IntermediateVector Database12 min read

Connect Azure AI Search to Gateco

Connect Azure AI Search to Gateco for policy-aware hybrid vector and BM25 search. Covers service creation, index design, API key setup, and identity-aware RAG access control.

Last updated: May 27, 2026

Prerequisites

  • An Azure account with permission to create Azure AI Search services
  • An Azure AI Search service (Basic tier or above for vector search)

Overview — 5 steps

  1. 1Create an Azure AI Search service
  2. 2Create an index with a vector field
  3. 3Get your service name and admin API key
  4. 4Add the connector in Gateco
  5. 5Configure search settings

Azure AI Search (formerly Azure Cognitive Search) is a fully managed cloud search service with built-in vector search, semantic ranking, and BM25 full-text search. Gateco connects to Azure AI Search as a Tier 2 connector — supporting vector, keyword, and hybrid search with native Azure RRF fusion — and adds identity-aware retrieval and vector database RBAC on top.

Azure AI Search integration requires Gateco Growth plan or above (vendor_iam and advanced connectors). Basic tier Azure AI Search is sufficient for the Gateco connector.

Grep (regex/substring) search is not supported for Azure AI Search in this release. Use vector, keyword, or hybrid search modes.

Step 1 — Create an Azure AI Search service

  1. In the Azure Portal, search for "AI Search" and click Create.
  2. Choose a resource group, region, and service name.
  3. Select Basic pricing tier (or above).
  4. Click Review + create → Create.
  5. Note your service name (e.g. my-search-service) — not the full URL.

Pass only the service name subdomain to Gateco (e.g. my-search-service), not the full URL. Gateco constructs https://{service_name}.search.windows.net internally.

Step 2 — Create an index with a vector field

json
{
  "name": "my-docs",
  "fields": [
    { "name": "id", "type": "Edm.String", "key": true, "filterable": true },
    { "name": "content", "type": "Edm.String", "searchable": true },
    { "name": "category", "type": "Edm.String", "filterable": true },
    {
      "name": "embedding",
      "type": "Collection(Edm.Single)",
      "dimensions": 1536,
      "vectorSearchProfile": "hnsw-profile"
    }
  ],
  "vectorSearch": {
    "profiles": [{ "name": "hnsw-profile", "algorithm": "hnsw-config" }],
    "algorithms": [{ "name": "hnsw-config", "kind": "hnsw" }]
  }
}

The api_version defaults to 2025-05-01-preview. Older versions (e.g. 2024-07-01) use incompatible vector query syntax. If you see vectorSearch field errors, verify you are using a preview or recent stable API version.

Step 3 — Get your service name and API key

  1. In your search service → Keys, copy the Primary admin key.
  2. Note your service name from the Overview page (not the full URL).

Step 4 — Add the connector in Gateco

  1. Navigate to Connectors → Add connector → Azure AI Search.
  2. Enter your service name and admin API key.
  3. Click Test connection.
  4. Click Save.
FieldExampleDescription
service_namemy-search-serviceAzure AI Search service name (subdomain only, no .search.windows.net)
api_keyyour-admin-api-keyAdmin or query key from Keys section
index_namemy-docsSearch index name
vector_fieldembeddingCollection(Edm.Single) field with dimensions configured
content_fieldcontentSearchable text field for keyword/hybrid search

Troubleshooting

ErrorCauseFix
403 ForbiddenInvalid API keyCopy the Primary admin key from your search service → Keys
Index not foundWrong index nameIndex names are lowercase and case-sensitive — check exact name
Invalid vectorSearch configurationOld API versionThe connector uses api_version 2025-05-01-preview by default. If you specified an older version, remove it to use the default.

Frequently asked questions

What tier of Azure AI Search do I need for vector search?

Vector search requires Basic tier or above. The Free tier (up to 50 MB storage) does not support vector fields. For production workloads with large corpora, Standard S1 or above provides more replicas and better query throughput. Basic tier is sufficient for development and small deployments.

Should I use an admin key or a query key?

Use an admin key for initial setup and the connection test (Gateco needs to read index schema). For production, you can switch to a query key if Gateco only runs retrieval — query keys have read-only access. Admin keys are required if Gateco will also ingest documents into the index. Manage keys under your search service → Keys in the Azure Portal.

How does Azure AI Search hybrid search work in Gateco?

Azure AI Search natively combines vector search and keyword (BM25) results using RRF (Reciprocal Rank Fusion). Gateco routes hybrid search requests to Azure's native RRF mode. The alpha parameter in Gateco's retrieval request is not yet mapped to Azure's vector weight — when alpha != 0.5, Gateco emits a warning and uses native Azure RRF unchanged. Full alpha support is planned for a future release.

Ready to add policy-aware retrieval?

Connect your Azure AI Search setup to Gateco in under 5 minutes.