Fix scheduling issues on Zeabur production deployment

[Copilot]

Problem

When deployed on Zeabur at 18:18 (6:18 PM), the bot would execute a few times and then stop working. The bot is configured to run every 1.5 hours from 6 AM to 11 PM (Buenos Aires time), but was failing to maintain this schedule in production.

Looking at the issue data, the last update was at 2025-10-17T08:16:31 even though the bot should have continued running throughout the day. The scheduling system was breaking down after just a few execution cycles.

Root Cause

The sleep calculation logic used integer hour arithmetic that didn't account for minutes and seconds:

# OLD CODE (BUGGY)
if hora_actual < 6:
    horas_hasta_inicio = 6 - hora_actual
else: 
    horas_hasta_inicio = (24 - hora_actual) + 6
time.sleep(horas_hasta_inicio * 3600)

This caused time drift that accumulated over multiple cycles. For example:

• Bot runs at 23:30 and needs to sleep until 6 AM
• Old calculation: (24-23)+6 = 7 hours → sleeps until 6:30 AM ❌
• Correct: Should sleep until exactly 6:00 AM ✅

After multiple cycles, these small errors compounded, causing the bot to drift outside its intended 6 AM - 11 PM operating window and eventually stop executing.

Additionally, the code had a hardcoded 6 instead of using the configurable horario_analisis_inicio, and mixed time.localtime() with time.time() which could cause timezone issues on UTC servers like Zeabur.

Solution

Replaced integer-based calculations with precise datetime timestamp arithmetic:

# NEW CODE (FIXED)
ahora = datetime.now(timezone_local)

if ahora.hour < horario_analisis_inicio:
    # Same day - calculate until start hour
    proxima_ejecucion = ahora.replace(hour=horario_analisis_inicio, minute=0, second=0, microsecond=0)
else: 
    # Next day - calculate until start hour tomorrow
    proxima_ejecucion = (ahora + timedelta(days=1)).replace(hour=horario_analisis_inicio, minute=0, second=0, microsecond=0)

segundos_hasta_inicio = (proxima_ejecucion - ahora).total_seconds()
if segundos_hasta_inicio > 0:
    time.sleep(segundos_hasta_inicio)

Changes

1. Precise time calculations: Uses full datetime objects with timedelta instead of integer hours
2. Configuration compliance: Replaced hardcoded 6 with horario_analisis_inicio from config
3. Timezone consistency: Uses timezone-aware datetime objects throughout (Buenos Aires UTC-3)
4. Better logging: Shows exact timestamps (HH:MM:SS format) instead of just hours
5. Defensive programming: Added check to prevent negative sleep times in edge cases

Testing

Comprehensive testing confirms the fix:

• ✅ 36-hour simulation: Bot executes every 1.5 hours during window, sleeps until exactly 6:00 AM
• ✅ Issue scenario: Deployment at 18:18 → runs at 18:18, 19:48, 21:18, 22:48, then sleeps until 06:00 next day
• ✅ Execution frequency: 11.3 per day (matches expected rate)
• ✅ Timezone handling: UTC-3 preserved correctly across all operations
• ✅ No time drift: All executions remain within 6 AM - 11 PM window

Expected Behavior

After deployment, the bot will:

• Execute every 1.5 hours precisely during 6 AM - 11 PM Buenos Aires time
• Sleep until exactly 6:00 AM (not drifted times)
• Continue running indefinitely without stopping
• Work correctly on UTC servers (Zeabur) thanks to timezone-aware datetime handling

The logs will now show precise timestamps for better monitoring:

Nos encontramos dentro del horario de analisis - Hora actual 18:18:00
Esperando hasta la próxima ejecución (2025-10-17 19:48:00)...

Or when outside the window:

Fuera del horario de análisis. Durmiendo hasta 2025-10-18 06:00:00 (5.70 horas)

Fixes the scheduling issue described where the bot would stop after running a few times.

Original prompt

---

This section details on the original issue you should resolve

<issue_title>scheduling issues on production (zeabur)</issue_title>
<issue_description>When deployed on Zeabur at 18:18 (6:18 PM), the bot would execute a few times and then stop working. The bot is configured to run from 6 AM to 11 PM (23:00) with a 1.5-hour interval between executions.
The bot configuration in config.py is marked as an integer, I think zeabur does not understand it cause locally that configuration works. Of you see subte_alerta.py the timestamps defined to mark the executions (1.5 hours) of difference works perfectly.
I think you can translate the horario_analisis_inicio and horario_analisis_fin into the same languague timestaps works. I attached files for you to see how is working the system(not good)

error: por algun motivo se deguto la ejecución:"{
"ultima_actualizacion": "2025-10-17T08:16:31.817681-03:00",
"estados_actuales": {
"Linea A": "Estación Río de Janeiro cerrada por obras de renovación integral.",
"Linea B": "Carlos Gardel y Uruguay cerradas por obras. Viernes y Sábado con horario extendido.",
"Linea C": "Normal",
"Linea D": "Plaza Italia y Agüero cerradas por obras de renovación integral.",
"Linea E": "Normal",
"Linea H": "Hoy horario extendido desde Caseros hasta 1:30 am.",
"Linea Premetro": "Normal"
},
"historial": {
"Linea A_obra": {
"estado": "Estación Río de Janeiro cerrada por obras de renovación integral",
"linea_original": "Linea A",
"tipo": "obra",
"contador": 12,
"primera_deteccion": "2025-10-16T09:45:30.272689-03:00",
"ultima_notificacion": "2025-10-16T09:45:30.272773-03:00",
"es_obra_programada": true,
"detectada_por_texto": true,
"activa": true,
"ya_notificada": true
},
"Linea B_obra": {
"estado": "Carlos Gardel y Uruguay cerradas por obras",
"linea_original": "Linea B",
"tipo": "obra",
"contador": 2,
"primera_deteccion": "2025-10-17T06:46:26.190151-03:00",
"ultima_notificacion": "2025-10-17T08:16:31.817671-03:00",
"es_obra_programada": true,
"detectada_por_texto": true,
"activa": true,
"ya_notificada": true
},
"Linea D_obra": {
"estado": "Plaza Italia y Agüero cerradas por obras de renovación integral",
"linea_original": "Linea D",
"tipo": "obra",
"contador": 8,
"primera_deteccion": "2025-10-16T15:45:54.437706-03:00",
"ultima_notificacion": "2025-10-16T15:45:54.437725-03:00",
"es_obra_programada": true,
"detectada_por_texto": true,
"activa": true,
"ya_notificada": true
},
"Linea B_problema": {
"estado": "Viernes y Sábado con horario extendido",
"linea_original": "Linea B",
"tipo": "problema",
"contador": 2,
"primera_deteccion": "2025-10-17T06:46:26.190163-03:00",
"ultima_notificacion": "2025-10-17T08:16:31.817671-03:00",
"es_obra_programada": false,
"detectada_por_texto": false,
"activa": true,
"ya_notificada": true
},
"Linea H_problema": {
"estado": "Hoy horario extendido desde Caseros hasta 1:30 am",
"linea_original": "Linea H",
"tipo": "problema",
"contador": 2,
"primera_deteccion": "2025-10-17T06:46:26.190205-03:00",
"ultima_notificacion": "2025-10-17T08:16:31.817671-03:00",
"es_obra_programada": false,
"detectada_por_texto": false,
"activa": true,
"ya_notificada": true
}
}
}</issue_description>

Comments on the Issue (you are @copilot in this section)

Fixes #8

---

💡 You can make Copilot smarter by setting up custom instructions, customizing its development environment and configuring Model Context Protocol (MCP) servers. Learn more Copilot coding agent tips in the docs.

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Preview	Comments	Updated (UTC)
bot-subte	Ready	Preview	Comment	Oct 18, 2025 1:44am

💡 Enable Vercel Agent with $100 free credit for automated AI reviews

[agmonetti]

buena idea