AI

Chatbot Message Persistence

Being able to store and load chat messages is crucial for most AI chatbots. In this guide, we'll show how to implement message persistence with useChat and streamText.

This guide does not cover authorization, error handling, or other real-world considerations. It is intended to be a simple example of how to implement message persistence.

Starting a new chat

When the user navigates to the chat page without providing a chat ID, we need to create a new chat and redirect to the chat page with the new chat ID.

app/chat/page.tsx
import { redirect } from 'next/navigation';
import { createChat } from '@tools/chat-store';


export default async function Page() {
  const id = await createChat(); // create a new chat
  redirect(`/chat/${id}`); // redirect to chat page, see below
}

Our example chat store implementation uses files to store the chat messages. In a real-world application, you would use a database or a cloud storage service, and get the chat ID from the database. That being said, the function interfaces are designed to be easily replaced with other implementations.

tools/chat-store.ts
import { generateId } from 'ai';
import { existsSync, mkdirSync } from 'fs';
import { writeFile } from 'fs/promises';
import path from 'path';


export async function createChat(): Promise<string> {
  const id = generateId(); // generate a unique chat ID
  await writeFile(getChatFile(id), '[]'); // create an empty chat file
  return id;
}


function getChatFile(id: string): string {
  const chatDir = path.join(process.cwd(), '.chats');
  if (!existsSync(chatDir)) mkdirSync(chatDir, { recursive: true });
  return path.join(chatDir, `${id}.json`);
}

Loading an existing chat

When the user navigates to the chat page with a chat ID, we need to load the chat messages and display them.

app/chat/[id]/page.tsx
import { loadChat } from '@tools/chat-store';
import Chat from '@ui/chat';


export default async function Page(props: { params: Promise<{ id: string }> }) {
  const { id } = await props.params; // get the chat ID from the URL
  const messages = await loadChat(id); // load the chat messages
  return <Chat id={id} initialMessages={messages} />; // display the chat
}

The loadChat function in our file-based chat store is implemented as follows:

tools/chat-store.ts
import { UIMessage } from 'ai';
import { readFile } from 'fs/promises';


export async function loadChat(id: string): Promise<UIMessage[]> {
  return JSON.parse(await readFile(getChatFile(id), 'utf8'));
}


// ... rest of the file

The display component is a simple chat component that uses the useChat hook to send and receive messages:

ui/chat.tsx
'use client';


import { UIMessage, useChat } from '@ai-sdk/react';
import { defaultChatStoreOptions } from 'ai';


export default function Chat({
  id,
  initialMessages,
}: { id?: string | undefined; initialMessages?: UIMessage[] } = {}) {
  const { input, handleInputChange, handleSubmit, messages } = useChat({
    chatId: id, // use the provided chat ID
    chatStore: defaultChatStoreOptions({
      api: '/api/chat',
      chats:
        initialMessages && id
          ? { [id]: { messages: initialMessages } }
          : undefined,
    }),
  });


  // simplified rendering code, extend as needed:
  return (
    <div>
      {messages.map(m => (
        <div key={m.id}>
          {m.role === 'user' ? 'User: ' : 'AI: '}
          {m.parts
            .map(part => (part.type === 'text' ? part.text : ''))
            .join('')}
        </div>
      ))}


      <form onSubmit={handleSubmit}>
        <input value={input} onChange={handleInputChange} />
      </form>
    </div>
  );
}

Storing messages

useChat sends the chat id and the messages to the backend.

The useChat message format is different from the ModelMessage format. The useChat message format is designed for frontend display, and contains additional fields such as id and createdAt. We recommend storing the messages in the useChat message format.

Storing messages is done in the onFinish callback of the streamText function. onFinish receives the messages from the frontend and backend as UIMessage[].

app/api/chat/route.ts
import { openai } from '@ai-sdk/openai';
import { saveChat } from '@tools/chat-store';
import { convertToModelMessages, streamText, UIMessage } from 'ai';


export async function POST(req: Request) {
  const { messages, chatId }: { messages: UIMessage[]; chatId: string } =
    await req.json();


  const result = streamText({
    model: openai('gpt-4o-mini'),
    messages: convertToModelMessages(messages),
  });


  return result.toUIMessageStreamResponse({
    originalMessages: messages,
    onFinish: ({ messages }) => {
      saveChat({ chatId, messages });
    },
  });
}

The actual storage of the messages is done in the saveChat function, which in our file-based chat store is implemented as follows:

tools/chat-store.ts
import { UIMessage } from 'ai';
import { writeFile } from 'fs/promises';


export async function saveChat({
  chatId,
  messages,
}: {
  chatId: string;
  messages: UIMessage[];
}): Promise<void> {
  const content = JSON.stringify(messages, null, 2);
  await writeFile(getChatFile(chatId), content);
}


// ... rest of the file

Message IDs

In addition to a chat ID, each message has an ID. You can use this message ID to e.g. manipulate individual messages.

The IDs for user messages are generated by the useChat hook on the client, and the IDs for AI response messages are generated by streamText.

You can control the ID format by providing ID generators (see createIdGenerator():

ui/chat.tsx
import { createIdGenerator } from 'ai';
import { useChat } from '@ai-sdk/react';
import { defaultChatStoreOptions } from 'ai';


const {
  // ...
} = useChat({
  // ...
  chatStore: defaultChatStoreOptions({
    // id format for client-side messages:
    generateId: createIdGenerator({
      prefix: 'msgc',
      size: 16,
    }),
  }),
});
app/api/chat/route.ts
import { createIdGenerator, streamText } from 'ai';


export async function POST(req: Request) {
  // ...
  const result = streamText({
    // ...
    // id format for server-side messages:
    experimental_generateMessageId: createIdGenerator({
      prefix: 'msgs',
      size: 16,
    }),
  });
  // ...
}

Sending only the last message

Once you have implemented message persistence, you might want to send only the last message to the server. This reduces the amount of data sent to the server on each request and can improve performance.

To achieve this, you can provide an prepareRequestBody function to the useChat hook. This function receives the messages and the chat ID, and returns the request body to be sent to the server.

ui/chat.tsx
import { useChat } from '@ai-sdk/react';
import { defaultChatStoreOptions } from 'ai';


const {
  // ...
} = useChat({
  // ...
  chatStore: defaultChatStoreOptions({
    // only send the last message to the server:
    prepareRequestBody({ messages, chatId }) {
      return { message: messages[messages.length - 1], chatId };
    },
  }),
});

On the server, you can then load the previous messages and append the new message to the previous messages:

app/api/chat/route.ts
import { appendClientMessage, convertToModelMessages } from 'ai';


export async function POST(req: Request) {
  // get the last message from the client:
  const { message, chatId } = await req.json();


  // load the previous messages from the server:
  const previousMessages = await loadChat(chatId);


  // append the new message to the previous messages:
  const messages = appendClientMessage({
    messages: previousMessages,
    message,
  });


  const result = streamText({
    // ...
    messages: convertToModelMessages(messages),
  });


  // ...
}

Handling client disconnects

By default, the AI SDK streamText function uses backpressure to the language model provider to prevent the consumption of tokens that are not yet requested.

However, this means that when the client disconnects, e.g. by closing the browser tab or because of a network issue, the stream from the LLM will be aborted and the conversation may end up in a broken state.

Assuming that you have a storage solution in place, you can use the consumeStream method to consume the stream on the backend, and then save the result as usual. consumeStream effectively removes the backpressure, meaning that the result is stored even when the client has already disconnected.

app/api/chat/route.ts
import { convertToModelMessages, streamText, UIMessage } from 'ai';
import { saveChat } from '@tools/chat-store';


export async function POST(req: Request) {
  const { messages, chatId }: { messages: UIMessage[]; chatId: string } =
    await req.json();


  const result = streamText({
    model,
    messages: convertToModelMessages(messages),
  });


  // consume the stream to ensure it runs to completion & triggers onFinish
  // even when the client response is aborted:
  result.consumeStream(); // no await


  return result.toUIMessageStreamResponse({
    originalMessages: messages,
    onFinish: ({ messages }) => {
      saveChat({ chatId, messages });
    },
  });
}

When the client reloads the page after a disconnect, the chat will be restored from the storage solution.

In production applications, you would also track the state of the request (in progress, complete) in your stored messages and use it on the client to cover the case where the client reloads the page after a disconnection, but the streaming is not yet complete.

Resuming ongoing streams

This feature is experimental and may change in future versions.

The useChat hook has experimental support for resuming an ongoing chat generation stream by any client, either after a network disconnect or by reloading the chat page. This can be useful for building applications that involve long-running conversations or for ensuring that messages are not lost in case of network failures.

The following are the pre-requisities for your chat application to support resumable streams:

  • Installing the resumable-stream package that helps create and manage the publisher/subscriber mechanism of the streams.
  • Creating a Redis instance to store the stream state.
  • Creating a table that tracks the stream IDs associated with a chat.

To resume a chat stream, you will use the experimental_resume function returned by the useChat hook. You will call this function during the initial mount of the hook inside the main chat component.

ui/chat.tsx
'use client';


import { useChat } from '@ai-sdk/react';
import { defaultChatStoreOptions, type UIMessage } from 'ai';
import { useEffect } from 'react';


export function Chat({
  chatId,
  autoResume,
  initialMessages = [],
}: {
  chatId: string;
  autoResume: boolean;
  initialMessages: UIMessage[];
}) {
  const {
    experimental_resume,
    // ... other useChat returns
  } = useChat({
    chatId,
    chatStore: defaultChatStoreOptions({
      api: '/api/chat',
      chats: {
        [chatId]: {
          messages: initialMessages,
        },
      },
    }),
  });


  useEffect(() => {
    if (autoResume) {
      experimental_resume();
    }
    // We want to disable the exhaustive deps rule here because we only want to run this effect once
    // eslint-disable-next-line react-hooks/exhaustive-deps
  }, []);


  return <div>{/* Your chat UI here */}</div>;
}

The experimental_resume function makes a GET request to your configured chat endpoint (or /api/chat by default) whenever your client calls it. If there’s an active stream, it will pick up where it left off, otherwise it simply finishes without error.

The GET request automatically appends the chatId query parameter to the URL to help identify the chat the request belongs to. Using the chatId, you can look up the most recent stream ID from the database and resume the stream.

GET /api/chat?chatId=<your-chat-id>

Earlier, you must've implemented the POST handler for the /api/chat route to create new chat generations. When using experimental_resume, you must also implement the GET handler for /api/chat route to resume streams.

1. Implement the GET handler

Add a GET method to /api/chat that:

  1. Reads chatId from the query string
  2. Validates it’s present
  3. Loads any stored stream IDs for that chat
  4. Returns the latest one to streamContext.resumableStream()
  5. Falls back to an empty stream if it’s already closed
app/api/chat/route.ts
import { loadStreams } from '@/util/chat-store';
import { createUIMessageStream, JsonToSseTransformStream } from 'ai';
import { after } from 'next/server';
import { createResumableStreamContext } from 'resumable-stream';


export async function GET(request: Request) {
  const streamContext = createResumableStreamContext({
    waitUntil: after,
  });


  const { searchParams } = new URL(request.url);
  const chatId = searchParams.get('chatId');


  if (!chatId) {
    return new Response('id is required', { status: 400 });
  }


  const streamIds = await loadStreams(chatId);


  if (!streamIds.length) {
    return new Response('No streams found', { status: 404 });
  }


  const recentStreamId = streamIds.at(-1);


  if (!recentStreamId) {
    return new Response('No recent stream found', { status: 404 });
  }


  const emptyDataStream = createUIMessageStream({
    execute: () => {},
  });


  return new Response(
    await streamContext.resumableStream(recentStreamId, () =>
      emptyDataStream.pipeThrough(new JsonToSseTransformStream()),
    ),
  );
}

After you've implemented the GET handler, you can update the POST handler to handle the creation of resumable streams.

2. Update the POST handler

When you create a brand-new chat completion, you must:

  1. Generate a fresh streamId
  2. Persist it alongside your chatId
  3. Kick off a createDataStream that pipes tokens as they arrive
  4. Hand that new stream to streamContext.resumableStream()
app/api/chat/route.ts
import {
  convertToModelMessages,
  createDataStream,
  generateId,
  streamText,
  UIMessage,
} from 'ai';
import { appendStreamId, saveChat } from '@/util/chat-store';
import { createResumableStreamContext } from 'resumable-stream';
import { openai } from '@ai-sdk/openai';


const streamContext = createResumableStreamContext({
  waitUntil: after,
});


async function POST(request: Request) {
  const { chatId, messages }: { chatId: string; messages: UIMessage[] } =
    await request.json();
  const streamId = generateId();


  // Record this new stream so we can resume later
  await appendStreamId({ chatId, streamId });


  // Build the data stream that will emit tokens
  const stream = createUIMessageStream({
    execute: ({ writer }) => {
      const result = streamText({
        model: openai('gpt-4o'),
        messages: convertToModelMessages(messages),
      });


      // Return a resumable stream to the client
      writer.merge(result.toUIMessageStream());
    },
  });


  const resumableStream = await streamContext.resumableStream(
    streamId,
    () => stream,
  );


  return resumableStream.toUIMessageStreamResponse({
    originalMessages: messages,
    onFinish: ({ messages }) => {
      saveChat({ chatId, messages });
    },
  });
}

With both handlers, your clients can now gracefully resume ongoing streams.