Skip to main content

Prisma

为了使用向量搜索增强 PostgreSQL 数据库中的现有模型，Langchain 支持将 Prisma 与 PostgreSQL 和 pgvector Postgres 扩展一起使用。

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

设置​

¥Setup

使用 Supabase 设置数据库实例​

¥Setup database instance with Supabase

请参阅 Prisma 和 Supabase 集成指南 使用 Supabase 和 Prisma 设置新的数据库实例。

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

安装 Prisma​

¥Install Prisma

npm install prisma

使用 docker-compose 设置 pgvector 自托管实例​

¥Setup pgvector self hosted instance with docker-compose

pgvector 提供了一个预构建的 Docker 镜像，可用于快速设置自托管的 Postgres 实例。

¥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:

创建新模式​

¥Create a new schema

假设你尚未创建模式，请创建一个包含 vector 字段和 Unsupported("vector") 类型的新模型：

¥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")?
}

之后，创建一个包含 --create-only 的新迁移，以避免直接运行迁移。

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

npx prisma migrate dev --create-only

如果尚未启用 pgvector 扩展，请将以下行添加到新创建的迁移中以启用它：

¥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;

之后运行迁移：

¥Run the migration afterwards:

npx prisma migrate dev

用法​

¥Usage

tip

请参阅 此部分包含有关安装集成包的一般说明。

¥See this section for general instructions on installing integration packages.

npm install @langchain/openai @langchain/community @langchain/core
danger

表名和列名（例如 tableName、vectorColumnName、columns 和 filter 等字段）直接传递到 SQL 查询中，无需参数化。这些字段必须事先进行清理，以避免 SQL 注入。

¥Table names and column names (in fields such as tableName, vectorColumnName, columns and filter) are passed into SQL queries directly without parametrisation. These fields must be sanitized beforehand to avoid SQL injection.

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

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

  // Use the `withModel` method to get proper type hints for `metadata` field:
  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);

  // create an instance with default filter
  const vectorStore2 = PrismaVectorStore.withModel<Document>(db).create(
    new OpenAIEmbeddings(),
    {
      prisma: Prisma,
      tableName: "Document",
      vectorColumnName: "vector",
      columns: {
        id: PrismaVectorStore.IdColumn,
        content: PrismaVectorStore.ContentColumn,
      },
      filter: {
        content: {
          equals: "default",
        },
      },
    }
  );

  await vectorStore2.addModels(
    await db.$transaction(
      texts.map((content) => db.document.create({ data: { content } }))
    )
  );

  // Use the default filter a.k.a {"content": "default"}
  const resultTwo = await vectorStore.similaritySearch("Hello world", 1);
  console.log(resultTwo);
};

API Reference:

以下 SQL 运算符可用作过滤器：equals, in, isNull, isNotNull, like, lt, lte, gt, gte, not.

¥The following SQL operators are available as filters: equals, in, isNull, isNotNull, like, lt, lte, gt, gte, not.

以上示例使用以下架构：

¥The samples 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
  namespace String?                @default("default")
  vector    Unsupported("vector")?
}

API Reference:

    如果你不需要 namespace，可以将其删除。

    ¥You can remove namespace if you don't need it.

    ¥Related