feat (ai/svelte): add useAssistant (#1593)

Co-authored-by: Jake Hall <jake.hallslm@gmail.com>

.changeset/thick-jars-hear.md

@@ -0,0 +1,5 @@
+---
+'ai': patch
+---
+
+feat (ai/svelte): add useAssistant

content/docs/05-ai-sdk-ui/03-openai-assistants.mdx

@@ -5,7 +5,11 @@ description: Learn how to use the useAssistant hook.
 
 # OpenAI Assistants
 
-The `useAssistant` hook allows you to handle the client state when interacting with an OpenAI compatible assistant API. This hook is useful when you want to integrate assistant capabilities into your application, with the UI updated automatically as the assistant is streaming its execution.
+The `useAssistant` hook allows you to handle the client state when interacting with an OpenAI compatible assistant API.
+This hook is useful when you want to integrate assistant capabilities into your application,
+with the UI updated automatically as the assistant is streaming its execution.
+
+The `useAssistant` hook is currently supported with `ai/react` and `ai/svelte`.
 
 ## Example
 

content/docs/07-reference/ai-sdk-ui/03-use-assistant.mdx

@@ -5,14 +5,22 @@ description: Reference documentation for the useAssistant hook in the AI SDK UI
 
 # `useAssistant`
 
-Allows you to handle the client state when interacting with an OpenAI compatible assistant API. This hook is useful when you want to integrate assistant capibilities into your application, with the UI updated automatically as the assistant is streaming its execution.
+Allows you to handle the client state when interacting with an OpenAI compatible assistant API.
+This hook is useful when you want to integrate assistant capibilities into your application,
+with the UI updated automatically as the assistant is streaming its execution.
 
-This works in conjunction with [`AssistantResponse`]() in the backend.
+This works in conjunction with [`AssistantResponse`](/docs/reference/stream-helpers/assistant-response) in the backend.
+
+`useAssistant` is currently supported with `ai/react` and `ai/svelte`.
 
 ## Import
 
 ### React
 
 <Snippet text={`import { useAssistant } from "ai/react"`} prompt={false} />
 
+### Svelte
+
+<Snippet text={`import { useAssistant } from "ai/svelte"`} prompt={false} />
+
 <ReferenceTable packageName="react" functionName="useAssistant" />

examples/sveltekit-openai/src/routes/api/assistant/+server.ts

@@ -0,0 +1,117 @@
+import type { RequestHandler } from './$types';
+
+import { env } from '$env/dynamic/private';
+
+import { AssistantResponse } from 'ai';
+import OpenAI from 'openai';
+
+const openai = new OpenAI({
+  apiKey: env.OPENAI_API_KEY || '',
+});
+
+const homeTemperatures = {
+  bedroom: 20,
+  'home office': 21,
+  'living room': 21,
+  kitchen: 22,
+  bathroom: 23,
+};
+
+export const POST = (async ({ request }) => {
+  // Parse the request body
+  const input: {
+    threadId: string | null;
+    message: string;
+  } = await request.json();
+
+  // Create a thread if needed
+  const threadId = input.threadId ?? (await openai.beta.threads.create({})).id;
+
+  // Add a message to the thread
+  const createdMessage = await openai.beta.threads.messages.create(threadId, {
+    role: 'user',
+    content: input.message,
+  });
+
+  return AssistantResponse(
+    { threadId, messageId: createdMessage.id },
+    async ({ forwardStream, sendDataMessage }) => {
+      // Run the assistant on the thread
+      const runStream = openai.beta.threads.runs.stream(threadId, {
+        assistant_id:
+          env.ASSISTANT_ID ??
+          (() => {
+            throw new Error('ASSISTANT_ID is not set');
+          })(),
+      });
+
+      // forward run status would stream message deltas
+      let runResult = await forwardStream(runStream);
+
+      // status can be: queued, in_progress, requires_action, cancelling, cancelled, failed, completed, or expired
+      while (
+        runResult?.status === 'requires_action' &&
+        runResult.required_action?.type === 'submit_tool_outputs'
+      ) {
+        const tool_outputs =
+          runResult.required_action.submit_tool_outputs.tool_calls.map(
+            (toolCall: any) => {
+              const parameters = JSON.parse(toolCall.function.arguments);
+
+              switch (toolCall.function.name) {
+                case 'getRoomTemperature': {
+                  const temperature =
+                    homeTemperatures[
+                      parameters.room as keyof typeof homeTemperatures
+                    ];
+
+                  return {
+                    tool_call_id: toolCall.id,
+                    output: temperature.toString(),
+                  };
+                }
+
+                case 'setRoomTemperature': {
+                  const oldTemperature =
+                    homeTemperatures[
+                      parameters.room as keyof typeof homeTemperatures
+                    ];
+
+                  homeTemperatures[
+                    parameters.room as keyof typeof homeTemperatures
+                  ] = parameters.temperature;
+
+                  sendDataMessage({
+                    role: 'data',
+                    data: {
+                      oldTemperature,
+                      newTemperature: parameters.temperature,
+                      description: `Temperature in ${parameters.room} changed from ${oldTemperature} to ${parameters.temperature}`,
+                    },
+                  });
+
+                  return {
+                    tool_call_id: toolCall.id,
+                    output: `temperature set successfully`,
+                  };
+                }
+
+                default:
+                  throw new Error(
+                    `Unknown tool call function: ${toolCall.function.name}`,
+                  );
+              }
+            },
+          );
+
+        runResult = await forwardStream(
+          openai.beta.threads.runs.submitToolOutputsStream(
+            threadId,
+            runResult.id,
+            { tool_outputs },
+          ),
+        );
+      }
+    },
+  );
+}) satisfies RequestHandler;

examples/sveltekit-openai/src/routes/api/assistant/assistant-setup.md

@@ -0,0 +1,63 @@
+# Home Automation Assistant Example
+
+## Setup
+
+### Create OpenAI Assistant
+
+[OpenAI Assistant Website](https://platform.openai.com/assistants)
+
+Create a new assistant. Enable Code interpreter. Add the following functions and instructions to the assistant.
+
+Then add the assistant id to the `.env` file as `ASSISTANT_ID=your-assistant-id`.
+
+### Instructions
+
+```
+You are an assistant with access to a home automation system. You can get and set the temperature in the bedroom, home office, living room, kitchen and bathroom.
+
+The system uses temperature in Celsius. If the user requests Fahrenheit, you should convert the temperature to Fahrenheit.
+```
+
+### getRoomTemperature function
+
+```json
+{
+  "name": "getRoomTemperature",
+  "description": "Get the temperature in a room",
+  "parameters": {
+    "type": "object",
+    "properties": {
+      "room": {
+        "type": "string",
+        "enum": ["bedroom", "home office", "living room", "kitchen", "bathroom"]
+      }
+    },
+    "required": ["room"]
+  }
+}
+```
+
+### setRoomTemperature function
+
+```json
+{
+  "name": "setRoomTemperature",
+  "description": "Set the temperature in a room",
+  "parameters": {
+    "type": "object",
+    "properties": {
+      "room": {
+        "type": "string",
+        "enum": ["bedroom", "home office", "living room", "kitchen", "bathroom"]
+      },
+      "temperature": { "type": "number" }
+    },
+    "required": ["room", "temperature"]
+  }
+}
+```
+
+## Run
+
+1. Run `pnpm dev` in `examples/sveltekit-openai`
+2. Go to http://localhost:5173/assistant

examples/sveltekit-openai/src/routes/assistant/+page.svelte

@@ -0,0 +1,46 @@
+<script>
+	import { useAssistant } from 'ai/svelte'
+	const { messages, input, submitMessage } = useAssistant({
+		api: '/api/assistant',
+	});
+</script>
+
+<svelte:head>
+	<title>Home</title>
+	<meta name="description" content="Svelte demo app" />
+</svelte:head>
+
+<section>
+	<h1>useAssistant</h1>
+	<ul>
+		{#each $messages as m}
+			<strong>{m.role}</strong> 
+			{#if m.role !== 'data'}
+			{m.content}
+			{/if}
+			{#if m.role === 'data'}
+			<pre>{JSON.stringify(m.data, null, 2)}}</pre>
+			{/if}
+			<br/>
+			<br/>
+		{/each}
+	</ul>
+	<form on:submit={submitMessage}>
+		<input bind:value={$input} />
+		<button type="submit">Send</button>
+	</form>
+</section>
+
+<style>
+	section {
+		display: flex;
+		flex-direction: column;
+		justify-content: center;
+		align-items: center;
+		flex: 0.6;
+	}
+
+	h1 {
+		width: 100%;
+	}
+</style>

packages/core/react/use-assistant.ts

@@ -4,10 +4,12 @@ import { isAbortError } from '@ai-sdk/provider-utils';
 import { useCallback, useRef, useState } from 'react';
 import { generateId } from '../shared/generate-id';
 import { readDataStream } from '../shared/read-data-stream';
-import { CreateMessage, Message } from '../shared/types';
-import { abort } from 'node:process';
-
-export type AssistantStatus = 'in_progress' | 'awaiting_message';
+import {
+  AssistantStatus,
+  CreateMessage,
+  Message,
+  UseAssistantOptions,
+} from '../shared/types';
 
 export type UseAssistantHelpers = {
   /**
@@ -16,7 +18,7 @@ export type UseAssistantHelpers = {
   messages: Message[];
 
   /**
-   * setState-powered method to update the messages array.
+   * Update the message store with a new array of messages.
    */
   setMessages: React.Dispatch<React.SetStateAction<Message[]>>;
 
@@ -83,42 +85,6 @@ Abort the current request immediately, keep the generated tokens if any.
   error: undefined | unknown;
 };
 
-export type UseAssistantOptions = {
-  /**
-   * The API endpoint that accepts a `{ threadId: string | null; message: string; }` object and returns an `AssistantResponse` stream.
-   * The threadId refers to an existing thread with messages (or is `null` to create a new thread).
-   * The message is the next message that should be appended to the thread and sent to the assistant.
-   */
-  api: string;
-
-  /**
-   * An optional string that represents the ID of an existing thread.
-   * If not provided, a new thread will be created.
-   */
-  threadId?: string;
-
-  /**
-   * An optional literal that sets the mode of credentials to be used on the request.
-   * Defaults to "same-origin".
-   */
-  credentials?: RequestCredentials;
-
-  /**
-   * An optional object of headers to be passed to the API endpoint.
-   */
-  headers?: Record<string, string> | Headers;
-
-  /**
-   * An optional, additional body object to be passed to the API endpoint.
-   */
-  body?: object;
-
-  /**
-   * An optional callback that will be called when the assistant encounters an error.
-   */
-  onError?: (error: Error) => void;
-};
-
 export function useAssistant({
   api,
   threadId: threadIdParam,
@@ -254,8 +220,7 @@ export function useAssistant({
           }
 
           case 'error': {
-            const errorObj = new Error(value);
-            setError(errorObj);
+            setError(new Error(value));
             break;
           }
         }

packages/core/shared/types.ts

@@ -1,6 +1,8 @@
 import { ToolCall as CoreToolCall } from '../core/generate-text/tool-call';
 import { ToolResult as CoreToolResult } from '../core/generate-text/tool-result';
 
+export * from './use-assistant-types';
+
 // https://github.com/openai/openai-node/blob/07b3504e1c40fd929f4aae1651b83afc19e3baf8/src/resources/chat/completions.ts#L146-L159
 export interface FunctionCall {
   /**

packages/core/shared/use-assistant-types.ts

@@ -0,0 +1,38 @@
+// Define a type for the assistant status
+export type AssistantStatus = 'in_progress' | 'awaiting_message';
+
+export type UseAssistantOptions = {
+  /**
+   * The API endpoint that accepts a `{ threadId: string | null; message: string; }` object and returns an `AssistantResponse` stream.
+   * The threadId refers to an existing thread with messages (or is `null` to create a new thread).
+   * The message is the next message that should be appended to the thread and sent to the assistant.
+   */
+  api: string;
+
+  /**
+   * An optional string that represents the ID of an existing thread.
+   * If not provided, a new thread will be created.
+   */
+  threadId?: string;
+
+  /**
+   * An optional literal that sets the mode of credentials to be used on the request.
+   * Defaults to "same-origin".
+   */
+  credentials?: RequestCredentials;
+
+  /**
+   * An optional object of headers to be passed to the API endpoint.
+   */
+  headers?: Record<string, string> | Headers;
+
+  /**
+   * An optional, additional body object to be passed to the API endpoint.
+   */
+  body?: object;
+
+  /**
+   * An optional callback that will be called when the assistant encounters an error.
+   */
+  onError?: (error: Error) => void;
+};