app icon
Milvus Vector Database
0.1.4

Complete Milvus vector database integration for Dify applications with collection management, vector search, data operations, and more

zeroz-lab/milvus3771 installs

Milvus Plugin for Dify

A plugin that integrates Milvus vector database with the Dify platform, providing vector operations for collection management, data insertion, search, and querying.

中文文档 | English

Features

🗂️ Collection Management

  • List Collections: View all available collections
  • Describe Collection: Get detailed collection information
  • Collection Stats: Retrieve collection statistics
  • Check Existence: Verify if collections exist

📥 Data Operations

  • Insert Data: Add vectors and metadata to collections (vectors must be precomputed by the caller)
  • Upsert Data: Insert or update existing data
  • Query Data: Retrieve data by ID or filter conditions (auto-detects the primary-key field)
  • Delete Data: Remove data from collections (accepts IDs or filter expressions)

When inserting, provide vectors directly in each entity, for example:

🔍 Vector Search

  • Similarity Search: Find similar vectors using various metrics
  • Filtered Search: Combine vector similarity with metadata filters
  • Multi-Vector Search: Search with multiple query vectors
  • Custom Parameters: Adjust search behavior parameters

🔀 Hybrid Search

  • Combine results from multiple vector fields in one request and rerank with a strategy.
  • Endpoint: (the plugin auto-injects ).
  • Tool name in Dify: Milvus Hybrid Search.

Parameters (Dify tool form)

  • (string, required)
  • (string, required): JSON array of search objects with precomputed vectors. Each object should include at least:
    • : list of embeddings (precomputed), e.g.
    • : target vector field name
    • : per-route top-K
    • Optional per-route keys supported by REST API: , , , , , ,
  • (select): or
  • (string or JSON object): For use ; for use
  • Optional top-level: , , (comma-separated string or JSON array), (comma-separated string or JSON array), , , , (boolean), (JSON)

Built-in validation

  • Each search item must have (non-empty) and (> 0 integer), and provide (non-empty array) with numeric vectors.
  • If , must be numeric and its length must equal the number of search routes.
  • If top-level is provided, ensure (API limit).
  • If or are provided as strings, they are split by comma after trimming whitespace.

Example

If you manage tool parameters via code, and can also be passed as native JSON structures (list/dict) instead of strings. The plugin will normalize both formats.

Installation & Configuration

Connection Configuration

Configure your Milvus connection in the Dify platform:

  • URI: Milvus server address (e.g., )
  • Token: Authentication token (optional, format: )
  • Database: Target database name (default: )

License

MIT License

CATEGORY
Tool
VERSION
0.1.4
zeroz-lab·10/28/2025 12:55 PM
REQUIREMENTS
Maximum memory
256MB