Overview
The bot uses node-cron to run the deal pipeline on a fixed schedule. Each time the cron fires, it fetches deals, runs both filter layers, and broadcasts the result to all subscribed users.
The schedule is controlled by the CRON_SCHEDULE environment variable and always runs in the America/Bogota (UTC-5) timezone.
Expressions follow the standard five-field cron syntax:
┌───────────── minute (0–59)
│ ┌─────────── hour (0–23)
│ │ ┌───────── day of month (1–31)
│ │ │ ┌─────── month (1–12)
│ │ │ │ ┌───── day of week (0–7, 0 and 7 = Sunday)
│ │ │ │ │
* * * * *
CRON_SCHEDULE=0 9 * * * # Every day at 9:00 AM Bogotá time
The cron expression is validated at startup. If CRON_SCHEDULE contains an invalid expression, the bot will throw an error and exit before connecting to Telegram.
Timezone
The scheduler is hardcoded to run in the America/Bogota timezone (UTC-5, no daylight saving time). This is set via the timezone option passed directly to node-cron in src/scheduler/cronJobs.ts:
cron.schedule(config.cron.schedule, async () => {
// ...
}, { timezone: 'America/Bogota' });
Changing the timezone requires modifying src/scheduler/cronJobs.ts directly. The CRON_SCHEDULE environment variable only controls the schedule expression — not the timezone. There is no environment variable for timezone.
Schedule examples
| Expression | Description |
|---|
0 9 * * * | Every day at 9:00 AM Bogotá time (default) |
0 18 * * * | Every day at 6:00 PM Bogotá time |
0 9 * * 1-5 | Weekdays only at 9:00 AM |
0 9,18 * * * | Twice daily at 9:00 AM and 6:00 PM |
Scheduler source
The full startScheduler function in src/scheduler/cronJobs.ts:
export function startScheduler(): void {
console.log(`⏰ Cron activado con schedule: "${config.cron.schedule}"`);
cron.schedule(config.cron.schedule, async () => {
console.log(`[${new Date().toISOString()}] 🔄 Ejecutando búsqueda de ofertas...`);
try {
const result = await fetchAndMarkDeals();
if (result.status === 'ai_error') {
// No enviar broadcast — un mensaje "no hay ofertas" sería engañoso.
// El error ya fue logueado en runPipeline(); aquí solo registramos el ciclo fallido.
console.error(`❌ Cron: broadcast omitido por fallo de IA. Razón: ${result.reason}`);
return;
}
if (result.status === 'no_deals' || result.deals.length === 0) {
// Sin ofertas que superen los filtros hoy — silencio correcto, no es un error.
console.log('📭 Cron: sin ofertas destacadas hoy. No se envía broadcast.');
return;
}
const copRate = await getExchangeRate();
const message = formatDealsMessage(result.deals, copRate);
await notifyAllUsers(message);
console.log(`✅ Broadcast enviado: ${result.deals.length} ofertas.`);
} catch (err) {
const msg = (err as Error).message ?? 'sin mensaje';
console.error(`❌ Error inesperado en cron: ${msg}`);
}
}, { timezone: 'America/Bogota' });
}
Error handling during cron runs
The scheduler handles two non-fatal outcomes silently without sending a message to users:
| Outcome | result.status | Behavior |
|---|
| AI curation failed | ai_error | Logs the error and skips the broadcast. Sending “no deals today” would be misleading — the pipeline didn’t finish. |
| No deals passed the filters | no_deals or empty array | Logs a notice and skips the broadcast. This is a correct silent outcome, not an error. |
try/catch and logged to the console without crashing the process. The scheduler remains active and will retry on the next scheduled run.