Why is the text field empty in results?

Vertex AI Vector Search stores only dense vectors — no raw document text. Retrieved documents will have an empty text field. Policy rules must use sidecar metadata set via the Gateco resource API at index time.

What is public_endpoint_domain?

The management API uses {location}-aiplatform.googleapis.com. The findNeighbors query API uses a per-endpoint public hostname like 123456789.us-central1-abc.vdb.vertexai.goog. This hostname is shown in the Cloud Console on the endpoint detail page under "Public endpoint domain."

Does Gateco support Vertex AI Vector Search private endpoints?

Not in v1. Private endpoint VPC peering is on the roadmap for the Private Data Plane release (Q3 2026). Current support is for public endpoints only.

Vector DBTier 2

Vertex AI Vector Search

Policy-enforced ANN retrieval on Vertex AI Vector Search for sub-100ms nearest-neighbor search at scale.

Search capabilities

Vector (ANN)	Keyword (ranked)	Keyword (filter)	Hybrid (native)	Grep (substring)	Grep (regex)

<100ms p95 ANN at 1M vectors on standard configuration. Policy overhead <25ms p95.

sidecar metadata

Vertex AI Vector Search (formerly Matching Engine) is Google Cloud's managed ANN search service. It handles vector index management, query routing, and replication. Gateco connects to it as a Tier 2 retrieval connector.

Vertex AI Vector Search stores dense vectors only — there is no native text field. The text content of matched vectors is always empty in raw responses. For policy evaluation, Gateco uses sidecar metadata resolution, which must be populated separately via the Gateco resource API.

The connector requires a public_endpoint_domain (the per-endpoint hostname, e.g. 123456789.us-central1-abc.vdb.vertexai.goog) in addition to the management API parameters. Pass the hostname only — not a full https:// URL. The distance_measure field must match the index configuration.

Retroactive registration is not available in Gateco v1: Vertex AI Vector Search has no listing API. Resources must be registered via the Gateco ingestion API before they can be policy-gated.

Sample search config

{
  "project_id": "my-project",
  "location": "us-central1",
  "index_endpoint_id": "1234567890123456789",
  "deployed_index_id": "my_deployed_index",
  "public_endpoint_domain": "123456789.us-central1-abc.vdb.vertexai.goog",
  "distance_measure": "COSINE"
}

Sample policy

{
  "name": "ML team access",
  "effect": "allow",
  "rules": [{
    "conditions": [
      {"field": "resource.classification", "operator": "in", "value": ["internal", "public"]},
      {"field": "principal.groups", "operator": "contains", "value": "ml-team"}
    ]
  }],
  "selectors": [{"connector_type": "vertex_ai_vector_search"}]
}

Policy conditions reference resource.* and principal.* fields. Policy reference →

Frequently asked questions

Why is the text field empty in results?: Vertex AI Vector Search stores only dense vectors — no raw document text. Retrieved documents will have an empty text field. Policy rules must use sidecar metadata set via the Gateco resource API at index time.
What is public_endpoint_domain?: The management API uses {location}-aiplatform.googleapis.com. The findNeighbors query API uses a per-endpoint public hostname like 123456789.us-central1-abc.vdb.vertexai.goog. This hostname is shown in the Cloud Console on the endpoint detail page under "Public endpoint domain."
Does Gateco support Vertex AI Vector Search private endpoints?: Not in v1. Private endpoint VPC peering is on the roadmap for the Private Data Plane release (Q3 2026). Current support is for public endpoints only.

Ready to connect Vertex AI Vector Search?

Follow the step-by-step setup guide or talk to the team for help with your specific configuration.

Setup guide →Talk to the team All integrations