All products

Analyse large tables with AI

Ask a question about a local CSV or Parquet file and get a supporting table with a repeatable Python calculation.

Analyse large tables with AI when a single total hides the detail you need. A lending library might want the proportion of items returned late for each branch, quarter by quarter. With polars, your agent can inspect the available dates and categories, agree on a definition of late, and return a small comparison table backed by a repeatable Python calculation.

A branch with three loans deserves a different reading from one with three thousand. Ask for the number of loans beneath each percentage. If the source has missing return dates, decide whether open loans belong in the denominator. The calculation can select the needed fields and build its filters and groups before requesting the result.

What it does

  • Compare periods using a stated denominator and complete date ranges.
  • Keep group counts next to averages or rates so small samples stand out.
  • Join a catalogue to a transaction file using the agreed record key.
  • Parse dates and mark missing values when reading a CSV.
  • Inspect a query plan when a calculation processes far more data than expected.

Install and use

  1. Use Python 3.10 or later with Polars installed. Supply an authorised table or a representative anonymised subset.
  2. Define the time window, the measure and your treatment of missing entries.
  3. Ask for the supporting counts, then cross-check one group before reusing the calculation.

FAQ

Source

Made by Polars. Original skill: polars, published on GitHub under the MIT licence.

View on GitHub