Daemon (argusd)

By default the file watcher and the alert detectors run inside argus start. Close it and ingestion stops. argusd moves that work into a long-running background process so your archive stays current whether or not a dashboard is open.

argusd does not serve the dashboard. Port 4242 is still bound only by argus start.

How the two fit together

When argusd is running, argus start notices its PID file and becomes a read-only viewer of the database the daemon keeps fresh. It logs argusd <pid> active — dashboard is read-only., the sidebar footer reads argusd daemon, and anything that would write, such as the indexing buttons in Settings, is refused with:

Argus is read-only because the argusd daemon is running. Manage indexing from the CLI (e.g. argus indexing enable) or stop the daemon with argus daemon stop.

When argusd is not running, argus start ingests in-process as usual. If the PID file is left over from a daemon that died, argus start logs Stale argusd PID file (process <pid> gone) — ignoring. and carries on.

Commands

argus daemon start    [--data-dir PATH]
argus daemon stop     [--data-dir PATH]
argus daemon restart  [--data-dir PATH]
argus daemon status   [--data-dir PATH]
argus daemon logs     [--data-dir PATH] [-n|--lines INTEGER] [-f|--follow]

start

Spawns a detached process and waits up to 3 seconds for it to write ~/.argus/argusd.pid. Prints argusd started, PID 12345. If a daemon is already running it prints daemon already running, PID 12345 and exits without error. On timeout it prints daemon did not start within timeout — check ~/.argus/argusd.log and exits with code 1.

The child is started as python -P -m argus.cli daemon run, not through the argus console script, so it is unaffected by the directory you launched it from.

stop

Prints argusd stopped. or argusd is not running. and removes the PID file either way, so it also cleans up a stale file.

restart

Stop, then start. Prints argusd restarted, PID 12345.

status

Always exits with code 0. One of:

argusd: not running.
argusd: not running (stale PID file for 12345).
argusd: running (PID 12345, uptime 3h07m21s).
Log: C:\Users\you\.argus\argusd.log

Uptime is measured from when the PID file was written, which is the moment the daemon registered itself.

logs

Tails ~/.argus/argusd.log, 40 lines by default. --follow polls every half second and survives log rotation, which a plain tail -f would not. Following starts at the end of the file. If there is no log yet it prints No log file yet at <path>.

What the daemon does

• Runs the same ingestion as argus start: a first pass over transcripts, then a filesystem watcher that reads only the new bytes of a changed file after a 100 ms quiet period.
• Runs the alert detectors once at startup, then every 10 minutes.
• Survives a missing ~/.claude/. It logs No agent data found (~/.claude/ is missing). Staying up and watching for it — ingest starts automatically once it appears. and checks every 30 seconds.
• Writes ~/.argus/argusd.log with full date stamps, rotating at 1,000,000 bytes with three backups.

Limits to know

• No auto-restart. A crashed daemon stays down until you start it again. The next start or stop cleans up the stale PID file.
• No OS-native autostart yet. The README lists an argus install command as a follow-up; it does not exist today. The hidden argus daemon run command is the blocking foreground form, meant for systemd, launchd or Windows Scheduled Tasks if you want to wire that up yourself.
• If argusd dies while a read-only dashboard is open, the footer keeps saying argusd daemon and ingestion stays paused until you restart the dashboard, which then resumes ingesting in-process.
• Idle cost is small. The README quotes roughly 40 to 70 MB of RAM and no idle CPU, since the watcher is event-driven and the scheduler wakes every 10 minutes.

Verified against argus-code 0.5.2.