IntermediateVector Database12 min read

Connect Vertex AI Search to Gateco

Connect Vertex AI Search (Discovery Engine) to Gateco for policy-aware enterprise search over unstructured data stores. Covers data store creation, service account setup, and identity-aware RAG access control.

Last updated: May 27, 2026

Prerequisites

  • A GCP project with the Discovery Engine API (discoveryengine.googleapis.com) enabled
  • A Vertex AI Search data store with at least one indexed document
  • A GCP service account with the Discovery Engine Viewer role

Overview — 5 steps

  1. 1Enable Discovery Engine API and create a data store
  2. 2Ingest documents into the data store
  3. 3Create a search app
  4. 4Create a service account
  5. 5Add the connector in Gateco

Vertex AI Search (powered by Google Discovery Engine API) is a fully managed enterprise search service with built-in document processing, semantic ranking, and auto-generated summaries. Gateco connects to Vertex AI Search as a Tier 2 connector — adding identity-aware retrieval and RAG access control on top of Google's semantic and keyword search. Documents are indexed and ranked by Google; Gateco gates which results each principal is allowed to see.

Vertex AI Search uses the Discovery Engine API (discoveryengine.googleapis.com), not the Vertex AI Platform API. This is a completely separate API surface from Vertex AI Vector Search.

Grep search is not supported for Vertex AI Search. Use vector, keyword, or hybrid search modes.

Step 1 — Enable the Discovery Engine API

bash
gcloud services enable discoveryengine.googleapis.com --project=YOUR_PROJECT_ID

Step 2 — Create a data store and ingest documents

  1. In GCP Console, navigate to Vertex AI → Search → Data stores.
  2. Click Create data store → Cloud Storage (or Web crawl, BigQuery, etc.).
  3. Select "Unstructured documents" as the content type.
  4. Set the location (typically "global").
  5. Complete the wizard. Note the data store ID from the data store details page.
  6. Go to Documents → Import to add your content.

The data store ID is shown on the data store overview page and in the GCP Console URL. It looks like a numeric ID or a slug.

Step 3 — Create a search app

  1. Navigate to Vertex AI → Search → Apps → Create app.
  2. Choose Search.
  3. Link the app to your data store.
  4. Note the serving_config_id (default: "default_serving_config") — only override if you have a custom configuration.

Step 4 — Create a service account

bash
gcloud iam service-accounts create gateco-vertex-search \
  --display-name="Gateco Vertex AI Search" \
  --project=YOUR_PROJECT_ID

gcloud projects add-iam-policy-binding YOUR_PROJECT_ID \
  --member="serviceAccount:gateco-vertex-search@YOUR_PROJECT_ID.iam.gserviceaccount.com" \
  --role="roles/discoveryengine.viewer"

gcloud iam service-accounts keys create gateco-vertex-search-key.json \
  --iam-account=gateco-vertex-search@YOUR_PROJECT_ID.iam.gserviceaccount.com

Step 5 — Add the connector in Gateco

  1. Navigate to Connectors → Add connector → Vertex AI Search.
  2. Enter your project ID, data store ID, location, and service account JSON.
  3. Click Test connection.
  4. Click Save.
FieldExampleDescription
project_idmy-gcp-projectGCP project ID
data_store_id123456789_my-data-storeData store ID from the GCP Console
locationglobalData store location (usually "global"; "us" or "eu" for regional stores)
serving_config_iddefault_serving_configSearch app serving config (default is fine for most setups)
service_account_json{"type":"service_account",...}Single-line service account JSON key from Step 4

Troubleshooting

ErrorCauseFix
PERMISSION_DENIEDService account lacks discoveryengine.viewer roleRun the gcloud add-iam-policy-binding command from Step 4
NOT_FOUND: data store not foundWrong data store ID or locationCopy the exact data store ID from GCP Console → Vertex AI Search → Data stores
0 results returnedData store not yet indexedCheck import status under the data store → Documents tab. Full indexing may take 15–30 minutes after import.
alpha parameter has no effectExpected — Vertex AI Search uses internal RRFGateco emits a warning when alpha != 0.5; native RRF fusion is always used regardless.

Frequently asked questions

What is the difference between Vertex AI Search and Vertex AI Vector Search?

Vertex AI Search (this guide) is a managed enterprise search service that handles document ingestion, chunking, indexing, and semantic + keyword ranking automatically. Vertex AI Vector Search is a low-level ANN service where you supply pre-computed embeddings. Use Vertex AI Search when you want Google to manage document processing; use Vertex AI Vector Search when you control embedding generation and need maximum throughput.

Does Vertex AI Search support hybrid search and grep?

Vertex AI Search uses Google's internal RRF fusion of keyword and vector signals automatically — all queries are effectively hybrid. Gateco routes both vector and hybrid search modes to Vertex AI Search's native endpoint. The alpha parameter is not mapped to a Vertex-side control in the current release. Grep search is not supported — Vertex AI Search is a semantic search engine and does not offer deterministic substring or regex matching.

What location should I use for my data store?

Most unstructured data stores use location: "global". Some enterprise features (e.g. CMEK, VPC Service Controls) require a regional location such as "us" or "eu". The default in Gateco connector config is "global". Check your data store's location in the GCP Console under Vertex AI Search → Data stores → your data store → Details.

Ready to add policy-aware retrieval?

Connect your Vertex AI setup to Gateco in under 5 minutes.

Get started freeView docs