status

name	status
description	Run the health monitor and report results. Auto-run when: session-context shows health is degraded or failed, after infrastructure changes (docker, systemd), when a service appears unreachable, or user asks about system health. Invoke proactively without asking.

Run the health monitor and report results. Unset LITELLM_API_KEY so the check falls through to pass (avoids stale shell env overriding the correct key):

cd ~/projects/hapax-council && env -u LITELLM_API_KEY uv run python -m agents.health_monitor 2>&1 | grep -v -e "LITELLM_API_KEY is not set" -e "warnings.warn("

Report the score, then list FAILED and DEGRADED items grouped by severity.

Triage

If --fix can handle it, suggest: uv run python -m agents.health_monitor --fix --yes

For items --fix cannot resolve, investigate using the patterns below.

auth.litellm FAILED (HTTP 0 or 401)

The most common cause is an env var overriding pass with a bad value. Check in order:

1. Fish universal variables — grep LITELLM ~/.config/fish/fish_variables. If the value is missing the sk- prefix, remove it: fish -c "set -Ue LITELLM_API_KEY"
2. direnv not loaded — direnv status in hapax-council. If .envrc is not loaded, the env may have a stale value from session init.
3. pass store — pass show litellm/master-key should return sk-... (67 chars). If not, the key itself needs updating.
4. LiteLLM container — curl -s -o /dev/null -w "%{http_code}" http://localhost:4000/health should return 200 (even without auth).

Stale sync agents (gdrive, gmail, youtube >24h)

Restart the stale services:

systemctl --user start gdrive-sync.service gmail-sync.service youtube-sync.service

If gdrive-sync fails with Invalid Value on pageToken, the sync state is corrupted. Reset the token in ~/.cache/gdrive-sync/state.json — set start_page_token to empty string, then restart the service.

profile-update.service failed

systemctl --user reset-failed profile-update.service && systemctl --user start profile-update.service

Check journal if it keeps failing: journalctl --user -eu profile-update.service --no-pager -n 30

axiom.ef_automated — missing timers

The check compares agent manifest systemd_unit values against active timers. If an agent's timer exists but under a different name, fix the manifest in agents/manifests/<agent>.yaml to match the actual unit name. Verify with:

systemctl --user list-timers --no-pager | grep <keyword>

connectivity.gdrive-sync stale

This follows from stale gdrive sync — fix the sync agent (above) and the connectivity warning clears.

systemd unit failed (timer will retry)

systemctl --user reset-failed <service> && systemctl --user start <service>

If the service has a MemoryMax drop-in and is OOM-killed, check journalctl --user -eu <service> for oom-kill entries.