awrise

Wake something on a schedule, let it do one thing, and put it back to sleep.

A simple scheduler for one-off jobs. Store schedules in a JSON file (~/.aither/awrise/jobs.json). Runs jobs that are due using run-due. Idempotent: running it twice in the same window won't execute the same job twice.

Installation

pip install -e .

Usage

Add a job

awrise add --name backup --every 1h --run "rsync -a src/ dest/"
awrise add --name check --every 15m --run "curl http://example.com/health"
awrise add --name deploy --every 2h --run "./deploy.sh"

Intervals: 15m (minutes), 2h (hours), 1d (days).

List jobs

awrise list

Output:

Name                 Interval     Command                              Last Run             Status
----                 --------     -------                              --------             ------
backup               1h           rsync -a src/ dest/                  never                pending
check                15m          curl http://example.com/health       never                pending
deploy               2h           ./deploy.sh                          never                pending

Run due jobs

awrise run-due

This is idempotent. Run it from a cron job, a systemd timer, or anywhere. It executes all jobs whose last_run timestamp is more than interval ago. Records the execution time so running it twice in the same window won't run the same job twice.

Jobs that fail record the failure and do NOT block other jobs.

awrise run-due --quiet  # Suppress output

Remove a job

awrise remove --name backup

Configuration

Store schedules in ~/.aither/awrise/jobs.json by default. Override with:

export AWRISE_HOME=/custom/path

Self-test

awrise --self-test

Verifies:

• Interval parsing (valid and invalid formats)
• Idempotency (running twice in same window doesn't execute twice)
• Empty command rejection
• Job list/add/remove

Design

The simplest useful scheduler: no daemon, no background process. Store state in JSON. The caller (cron, systemd, your own loop) decides when to call run-due.

• Idempotent: Last-run timestamp prevents duplicate executions
• Resilient: Failed jobs don't block other jobs
• Honest: Refuses unparseable intervals, empty commands, nonexistent jobs