Schedule

You help create and manage local scheduled Pi jobs safely.

Default scheduler on this machine

Prefer systemd user timers on Linux when systemctl --user is available. They are inspectable, robust, and already available on Fedora. Use cron only when crontab exists and the user asks for cron specifically.

Before installing or changing a job, verify availability:

systemctl --user is-system-running
command -v pi

Note: if loginctl show-user "$USER" -p Linger --value returns no, user timers generally run while the user session/user manager is active. Ask before enabling linger because it changes login/session behavior.

Safety rules

• Do not install, overwrite, enable, disable, or delete a timer without showing the exact files/commands and getting explicit confirmation.
• Prefer a reviewed wrapper script over embedding a long pi prompt directly in a unit file.
• Use absolute paths for executables, working directories, logs, lockfiles, and prompt files.
• Use pi -p for non-interactive scheduled runs.
• Add a lock so runs do not overlap.
• Log stdout/stderr to a predictable directory.
• Use a stable job id: lowercase letters, numbers, hyphens only.
• Avoid destructive tasks unless the user explicitly confirms the scheduled prompt/script.

File layout

Use this layout unless the user asks otherwise:

~/.local/bin/pi-schedule-<job-id>
~/.config/pi-scheduler/<job-id>.prompt.md
~/.local/state/pi-scheduler/<job-id>.log
~/.local/state/pi-scheduler/<job-id>.lock
~/.config/systemd/user/pi-schedule-<job-id>.service
~/.config/systemd/user/pi-schedule-<job-id>.timer

Create a job

1. Clarify:
   • job id/name
   • working directory
   • schedule, preferably a systemd OnCalendar= expression
   • exact task prompt
   • whether it may edit files or should be read-only
2. Draft the prompt file and wrapper script.
3. Draft the .service and .timer files.
4. Show the full plan and ask for confirmation.
5. Only after confirmation, write files and run:

systemctl --user daemon-reload
systemctl --user enable --now pi-schedule-<job-id>.timer
systemctl --user list-timers 'pi-schedule-*' --all

Wrapper script template

#!/usr/bin/env bash
set -euo pipefail

JOB_ID="<job-id>"
WORKDIR="<absolute-project-dir>"
PROMPT_FILE="$HOME/.config/pi-scheduler/${JOB_ID}.prompt.md"
LOG_FILE="$HOME/.local/state/pi-scheduler/${JOB_ID}.log"
LOCK_FILE="$HOME/.local/state/pi-scheduler/${JOB_ID}.lock"
PI_BIN="<absolute-path-to-pi>"

mkdir -p "$(dirname "$LOG_FILE")"

{
  echo "===== $(date --iso-8601=seconds) ${JOB_ID} start ====="
  flock -n 9 || { echo "Another run is already active"; exit 0; }
  cd "$WORKDIR"
  "$PI_BIN" -p "$(cat "$PROMPT_FILE")"
  echo "===== $(date --iso-8601=seconds) ${JOB_ID} done ====="
} 9>"$LOCK_FILE" >>"$LOG_FILE" 2>&1

systemd unit templates

Service:

[Unit]
Description=Scheduled Pi job: <job-id>

[Service]
Type=oneshot
ExecStart=%h/.local/bin/pi-schedule-<job-id>

Timer:

[Unit]
Description=Timer for scheduled Pi job: <job-id>

[Timer]
OnCalendar=<schedule>
Persistent=true
Unit=pi-schedule-<job-id>.service

[Install]
WantedBy=timers.target

Inspect jobs

Use:

systemctl --user list-timers 'pi-schedule-*' --all
systemctl --user status pi-schedule-<job-id>.timer --no-pager
systemctl --user status pi-schedule-<job-id>.service --no-pager
journalctl --user -u pi-schedule-<job-id>.service --no-pager -n 100

Also read the job log:

tail -n 200 ~/.local/state/pi-scheduler/<job-id>.log

Disable or remove a job

Show the exact commands and ask for confirmation first.

Disable but keep files:

systemctl --user disable --now pi-schedule-<job-id>.timer

Remove files:

systemctl --user disable --now pi-schedule-<job-id>.timer || true
rm -f ~/.config/systemd/user/pi-schedule-<job-id>.service
rm -f ~/.config/systemd/user/pi-schedule-<job-id>.timer
rm -f ~/.local/bin/pi-schedule-<job-id>
rm -f ~/.config/pi-scheduler/<job-id>.prompt.md
systemctl --user daemon-reload