n8n-nodes-milvus-enhanced

An n8n community node for Milvus vector database with enhanced metric type switching capabilities.

Features

• New Credential: Uses a new milvusNew credential separate from built-in n8n Milvus nodes
• AI Tool Integration: Use as an AI tool for document retrieval in LangChain workflows
• Enhanced Metric Type Support: Switch between different similarity metrics (L2, IP, Cosine, Hamming, Jaccard)
• Complete CRUD Operations: Insert, search, query, and delete operations
• Collection Management: Create collections with configurable metric types
• Advanced Search: Support for vector similarity search with customizable parameters
• Filtering: Boolean expression filtering for precise results
• Flexible Configuration: Configurable vector dimensions and search parameters
• Resource Locator: Easy collection selection with dropdown list

Installation

To install this community node in your n8n instance:

Manual Installation

1. Navigate to your n8n installation directory
2. Install the package:

   npm install n8n-nodes-milvus-enhanced

3. Restart n8n

Using n8n Editor UI

1. Go to Settings → Community Nodes
2. Click Install a community node
3. Enter: n8n-nodes-milvus-enhanced
4. Click Install

Troubleshooting Installation

If you encounter the error "The specified package could not be loaded":

1. Check Package Structure: Ensure the package includes all required files:

   npm run build
   npm run lint

2. Verify n8n Version: Ensure you're running n8n version 1.0.0 or higher

3. Manual Installation: If UI installation fails, try manual installation:

   npm install n8n-nodes-milvus-enhanced
   # Restart n8n

4. Clear Cache: Clear n8n's cache and restart:

   # Stop n8n, then:
   rm -rf ~/.n8n/cache
   # Start n8n again

Configuration

Credentials Setup

1. Create a new "Milvus New API" credential in n8n
2. Configure the following:
   • Base URL: Full Milvus server URL (e.g., http://10.42.0.230:19530)
   • Username: Authentication username
   • Password: Authentication password

Operations

Retrieve as Tool (AI Integration)

Use the node as an AI tool for document retrieval with embedding support:

• Tool Description: Custom description for the AI agent
• Collection: Select from existing collections using resource locator
• Top K: Number of documents to retrieve
• Search Metric Type: Override metric type for search operations (L2, IP, Cosine, Hamming, Jaccard)
• Takes ai_embedding input from embedding models
• Provides ai_tool output to AI agents

Insert Documents

Insert documents and vectors into a collection:

• Collection: Target collection (resource locator)
• Data: JSON array of documents to insert

Get Many

Retrieve documents from a collection:

• Collection: Source collection (resource locator)
• Top K: Number of documents to return
• Search Parameters: Advanced retrieval options

Create Collection

Create a new collection with a specified metric type:

• Collection Name: Name of the collection to create
• Metric Type: Choose from L2, IP, Cosine, Hamming, or Jaccard
• Vector Dimension: Dimension of vector fields
• Description: Optional collection description

Insert Vectors

Insert vectors into a collection:

• Collection Name: Target collection
• Data: JSON array of data objects to insert

Search Vectors

Search for similar vectors with configurable metric types:

• Collection Name: Collection to search in
• Metric Type: Similarity metric (L2, IP, Cosine, Hamming, Jaccard)
• Vectors: Query vectors as JSON array
• Top K: Number of results to return
• Search Parameters: Advanced search parameters
• Output Fields: Fields to include in results
• Filter Expression: Boolean filter expression

Query Vectors

Query vectors by filter expressions:

• Collection Name: Target collection
• Filter Expression: Boolean expression for filtering
• Output Fields: Fields to include in results

Delete Vectors

Delete vectors based on filter expressions:

• Collection Name: Target collection
• Filter Expression: Boolean expression for deletion criteria

List Collections

List all available collections in the database.

Supported Metric Types

| Metric Type | Description | Use Case | |-------------|-------------|----------| | L2 | Euclidean distance | General purpose, normalized vectors | | IP | Inner Product | Cosine similarity for normalized vectors | | Cosine | Cosine similarity | Text embeddings, semantic search | | Hamming | Hamming distance | Binary vectors, fingerprints | | Jaccard | Jaccard distance | Set similarity, sparse vectors |

Example Workflows

AI Tool with Embeddings Integration

{
  "nodes": [
    {
      "parameters": {
        "model": "text-embedding-3-large",
        "options": {
          "dimensions": 3072
        }
      },
      "type": "@n8n/n8n-nodes-langchain.embeddingsOpenAi",
      "typeVersion": 1.2,
      "position": [784, 624],
      "id": "embeddings",
      "name": "Embeddings OpenAI"
    },
    {
      "parameters": {
        "mode": "retrieve-as-tool",
        "toolDescription": "repository source",
        "milvusCollection": {
          "__rl": true,
          "value": "gitlab_stemx_source",
          "mode": "name"
        },
        "topK": 10,
        "searchMetricType": "COSINE"
      },
      "type": "n8n-nodes-milvus-enhanced.milvusEnhanced",
      "typeVersion": 1,
      "position": [864, 384],
      "id": "milvus-enhanced",
      "name": "Milvus Enhanced",
      "credentials": {
        "milvusNew": {
          "id": "your-milvus-new-credential-id",
          "name": "Milvus New account"
        }
      }
    },
    {
      "parameters": {
        "authentication": "headerAuth",
        "path": "webhook-path"
      },
      "type": "@n8n/n8n-nodes-langchain.mcpTrigger",
      "typeVersion": 2,
      "position": [816, 32],
      "id": "mcp-trigger",
      "name": "MCP Server Trigger"
    }
  ],
  "connections": {
    "Embeddings OpenAI": {
      "ai_embedding": [
        [
          {
            "node": "Milvus Enhanced",
            "type": "ai_embedding",
            "index": 0
          }
        ]
      ]
    },
    "Milvus Enhanced": {
      "ai_tool": [
        [
          {
            "node": "MCP Server Trigger",
            "type": "ai_tool",
            "index": 0
          }
        ]
      ]
    }
  }
}

Insert Documents

{
  "parameters": {
    "mode": "insert",
    "milvusCollection": {
      "__rl": true,
      "value": "my_collection",
      "mode": "list"
    },
    "data": [
      {
        "id": 1,
        "vector": [0.1, 0.2, 0.3],
        "text": "Sample document"
      }
    ]
  }
}

Create Collection with Custom Metric

{
  "parameters": {
    "operation": "createCollection",
    "collection": "embeddings_collection",
    "metricType": "COSINE",
    "dimension": 768,
    "description": "Text embeddings with cosine similarity"
  }
}

Development

Prerequisites

• Node.js 18+
• n8n development environment

Local Development

# Clone repository
git clone https://gitlab.com/interticket/it-stemx/ai-source/milvus-n8n-node.git
cd milvus-n8n-node

# Install dependencies
npm install

# Build
npm run build

# Lint
npm run lint

Project Structure

├── credentials/
│   └── MilvusApi.credentials.ts    # Credential configuration
├── nodes/
│   └── MilvusEnhanced/
│       ├── MilvusEnhanced.node.ts  # Main node implementation
│       └── milvus.svg              # Node icon
├── dist/                           # Compiled output
├── package.json                    # Package configuration
└── README.md                       # Documentation

Requirements

• n8n version 1.0.0 or higher
• Milvus server 2.3.0 or higher

License

MIT License - see LICENSE file for details.

Support

For issues, questions, or contributions:

• Create an issue in the GitLab repository
• Check n8n community forums for general n8n questions

Changelog

1.0.0

• Initial release
• Support for all major Milvus operations
• Enhanced metric type switching
• Complete CRUD functionality