Query data

Choose the query that matches your question. Each method has its own setup, request fields and response format. Prepare the data before calling its endpoint.

SQL — Filter rows, combine tables and calculate totals.

Vector search — Find records with similar meaning using stored numeric representations called embeddings.

Full-text search — Find words and phrases, or rank matches. Check the API’s supported text analysis; engine-only language features are not automatically available through HTTP.

Graph — Follow relationships, find reachable records or inspect a path.

Ask — Ask a focused question in plain language, then check the generated plan and returned rows.

1. 01
   Identify the prepared modelCheck which tables, vectors, text indexes or relations exist.
2. 02
   Use its query interfaceSupply the input expected by that endpoint.
3. 03
   Inspect the response shapeRead rows, ranked IDs, word matches or graph paths.
4. 04
   Resolve the source when neededUse matching IDs to retrieve the original record.

Shared mechanics.Copy Anchor Link

Use the options documented for the endpoint you call. Explain output, timeouts and cancellation differ between APIs. The current guide explains shared timeout and query-ID headers; individual endpoint references describe their fields.

A query plan describes the work the database performs: scanning a table, selecting rows, joining records or calculating totals. See SQL for supported query shapes and Ask for inspecting a generated plan.