Adds cycles-until-due and last-cleaned sensors decoded from the same
options[] array the course selector already reads (DrumCleanProposal_N -
WashingTimes_N for cycles remaining, DrumCleanLog_<iso> for last-cleaned),
verified byte-for-byte against a live app screenshot ("Potreba cistenia po
37 cykloch" / "Naposledy cistene pred 10 dnami").
Course selection is now a writable select entity sourced from each
device's own x.com.samsung.da.editCourseList (via a new options-as-callable
form on SelectDesc, reading the coordinator's full resource snapshot
instead of just the entity's own href), rather than a hardcoded course
table baked into Python. A MostUsed_ field on the same resource was
considered as a fallback but rejected after byte-level analysis showed it
doesn't reliably encode a course list. When a device never populates
editCourseList, the selector isn't created at all (exists_fn now takes
(rep, resources) to check a sibling href, matching the existing
match_fn(rep, resources) pattern).
Display names moved out of Python into strings.json/translations under
entity.select.{washer,dishwasher}_cycle.state.*, matching this
integration's existing translation_key convention (fridge.py's ice_type,
flex_zone_mode, etc.) instead of hardcoding English names in code.
discover() flagged /cycleinterface/vs/0, /drlc/0, and /operational/state/0
as unbound on real washer dumps, which would surface a spurious device
coverage gap for washer owners. All three are noise: cycleinterface is
empty on every dump seen, drlc/0 is an OCF-native duplicate of the
already-ignored /drlc/vs/0, and operational/state/0 is a read-only
OCF-native duplicate of /operational/state/vs/0 (already modeled by the
richer, write-capable operational.OPERATIONAL_STATE).
_normalize() lowercased every select's options/state for display, but
that's only needed for entities with a translation_key (whose
strings.json lookup requires lowercase keys, per hassfest). Untranslated
selects (Cycle, Smart Dry, Sound mode, LED brightness) had no lookup to
protect and were just getting mangled -- "AI Wash" became "ai wash",
"ExtraHigh" became "extrahigh", etc.
Replace with _display(), which only lowercases for translation_key
entities and otherwise passes the device's own casing through, with two
cosmetic fixups: a fully lowercase wire value (e.g. "voice") is
title-cased, and a PascalCase value (e.g. "ExtraHigh") gets a space at
the case boundary. Already human-friendly values pass through
untouched. Avoids hand-authoring strings.json translations for
open-ended, per-model option lists (e.g. dishwasher cycle names) that
would silently regress on any value we didn't enumerate.
x.com.samsung.da.delayStartTime is HH:MM:SS until the cycle starts
(e.g. "01:00" means 1 hour from pressing start), not a time of day. It
was wired up as a time entity with a suspicious hour % 24 wrap -- a
sign it was never really a clock time. Replace with a number entity
(0-24h, 1h steps) that reads/writes the field as elapsed hours.
Home Assistant 2026.3+ serves brand icons/logos for custom integrations
from a local brand/ directory before falling back to the CDN (see
developers.home-assistant.io/blog/2026/02/24/brands-proxy-api). Copied
Samsung's existing icon/logo assets from home-assistant/brands'
core_brands/samsung so the integration shows proper branding instead of
a placeholder. No manifest.json change needed -- has_branding is
auto-detected from the directory's presence.
Exposes whether a device is currently in observe (push) or poll mode,
for troubleshooting the observe-mode feature without digging through
logs. Disabled by default since it's not everyday-use information.
Replace the breadth/silence downgrade heuristic with a simpler rule: a
still-live OBSERVE session is never torn down on a sweep/cache mismatch
(the 30s sweep already corrects the cache regardless of mode) -- only a
proven reconnect invalidates subscriptions. A mismatch instead triggers
extra hot/warm subpolls this cycle as a bounded fallback.
Also stop treating a summary-poll block-level ACK timeout as session
death: distinguish TimeoutError (transfer was progressing, session
likely alive) from ConnectionError (session actually closed), backed by
a consecutive-timeout counter so a genuinely dead channel still
recovers. This was causing a flaky/slow device (e.g. a dishwasher) to
flap observe<->poll every ~45s on nothing but a slow blockwise GET.
Also remove the dead is_active/active_when scaffolding (never wired
into the coordinator) and add DEBUG logging for each OBSERVE notify
received, including its href.
A poll-failure-triggered reconnect while in observe mode left OBSERVE
silently dead: the new session had zero subscriptions, mode stayed
'observe' so the hot/warm sub-poll fallback stayed disabled, and the
ObserveRefreshTask kept retrying against the closed session instead of
the live one. Downgrading to poll mode on reconnect hands recovery to
the existing poll-mode retry path, which re-subscribes on the new
session.
Also snapshot ObserveManager._notified before intersecting it against
the subscribed set, closing a rare set-mutated-during-iteration race
between the DTLS reader thread (on_notification) and the executor
thread computing the grace-period success fraction.
Replace the ad-hoc option_names dicts (hardcoded English, manual
reverse-mapping) with the idiomatic HA approach: normalize raw Samsung
enum values to lowercase at the select/sensor boundary and let
strings.json/translations/en.json supply the display text, keyed by
the normalized value. select.py maps the chosen (lowercased) option
back to the device's original casing before writing. The ice-making
status sensor now also declares device_class=enum with an options
list, as HA expects for enum sensor state translations.
Also drop the hardcoded _BZONE_MODES list in favor of discovering
beverage zone modes from the device's own roomSupportedModes field,
matching how flex_zone_mode and ice_type already work.
hassfest rejects translation keys containing uppercase letters, but the
raw Samsung enum values for flex zone, beverage zone, ice type, and ice
making status are uppercase (e.g. CV_TTYPE_RF9000A_FREEZE). Follow the
existing dryer-course pattern instead: map raw device values to
human-readable names in Python via a new SelectDesc.option_names field
(mirrored read/write in select.py) and a value_fn for the sensor, and
drop the now-unused translation_key entries from strings.json/en.json.
- Add hacs.json and MIT LICENSE required for HACS/default-store inclusion
- Fill required manifest.json keys (documentation, issue_tracker,
codeowners) and reorder per hassfest's key-ordering rule
- Add validate.yml workflow running hassfest and HACS validation
- Update README install note now that hacs.json is checked in
Maps the fridge/dishwasher hrefs flagged by last session's coverage-gap
Repairs issue: defrost delay (switch) + block status (diagnostic binary
sensor), master ice-maker enable, self-check trigger/status, dishwasher
diagnosis trigger/status, and last-operation-source. Adds fallback
capabilities for /doors, /temperatures, and /icemaker/status that only
bind when the richer per-instance hrefs they duplicate are absent, so
simpler devices without those hrefs still get the data. Everything else
in the original gap list moves to capabilities.ignored (Bixby audio
feedback, inert DR/energy-planner resources, redundant metadata).
Fixes discover() reporting a known href as an unregistered gap whenever
its match_fn declined to bind for that device (e.g. a filter capability
on hardware without that filter) — it's now only a gap when no capability
is registered for the href at all.
Device naming now uses x.com.samsung.da.modelNum (not the OCF /oic/p,d
metadata, which produced inconsistent names like "[dishwasher] Samsung")
to build "Samsung <Type> (<model>)".
Completes the device-capability-diagnostics spec:
- diagnostics.py: the standard HA diagnostics hook, returning device type,
one_ui_version, unbound hrefs, and the redacted raw resource tree, plus
integration and smartthings-local version numbers. This is what the
Repairs issue (added in the previous commit) points users at, and what
they attach to a device-support issue.
- config_flow.py: _probe_and_validate now also reports oneUiVersion and
whether the device type is recognized. Recognized types are unaffected;
an unrecognized type shows a new confirm_unknown_type step explaining
that only common capabilities will be available before creating the
entry, so expectations are set at setup time rather than only after the
fact via Repairs.
- .github/ISSUE_TEMPLATE/device-support.yml: structured template for
filing a capability gap, linked from both the Repairs issue and the
config-flow confirmation step.
- README: short section pointing at this whole mechanism.
Companion to the git-filter-repo pass that just stripped every prior
version of these two files (which carried a real Samsung account email,
Bixby access token/device-ID hash, and real WiFi/BLE MAC addresses) out of
history entirely. This commit reintroduces only the already-redacted
content, so the real values are no longer reachable from any commit.
Implements the first half of the device-capability-diagnostics spec:
- registry/capabilities/ignored.py: known-noise hrefs (Bixby/voice
provisioning, WiFi/BLE info, OTA/region housekeeping, and a couple of
redundant hrefs) declared as no-entity Capability objects, so
discover()'s existing unknown-resource reporting treats them as covered
instead of flagging every device as having gaps. Folded into each
by_type registry and into the global fallback CAPABILITIES set.
- discovery.py: log() callback now passes the raw href instead of a
formatted message, so callers can collect a clean unbound-hrefs list.
- coordinator.py: wires that callback into self._unbound_hrefs, tracks
device_type_name/one_ui_version, and raises (or clears) a Repairs issue
when a device's type is unrecognized or it has genuinely unmodeled
hrefs left over.
- registry/redact.py: recursive, substring-keyed redaction for anything
that looks like account/identity data, ahead of the diagnostics.py
platform that will consume it.
Verified against the real dishwasher/refrigerator fixtures: known-noise
hrefs no longer show up as gaps, while genuine gaps (e.g. /bespoke/vs/0
on the fridge) still do.
The DTLS/CoAP transport code that made "ocf" an accurate name moved out to
the smartthings-local package. What's left here (capability.py, entities.py,
discovery.py, adapter.py, identity.py, capabilities/, by_type/, plus the
/device/0 batch parser) is entirely the device capability registry, so name
the package for what it does.
Flattened the redundant ocf/registry/ nesting into a single top-level
registry/ package and updated every import across the platform modules and
test suite accordingly. Verified: full test suite (80/80) passes, and the
Docker dev container reconnects to both live appliances and rediscovers
their entities cleanly after the rename.
requirements.txt (cbor2/paho-mqtt/pyOpenSSL for the old standalone bridge),
.env.example (APPLIANCE_*/MQTT_*/deploy.sh env vars), and deploy.sh (ssh+tar
deploy for the old bridge container) all referenced main.py and
samsung_appliance/, which no longer exist. Nothing in the current HA
integration, its tests, or the README depends on them; config now happens
entirely through the HA config flow, and manifest.json/requirements-dev.txt
own the Python dependencies.
Also dropped the .gitignore exception for ocf_root_ca.pem (that vendored
file was deleted along with the rest of the vendored transport) and the
now-inapplicable "Bridge runtime" log ignores.
The dev container was relying on HA's runtime pip-install of manifest.json
requirements, which only fires when the integration is set up and needs
outbound network access at that exact moment; a container that had been
running since before the smartthings-local migration kept the old code
loaded in memory and never went through that install path, so restarting
it failed once it picked up the new manifest.json.
Repurpose the stale MQTT-bridge-era Dockerfile (its own code was already
deleted from this repo) to build on the official HA image with
smartthings-local pre-installed, and point docker-compose.yml at it via
`build: .`. Verified end-to-end: rebuilt the image, recreated the
container, and confirmed both live appliances (fridge, dishwasher)
reconnect and discover entities with no runtime install needed.
Replaced all em-dashes with plain punctuation, broke up the mechanical
bold-lead-in bullet list in "What you get" into prose paragraphs, and
removed the "X, not Y" contrastive framing and decorative arrows.
Part 2 pointed readers to run smartthings-local's setup_cert.py directly;
reword it as an example of how to obtain the CA cert/key rather than a
step to execute, since this repo makes no claim about how that script
behaves or is maintained upstream.
Getting the AC14K_M CA cert+key is a protocol-layer concern, not an
HA-integration concern, and the two copies here had already drifted from
each other. Drop root setup_cert.py + requirements-bootstrap.txt and have
Part 2 explain why the CA cert is needed, then link to the smartthings-local
project's setup_cert.py as the canonical way to obtain it.
Now that mbillow/localthings#8's protocol fixes are merged upstream and
smartthings-local 0.1.0 is on PyPI, drop the vendored ocf/coap_dtls.py +
ocf/observe_refresh.py transport (dead code, never instantiated) and the
duplicated ocf_root_ca.pem in favor of the real package. manifest.json and
requirements-dev.txt now pin smartthings-local>=0.1.0 instead of the
unmerged git branch.
README rewritten from scratch — it still described the old standalone
MQTT-bridge (samsung_appliance/, main.py, docker-compose bridge) that this
repo moved off of; it now documents the real architecture: a native HA
custom component with config-flow-driven cert minting and a per-device-type
capability registry, with the protocol layer split out to smartthings-local.
Dishwasher:
- Revert remote_control to BinarySensorDesc ("Smart Control", read-only)
- Add cycle SelectDesc on /course/vs/0 via options-array RMW (AI Wash,
Pre blast, Self clean, Normal, Express 60, Heavy, Pots and pans,
Delicate, Plastic, Baby Care)
- Add Storm Wash+ and Auto release dry SwitchDescs on /course/vs/0
- Add Sanitize SwitchDesc and Smart Dry SelectDesc on /dishwasher/vs/0
- Add Sound volume NumberDesc (0–15 step 5) on /settings/sound/volume/vs/0
- Add Door LED night brightness, night start/end time to DOOR_LED capability
- Add SOUND_VOLUME capability to laundry module
Stale-state fixes (firmware leaves values set after cycle ends):
- finish_time: suppress when machine state is not active
- progress: return 'Idle' when machine state is not active
- water filter: gate WATER_FILTER capability on filterStatus != 'notused'
Translations:
- Add ICESTATUS_RUN → "Making ice" to strings.json and en.json
Tests:
- Update dishwasher and refrigerator golden fixtures to reflect current entity set
- Remove dead test_project_extrapolates_remaining_time (on_observation removed)
Child lock and remote control promoted from read-only binary sensors to
writable switches. Delay start time promoted from a raw string sensor to
a writable TimeEntity.
On persistent poll failure (both initial attempt and reconnect), fall back
to the last known resource values instead of raising UpdateFailed. This
keeps entities available with stale data rather than flooding history with
Unavailable gaps whenever the device drops a /device/0 block fetch.
UpdateFailed is still raised on the very first poll before any data exists.
Removes two redundant sensors (a raw H:MM:SS string and a duration integer)
in favour of a single SensorDeviceClass.TIMESTAMP that shows the absolute
estimated finish time. The HA frontend auto-renders this as "in X minutes"
and keeps it current without re-polling.
Also removes dead _project / _on_observation / _rem_minutes code from both
operational.py and oven.py — these hooks were declared on the Capability but
never called anywhere in the coordinator or adapter.
Two fixes for the alarm_code diagnostic sensor:
1. _is_included now treats an empty rep ({}) from /device/0 as a stub
meaning the resource exists but hasn't been individually fetched yet.
Previously the field-presence check against {} returned False, leaving
the entity unregistered and showing Unavailable from the entity registry.
2. Replace _last_alarm_code (silently dropped all but the final alarm) with
_active_alarm_codes which joins all codes in the items list. Shows 'none'
when no alarms are active.
Samsung appliances silently drop CoAP requests when hit faster than their
firmware ceiling (~8–14 req/s measured; dishwasher unknown). Multi-block
Block2 fetches on LAN hit ~100 req/s instantaneously, causing the observed
burst timeouts on block 4 and block 10 of /device/0.
Adds rate_limit_rps (default 5 req/s / 200 ms) to DtlsCoapSession with a
pace() method that enforces the inter-block delay via _stop.wait() so session
teardown interrupts the sleep cleanly. Coordinator calls pace() between hrefs
in _poll_hrefs_blocking. Also adds device IP to all GET log lines for easier
per-device filtering.
_is_included now checks that an entity's field exists in the resource
rep before registering it, preventing phantom entities when optional
fields are absent on a shared resource. Explicit exists_fn still takes
priority for cases requiring custom logic. All platforms use the shared
helper; select.py's inline duplicate removed.
Marc Billow