Data Formulator
An open-source AI data analysis workspace that runs locally in your browser or Docker, with Ollama support and editable charts. MIT licensed.
Alternative to:
Data Formulator is a local AI data analysis workspace for analysts who want to explore questions through charts and keep track of alternative investigations. This Microsoft Research project combines natural language requests with direct chart editing. You can revisit an earlier result and branch into a different analysis without losing the original path.
The app runs locally in a browser and supports Docker. It's open source under the MIT license. Model connections include Ollama, OpenAI, Azure and Anthropic through LiteLLM, so you can choose a local model backend or a cloud service. Cloud model connections send requests to the selected provider.
Its Data Thread keeps explanations, tables and editable charts together as an analysis develops. You can refine a result yourself or ask the AI analyst to investigate further. The analyst can also find data sources, clarify a request and propose what to load, with a review step before data enters the workspace.
Inputs include CSV, TSV, Excel and JSON, plus screenshots and text. Database connectors include MySQL, PostgreSQL, MSSQL and Azure Data Explorer; S3 and Azure Blob connections cover stored data. Data memory records relationships between sources, while automatic joins support work across multiple tables. DuckDB integration supports larger datasets.
Flint powers the charts, and a style refinement agent helps adjust their appearance. Sessions and workspaces persist between visits, and you can assemble analysis results into editable reports to share.
