flutter_ai_client 0.3.0

Provider-agnostic chat controller for flutter_ai: wraps any LlmProvider with optimistic send, cancellation, regeneration, and frame-batched streaming.

flutter_ai_client

The useChat controller for Flutter — wrap any LlmProvider and get optimistic send, batched streaming, cancel, and regenerate as a plain Listenable. No state-manager lock-in.

Family: flutter_ai · core · elements · openai · anthropic · gemini · tools · mcp · voice
Recipes · Migrating from the Vercel AI SDK

_{The transcript above is driven by this package's UseChatController (rendered with flutter_ai_elements).}

---

Provider-agnostic chat controller for the flutter_ai family.

UseChatController wraps any LlmProvider (from flutter_ai_core) and exposes conversation state as a plain Listenable — so it drops into ListenableBuilder and adapts cleanly to Bloc, Riverpod, or Provider. It bundles no state-manager of its own.

Features

• Optimistic send — the user's message paints synchronously, before the request is dispatched.
• Streaming, batched — events are folded by flutter_ai_core's MessageProcessor; notifications are coalesced (injectable scheduler) so high token rates don't drop frames.
• Full control — sendText, submit, stop, regenerate, clear.
• Provider/model switching — setProvider, setOptions, setTools take effect on the next turn without touching the UI.
• Escape hatch — a raw events stream for custom state layers.

Usage

final controller = UseChatController(
  provider: myProvider,                 // any LlmProvider
  options: const AiRequestOptions(model: 'gpt-4o'),
);

// Bind to the UI — rebuilds when the conversation changes.
ListenableBuilder(
  listenable: controller,
  builder: (context, _) => ListView(
    children: [
      for (final m in controller.messages) Text('${m.role.name}: ${m.text}'),
    ],
  ),
);

// Send / stop.
controller.sendText('Hello');
if (controller.status.isBusy) controller.stop();

// Switch model live.
controller.setOptions(const AiRequestOptions(model: 'gpt-4o-mini'));

See example/ for a minimal end-to-end widget.

Implementing a provider

A provider maps your backend's stream onto flutter_ai_core's AiStreamEvents:

class MyProvider implements LlmProvider {
  @override
  Stream<AiStreamEvent> send(
    AiConversation conversation, {
    List<ToolDefinition>? tools,
    AiRequestOptions? options,
  }) async* {
    yield const MessageStarted(messageId: 'a1', role: AiRole.assistant);
    yield const TextDelta(messageId: 'a1', delta: 'Hello!');
    yield const MessageFinished(messageId: 'a1', reason: FinishReason.stop);
  }
}

Status

Published on pub.dev (see the CHANGELOG for versions); depends on flutter_ai_core.

If flutter_ai saves you time, you can buy me a coffee ☕.