Weaviate Plugin for Dify

A comprehensive Dify plugin that provides seamless integration with Weaviate vector database, enabling powerful vector search, data management, and text embedding capabilities.

Features

🔍 Search & Query Tools

• Vector Search: Perform similarity search using vector embeddings
• Hybrid Search: Combine vector similarity and keyword search for comprehensive results
• Keyword Search: BM25-based keyword search for exact text matching
• Generative Search: RAG-powered search with LLM-generated responses
• Query Agent: Natural language query interface for intelligent operations
• Data Management: Full CRUD operations for Weaviate objects
• Schema Management: Create, delete, and manage collection schemas

🤖 Text Embedding Model

• Multiple Vectorizers: Support for OpenAI, Cohere, Hugging Face, and more
• Configurable Dimensions: Customize embedding vector dimensions
• Flexible Models: Use different embedding models based on your needs

Installation

1. Configure your Weaviate instance credentials in Dify

Configuration

Provider Credentials

• Weaviate URL: Your Weaviate instance URL (required)
• API Key: Authentication key (optional for open instances)

Text Embedding Model Parameters

• Dimensions: Number of vector dimensions (default: 1536)
• Vectorizer: Choose from text2vec-openai, text2vec-cohere, text2vec-huggingface, etc.
• Model Name: Specific model name for the selected vectorizer

Usage

Vector Search

Search for similar vectors in your collections:

{
  "collection_name": "MyCollection",
  "query_vector": "0.1,0.2,0.3,...",
  "limit": 10,
  "where_filter": "{\"path\": [\"category\"], \"operator\": \"Equal\", \"valueText\": \"AI\"}"
}

Hybrid Search

Combine vector and keyword search:

{
  "collection_name": "MyCollection",
  "query": "artificial intelligence",
  "query_vector": "0.1,0.2,0.3,...",
  "alpha": 0.7,
  "limit": 10
}

Keyword Search

Perform BM25 keyword search:

{
  "collection_name": "MyCollection",
  "query": "machine learning algorithms",
  "limit": 10,
  "search_properties": "title,content"
}

Generative Search

RAG-powered search with LLM responses:

{
  "collection_name": "MyCollection",
  "query": "What are the benefits of AI?",
  "query_vector": "0.1,0.2,0.3,...",
  "limit": 5,
  "llm_provider": "openai",
  "llm_model": "gpt-3.5-turbo"
}

Query Agent

Natural language query interface:

{
  "query": "Show me all documents about machine learning",
  "collection_name": "MyCollection",
  "max_results": 10
}

Data Management

Insert, update, delete, or retrieve objects:

{
  "operation": "insert",
  "collection_name": "MyCollection",
  "object_data": "{\"text\": \"Hello World\", \"category\": \"greeting\"}"
}

Schema Management

Create and manage collection schemas:

{
  "operation": "create_collection",
  "collection_name": "MyCollection",
  "properties": "[{\"name\": \"text\", \"data_type\": \"TEXT\"}, {\"name\": \"category\", \"data_type\": \"TEXT\"}]"
}

File Structure

weaviate_plugin/
├── _assets/                    # Plugin icons
├── models/
│   └── text_embedding/         # Text embedding model
├── provider/                   # Provider configuration
├── tools/                      # Tool implementations
│   ├── vector_search.py        # Vector similarity search
│   ├── hybrid_search.py        # Hybrid search
│   ├── keyword_search.py       # BM25 keyword search
│   ├── generative_search.py    # RAG-powered search
│   ├── query_agent.py          # Natural language query agent
│   ├── data_management.py      # CRUD operations
│   └── schema_management.py    # Schema operations
├── utils/                      # Utility functions
│   ├── client.py              # Weaviate client
│   ├── validators.py          # Input validation
│   └── helpers.py             # Helper functions
├── main.py                    # Plugin entry point
├── manifest.yaml              # Plugin manifest
└── requirements.txt           # Dependencies

Supported Operations

Vector Search

• Similarity search using vector embeddings
• Configurable result limits and filters
• Metadata and property selection

Hybrid Search

• Combines vector similarity and keyword search
• Adjustable alpha parameter for weighting
• Advanced filtering capabilities

Keyword Search

• BM25-based keyword matching
• Configurable search properties
• Exact text matching capabilities

Generative Search

• RAG-powered search with LLM integration
• Context-aware response generation
• Support for multiple LLM providers

Query Agent

• Natural language query interpretation
• Intelligent operation selection
• Conversational response generation

Data Management

• Insert single or multiple objects
• Update existing objects by UUID
• Delete objects by UUID
• Retrieve objects with property selection
• List all collections

Schema Management

• Create collections with custom properties
• Delete collections
• Retrieve collection schemas
• Get collection statistics
• List all collections

Error Handling

The plugin includes comprehensive error handling for:

• Invalid credentials
• Network connectivity issues
• Malformed input data
• Weaviate API errors
• Validation failures

Development

To extend the plugin:

1. Add new tools in the tools/ directory
2. Create corresponding YAML configurations
3. Register tools in main.py
4. Update the provider YAML to include new tools

License

This plugin is provided by Weaviate for integration with Dify.

Author: weaviate Version: 0.0.1 Type: tool

Repository

Source code: https://github.com/weaviate/Dify-Weaviate-Plugin