Varn1t / EDAgent

What is EDAgent?

EDAgent is a Python tool that drops a CSV into a web interface and spits out a complete exploratory data analysis. Under the hood, it runs nine specialized AI agents in sequence—each one examining schema, quality, statistics, outliers, correlations, feature importance, and more—before synthesizing everything into plain-English insights. It uses LangGraph to orchestrate the pipeline and Ollama to run the language model entirely on your machine, so nothing leaves your computer. You get an interactive Streamlit dashboard, a downloadable HTML report, and a correlation heatmap.

Why is it gaining traction?

The hook is privacy and simplicity: no API keys, no cloud, no data leaving your machine. The nine-agent architecture is also genuinely interesting—each stage runs statistical computations first, then hands the results to the LLM for human-readable summarization. This hybrid approach tries to balance statistical rigor with natural language insights. The feature engineering agent even has a sandbox with error feedback, attempting to self-correct code suggestions. For teams already using Ollama, this plugs right in with minimal friction.

Who should use this?

Data scientists who want a quick first-pass EDA on a new dataset without writing boilerplate code. Analysts who need to share findings with non-technical stakeholders via the polished HTML report. Python developers exploring tabular data for the first time and wanting guidance on what to look for. It's less useful if you need production-grade pipelines or already have mature analysis workflows.

Verdict

The concept is solid and the local-first approach is increasingly appealing. However, with only 26 stars, this is early-stage software. Documentation is limited to the README, and there's no visible test suite. The credibility score sits around 0.9%, reflecting the project's youth. Worth trying for personal projects or experimentation, but don't rely on it for mission-critical work until it matures.