View only. Sign in to run this demo.

Sign in

Tabular Agent

Ask your own CSV or Excel file questions in plain language — the agent writes and runs read-only SQL

How to use the Tabular Agent

  1. 1Load a fileA CSV, Excel, Parquet or JSON file, or one of the examples. Each Excel sheet becomes its own table.
  2. 2Check what the agent seesThe agent profiles every column: its type, its gaps and the values it really holds.
  3. 3Ask in plain languageAsk about totals, rankings, comparisons or trends, in any language the model reads.
  4. 4Check the answerRead the answer, then the SQL behind it and the rows it came from. Sort them, chart them, download them.
  • The agent writes read-only SQL and runs it on your file in memory. It cannot change the file, and it does no calculations outside SQL. If the SQL fails it sees the error and tries again.
  • Every question is answered on its own: the agent does not remember the earlier ones, so name what you mean. Your file is deleted when you close the session or leave the page.

Start from an example file

1. The data
Upload your own file, or use an example above. Messy report sheets (title rows, subtotals, notes) are cleaned up.
2. Ask a question
Load a file first.

No file yet

Pick an example file, or upload a CSV or Excel file, to get started.

GAIK Toolkit

Generative AI building blocks for knowledge capture, access and synthesis.

Co-funded by the European Union

Demos

Use CasesSoftware ModulesSoftware ComponentsNo-code Assets

Resources

GitHubDocsPyPI

Project

gaik.aiRequest AccessPrivacy

Disclaimer: The demo website uses Microsoft Azure APIs for OpenAI's and Anthropic's models, and Google's Vertex AI for Gemini models. No user data is retained, and neither Azure nor Vertex AI uses any user data for model training.

© 2026 GAIK ProjectThe gaik package is open source under the MIT licence.
GAIK