Data Formulator: Vibe with data, in control

🪄 Turn data into insights with AI Agents, with the exploration paths you choose. Try Data Formulator now!

• 🤖 New in v0.5: agent model + interative control (video)
• 🔥🔥🔥 Try our online demo at https://data-formulator.ai
• Any questions, thoughts? Discuss in the Discord channel!

https://github.com/user-attachments/assets/8ca57b68-4d7a-42cb-bcce-43f8b1681ce2

News 🔥🔥🔥

[11-07-2025] Data Formulator 0.5: Vibe with your data, in control

• 📊 Load (almost) any data: load structured data, extract data from screenshots, from messy text blocks, or connect to databases.
• 🤖 Explore data with AI agents:
  • In agent mode, provide a high-level goal and ask agents to explore data for you.
  • To stay in control, directly interact with agents: ask for recommendations or specify chart designs with UI + NL inputs, and AI agents will formulate data to realize your design.
  • Use data threads to control branching exploration paths: backtrack, branch, or follow up.
• ✅ Verify AI generated results: interact with charts and inspect data, formulas, explanations, and code.
• 📝 Create reports to share insights: choose charts you want to share, and ask agents to create reports grounded in data formulated throughout exploration.

Previous Updates

Here are milestones that lead to the current design:

• v0.2.2 (Demo): Goal-driven exploration with agent recommendations and performance improvements
• v0.2.1.3/4 (Readme | Demo): External data loaders (MySQL, PostgreSQL, MSSQL, Azure Data Explorer, S3, Azure Blob)
• v0.2 (Demos): Large data support with DuckDB integration
• v0.1.7 (Demos): Dataset anchoring for cleaner workflows
• v0.1.6 (Demo): Multi-table support with automatic joins
• Model Support: OpenAI, Azure, Ollama, Anthropic via LiteLLM (feedback)
• Python Package: Easy local installation (try it)
• Visualization Challenges: Test your skills (challenges)
• Data Extraction: Parse data from images and text (demo)
• Initial Release: Blog | Video

Overview

Data Formulator is an application from Microsoft Research that uses AI agents to make it easier to turn data into insights.

Data Formulator is an AI-powered tool for analysts to iteratively explore and visualize data. Started with data in any format (screenshot, text, csv, or database), users can work with AI agents with a novel blended interface that combines user interface interactions (UI) and natural language (NL) inputs to communicate their intents, control branching exploration directions, and create reports to share their insights.

Get Started

Play with Data Formulator with one of the following options:

• Option 1: Install via Python PIP

  Use Python PIP for an easy setup experience, running locally (recommend: install it in a virtual environment).

  # install data_formulator
  pip install data_formulator
  
  # Run data formulator with this command
  python -m data_formulator
  

  Data Formulator will be automatically opened in the browser at http://localhost:5000.

  you can specify the port number (e.g., 8080) by python -m data_formulator --port 8080 if the default port is occupied.

• Option 2: Codespaces (5 minutes)

  You can also run Data Formulator in Codespaces; we have everything pre-configured. For more details, see CODESPACES.md.

  

• Option 3: Working in the developer mode

  You can build Data Formulator locally if you prefer full control over your development environment and develop your own version on top. For detailed instructions, refer to DEVELOPMENT.md.

Using Data Formulator

Load Data

Besides uploading csv, tsv or xlsx files that contain structured data, you can ask Data Formulator to extract data from screenshots, text blocks or websites, or load data from databases use connectors. Then you are ready to explore.

Explore Data

There are four levels to explore data based depending on whether you want more vibe or more control:

• Level 1 (most control): Create charts with UI via drag-and-drop, if all fields to be visualized are already in the data.
• Level 2: Specify chart designs with natural language + NL. Describe how new fields should be visualized in your chart, AI will automatically transform data to realize the design.
• Level 3: Get recommendations: Ask AI agents to recommend charts directly from NL descriptions, or even directly ask for exploration ideas.
• Level 4 (most vibe): In agent mode, provide a high-level goal and let AI agents automatically plan and explore data in multiple turns. Exploration threads will be created automatically.

https://github.com/user-attachments/assets/164aff58-9f93-4792-b8ed-9944578fbb72

• Level 5: In practice, leverage all of them to keep up with both vibe and control!

Create Reports

Use the report builder to compose a report of the style you like, based on selected charts. Then share the reports to others!

Developers' Guide

Follow the developers' instructions to build your new data analysis tools on top of Data Formulator.

Help wanted:

• Add more database connectors (https://github.com/microsoft/data-formulator/issues/156)
• Scaling up messy data extractor: more document types and larger files.
• Adding more chart templates (e.g., maps).
• other ideas?

Research Papers

• Data Formulator 2: Iteratively Creating Rich Visualizations with AI

@article{wang2024dataformulator2iteratively,
      title={Data Formulator 2: Iteratively Creating Rich Visualizations with AI}, 
      author={Chenglong Wang and Bongshin Lee and Steven Drucker and Dan Marshall and Jianfeng Gao},
      year={2024},
      booktitle={ArXiv preprint arXiv:2408.16119},
}

• Data Formulator: AI-powered Concept-driven Visualization Authoring

@article{wang2023data,
  title={Data Formulator: AI-powered Concept-driven Visualization Authoring},
  author={Wang, Chenglong and Thompson, John and Lee, Bongshin},
  journal={IEEE Transactions on Visualization and Computer Graphics},
  year={2023},
  publisher={IEEE}
}

Contributing

This project welcomes contributions and suggestions. Most contributions require you to agree to a Contributor License Agreement (CLA) declaring that you have the right to, and actually do, grant us the rights to use your contribution. For details, visit https://cla.microsoft.com.

When you submit a pull request, a CLA-bot will automatically determine whether you need to provide a CLA and decorate the PR appropriately (e.g., label, comment). Simply follow the instructions provided by the bot. You will only need to do this once across all repositories using our CLA.

This project has adopted the Microsoft Open Source Code of Conduct. For more information see the Code of Conduct FAQ or contact opencode@microsoft.com with any additional questions or comments.

Trademarks

This project may contain trademarks or logos for projects, products, or services. Authorized use of Microsoft trademarks or logos is subject to and must follow Microsoft's Trademark & Brand Guidelines. Use of Microsoft trademarks or logos in modified versions of this project must not cause confusion or imply Microsoft sponsorship. Any use of third-party trademarks or logos are subject to those third-party's policies.