Vertex AI Vector Search

View as Markdown

A tool node for interacting with Google Cloud Vertex AI Vector Search (formerly Matching Engine).

This node exposes a search tool only. It is not a full vector store adapter (no add/upsert via DocumentStoreBase).

Authentication

This node uses the shared RocketRide Google Cloud authentication logic. You can authenticate by either:

1. Uploading a Service Account JSON Key.
2. Leaving the key blank and relying on Application Default Credentials (ADC) if your RocketRide server is hosted on Google Cloud Run, GKE, or a Compute Engine VM with attached service accounts.

Usage

When placed in a pipeline as a tool node, it provides the search tool for finding K-nearest neighbors to a given embedding vector, honoring top_k and score_threshold.

score_threshold assumes similarity semantics (higher is better), such as DOT_PRODUCT_DISTANCE. It is not correct for SQUARED_L2_DISTANCE (where lower distance is better).

Schema

Field	Type	Description	Default
vertex.deployedIndexId	string	Deployed Index ID The ID of the deployed index.
vertex.indexEndpointId	string	Index Endpoint ID The ID of the Vector Search Index Endpoint.
vertex.location	string	Location The Google Cloud region, e.g. us-central1.	"us-central1"
vertex.profile	string		"default"

Dependencies

• google-auth >=2.23.3
• google-cloud-aiplatform >=1.40.0