Skip to content
Source
Sign inSign up
TanStack/ai
Open History to see the last commit

README.md

172 lines6.6 KB

TanStack Angular AI

TanStack AI — Angular

Angular signal-based bindings for TanStack AI — streaming chat, tool-calling agents, and media generation built on Angular's native reactivity model.

Read the docs ->

Install

pnpm add @tanstack/ai-angular @tanstack/ai
npm install @tanstack/ai-angular @tanstack/ai

Minimal Usage

Important: All inject* functions use Angular's dependency injection system and must be called within an Angular injection context — a component or directive class field initializer, the constructor, or inside runInInjectionContext. Calling them outside an injection context will throw a runtime error.

The example below shows a standalone component that streams chat messages from a server endpoint via SSE:

import { Component } from '@angular/core'
import { CommonModule } from '@angular/common'
import { injectChat } from '@tanstack/ai-angular'
import { fetchServerSentEvents } from '@tanstack/ai-client'

@Component({
  selector: 'app-chat',
  standalone: true,
  imports: [CommonModule],
  template: `
    <ul>
      @for (message of chat.messages(); track message.id) {
        <li>{{ message.role }}: {{ message.content }}</li>
      }
    </ul>
    <input #input placeholder="Type a message..." />
    <button (click)="chat.sendMessage(input.value); input.value = ''">
      Send
    </button>
    @if (chat.isLoading()) {
      <p>Thinking...</p>
    }
  `,
})
export class ChatComponent {
  // injectChat is called in a field initializer — this is a valid injection context.
  chat = injectChat({
    connection: fetchServerSentEvents('/api/chat'),
  })
}

All state is exposed as Angular Signals. Read them by calling them as functions:

SignalTypeDescription
chat.messages()UIMessage[]Current message list
chat.isLoading()booleanWhether a response is streaming
chat.error()Error | undefinedLast error, if any
chat.status()ChatClientState'ready', 'streaming', 'error', ...
chat.isSubscribed()booleanWhether a live (SSE push) session is active
chat.connectionStatus()ConnectionStatusTransport connection status

Available methods on the return value: sendMessage, append, reload, stop, clear, setMessages, addToolResult, addToolApprovalResponse.

Server endpoint

The client pairs with any endpoint that returns a TanStack AI SSE stream:

import { chat, toServerSentEventsResponse } from '@tanstack/ai'
import { openaiText } from '@tanstack/ai-openai'

export async function POST(request: Request) {
  const body = await request.json()

  const stream = chat({
    adapter: openaiText('gpt-5.5'),
    messages: body.messages,
  })

  return toServerSentEventsResponse(stream)
}

Available Functions

FunctionDescription
injectChatStreaming chat with messages, tool calls, and structured outputs
injectGenerationGeneric generation client (streaming or one-shot)
injectGenerateImageImage generation
injectGenerateAudioAudio generation
injectGenerateVideoVideo generation
injectGenerateSpeechText-to-speech
injectSummarizeSummarization
injectTranscriptionAudio transcription

All generation functions return signals (result, isLoading, error, status) and methods (generate, stop, reset).

Injection Context

Angular's DI system requires that inject() is called during component construction. Every inject* function in this package calls inject() internally. Valid call sites:

// Field initializer (recommended)
export class MyComponent {
  chat = injectChat({ connection: fetchServerSentEvents('/api/chat') })
}

// Constructor
export class MyComponent {
  chat: ReturnType<typeof injectChat>
  constructor() {
    this.chat = injectChat({ connection: fetchServerSentEvents('/api/chat') })
  }
}

// Inside runInInjectionContext
const chat = runInInjectionContext(injector, () =>
  injectChat({ connection: fetchServerSentEvents('/api/chat') }),
)

Get Involved