Text2SQL Query Caching

Solid caches Text2SQL results to reduce latency on repeated questions.

Solid caches Text2SQL results keyed to the original end-user question, so repeated questions return faster without re-generating SQL from scratch.

How it works

When a user asks a question that has been asked before (or is semantically equivalent), Solid can return the cached SQL result instead of running the full generation pipeline. The cache is keyed to the original question text so the match is precise.

What this means for users

  • Faster responses for frequently asked questions
  • Consistent SQL — the same question returns the same SQL until the semantic model is updated
  • No change to behavior — cached results go through the same response flow; users see no difference

Cache invalidation

The cache is cleared automatically when:

  • The semantic model is updated (new tables, metrics, or instructions)
  • An admin manually clears the cache
  • A cached result receives negative feedback via submit_text2sql_feedback

Note: Query caching is applied automatically — no configuration is required.


Did this page help you?