Quick Start¶
This page shows the smallest working ChatLinux fleet-status flow: initialize a track config, refresh the Ansible-backed cache once, then read the cached status quickly from a shell.
Flow Entry Points¶
-
Install and Verify
Install the package and confirm
chatlinux --versionpluschatlinux fleet --helpwork. -
Initialize Internal State
Write
fleet.jsonand preparecache/plusruntime/. By default all three live under~/.chatarch/chatlinux/. -
Refresh Remote Status
Use Ansible to generate inventory/probe files, collect common read-only metrics, and write the local cache.
-
Read Cache / Safety Boundary
show/statusread cache by default; remote sampling is explicit throughrefreshorstatus --refresh.
Install and Verify¶
From a development checkout:
Initialize the .cube Track¶
The default state directory lives inside ChatArch home: CHATLINUX_HOME when set, otherwise $CHATARCH_HOME/chatlinux, and finally ~/.chatarch/chatlinux/. Write the sample config first:
The JSON output includes the full state_paths object, making every local state path auditable under the same ChatArch state root. This creates or uses:
runtime/ stores the generated fleet-inventory.ini and fleet_probe.py for refresh runs. These files still live inside ChatArch home and are not written to /etc, /usr, system service directories, or the task repository.
For task-local or CI state, override the home directory:
Refresh Status¶
refresh reads the track config, renders an Ansible inventory and a read-only probe script, runs Ansible, and writes a normalized cache.
Machine-readable output:
The current sample track defaults to:
and uses the remote Python validated during the .cube audit:
Read the Last Cache¶
Reading cached status does not contact hosts, so it is fast for shell checks:
Refresh first and then display:
Current Metrics¶
The MVP collects only common, easy, read-only metrics:
- hostname, sample time, uptime, CPU count, and load average;
- memory total/available and swap total/used;
- filesystem capacity from
df -P -T; - GPU count/model/memory/utilization when
nvidia-smiis available; - failed systemd unit names from
systemctl --failed.
Safety Boundary¶
- No sudo.
- No cleanup, restart, or remote repair.
show/statusreads cache by default; onlyrefreshorstatus --refreshcontacts hosts.- SMART/NVMe lifetime data is not collected yet; it needs an explicit fleet helper or read-only sudo authorization.