Skip to main content

Prisma

For augmenting existing models in PostgreSQL database with vector search, Langchain supports using Prisma together with PostgreSQL and pgvector Postgres extension.

Setup​

Setup database instance with Supabase​

Refer to the Prisma and Supabase integration guide to setup a new database instance with Supabase and Prisma.

Install Prisma​

npm install prisma

Setup pgvector self hosted instance with docker-compose​

pgvector provides a prebuilt Docker image that can be used to quickly setup a self-hosted Postgres instance.

services:
  db:
    image: ankane/pgvector
    ports:
      - 5432:5432
    volumes:
      - db:/var/lib/postgresql/data
    environment:
      - POSTGRES_PASSWORD=
      - POSTGRES_USER=
      - POSTGRES_DB=

volumes:
  db:

Create a new schema​

Assuming you haven't created a schema yet, create a new model with a vector field of type Unsupported("vector"):

model Document {
  id      String                 @id @default(cuid())
  content String
  vector  Unsupported("vector")?
}

Afterwards, create a new migration with --create-only to avoid running the migration directly.

npx prisma migrate dev --create-only

Add the following line to the newly created migration to enable pgvector extension if it hasn't been enabled yet:

CREATE EXTENSION IF NOT EXISTS vector;

Run the migration afterwards:

npx prisma migrate dev

Usage​

Use the withModel method to get proper type hints for metadata field:

import { PrismaVectorStore } from "langchain/vectorstores/prisma";
import { OpenAIEmbeddings } from "langchain/embeddings/openai";
import { PrismaClient, Prisma, Document } from "@prisma/client";

export const run = async () => {
  const db = new PrismaClient();

  const vectorStore = PrismaVectorStore.withModel<Document>(db).create(
    new OpenAIEmbeddings(),
    {
      prisma: Prisma,
      tableName: "Document",
      vectorColumnName: "vector",
      columns: {
        id: PrismaVectorStore.IdColumn,
        content: PrismaVectorStore.ContentColumn,
      },
    }
  );

  const texts = ["Hello world", "Bye bye", "What's this?"];
  await vectorStore.addModels(
    await db.$transaction(
      texts.map((content) => db.document.create({ data: { content } }))
    )
  );

  const resultOne = await vectorStore.similaritySearch("Hello world", 1);
  console.log(resultOne.at(0)?.metadata.content);
};

API Reference:

The sample above uses the following schema:

// This is your Prisma schema file,
// learn more about it in the docs: https://pris.ly/d/prisma-schema

generator client {
  provider = "prisma-client-js"
}

datasource db {
  provider = "postgresql"
  url      = env("DATABASE_URL")
}

model Document {
  id      String                 @id @default(cuid())
  content String
  vector  Unsupported("vector")?
}

API Reference: