tramai-vectorstore-pgvector

Version: 0.3.1
Status: Stable
Role: PostgreSQL pgvector implementation of the vector store SPI.

Purpose

This module provides a concrete implementation of VectorStore that connects to a PostgreSQL database with the pgvector extension installed. It allows you to store embeddings and perform nearest-neighbor searches directly within your relational database.

Dependencies

// build.gradle.kts
dependencies {
    implementation("dev.tramai:tramai-vectorstore-pgvector:0.3.1")
    // Depending on your stack, you'll need a JDBC driver (e.g. org.postgresql:postgresql)
}

Quick Start

import dev.tramai.vectorstore.pgvector.PgVectorStore
import dev.tramai.embedding.openai.OpenAiEmbeddingModel
import javax.sql.DataSource

val dataSource: DataSource = // ... your DB configuration ...

val store = PgVectorStore(
    dataSource = dataSource,
    embeddingModel = OpenAiEmbeddingModel(apiKey = "sk-..."),
    tableName = "document_embeddings",
    vectorDimensions = 1536 // Matches openai text-embedding-3-small
)

// The store can auto-initialize the table and vector extension if configured
store.initSchema() 

// Search
val results = store.similaritySearch("Find documents about PostgreSQL", k = 3)

When to use this module

You are already using PostgreSQL as your primary database.
You prefer to keep relational data and vector data in the same transactional boundary.
You do not want to manage a separate standalone vector database infrastructure.

When NOT to use this module

Your Postgres instance does not support the pgvector extension.
You are operating at an extremely massive vector scale where specialized cluster DBs (like Milvus or Pinecone) might be required.