Zep Collection

The Zep Collection node is a vector store implementation using Zep, an open-source platform for building LLM apps. It allows for upsert of embedded data and performing similarity or MMR (Maximal Marginal Relevance) search upon query.

Node Details

• Name: zep

• Type: Zep

• Version: 2.0

• Category: Vector Stores

Base Classes

• Zep

• VectorStoreRetriever

• BaseRetriever

Parameters

Input Parameters

1. Document (optional)

   • Type: Document

   • List: true

   • Description: The documents to be stored in the vector store.

2. Embeddings

   • Type: Embeddings

   • Description: The embedding model to use for vectorizing the documents.

3. Base URL

   • Type: string

   • Default: “http://127.0.0.1:8000”

   • Description: The base URL of the Zep instance.

4. Zep Collection

   • Type: string

   • Description: The name of the Zep collection to use.

5. Zep Metadata Filter (optional)

   • Type: json

   • Description: A JSON object to filter documents based on metadata.

6. Embedding Dimension

   • Type: number

   • Default: 1536

   • Description: The dimension of the embedding vectors.

7. Top K (optional)

   • Type: number

   • Default: 4

   • Description: The number of top results to fetch.

8. MMR-related parameters (added through addMMRInputParams function)

Credential Parameters

• API Key (optional)

  • Type: credential

  • Description: JWT authentication for the Zep instance.

Outputs

1. Zep Retriever

   • Type: retriever

   • Base Classes: [Zep, VectorStoreRetriever, BaseRetriever]

2. Zep Vector Store

   • Type: vectorStore

   • Base Classes: [Zep, ZepVectorStore]

Functionality

The Zep_VectorStores node provides two main functionalities:

1. Upsert: Allows adding new documents to the Zep collection. It processes the input documents, applies the specified embeddings, and stores them in the Zep vector store.

2. Similarity Search: Enables querying the vector store for similar documents. It supports both standard similarity search and MMR search for diverse results.

Usage

This node is particularly useful for:

• Building knowledge bases or document retrieval systems

• Implementing semantic search functionality in applications

• Creating question-answering systems with context retrieval

The node can be integrated into larger workflows where document storage, retrieval, and similarity search are required, making it a versatile component for various LLM-powered applications.