v0.9.0

Added

• runwisp import cron / runwisp import supervisord. Convert an existing crontab or supervisord config into an annotated runwisp.toml, with inline # TODOs for anything that needs a human. See Migrating from cron and Migrating from supervisord.
• depends_on on [services.*]. Gate a service’s boot on other services becoming healthy, with reverse-order teardown on shutdown — boot ordering only, never a workflow DAG, and it never deadlocks (a dependency that won’t come up starts the dependent anyway with a warning). See Boot order with depends_on.
• include on [daemon]. Glob extra TOML files (e.g. conf.d/*.toml) into the config so tasks can be split across files; collections accumulate, names stay unique across files, and singleton tables stay in the root. See Splitting config across files.
• runwisp reload (and SIGHUP) for explicit, validated config reload. Pick up runwisp.toml task add/change/remove edits in a running daemon without a restart; validate-first, so a bad edit or a restart-only change is rejected and the live set is untouched. See Reload.
• jitter on [tasks.*] / [defaults]. Pace cron tasks that share a fire time through a daemon-wide one-at-a-time gate so they take turns instead of stampeding — each runs as soon as the gate frees and slips up to its window only under contention. See jitter.
• Missed-run alerts. A scheduled run skipped because the daemon was down is detected on restart, recorded as a browsable missed run, and raised as a failure-level alert — silence it per task or globally with notify_on_missed. See Missed ticks.
• working_dir and shell on [tasks.*] / [services.*]. Run a command in a specific directory and under a chosen interpreter (e.g. /bin/bash); shell also takes a [defaults] value. See Working directory & shell.
• exit_codes on [tasks.*] / [services.*] / [defaults]. List the exit codes treated as success (default [0]) so a non-zero “nothing to do” code doesn’t trip retries or alerts. See Retries & timeout.
• stop_signal on [tasks.*] / [services.*] / [defaults]. Choose the signal that opens the stop ladder before SIGKILL (default SIGTERM); SIGKILL skips the grace window. See Retries & timeout.
• priority and autostart on [services.*]. priority fixes the boot start order (lowest first, deterministic — not a dependency); autostart = false defines a service that boots stopped until you start it from the UI/API. See Startup.
• umask on [tasks.*] / [services.*]. Set the octal file-creation mask for a run (e.g. "027"); applied per-process so concurrent runs never affect each other. See Working directory & shell.
• run_on_start on [tasks.*]. Fire a task once at daemon boot, on top of any cron — the @reboot equivalent, tagged triggered_by = startup. See Scheduling.
• user on [tasks.*] / [services.*]. Run as another OS user (user or user:group, name or numeric id) when the daemon runs as root. See Working directory & shell.
• start_retries on [services.*] / [defaults], plus FATAL services. A service that keeps fast-failing (exits below healthy_after) now tolerates start_retries failed starts in a row before going FATAL — it stops restarting, records a start_failed run, and rings the bell instead of flapping forever. See When a service can’t start.
• Six-field cron with second precision. cron now accepts an optional leading seconds field, so */30 * * * * * fires every 30 seconds aligned to the wall clock. Existing 5-field specs are unchanged. See Scheduling.

Changed

• backoff_reset_after renamed to healthy_after, which now governs both the restart-backoff reset and the failed-start/FATAL threshold. See Healthy threshold.

Fixed

• Spurious “managed by systemd” quit prompt. A daemon spawned by runwisp / runwisp demo from a systemd-scoped terminal no longer inherits INVOCATION_ID and falsely self-reports as service-managed, so the TUI’s quit dialog offers “Shut Down” again.