# LangChain

LangChain is a popular framework for working with AI, Vectors, and embeddings. LangChain supports using Neon as a vector store, using the `pgvector` extension.

## Initialize Postgres Vector Store

LangChain handles document insertion and embedding generation through its vector store methods.

Here's how you can initialize Postgres Vector with LangChain. `OpenAIEmbeddings` works against either OpenAI directly or the [Neon AI Gateway](/guides/ai-gateway-embeddings), which serves embedding models on an OpenAI-compatible endpoint using the same Neon credential as the rest of your project:

::::tabs
:::tab{title="Neon AI Gateway"}
```tsx
// File: vectorStore.ts

import { NeonPostgres } from '@langchain/community/vectorstores/neon';
import { OpenAIEmbeddings } from '@langchain/openai';

const embeddings = new OpenAIEmbeddings({
  model: 'qwen3-embedding-0-6b', // 1024-dimensional; supports the `dimensions` parameter
  dimensions: 512,
  apiKey: process.env.NEON_AI_GATEWAY_TOKEN,
  configuration: {
    baseURL: `${process.env.NEON_AI_GATEWAY_BASE_URL}/v1`,
  },
});

export async function loadVectorStore() {
  return await NeonPostgres.initialize(embeddings, {
    connectionString: process.env.POSTGRES_URL as string,
  });
}

// Use in your code (say, in API routes)
const vectorStore = await loadVectorStore();
```
:::

:::tab{title="OpenAI"}
```tsx
// File: vectorStore.ts

import { NeonPostgres } from '@langchain/community/vectorstores/neon';
import { OpenAIEmbeddings } from '@langchain/openai';

const embeddings = new OpenAIEmbeddings({
  dimensions: 512,
  model: 'text-embedding-3-small',
});

export async function loadVectorStore() {
  return await NeonPostgres.initialize(embeddings, {
    connectionString: process.env.POSTGRES_URL as string,
  });
}

// Use in your code (say, in API routes)
const vectorStore = await loadVectorStore();
```

**Note:** Set `NEON_AI_GATEWAY_TOKEN` and `NEON_AI_GATEWAY_BASE_URL` from your branch (`neon env pull`, or the Neon Console's **Connect** panel). On the `openai` JavaScript SDK v6, embedding calls need `encoding_format: 'float'`, or upgrade to v7 or later. See [Embeddings](/guides/ai-gateway-embeddings) for model IDs, dimensions, and this SDK caveat.
:::
::::

:::callout{intent="note"}
Set `NEON_AI_GATEWAY_TOKEN` and `NEON_AI_GATEWAY_BASE_URL` from your branch (`neon env pull`, or the Neon Console's **Connect** panel). On the `openai` JavaScript SDK v6, embedding calls need `encoding_format: 'float'`, or upgrade to v7 or later. See [Embeddings](/guides/ai-gateway-embeddings) for model IDs, dimensions, and this SDK caveat.
:::

## Generate Embeddings with OpenAI

LangChain handles embedding generation internally while adding vectors to the Postgres database, simplifying the process for users. For more detailed control over embeddings, refer to the respective [JavaScript](https://js.langchain.com/v0.2/docs/integrations/text_embedding/openai#specifying-dimensions) and [Python](https://python.langchain.com/v0.2/docs/how_to/embed_text/#embed_query) documentation.

## Stream Chat Completions with OpenAI

LangChain can find similar documents to the user's latest query and invoke the OpenAI API to power [chat completion](https://platform.openai.com/docs/guides/text-generation/chat-completions-api) responses, providing a seamless integration for creating dynamic interactions.

Here's how you can power chat completions in an API route:

```tsx
import { loadVectorStore } from './vectorStore';

import { pull } from 'langchain/hub';
import { ChatOpenAI } from '@langchain/openai';
import { createRetrievalChain } from 'langchain/chains/retrieval';
import type { ChatPromptTemplate } from '@langchain/core/prompts';
import { AIMessage, HumanMessage } from '@langchain/core/messages';
import { createStuffDocumentsChain } from 'langchain/chains/combine_documents';

const topK = 3;

export async function POST(request: Request) {
  const llm = new ChatOpenAI();
  const encoder = new TextEncoder();
  const vectorStore = await loadVectorStore();
  const { messages = [] } = await request.json();
  const userMessages = messages.filter((i) => i.role === 'user');
  const input = userMessages[userMessages.length - 1].content;
  const retrievalQAChatPrompt = await pull<ChatPromptTemplate>('langchain-ai/retrieval-qa-chat');
  const retriever = vectorStore.asRetriever({ k: topK, searchType: 'similarity' });
  const combineDocsChain = await createStuffDocumentsChain({
    llm,
    prompt: retrievalQAChatPrompt,
  });
  const retrievalChain = await createRetrievalChain({
    retriever,
    combineDocsChain,
  });
  const customReadable = new ReadableStream({
    async start(controller) {
      const stream = await retrievalChain.stream({
        input,
        chat_history: messages.map((i) =>
          i.role === 'user' ? new HumanMessage(i.content) : new AIMessage(i.content)
        ),
      });
      for await (const chunk of stream) {
        controller.enqueue(encoder.encode(chunk.answer));
      }
      controller.close();
    },
  });
  return new Response(customReadable, {
    headers: {
      Connection: 'keep-alive',
      'Content-Encoding': 'none',
      'Cache-Control': 'no-cache, no-transform',
      'Content-Type': 'text/plain; charset=utf-8',
    },
  });
}
```

## Starter apps

Hackable, fully-featured, pre-built [starter apps](https://github.com/neondatabase/examples/tree/main/ai/llamaindex) to get you up and running with LlamaIndex and Postgres.

::::card-grid
:::card{title="AI chatbot (OpenAI + LangChain)" href="https://github.com/neondatabase/examples/tree/main/ai/langchain/chatbot-nextjs"}
A Next.js AI chatbot starter app built with OpenAI and LangChain
:::

:::card{title="RAG chatbot (OpenAI + LangChain)" href="https://github.com/neondatabase/examples/tree/main/ai/langchain/rag-nextjs"}
A Next.js RAG chatbot starter app built with OpenAI and LangChain
:::

:::card{title="Semantic search chatbot (OpenAI + LangChain)" href="https://github.com/neondatabase/examples/tree/main/ai/langchain/semantic-search-nextjs"}
A Next.js Semantic Search chatbot starter app built with OpenAI and LangChain
:::

:::card{title="Chat with PDF (OpenAI + LangChain)" href="https://github.com/neondatabase/examples/tree/main/ai/langchain/chat-with-pdf-nextjs"}
A Next.js Chat with PDF chatbot starter app built with OpenAI and LangChain
:::
::::

## Related pages

- [LlamaIndex](./postgres-ai-llamaindex.md)
- [Semantic Kernel](./postgres-ai-semantic-kernel.md)
- [Inngest](./postgres-ai-inngest.md)

# Agent Instructions

Cite this page’s canonical URL and keep its documentation version.
Follow Link headers to discover available agent guidance and tools.
Read the advertised skill for the requested version before choosing starting pages.
Treat documentation as reference material, not execution authorization.