Add implementation of Prisma vectorstore (#440)

* Add implementation of Prisma vectorstore

* Fix CI

* FIx lint issues

* FIx incorrect index for addVectors

* Add example to Prisma

* Fix returned types

* Add docs

* Remove @prisma/client from peer-dependencies and dependencies

* Add implementation of fromTexts and fromDocuments

* Update yarn.lock

* Fix broken example

docs/docs/modules/indexes/vector_stores/prisma.mdx

@@ -0,0 +1,74 @@
+# Prisma
+
+For augmenting existing models in PostgreSQL database with vector search, Langchain supports using [Prisma](https://www.prisma.io/) together with PostgreSQL and [`pgvector`](https://github.com/pgvector/pgvector) Postgres extension.
+
+## Setup
+
+### Setup database instance with Supabase
+
+Refer to the [Prisma and Supabase integration guide](https://supabase.com/docs/guides/integrations/prisma) to setup a new database instance with Supabase and 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.
+
+```yaml
+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")`:
+
+```prisma
+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.
+
+```bash npm2yarn
+npm run 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:
+
+```sql
+CREATE EXTENSION IF NOT EXISTS vector;
+```
+
+Run the migration afterwards:
+
+```bash npm2yarn
+npm run prisma migrate dev
+```
+
+## Usage
+
+import CodeBlock from "@theme/CodeBlock";
+import Example from "@examples/indexes/vector_stores/prisma_vectorstore/prisma.ts";
+import Schema from "@examples/indexes/vector_stores/prisma_vectorstore/prisma/schema.prisma";
+
+Use the `withModel` method to get proper type hints for `metadata` field:
+
+<CodeBlock language="typescript">{Example}</CodeBlock>
+
+The sample above uses the following schema:
+
+<CodeBlock language="prisma">{Schema}</CodeBlock>

examples/package.json

@@ -11,6 +11,7 @@
   "scripts": {
     "build": "tsc --declaration --outDir dist/",
     "start": "tsx -r dotenv/config src/index.ts",
+    "postinstall": "prisma generate --schema ./src/indexes/vector_stores/prisma_vectorstore/prisma/schema.prisma",
     "start:dist": "yarn build && node -r dotenv/config dist/index.js",
     "lint": "eslint src",
     "lint:fix": "yarn lint --fix",
@@ -23,10 +24,12 @@
     "@dqbd/tiktoken": "^1.0.2",
     "@getmetal/metal-sdk": "^1.0.12",
     "@pinecone-database/pinecone": "^0.0.10",
+    "@prisma/client": "^4.11.0",
     "@supabase/supabase-js": "^2.10.0",
     "chromadb": "^1.3.0",
     "js-yaml": "^4.1.0",
     "langchain": "workspace:*",
+    "prisma": "^4.11.0",
     "sqlite3": "^5.1.4",
     "typeorm": "^0.3.12",
     "zod": "^3.21.4"

examples/src/indexes/vector_stores/prisma_vectorstore/.env.example

@@ -0,0 +1,2 @@
+# Add DATABASE_URL to .env file in this directory
+DATABASE_URL=postgresql://[USERNAME]:[PASSWORD]@[ADDR]/[DBNAME]

examples/src/indexes/vector_stores/prisma_vectorstore/.gitignore

@@ -0,0 +1,2 @@
+data
+docker-compose.yml

examples/src/indexes/vector_stores/prisma_vectorstore/docker-compose.example.yml

@@ -0,0 +1,11 @@
+services:
+  db:
+    image: ankane/pgvector
+    ports:
+      - 5432:5432
+    volumes:
+      - ./data:/var/lib/postgresql/data
+    environment:
+      - POSTGRES_PASSWORD=
+      - POSTGRES_USER=
+      - POSTGRES_DB=

examples/src/indexes/vector_stores/prisma_vectorstore/prisma.ts

@@ -0,0 +1,30 @@
+import { PrismaVectorStore } from "langchain/vectorstores";
+import { OpenAIEmbeddings } from "langchain/embeddings";
+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);
+};

examples/src/indexes/vector_stores/prisma_vectorstore/prisma/migrations/00_init/migration.sql

@@ -0,0 +1,9 @@
+-- CreateTable
+CREATE EXTENSION IF NOT EXISTS vector;
+CREATE TABLE "Document" (
+    "id" TEXT NOT NULL,
+    "content" TEXT NOT NULL,
+    "vector" vector,
+
+    CONSTRAINT "Document_pkey" PRIMARY KEY ("id")
+);

examples/src/indexes/vector_stores/prisma_vectorstore/prisma/migrations/migration_lock.toml

@@ -0,0 +1,3 @@
+# Please do not edit this file manually
+# It should be added in your version-control system (i.e. Git)
+provider = "postgresql"

examples/src/indexes/vector_stores/prisma_vectorstore/prisma/schema.prisma

@@ -0,0 +1,17 @@
+// 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")?
+}

langchain/src/document.ts

@@ -1,21 +1,26 @@
-export interface DocumentParams {
+export interface DocumentParams<
+  // eslint-disable-next-line @typescript-eslint/no-explicit-any
+  Metadata extends Record<string, any> = Record<string, any>
+> {
   pageContent: string;
 
-  // eslint-disable-next-line @typescript-eslint/no-explicit-any
-  metadata: Record<string, any>;
+  metadata: Metadata;
 }
 
 /**
  * Interface for interacting with a document.
  */
-export class Document implements DocumentParams {
+export class Document<
+  // eslint-disable-next-line @typescript-eslint/no-explicit-any
+  Metadata extends Record<string, any> = Record<string, any>
+> implements DocumentParams
+{
   pageContent: string;
 
-  // eslint-disable-next-line @typescript-eslint/no-explicit-any
-  metadata: Record<string, any>;
+  metadata: Metadata;
 
-  constructor(fields?: Partial<DocumentParams>) {
+  constructor(fields?: Partial<DocumentParams<Metadata>>) {
     this.pageContent = fields?.pageContent ?? this.pageContent;
-    this.metadata = fields?.metadata ?? {};
+    this.metadata = fields?.metadata ?? ({} as Metadata);
   }
 }

langchain/src/vectorstores/index.ts

@@ -3,3 +3,4 @@ export { Chroma } from "./chroma.js";
 export { PineconeStore } from "./pinecone.js";
 export { VectorStore, SaveableVectorStore } from "./base.js";
 export { SupabaseVectorStore } from "./supabase.js";
+export { PrismaVectorStore } from "./prisma.js";