Zilliz RAG Knowledge Base icon

Zilliz RAG Knowledge Base

Build and manage RAG knowledge base with Zilliz vector database

Actions4

Overview

This node integrates with the Zilliz vector database to build and manage Retrieval-Augmented Generation (RAG) knowledge bases. It supports creating new knowledge base collections, processing and storing documents by chunking and embedding them, performing semantic searches, and querying with context for AI agent workflows.

Typical use cases include:

  • Creating a new vector collection in Zilliz to store document embeddings.
  • Processing large documents by splitting them into chunks, cleaning text, generating embeddings, and storing them for fast similarity search.
  • Performing semantic search queries over stored documents using natural language or precomputed query vectors.
  • Retrieving relevant contextual information to enhance AI agent responses.

For example, a user can create a knowledge base collection, then process and store product manuals as embedded chunks, and later perform semantic searches to find relevant instructions based on user queries.

Properties

Name Meaning
Database Name Name of the Zilliz database to use (default is "default").
Collection Name Name of the Zilliz collection where data will be stored or queried. Must start with a letter and contain only letters, numbers, and underscores.
Embedding Settings Settings related to vector embeddings:
- Vector Dimension: Size of embedding vectors (e.g., 1536).
- Embedding Field: Field name containing vector embeddings.
- Metric Type: Distance metric for similarity calculation; options are Cosine, Euclidean (L2), Inner Product (IP).

Output

The output JSON structure varies by operation:

  • Create Knowledge Base: Returns success status, confirmation message, collection and database names, embedding dimension, and metric type used.

    {
  "success": true,
  "message": "Knowledge base collection 'collection_name' created successfully",
  "collection": "collection_name",
  "database": "database_name",
  "dimension": 1536,
  "metric_type": "COSINE"
}

  • Other operations produce outputs related to processed documents, search results, or context retrieval, but for this operation specifically, the output confirms creation success.

Dependencies

  • Requires an API key credential and cluster endpoint configuration for authenticating with the Zilliz vector database service.
  • The node depends on the ZillizClient class from bundled shared code to interact with the Zilliz API.
  • No additional external services beyond Zilliz are required for this operation.

Troubleshooting

  • Invalid Collection Name: Collection names must start with a letter and contain only letters, numbers, and underscores. Violations cause validation errors.
  • API Authentication Errors: Ensure that the API key and cluster endpoint credentials are correctly configured and valid.
  • Dimension Mismatch: The vector dimension specified must match the embedding model used elsewhere in the workflow.
  • Unknown Operation Error: If an unsupported operation is selected, the node throws an error indicating the unknown operation.

Links and References

Discussion