Postgres Vector Store

The Postgres Vector Store node is a component that allows for the storage and retrieval of vector embeddings using PostgreSQL with the pgvector extension. It’s designed to work within a larger system for document storage, indexing, and similarity search operations.

Node Details

• Name: Postgres_VectorStores

• Type: Postgres

• Version: 6.0

• Category: Vector Stores

Base Classes

• Postgres

• VectorStoreRetriever

• BaseRetriever

Credential

• Type: PostgresApi

• Required Fields:

  • User

  • Password

Input Parameters

1. Document (optional, list)

   • Type: Document

   • Description: List of documents to be stored in the vector store

2. Embeddings

   • Type: Embeddings

   • Description: Embedding model to use for vectorizing documents

3. Record Manager (optional)

   • Type: RecordManager

   • Description: Keeps track of records to prevent duplication

4. Host

   • Type: string

   • Description: PostgreSQL server host

5. Database

   • Type: string

   • Description: Name of the PostgreSQL database

6. Port (optional)

   • Type: number

   • Default: 6432

   • Description: PostgreSQL server port

7. Table Name (optional)

   • Type: string

   • Default: “documents”

   • Description: Name of the table to store vectors

8. File Upload (optional)

   • Type: boolean

   • Description: Enables file upload functionality in the chat

9. Additional Configuration (optional)

   • Type: JSON

   • Description: Additional PostgreSQL connection options

10. Top K (optional)

    • Type: number

    • Default: 4

    • Description: Number of top results to fetch in similarity search

11. Postgres Metadata Filter (optional)

    • Type: JSON

    • Description: Filter to apply on metadata during similarity search

Outputs

1. Postgres Retriever

   • Type: VectorStoreRetriever

   • Description: A retriever object for similarity search operations

2. Postgres Vector Store

   • Type: TypeORMVectorStore

   • Description: The vector store object for direct interactions

Functionality

Initialization

• Establishes a connection to the PostgreSQL database

• Sets up the vector store with the specified table and embeddings

• Configures similarity search function to use pg pool for better performance

Upsert Method

• Allows adding new documents to the vector store

• Handles file upload scenarios by adding chat ID to metadata

• Supports using a record manager to prevent duplicates

Delete Method

• Enables deletion of documents from the vector store

• Supports deletion using record manager or direct ID-based deletion

Similarity Search

• Performs similarity search on stored vectors

• Applies metadata filters if specified

• Returns top K results based on cosine similarity

Use Cases

• Document storage and retrieval in AI-powered applications

• Semantic search functionality in large document collections

• Building knowledge bases for question-answering systems

Notes

• Requires PostgreSQL with pgvector extension installed

• Optimized for performance with connection pooling

• Supports additional configuration for fine-tuning PostgreSQL connection

• Integrates with file upload functionality for chat-based document ingestion