Couchbase Search Vector Store icon

Couchbase Search Vector Store

Work with your data using the Couchbase Search Vector Store. This node conducts vector retrievals using the Search service.

Couchbase Search Vector Store

AI-generated

Summary

Searches a Couchbase Search vector index or exposes it for AI retrieval, and can also insert or update documents. Choose an operation mode and configure the Couchbase bucket, scope, collection, vector index, embedding field, and text field; an Embedding connection is required.

Inputs

  • Operation Mode (required) — Choose Get Many, Insert Documents, Retrieve as Vector Store, Retrieve as Tool, or Update Documents.
  • Couchbase Bucket, Couchbase Scope, Couchbase Collection, Vector Index Name (required) — Select from the available lists or enter names. Use Scoped Index to choose between a scoped and global search index.
  • Embedding Field Key, Text Field Key (required) — Name the Couchbase document fields that contain the embedding array and raw text.
  • Embedding connection (required) — Connect an embedding model for vector search and document insertion or updates.
  • Document connection (required) — For Insert Documents, connect a document source. The node embeds and stores the connected documents; Embedding Batch Size controls the batch size (default 200). Optional insert settings accept comma-separated IDs and a JSON array of metadata objects.
  • Prompt (required) — For Get Many, provide the query text used for similarity-ranked search.
  • Limit — For Get Many and Retrieve as Tool, set the maximum number of results (default 4).
  • Include Metadata — For Get Many and Retrieve as Tool, choose whether returned documents include metadata (default true).
  • Metadata Filter — For Get Many and either retrieval mode, provide metadata name/value filters.
  • Rerank Results — For Get Many or either retrieval mode, enable reranking (default false); when enabled, connect a reranker.
  • Tool Description (required) — For Retrieve as Tool, describe the search tool for the AI agent. The tool accepts an input query. Older node versions also expose a required Name field.
  • ID (required) — For Update Documents, provide the ID of the embedding entry to update. Supply one document per input item.

Output shape

Main output items for Get Many, Insert Documents, Update Documents, and direct Retrieve as Tool execution; an AI Vector Store or AI Tool connection for the corresponding retrieval modes.

Get Many returns items containing each document's pageContent, optional metadata, and similarity score. Insert and Update return serialized document items; inserted items include documentId when Couchbase returns IDs. Direct Retrieve as Tool execution reads the search query from each input item's json.input and returns a response containing text results. Retrieve as Vector Store supplies a vector store for connected AI nodes rather than ordinary output items. Retrieval options such as Limit and Include Metadata are not exposed in Retrieve as Vector Store mode.

Links

Discussion