BeginnerVector Database10 min read

Connect Milvus to Gateco

Connect Milvus (Zilliz Cloud or self-hosted) to Gateco for policy-aware vector retrieval. Covers authentication, collection setup, and identity-aware RAG access control.

Last updated: May 27, 2026

Prerequisites

  • A Milvus instance (self-hosted v2.3+) or a Zilliz Cloud cluster
  • At least one Milvus collection (or permission to create one)

Overview — 4 steps

  1. 1Get your Milvus endpoint and credentials
  2. 2Create a collection
  3. 3Add the connector in Gateco
  4. 4Configure search settings

Milvus is an open-source vector database built for scalable similarity search. Gateco connects to Milvus as a Tier 2 connector, adding identity-aware retrieval and RAG access control to your Milvus collections without modifying your existing schema.

Milvus keyword search is filter-only (no BM25 ranking). Matches return uniform score=1.0. For ranked text search, use a Postgres-based connector (Neon, Supabase, pgvector).

Step 1 — Get your endpoint and credentials

For Zilliz Cloud:

  1. Log in to cloud.zilliz.com and open your cluster.
  2. Copy the Public Endpoint URL (e.g. https://in01-abc123.serverless.gcp-us-west1.cloud.zilliz.com).
  3. Navigate to API Keys → Create API key. Copy the key — it is shown only once.

For self-hosted Milvus, your endpoint is typically http://your-host:19530. If RBAC is disabled (default on fresh installs), username and password are not required.

Step 2 — Create a collection

If you already have a collection, skip this step.

python
from pymilvus import MilvusClient, DataType

client = MilvusClient(uri="http://localhost:19530")

schema = MilvusClient.create_schema(auto_id=False, enable_dynamic_field=True)
schema.add_field("id", DataType.VARCHAR, max_length=256, is_primary=True)
schema.add_field("embedding", DataType.FLOAT_VECTOR, dim=1536)
schema.add_field("content", DataType.VARCHAR, max_length=65535)

client.create_collection("my_docs", schema=schema)

Step 3 — Add the connector in Gateco

  1. Navigate to Connectors → Add connector → Milvus.
  2. Enter your endpoint URL, collection name, and credentials.
  3. Click Test connection.
  4. Click Save.
FieldExampleDescription
hostin01-abc123.serverless.gcp-us-west1.cloud.zilliz.comMilvus/Zilliz endpoint hostname (no https://)
port19530Default Milvus port (Zilliz Cloud uses 443)
collection_namemy_docsName of your Milvus collection
api_keyyour-zilliz-api-tokenZilliz Cloud API token (leave blank for unauthenticated self-hosted)
usernamerootMilvus RBAC username (self-hosted only, leave blank if RBAC disabled)
passwordyour-passwordMilvus RBAC password (self-hosted only)

Step 4 — Configure search settings

  1. Click your Milvus connector → Search config.
  2. Set the Vector field (e.g. embedding).
  3. Set the Content field for the document text (e.g. content).
  4. Save.

Troubleshooting

ErrorCauseFix
connection refusedWrong port or firewallVerify port 19530 (self-hosted) or 443 (Zilliz Cloud) is accessible from Gateco
collection not foundWrong collection name (case-sensitive)Check exact collection name in your Milvus client or console
unauthenticatedRBAC enabled but credentials missingAdd username and password for self-hosted RBAC, or API key for Zilliz Cloud

Frequently asked questions

Does Gateco support both Zilliz Cloud and self-hosted Milvus?

Yes. For Zilliz Cloud, enter the cluster endpoint (e.g. https://in01-abc123.serverless.gcp-us-west1.cloud.zilliz.com) and an API token. For self-hosted Milvus, use your host and port (e.g. localhost:19530). If your self-hosted instance has authentication disabled, leave the username and password fields empty.

Does Milvus support ranked keyword search?

No. Milvus keyword search uses payload text matching (like "%query%") and returns uniform score=1.0 for all matches — there is no BM25 ranking. For ranked keyword search in a managed Postgres-compatible environment, consider using Neon or Supabase. For Milvus, hybrid search runs a vector search and uses the vector similarity scores for ranking.

Can I use Milvus partitions or partition keys with Gateco?

Gateco connects at the collection level. If you use Milvus partition keys for multi-tenancy, Gateco's identity-aware retrieval policies provide an additional layer: access is gated by organizational identity policies after the vector search returns results from the partition. To gate access before search (partition-level isolation), use the metadata_resolution_mode: inline option with a field that maps to your partition key, and write policies against that field.

Ready to add policy-aware retrieval?

Connect your Milvus setup to Gateco in under 5 minutes.