Vertex Vector SearchInstall

Summary

Query Google Cloud Vertex AI Vector Search with a given embedding vector to find similar documents from a deployed index.

Inputs

• Project Number (required) — The Google Cloud Project Number, found in the GCP Console under Project Info.
• Location (required) — The Google Cloud region where the Vertex AI Vector Search index endpoint is deployed.
• Index Endpoint ID (required) — The Vertex AI Vector Search Index Endpoint ID to query.
• Deployed Index ID (required) — The specific deployed index ID within the index endpoint to search.
• Query Embedding (required) — The embedding vector (array of numbers) representing the query, typically from a previous node.
• Number of Results — Number of similar documents (neighbors) to return, between 1 and 100 (default 5).
• Options — Additional optional parameters including GCS path to ID-to-text map JSON, metadata field names for text and source, context separator, and flags to include metadata and raw API response for debugging.

Output shape

A list of items matching the input length, each containing a JSON object with the retrieved context text string, an array of similar documents with metadata and distances, count of documents, and optionally the raw API response.

If a GCS ID-to-text mapping is provided, the node uses it to resolve document texts. Including metadata is enabled by default but can be disabled. Errors for individual items can be included if 'Continue On Fail' is enabled. The raw API response can be included for debugging if requested.