search, calculators, etc.)
• Clear responsibilities within the workflow

Usage

Prerequisites

• Python 3.9+
• OpenAI API key (required)
• Tavily API key (for search functionality)
• LangChain API key (optional, for run tracking)

Installation

# Clone the repository
git clone https://github.com/yourusername/deep_researcher.git
cd deep_researcher

# Install dependencies
pip install uv
uv sync --dev

Running the Application

# Start the LangGraph development server
langgraph dev --no-browser

# Or using the provided script
./run_tests.sh dev

Then visit http://localhost:2024 to interact with DeepResearcher.

Running Tests

# Run agent evaluations
./run_tests.sh eval

# Or directly with Python
python run_evals.py

Customization

Modifying Agent Prompts

Agent prompts are stored in a procedural memory system and can be optimized over time based on feedback:

# Example: Updating the research agent prompt
memory_store.put(
    (("instructions",)), 
    key="research_agent", 
    value={
        "prompt": "Your custom research agent prompt here..."
    }
)

Adding New Tools

Extend DeepResearcher's capabilities by adding custom tools to relevant agents:

# Example: Adding a new tool to the research agent
@tool
def custom_research_tool(query: str) -> str:
    """Description of what the tool does"""
    # Tool implementation
    return result

# Add to research tools list
research_tools.append(custom_research_tool)

Workflow Customization

Modify the workflow by adjusting the StateGraph configuration:

# Example: Adding a new node to the research workflow
research_workflow.add_node("my_custom_node", my_custom_function)
research_workflow.add_edge("research_agent", "my_custom_node")
research_workflow.add_edge("my_custom_node", "math_agent")

Technical Details

DeepResearcher is built using:

• LangGraph: For defining agent workflows and state management
• LangChain: For tool integration and LLM interaction
• OpenAI: For powerful LLM capabilities (default is gpt-4o-mini)
• Fast API / FastHTML: For web interface (when applicable)

License

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

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

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