A lightweight in-memory Python job scheduler that loads JSON job definitions from a watched directory, runs tasks on cron or one-time schedules with concurrency, and updates the schedule when job files are added, modified, or deleted - no database, no restart.

To clone the repo, set up a virtual environment, install dependencies, start the scheduler (python main.py), and run the test suite, follow RUNNING.md.

Modern applications need to run background tasks on a predefined schedule. These tasks range from sending daily email digests and generating reports to performing routine data cleanup. A reliable and dynamic job scheduler is a critical component for automating these workflows.

Design and implement a lightweight, in-memory job scheduling service. The service will read job definitions from a configuration directory, execute the defined tasks according to their specified schedules, and log the outcomes.

The scheduler must monitor a directory (e.g., /etc/scheduler/jobs.d/) for job definition files in JSON format. It must react to file system changes without requiring a service restart:

The service must support two types of time-based schedules:

• One-Time — a specific future timestamp in ISO 8601 format (e.g., 2025-10-26T15:00:00Z)
• Recurring — a cron string (e.g., */5 * * * * for every 5 minutes)

The system must be designed to support different kinds of tasks. For this implementation, only one task type is required:

• execute_command — runs a shell command

However, the design must make it straightforward to add new task types in the future (e.g., http_request) without altering the core scheduling logic.

• The scheduler must be able to run multiple jobs concurrently if their schedules overlap.
• Each job execution must produce a simple log entry recording:
  • Unique job ID
  • Execution timestamp
  • Status: SUCCESS or FAILURE

The scheduler does not need a persistent database. All scheduling and job state is managed in memory and can be rebuilt from the configuration files upon startup.

Job definitions are JSON files placed in the watched directory.

{
  "job_id": "report-001",
  "description": "Generate the daily sales report.",
  "schedule": "0 1 * * *",
  "task": {
    "type": "execute_command",
    "command": "/usr/bin/python /opt/scripts/generate_sales_report.py"
  }
}

{
  "job_id": "cleanup-xyz",
  "description": "Perform a one-time cleanup of temporary files.",
  "schedule": "2025-09-20T02:00:00Z",
  "task": {
    "type": "execute_command",
    "command": "rm -rf /tmp/stale-data/*"
  }
}