Aviary

Hello chat

The smallest end-to-end AI chat — a NestJS server in about six lines and a React client in about five, no login, no database.

A working, public, login-less chat: every visitor gets their own conversation history, nothing to configure but the model.

Server

pnpm add @dudousxd/nestjs-agent @dudousxd/nestjs-agent-ai-sdk @ai-sdk/openai
// src/app.module.ts
import { Module } from '@nestjs/common';
import { AgentModule } from '@dudousxd/nestjs-agent';
import { aiSdkModel } from '@dudousxd/nestjs-agent-ai-sdk';
import { openai } from '@ai-sdk/openai';

@Module({ imports: [AgentModule.forRoot({ model: aiSdkModel(openai('gpt-5-mini')) })] })
export class AppModule {}

Set OPENAI_API_KEY, start the app, and /agent/chat, /agent/threads, … are live. The two boot notices (in-memory store, public endpoints) are the defaults you change for production — see Getting Started → the defaults.

React

pnpm add @dudousxd/nestjs-agent-react ai @ai-sdk/react
import { ChatInput, MessageList, useAgentChat } from '@dudousxd/nestjs-agent-react';

export function Chat() {
  const chat = useAgentChat();
  return (
    <>
      <MessageList messages={chat.messages} status={chat.status} queue={chat.queue} />
      <ChatInput onSubmit={(text) => chat.sendMessage({ text })} />
    </>
  );
}

Keep typing while it answers: a message sent mid-turn waits in the thread's queue (server-side) and runs when the answer finishes — queue={chat.queue} shows what is waiting. See Typing ahead.

The hook talks to the same origin under /agent. If the API lives elsewhere, wrap the app once:

<AgentProvider baseUrl="https://api.example.com" credentials="include">
  <Chat />
</AgentProvider>

(a cross-site API also needs AnonymousActorResolver's cookie sent cross-site — actorResolver: new AnonymousActorResolver({ sameSite: 'none' }) on the server, over HTTPS).

What you get without writing it

Each browser is its own anonymous actor (an HttpOnly cookie), so useThreads() lists only that visitor's conversations. Components ship no styles: pass classNames, or render your own markup off chat.transcript (Bring your own UI).

MessageList does not draw tool calls or questions

MessageList renders text, reasoning, files and pushed components (through renderUi or a <GenuiProvider>). A tool call appears only through your renderToolPart / renderToolGroup, which get the raw part, and a question set not at all — so an action tool's approval has no button there unless your renderToolPart draws one (chat.approve({ toolCallId })). chat.transcript carries both with approve / reject / answer / skip already bound to the backend: render them from it, or install the registry's AgentChat (Chat components), which draws them.

Next

Getting Started

Stand up an AI chat in a NestJS app — install, register the module with a model, declare a tool, and stream your first turn.

Full-featured chat

Login, a persistent store, a model picker, quota, attachments, approvals and generative UI — one AgentProvider and one useAgentChat, with your own markup.

On this page

ServerReactNext