Azure PostgreSQL Vector Database

This repository contains the simplest way to setup and interact with a vector database in Azure PostgreSQL using the pgvector extension. It's designed to essentially mimic Pinecone's Vector Database but with Azure PostgreSQL.

Features

• Create Vector Tables: Easily create tables with vector and metadata columns.
• Upsert Vectors: Insert or update vectors with associated metadata.
• Query Vectors: Perform efficient nearest neighbor searches on vectors.

Getting started

Prerequisites

• Azure PostgreSQL database with pgvector extension enabled.
• Python environment with the required packages installed.

Installation

pip install -r requirements.txt

Configuration

1. Open the vector_database.py file.
2. Fill in the DB_PARAMS dictionary with your database credentials.

Create an index

Creates an index for storing vectors and metadata.

from vector_database import create_index

# Create an index with dimension of 1536
table_name = create_index(name='my_vector_index', dimension_size=1536)

Upsert vectors

Inserts or updates vectors with associated metadata.

from vector_database import upsert_vectors

# Upsert vectors to simulate OpenAI's embedding vectors
example_vectors = [
    (1, [0.1] * 1536, json.dumps({"description": "Vector 1"})),
    (2, [0.2] * 1536, json.dumps({"description": "Vector 2"})),
    (3, [0.3] * 1536, json.dumps({"description": "Vector 3"})),
    (4, [0.4] * 1536, json.dumps({"description": "Vector 4"}))
]
upsert_vectors(table_name, example_vectors)

Query vectors

from vector_database import query_vectors

# Query vectors
query_results = query_vector(table_name, [0.15] * 1536, limit=1)