FluxyChat

Cookbook

Transport fallback (WebSocket → SSE → polling)

`useChat` and `createFluxyRoomSession` automatically degrade transport when the room WebSocket cannot stay connected.

Order

  1. WebSocket — FluxyChatRoomConnection with exponential backoff reconnect (default 6 attempts in useChat).
  2. SSE — GET /rooms/:id/stream when reconnect attempts are exhausted and a member JWT is present.
  3. Polling — REST GET /api/messages every 4s if SSE is unavailable.

When WebSocket reconnects, SSE/polling stops and live delivery resumes on the socket.

UI

Use connectionState from the SDK (not raw connectionStatus alone):

const { connectionState } = useChat({ roomId });

if (connectionState.status === "reconnecting" && connectionState.nextRetryAt) {
  const seconds = Math.max(
    0,
    Math.ceil((new Date(connectionState.nextRetryAt).getTime() - Date.now()) / 1000),
  );
  return <span>Reconnecting in {seconds}s…</span>;
}

if (connectionState.transport === "sse") {
  return <span>Live updates via SSE</span>;
}

if (connectionState.transport === "polling") {
  return <span>Updating every few seconds</span>;
}

Vanilla (no React)

import { createFluxyRoomSession } from "@fluxy-chat/sdk";

const { store, stop } = createFluxyRoomSession({
  roomId: "my-room",
  client,
});

store.subscribe((state) => {
  console.log(state.connectionState);
});

// later: stop();

Public demo & guest token hardening

The dashboard route `/demo` and Worker `GET|POST /demo/session` mint short-lived guest JWTs for a shared room. Without guards, bots can burn quota and pollute t

Use case: Assistant room (AI agent)

Goal: a room where an AI agent can be invoked (manually or via mentions), with run history and cost visibility.

On this page

OrderUIVanilla (no React)Related