zi-yue-1129/DATAGEN

What it solves

DATAGEN is designed to automate the complex process of data analysis and research. It removes the manual effort of generating hypotheses, cleaning data, writing analysis code, creating visualizations, and compiling final research reports.

How it works

The system uses a multi-agent architecture powered by LangGraph and LangChain, where specialized agents collaborate to execute a research workflow. A hypothesis_agent starts by proposing research directions, which are then processed by a process_agent that coordinates other specialized roles like a code_agent for analysis, a visualization_agent for graphics, and a searcher_agent for external research. A note_agent tracks the state and context across the entire process, while a quality_review_agent ensures the final output meets standards before the report is finalized by the report_agent.

Who it’s for

It is intended for researchers and data analysts who need to transform raw data files (like CSVs) into comprehensive, graphical research reports using AI-driven automation.

Highlights

• Multi-Agent Coordination: Uses specialized agents for hypothesis generation, coding, visualization, and reporting.
• State Management: Features a dedicated Note Taker agent for efficient context retention and state tracking.
• Flexible Model Configuration: Allows users to mix and match LLM providers (OpenAI, Google, Anthropic, etc.) for different agents via a YAML configuration.
• External Integration: Supports the Model Context Protocol (MCP) for integrating external servers like GitHub and web search tools.
• Adaptive Pipeline: Implements a progressive disclosure architecture to optimize the context window and resource usage.