DIGITAL STEALTH CARE · CHANGELOG

Digital Stealth Care, release by release.

Every release and feature, dated, newest first. The bugs, the incidents told in full and the lessons behind them are one filter away, in the state they are really in.

● bug states shown as they are, never rounded up

● releases, features and fixes on one timeline

How fixes are checked

By using the software. A fix is checked in daily use on the real hardware; if it doesn’t hold, the bug gets a new report rather than a quiet edit, so a relapse stays on the record. Every row is shown in the state it is in.

Changelog · every release and feature, dated · fixes one filter away

Releases, features and milestones show first. Switch on bugs, incidents and lessons with the filters to see the whole record: bugs are shown in the state they are in, and nothing here is rounded up.

loading entries

Below the timeline: the newest 60 entries as a plain list. Every entry and write-up also has a page of its own.

CHANGELOG · THE NEWEST 60, AS A PLAIN LIST

Newest first, for when the timeline above is not running. Use the timeline’s search and filters for everything else.

18 Sep 2026 · Milestone · CannaLib, DSC, Site · Released

Journey and log become one timeline

Write-ups are now journey entries. One vertical timeline, newest first, with a column per product; /log/ opens it filtered to write-ups.

DIGITAL STEALTH CARE · CHANGELOG

Digital Stealth Care, release by release.

CHANGELOG · THE NEWEST 60, AS A PLAIN LIST

Journey and log become one timeline

75 held entries published, details withheld

55 bench logs from the last ten days

Copy pass; old DSC help retired

/dsc/ rebuilt as a product page

/dsc/supported/: the equipment directory

Firmware page listed bugs instead of releases

User guide re-verified for 9.0.0

Release manifest and library hosting

DSC alpha sign-up; releases hosted here

The DSC user guide lives on the site

Version stamps follow the test kit

Pre-existing pytest teardown flakiness (paho-mqtt loop_stop)

‘Update available’ false positive: nags to update 8.0.0 to v8.0.0-AlphaPi (same version)

Settings is reachable only through an unlabelled 16px gear tucked behind the status chips — operator could not find it

Hub grow-mat root-zone vote reads past its arrays: probes[2]/vote[2] looped to i < 4 (undefined behaviour, phantom probe can drive the mat)

Four brain tests fail on master independent of any branch: test_alert_correctness (humidifier/heater ineffective) and test_stale_hub_alerts (live hub still rais

Settings > System transparency cards: failover state, brain routes, developer switches, setup profile, resets (S4)

Nutrient and medium catalog rows carry no dose / EC / pH data — feed targets cannot come from CannaLib yet

DashHomeSections grow-stage chip hardcodes tone=”ok”

Settings > System: OS/journal log tab surfaces raw “[Errno 2] … ‘journalctl'” python error

StatusChip tone=teal is not a valid tone – two type errors in MediaStorageCard

Fleet-heal RF Status lambda: dead first snprintf truncates 4 hex chars into a 4-byte buffer (GCC -Wformat-truncation on every hub build)

Deploy hot-synced SPA static (bundle index-B2R-r53j.js) is not in the dsc-hub-brain:8.3.0 image – the next container recreate reverts the span-picker popover fi

Extension: turn the reading-based ‘probe dark’ timer into an actual alert (probe dark > N h) – Runtz Punch probe has been DARK 24h+ on a live D62 flowering plan

3D twin had no marker for placed climate sensors — the Zigbee/Tuya hygrometers behind the airflow deltas were invisible in the scene

Snapshot chips omit missing values but don’t explain why

/setup/state stuck at phase ‘usb_flash’ with debt ‘not_flashed:control’ on a fully-online fleet

SF1000 / Twin brightness history is recorded but unreadable through /history

Mobile: desks and help tips fall off phone screens

Live plants’ flip date is a calendar stand-in (day 49) — enter the real flip on the plant panel after deploy

Light catch-up is opaque: no warning that releasing manual hold turns the lamp ON, no way to cancel a catch-up, and no stated next scheduled light change

Version statements disagree: README says surface 8.2.0, RELEASE.md says 8.3.0 expects firmware 8.1.0.0, paths.py EXPECTED_FIRMWARE is 8.2.0.0

Settings -> General tab is static docs, not configurable settings

GET /history?entity_id= (empty) returned 422 instead of an honest empty series

lights_on_only never suppresses a capture — it reads a hub key that is not in hub values

Retired pot3/pot4 came back as offline seats after the one-shot removal gate; in-service dark probes vanished from /fleet

Progressive disclosure for collapsed system rows

Fan demand scale is ONE global multiplier applied to four physically different fans

DSC v8.3.0: Ingest orchestrator

Root page “thereabouts” idle-probe reading has no independent staleness check

flash-fleet-remote.sh and bake-firmware.sh still use docker exec dsc-hub-esphome (container removed by deploy)

master does not typecheck — MediaStorageCard uses a badge tone that does not exist

Inspector ‘Limited history’ chip ignores the brain’s recorded:false – a series that has never been written is labelled as thin instead of ‘Not recorded by the b

Explicit “show/hide system rows” toggle on journal views

No lamp role for a plug — adopting the 4×8 light plug has nothing to bind to

Possible leftover “pots” language in CropScheduler copy (rendered on Climate)

Settings > General is prose; subtitle promises ‘kit language and operator notes’ that do not exist

The brain test suite reaches out to the live hub at 10.42.0.10

Probe N/P/K and *_raw history is unreachable: recorder writes pot metric ‘nitrogen’ / ‘moisture_pct_raw’, /history resolves ‘soil_nitrogen’ / ‘soil_moisture_raw

Formalize the snapshot-backfill admin endpoint, or retire it

Operator-chosen tent pressure: a slider for the target regime (positive / neutral / negative) and how strong the pressure difference should be, per tent or syst

Pot-moisture “dry” threshold line is a bare hardcoded constant with no live setting to bind to

History audit: 7-day charts collapsed to the newest ~25 h (list_history LIMIT 2000) — /history now buckets the whole window

An imported-but-unconfigured Tuya device reads OFFLINE, which looks like a network fault

Zigbee canopy and Tuya tank readings are never written to fleet_history by their lanes – they only reach history via the 60 s computed recorder (change or 5 min

Surface the live SPA bundle hash/version in Settings

VPD charts with history that read against correct VPD ranges (sleek): reference zones, time-in-zone, temperature × RH VPD map

The journey opens

Trend-chart threshold bands are inconsistently drawn – band prop only recolors the line, doesn’t show a visible band