GitHub Repository

You can find the project source code on GitHub.

This guide provides detailed, step-by-step instructions on how to use Upstash RAG Chat with Next.js. You can also explore our Next.js Chat to Website example for detailed, end-to-end examples and best practices.

​
Project Setup

Create a new Next.js application and install @upstash/rag-chat package.

npx create-next-app@latest
cd my-app
npm install @upstash/rag-chat

​
Setup Upstash Redis

Create a Redis database using Upstash Console or Upstash CLI and copy the UPSTASH_REDIS_REST_URL and UPSTASH_REDIS_REST_TOKEN into your .env file.

.env
UPSTASH_REDIS_REST_URL=<YOUR_URL>
UPSTASH_REDIS_REST_TOKEN=<YOUR_TOKEN>

​
Setup Upstash Vector

Create a Vector index using Upstash Console or Upstash CLI and copy the UPSTASH_VECTOR_REST_URL and UPSTASH_VECTOR_REST_TOKEN into your .env file.

.env
UPSTASH_VECTOR_REST_URL=<YOUR_URL>
UPSTASH_VECTOR_REST_TOKEN=<YOUR_TOKEN>

​
Setup QStash LLM

Navigate to QStash Console and copy the QSTASH_TOKEN into your .env file.

.env
QSTASH_TOKEN=<YOUR_TOKEN>

​
Configure Middleware for Cookies

Create a middleware in /app/middleware.ts to generate a session ID cookie. This cookie will be used to track the user’s chat session.

middleware.ts
import { NextRequest, NextResponse } from "next/server";

export function middleware(req: NextRequest) {
  const res = NextResponse.next();

  const cookie = req.cookies.get("sessionId");

  if (!cookie) {
    res.cookies.set("sessionId", crypto.randomUUID());
  }

  return res;
}

​
Create a Chat Streaming API Route

We will create a RAG Chat instance in /app/lib/rag-chat.ts to later use in our API route and Home page.

/app/lib/rag-chat.ts
import { RAGChat, upstash } from "@upstash/rag-chat";
import { Redis } from "@upstash/redis";

const redis = Redis.fromEnv();

export const ragChat = new RAGChat({
  model: upstash("meta-llama/Meta-Llama-3-8B-Instruct"),
  redis: redis,
});

Using the RAG Chat instance we created, we can now create a chat streaming API route in /app/api/chat-stream/route.ts.

/app/api/chat-stream/route.ts
import { ragChat } from "@/app/lib/rag-chat";
import { aiUseChatAdapter } from "@upstash/rag-chat/nextjs";
import { NextRequest } from "next/server";

export const POST = async (req: NextRequest) => {
  const { messages, sessionId } = await req.json();

  const lastMessage = messages[messages.length - 1].content;

  const response = await ragChat.chat(lastMessage, {
    streaming: true,
    sessionId,
  });

  return aiUseChatAdapter(response);
};

​
Create a Chat Component

We can now create a chat component in /app/components/chat.tsx using the chat streaming API route we created.

/app/components/chat.tsx
"use client";

import { useChat, Message } from "ai/react";

export const Chat = ({
  sessionId,
  initialMessages,
}: {
  sessionId: string;
  initialMessages: Message[];
}) => {
  const { messages, handleInputChange, handleSubmit, input, setInput } =
    useChat({
      api: "/api/chat-stream",
      body: { sessionId },
      initialMessages,
    });

  return (
    <div>
      <div className="flex h-[calc(100vh-64px)] flex-col overflow-y-auto items-center">
        {messages.length ? (
          messages.map((message, i) => (
            <div className="w-[60vw] flex flex-row gap-3 p-6" key={i}>
              <div className="min-w-12">
                {message.role === "user" ? "User" : "Bot"}
              </div>
              <p className="text-sm font-normal text-gray-900">
                {message.content}
              </p>
            </div>
          ))
        ) : (
          <div className="text-zinc-500 text-sm pt-6">
            Ask your first question to get started.
          </div>
        )}
      </div>
      <form onSubmit={handleSubmit} className="relative">
        <div className="flex flex-row gap-2 p-3 h-16 items-center">
          <input
            onChange={handleInputChange}
            value={input}
            onKeyDown={(e) => {
              if (e.key === "Enter" && !e.shiftKey) {
                e.preventDefault();
                handleSubmit();
                setInput("");
              }
            }}
            placeholder="Enter your question..."
            className="border rounded-xl w-full py-1 px-3"
          />

          <button type="submit" className="border rounded-xl py-1 px-3">
            Send
          </button>
        </div>
      </form>
    </div>
  );
};

​
Update Home Page

Update /app/page.tsx using the Chat component we created.

/app/page.tsx
import { Chat } from "./components/chat";
import { cookies } from "next/headers";
import { ragChat } from "@/app/lib/rag-chat";

export default async function Home() {
  const sessionId = cookies().get("sessionId")?.value;
  const initialMessages = await ragChat.history.getMessages({
    amount: 10,
    sessionId,
  });
  return <Chat sessionId={sessionId!} initialMessages={initialMessages} />;
}

​
Run & Deploy

Run the app locally with npm run dev, check http://localhost:3000/

Deploy your app with vercel

Was this page helpful?

OverviewNext.js Server Actions