Features/AI Bots/Streaming Chat

Watch AI Think In Real Time.

Bot responses stream token-by-token directly into chat bubbles. See the artificial intelligence think, compose, and respond as it happens.

Streaming Capabilities

stream

Token-by-Token

Every word appears as the model generates it. No loading spinners, no black boxes — just pure, visible intelligence.

chat_bubble

In-World Bubbles

Streams appear directly in proximity-based chat bubbles above the bot. The conversation feels alive and immediate.

bolt

Low Latency

Optimized WebSocket pipeline keeps response delay to a minimum. First-token latency under 200ms for supported providers.

Streaming Architecture

Our streaming pipeline uses server-sent events and WebSocket connections to push tokens from the AI provider directly into the spatial chat system with minimal overhead.

  • check_circleDirect provider-to-browser streaming
  • check_circleAutomatic reconnection on interruption
  • check_circleMulti-model streaming support
stream_handler.ts
const stream = await bot.chat.stream({
  message: userInput,
  onToken: (token) => {
    bubble.append(token);
  },
});

// Each token appears in the bubble instantly

Ready to see AI think in real-time?

Enable streaming on your bots through the provider config.