Skip to content
Documentation
Cookbook
Capabilities: RAG & Retrieval
SQL queries

SQL queries

Use this page when you already know SQL is the right interface and want practical patterns for making generation safer and more repeatable.

Compared with general Text-to-SQL discussions, this example is more operational: how to present schema context, how to keep execution bounded, and how to make outputs usable in real workflows.

What to focus on

  • How to supply schema context without leaking sensitive data
  • Guardrails for execution (read-only, limits, parameterization)
  • Making SQL output deterministic enough for automation

When it works well

  • Your team already uses SQL heavily in analysis or reporting
  • You need practical safeguards more than model novelty
  • Generated queries will feed into scripts, dashboards, or automated jobs

If you want to reproduce it locally

After your local Cookbook environment is ready, you can validate the matching example with:

make test-notebooks NOTEBOOK=misc/how_to_make_sql_queries.ipynb
ClassificationPDF upload summarization