DATAGEN

Why we included this project

Point DATAGEN at a data file with a plain-English instruction and it runs the research loop for you: a hypothesis agent proposes what to test, another writes the analysis code, a third builds the visualizations, and a dedicated writer produces the report before a reviewer agent checks the output and flags revisions. The orchestration runs on LangGraph, so the agent flow is explicit and inspectable rather than a black box, and a note-taking agent carries context between stages so each step does not start from scratch. You can edit a YAML file to assign different model providers and models to each agent, which also makes it straightforward to swap between dev and production setups. This is best treated as an end-to-end automation tool for exploratory research, not a library: it pays off when you trust the agent loop and mainly want the finished document back, rather than hands-on control of every step.