openGauss DataVec Store icon

openGauss DataVec Store

Work with openGauss DataVec vector database

openGauss DataVec Store

AI-generated

Summary

Retrieve matching documents from an openGauss DataVec vector store (default mode).

Inputs

  • mode (required) — Operation mode: load, insert, retrieve, or retrieve-as-tool
  • schema (required) — Database schema name (default: public)
  • table (required) — Name of the vector table
  • distanceStrategy — Distance metric for similarity search (l2, cosine, inner_product, manhattan)
  • prompt (required) — Search query prompt used to embed and find similar vectors
  • toolDescription (required) — Description of the tool for AI agent mode
  • topK — Maximum number of results to return
  • dimensions — Vector dimension size, needed when creating a new table
  • metadataFilter — JSON filter applied to document metadata during search

Output shape

array of objects with fields pageContent, metadata, and distance (similarity score)

Returns up to topK matching documents in a single response; no pagination; distance reflects the selected distance strategy.

Examples

Example 1: Load many ranked documents for pipeline use

mode=load, prompt='summarize recent AI trends', topK=10, metadataFilter='{"field":"technology"}', distanceStrategy='cosine'

Example 2: Retrieve as AI agent tool

mode='retrieve-as-tool', table='public.rag_docs', toolDescription='Search openGauss vector store for user query', topK=5, metadataFilter='{"status":"active"}'

Example 3: Insert new documents into vector store

mode=insert, dimensions=768, provides document payload via main input, inserts records and returns insertedCount

Links

Discussion