Typing Indicators

await chat.startTyping();
await chat.stopTyping();
 
chat.on('typing', (event) => {
  console.log(`${event.userId} ${event.isTyping ? 'started' : 'stopped'} typing`);
});

const { typingUsers, onInput } = useTyping();
// <input onChange={onInput} /> — call on every keystroke

await raven.chat!.startTyping();
await raven.chat!.stopTyping();

await chat.startTyping();
await chat.stopTyping();
 
chat.typing.listen((event) {
  print('${event.userId} ${event.isTyping ? "started" : "stopped"} typing');
});

Safe to call startTyping() on every keystroke.

Never persisted

Typing events never reach Postgres — a signal meaningless seven seconds after it happens, and writing one per keystroke would be the single most wasteful thing in the system. Redis, 7-second TTL, fan-out over pub/sub. That's the whole implementation.

Stale indicators

"Alice is typing…" that never goes away is the classic bug here. It happens when a client crashes, closes a tab, or loses its network mid-sentence and the stop signal never arrives. Three independent defenses, because relying on any one alone is how the bug gets back in:

1. Server-side TTL — the Redis key expires after 7 seconds regardless of what any client sends.
2. Client-side timeout — @corvidhq/chat arms a local timer on startTyping() and stops automatically after a pause.
3. Receiver-side expiry — @corvidhq/react's store expires a typing user locally after 8 seconds even if the stop frame is lost in transit.

You never see your own typing event echoed back — there's no reason a client would need to be told it's typing.

Next

• Presence
• Reactions