Dashboards

nginx.service, platform/sbl7/

/ -> website (3417), /wifi/ -> netwatch console (8799), /healthz. Installed by platform/sbl7/install.sh.

stretch-dashboard.service, dashboard/

The website, served through nginx at sbl7.local/. `npm run dev` because output:export needs the dev server for live API routes. User unit for curious (linger on). Moved from sbl6 2026-09-04.

netwatch.service, services/netwatch/

Wired observer on sbl7 since 2026-09-04 (iface null): AP, gateway and WAN hops probed separately. Source adopted into services/netwatch the same day; sbl7 still runs the ~/wifi_watch copy until deploy.py --apply is run there. run.env and *.creds are per-host and not in git.

netwatch-ui.service, services/netwatch/netwatch_ui.py

Team URL http://sbl7.local/wifi/ (nginx on sbl7 in front of 127.0.0.1:8799). stretch-lab.local:8799 and /wifi redirect there.

netwatch-unifi.service, services/netwatch/unifi_poll.py

Reads the UniFi controller classic API every 30 s with a read-only API key (unifi.creds, per host). Writes clients/radios/neighbours CSVs and the controller event feed into the netwatch run dir.

netwatch-rotate.timer, services/netwatch/rotate.sh

Mon 04:07: close the run, write its report, gzip CSVs, prune (KEEP_RUNS=52 then BUDGET_GB=15), start the next week. deploy.py installs the timer and its service together.

netwatch-db.service, services/netwatch/unifi_db.py

Every 30-60 s: all per-client / per-radio / per-SSID / gateway / path-monitor telemetry the controller exposes, the app's 5-min WAN and airtime series, events, known clients and netwatch's own probe rollups into ~/lab-var/netwatch/netwatch.db (SQLite, WAL). Read-only against the controller. Schema and tiers in services/netwatch/netdb.py.

netwatch-db-maintain.timer, services/netwatch/netdb.py

Hourly at :07: rolls raw samples up into forever-kept hourly tables, exports each completed month of raw rows to ~/lab-var/netwatch/archive/*.sqlite.gz, prunes raw oldest-day-first past NETWATCH_RAW_GB (15) or NETWATCH_RAW_DAYS (365). netdb.py stats shows growth per month.

nginx.service, platform/sbl6/

Index of the bench tools with a lamp per tool; /discharge/ -> 8793, /bms/ -> 8792, /motor-load/ -> 8794, /pimu-cycle/ -> 8795, /rig/ -> 8798. Installed by platform/sbl6/install.sh.

nginx.service + stretch-lab-mdns.service, registered stale, platform/portal/

Since 2026-09-04 every path 302s to http://sbl7.local/ so bookmarks land. Retire the site file and the stretch-lab-mdns publisher on 2026-10-04. /_portal/healthz still answers locally.

stretch-dashboard.service, registered on-demand, dashboard/

Same unit as sbl7's, run while developing on sbl6. Not the public site.

station-archiver.service, ~/.local/share/station-archiver (detached worktree of stretch4_battery_station)

Writes ~/stretch_station_logs/station.db. Runs from a detached worktree on purpose so branch switching in the dev clone cannot change the importer.

station-runwatch.service, ~/.local/bin/station_runwatch.py

Announces when an unattended station run ends. Candidate to adopt into services/.

scripts/stretch_lab_desktop.py, registered on-demand, services/pimu-bms-reader/

JSON API only (no page of its own); the dashboard BMS page reads it. Reached as sbl6.local/bms/api/bms/latest.

scripts/stretch_lab_desktop.py, registered on-demand, services/battery-discharge/

Reached as sbl6.local/discharge/ (GUI fetches are relative since 2026-09-04). Technician GUI was also seen on 8796 in earlier sessions; confirm which port the current build uses.

scripts/stretch_lab_desktop.py, registered on-demand, services/battery-cycle-motor-load/

API only; the page is the website's /motor-load, which fetches http://sbl6.local/motor-load by default.

scripts/stretch_lab_desktop.py, registered on-demand, dashboard/src/lib/pimu-cycle-api.ts

API only; the page is the website's /pimu-cycle, which fetches http://sbl6.local/pimu-cycle by default.

stretch_body_server.service, registered disabled, ~/repos/stretch4_body

Disabled 2026-09-04 after 7277 auto-restarts with no robot attached to sbl6. Re-enable (systemctl --user enable --now stretch_body_server) when a robot body is plugged into this machine.

stretch-ros2-zenoh-router.service, ~/ament_ws

No established clients on :7447 when checked 2026-09-04. Robots run their own ROS 2 stack; if nothing uses this router by 2026-09-18, disable it.

voice-stt-client.service, ~/.config/systemd/user/voice-stt-client.service

Backend is sbl7 (10.1.10.23): whisper-server.service (whisper.cpp CUDA) + stt-adapter.service on :9878, rebuilt 2026-09-03.

gnome-remote-desktop + rdp-guard.service, /etc/systemd/system/rdp-guard.service

rdp-guard restricts 3389 to loopback and the tailnet.

ideogram-portal.service, ~/ideogram-harness/portal/

Personal side project (history.too.foo). Not part of the Hello Robot portal; listed so the port is not reused. Weights are 26 GB and re-downloadable.

psf-console.service, registered disabled, services/power-state-fmea/scripts/station.py

Rig NUC (S4RIG_v3) is stretch-se4-4000 (10.1.10.7, decided 2026-09-04): the relay controller, sense board, cameras and Shellys hang off it, so the console runs there, not on sbl6. Disabled until the v3 rig is wired. SAFETY: TimeoutStopSec=90 so relays park on stop. Page still uses absolute /api/ paths; convert before putting it under /rig/.

no unit, registered external, ~/repos/stretch4_battery_station

Separate device on Hello Robot 2.4 wifi; not proxied (mDNS name, nginx cannot resolve it at start). Link only.

stretch-bridge.service, registered on-demand, services/stretch-bridge/

Runs on the robot NUC; the nuc-powerd and robot pages talk to it.

stretch-usb-power-logger.service / stretch-usb-power-viewer.service, registered on-demand, services/usb-power-monitor/

Recorder runs on robots; collected/ is a symlink to ~/lab-var/stretch_lab/usb-power-monitor/collected (5.2 GB after merging the Jetson copy 2026-09-03).