Milvus

The vector database used for retrieval and semantic search in AI.

Agentic Friendly

Component Category

Vector database

Component Description

Milvus is the vector database used by AI for storing embeddings and supporting semantic retrieval workflows. It is built for large-scale vector search and similarity matching across embedding-driven applications. BullSequana AI uses one shared Milvus service with platform-managed collections scoped to individual tenants.

Why It Is Used

In BullSequana AI, Milvus provides the vector storage and similarity search capabilities needed for retrieval-augmented generation and other embedding-based experiences. Tenant-scoped collection names let the BSQAI API enforce the active tenant boundary before it reads, searches, or changes vector data.

Learn More

Tenant isolation

The BSQAI API gives each collection the tenant's prefix and a generated suffix. A tenant's first embedding-model save creates its collection. A reindex fills a new collection, switches search to it only when the reindex completes, and then deletes the previous one.

Collection operations expose only collections in the active tenant's namespace. Search uses the active collection recorded for that tenant, so a foreign or inactive collection identifier is returned as not found.

Deployment notes

BullSequana AI deploys Milvus 2.6.25, with Helm chart 5.0.30, in cluster mode with a streaming node and a mixed coordinator. A three-member etcd cluster stores coordination metadata. The write-ahead log uses the platform's Strimzi Kafka cluster over TLS with SCRAM-SHA-512 authentication; the Kafka, Pulsar, and Woodpecker services that the Milvus chart can bundle are off. The configured S3-compatible object store persists Milvus data in the milvus bucket, in Rook-Ceph RGW or in the configured external provider.

Interacts With

  • BSQAI API, for vector queries and retrieval logic, and for the embeddings that its file-event worker writes.
  • Kafka (Strimzi), the platform Kafka cluster that holds the Milvus write-ahead log, reached over TLS with SCRAM-SHA-512.
  • etcd, which stores Milvus coordination metadata in a three-member cluster.
  • Rook-Ceph or external S3-compatible storage, which persists Milvus data.
  • Docling, whose document processing outputs feed vector embedding workflows.
  • AI Web Portal, where administrators inspect platform-managed collections for the selected tenant.

On this page