Lumen

Lumen is an open source Python framework from the HoloViz ecosystem for building AI data applications that let users chat with their data. Questions asked in natural language are turned into explicit, reproducible Python and SQL execution, so results stay inspectable rather than being produced by an opaque model. It is aimed at data teams and developers who want conversational data exploration while keeping control over logic and execution. Users bring their own language model, with documented examples using OpenAI, Amazon Bedrock through LiteLLM, Google, Mistral, Ollama and local models through LlamaCpp. Data sources include CSV and Parquet files, DuckDB, SQLAlchemy databases such as PostgreSQL, BigQuery and Snowflake. A command line entry point can serve a data file directly, and the ExplorerUI component can be embedded in a custom application. Developers can plug in custom data sources, Python analyses, tools and agents, and encode business terminology and constraints directly in code. Beyond chat, Lumen provides pipelines, transforms and views for building dashboards, and a Reports feature that defines queries and visualizations once so they can be rerun on demand and exported as Jupyter notebooks or deployed as web apps. Tutorials include an executive dashboard and weather and census data explorers. Lumen is free to use, and the documentation site lists no paid plans. Costs for the chosen language model provider are separate.