Milvus

Data

Neotask connects OpenClaw to your Milvus vector database — search, index, and manage embeddings through natural conversation instead of writing query code.

• Run text, vector, hybrid, and multi-vector searches across your collections without writing code
• Create collections, insert data, and build indexes through plain English instructions
• Monitor collection stats and manage your vector infrastructure conversationally

What You Can Do

The Milvus integration provides 11 actions covering search, data management, and collection administration:

Search Operations

Text search — full-text search across collection fields for keyword matching

Vector search — similarity search using embeddings for semantic retrieval

Hybrid search — combine text and vector search for best-of-both-worlds results

Multi-vector search — search across multiple vector fields simultaneously

Query — filter and retrieve records using boolean expressions

Count — count records matching filter conditions

Collection Management

List all collections in your instance

Get schema, index, and statistics for any collection

Create new collections with defined schemas

Insert data records into collections

Build indexes on vector or scalar fields

Every action runs autonomously or requires your approval — you decide.

Try Asking

"Search our product catalog for items similar to this description"

"How many records are in the customer embeddings collection?"

"Create a new collection called 'support_tickets' with a 768-dimension vector field"

"Run a hybrid search combining keyword 'pricing' with the semantic meaning of 'cost reduction'"

"Show me the schema and index details for the knowledge_base collection"

"Insert these 50 product records into the catalog collection"

Pro Tips

Hybrid search typically outperforms pure vector or pure text search — default to it for production retrieval use cases.

Build indexes before running large-scale searches; unindexed collections scan every record.

Use multi-vector search when your records have separate embeddings for title, body, and metadata.

Pair Milvus with your content management system in an app group so agents can automatically embed and store new documents as they are published.

Works Well With

• openai - Connect Milvus and OpenAI with Neotask to build AI-powered semantic search, vector storage, and retrieval-augmented gene...
• synapseorg - Connect Milvus vector search with Synapse biomedical datasets. Index research embeddings, run similarity search across p...