zi-yue-1129/DATAGEN

What it solves

DATAGEN automates the complex process of data research and analysis. It replaces manual data cleaning, hypothesis testing, and report writing with a coordinated system of AI agents that can handle the entire pipeline from raw data to a final graphical report.

How it works

The platform uses a multi-agent architecture powered by LangChain and LangGraph to manage a state-based workflow. Specialized agents are assigned specific roles—such as generating hypotheses, writing analysis code, creating visualizations, and conducting web searches—while a "Note Taker" agent maintains context and state across the process. The system follows a structured pipeline: generating a hypothesis, awaiting human approval, executing the analysis and visualization, and performing a final quality review.

Who it’s for

It is designed for researchers and data analysts who need to perform automated, scalable data analysis and generate professional reports without manually writing every step of the processing code.

Highlights

• Multi-Agent Specialization: Dedicated agents for hypothesis generation, coding, visualization, searching, and reporting.
• Flexible Model Configuration: Supports multiple LLM providers including OpenAI, Anthropic, Google, and others via a YAML configuration.
• State Management: Uses a specialized Note Taker agent and LangGraph to ensure context is retained across different analysis phases.
• External Integration: Integrates with the Model Context Protocol (MCP) for filesystem access, GitHub integration, and web searching.