- Test box: restart + start-up capture + online InfluxDB check (data newer
than the restart), with a status strip; checklist and findings are the
main view, live log / guided troubleshooting / box details are tabs,
rarely used actions in a More menu
- light palette on all systems (macOS dark mode made the colours unreadable)
- new rule influx.invalid_value: -inf from a silent microphone makes
InfluxDB reject every upload (HTTP 400); guide end names it
- exclusive serial port on macOS/Linux: two tool windows no longer split
the log between them
- InfluxDB check: `since` filter, neutral wording when nothing new arrives
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
New guided check "Sensor data in the database (InfluxDB)", also the last
step of the Quick check on the green path. It queries InfluxDB (read
token) for the newest point per measurement of the box and tells apart:
sensor data arrives / status arrives without sensor data / offline now /
never seen / check not possible. Polls every 30 s for up to 4 minutes.
With a stopped RTC in the log it concludes that sensor readings are lost
(stamped 1 Jan 2000): box A4:CB:8F:C9:47:3C sent only device_status and
microphone.
- service/influx.py: query, CSV parsing, classification (no new deps)
- guides: new step type "influx" (host answers via influx_result)
- Settings > InfluxDB access (or KOTA_INFLUX_TOKEN); token never logged
- service status mapping for tech.influx.* findings
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
On macOS/Linux the OS asserts DTR and RTS when the port is opened and
pyserial then writes DTR before RTS, passing through RTS=1/DTR=0 - the
USB-Serial-JTAG reset state. Every Connect restarted the box. Drop RTS
before open and DTR right after it on posix; Windows is unchanged.
Verified on macOS 27 / arm64 with box A4:CB:8F:C9:47:3C (fw 1.0.16,
build 3ae273ec): connect keeps the box running, Reset & capture still
restarts it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
