REST usage

Send contextual events

Enrich ai_licia’s context (max 1,000 chars). Does not force an immediate response.

await client.sendEvent('Player health: 85%, Position: X:145 Y:230, Kills: 12');
await client.sendEvent('Current song: Never Gonna Give You Up');

With TTL (seconds):

await client.sendEvent('Lights: blue', 300);

Trigger a reaction

Ask ai_licia to react now (max 300 chars). Returns generation metadata.

const res = await client.triggerGeneration('Player just defeated the final boss!');
console.log(res);

For a brief, time-sensitive reaction, select fast model execution while keeping TTS independently controllable:

import { GenerationMode } from 'ai_licia-client';

await client.triggerGeneration(
  'Tyre puncture. Briefly tell the driver to pit safely.',
  {
    mode: GenerationMode.FAST,
    tts: true,
    ttl: 8,
  },
);

Fast mode retains the active persona, configured language, and compact stream environment. It excludes history, memory, knowledge-base content, goals, mission, vision, web search, and commands, and makes one model attempt without fallback. It does not change the accepted content you provide. TTS remains independently controllable.

STANDARD and FAST use independent per-key rate windows: roughly 25 seconds for STANDARD, and 1 second for FAST. ttl is optional and measured in seconds. It prevents a stale request from being processed after its deadline.

Use the default STANDARD mode when minimum latency is not required or the response needs any excluded context or capability.

Channel helpers

const characters = await client.listCharacters();
await client.setActiveCharacter(characters[0].id);

const join = await client.requestStreamJoin();
if (!join.success) {
  console.warn('Join failed', join.message);
}

await client.requestStreamLeave('backupChannel');