I built a self-hostable AI data analyst — 4 agents, sandboxed code execution, your choice of LLM

Developer Laban has released Insight Orchestra, an open-source, self-hostable AI data analysis platform designed to replace cloud-based alternatives like ChatGPT's Code Interpreter. The tool allows users to connect various data sources, including CSV, Excel, and SQL databases, and perform analysis using a multi-agent pipeline. The system utilizes four specialized agents—Data Janitor, Hypothesis Bot, Debate Manager, and Viz Whiz—to clean data, generate insights, reach consensus, and create visualizations. A key feature is its focus on security; the platform uses RestrictedPython to sandbox LLM-generated code, preventing unauthorized filesystem or network access. Users can choose between cloud-based LLMs like OpenAI and Anthropic or run models locally via Ollama. Insight Orchestra is available on GitHub under an MIT license, offering a private, flexible solution for data professionals who require control over their infrastructure and data privacy.