// PYTHON LIBRARY

ask()

`df.omna.ask(question)` answers a plain-English question about your DataFrame using Claude. It is the one method that makes a network call — and even then, the sample rows in the prompt are masked before they leave your machine.

1 min readPlain-English queries via Claude

Signature

python

df.omna.ask(question, model=..., mask_rows=True)

Parameter	Description
`question`	The question to ask, in plain English
`model`	Override the Claude model. Default: `claude-haiku-4-5-20251001`
`mask_rows`	Mask PII in the sampled rows before they leave your machine (default `True`). Pass `False` for synthetic or public data

ask() sends the schema plus up to 20 sample rows to Claude and requires an ANTHROPIC_API_KEY.

bash

export ANTHROPIC_API_KEY=sk-ant-...

Example

python

import polars as pl
import omna

df = pl.read_csv("documents.csv")

results = df.omna.search("insurance claim denied", on="text", k=5)
results.omna.ask("What personal data do these documents expose?")
# → "These insurance documents expose SSNs, medical record numbers,
#    dates of birth, health plan numbers, and claimant identifiers."

# Override the model
results.omna.ask("Summarise the key themes", model="claude-sonnet-4-6")

Note: Privacy by default — the sampled rows in the prompt are masked before they leave your machine (PII replaced with tokens, secrets redacted). Pass mask_rows=False only for synthetic or public data.

What's next

DocsMask PII — mask_pii() →DocsLibrary overview →