Skip to content

Scheduled Tasks in Backend API

In our backend API, we rely on scheduled tasks (commonly referred to as cron jobs) to automate recurring operations. These tasks are executed by Celery, a distributed task queue, and are often scheduled via periodic triggers (e.g. every minute, hour, or day) to keep the system clean, updated, and responsive to user interactions.

Below is a list of the key cron jobs, along with brief descriptions of what each one does.

List of Scheduled Tasks

recompute_computed_fields (Marketplace)

Purpose: Recomputes marketplace listing data (like available weight or pricing) when the listing hasn't been updated today. Ensures fresh computed values for buyers and sellers.

Frequency: Daily at 1 AM (only processes stale or uncomputed records)

expire_unpaid_orders

Purpose: Identifies orders in PAYMENT_PENDING status older than 1 hour and marks them as PAYMENT_EXPIRED. It also triggers recomputation of the affected order data, releasing the locked inventory (weight) in the market.

Frequency: Every minute

recompute_computed_fields (Storage)

Purpose: Recomputes crate- and produce-level computed fields if they have not been updated since the beginning of the current day.

Frequency: Daily at 1 AM (only processes stale or uncomputed records)

recompute_digital_twin

Purpose: Recalculates the digital twin shelf-life prediction for crates under active monitoring. Skips crates that were already updated in the past 5.5 hours. Sends an email alert in case of processing failures.

Frequency: Every 12 hours

update_temperature

Purpose: Fetches temperature data from third-party sensors (e.g., Ubibot, Figorr, Victron, Ecozen) for cold rooms and stores the updates.

Frequency: Every 4 hours

time_to_pick_up_notifications

Purpose: Sends notifications (and SMS for non-smartphone users) when crates in storage have less than 2 days of shelf life remaining. Also removes the notifications if crates are no longer checked in.

Frequency: Every 12 hours

market_survey_checks

Purpose: - Marks outdated MarketsurveyPreprocessing entries as inactive - Creates new entries for recent checkouts - Sends notifications to farmers and operators - Deletes market survey notifications older than 14 days

Frequency: Weekly on Mondays at 1 AM

prediction_calls

Purpose: Calls the ML prediction API for supported commodities in India. Stores 14-day forecast results in the database for each market-commodity combination.

Frequency: Daily at 3:30 AM

prediction_calls_ng

Purpose: Calls the Nigerian prediction API for selected commodities across supported states. Stores 8-week forecasts and flags interpolated data if real data is unavailable.

Frequency: Daily at 2:30 AM

Timezone Notes

Important: Celery is currently not configured with an explicit timezone (CELERY_TIMEZONE is not set). As a result, all scheduled tasks run according to the local system timezone of the environment where the Celery worker and beat are running.

This means task timing (e.g., “Daily at 1 AM”) may vary depending on the server’s configured timezone. If you’re deploying across multiple environments or regions, be aware that scheduled times could shift unexpectedly.

How to Set a Timezone (Optional Configuration)

To enforce a consistent timezone for all scheduled tasks, you can configure the following in your Celery settings:

CELERY_ENABLE_UTC = True
CELERY_TIMEZONE = "UTC"  # Or your preferred timezone, e.g. "Africa/Lagos"

  • With CELERY_ENABLE_UTC = True, Celery uses UTC internally but interprets crontab schedules in the timezone you specify.

  • You can use any valid timezone string supported by Python’s pytz library.