🎉 MyHOME v2.0.0b14 Pre-Release: built on your bus traces

[GreenGrassBlueOcean]

🎉 MyHOME v2.0.0b14 is out

MyHOME v2.0.0b14 is merged and ready for field testing. It is a big one: 105 pull requests since b13, on top of the new OWNd 2.0.0b9. Release notes and myhome.zip: v2.0.0b14.

Before the list of changes, the most important part of this post.

---

❤️ Thank you for the traces

When we opened the call for real bus traces in #466, we hoped for a few captures to fill the gaps in the test corpus. What we got was dozens of sweeps and traces from gateways we do not own, from people who took the time to press buttons, run a sweep, export the file and describe what they did.

Every one of those traces went into the test suite. When we replayed them, they caught bugs that nobody had run into yet. Several of the fixes below were made before anyone saw the problem at home:

• @gdluck (MyHomeServer1, 36 heating zones): replaying your capture showed that temperature probe 169 was routed to heating zone 69 and could create a phantom zone (Climate: a temperature probe (WHERE >= 100) is routed to zone WHERE-100 and can discover a phantom zone #549 → fix(climate): a temperature probe is not zone WHERE-100 #550). It also showed that a setpoint change echoed as dimension 12 mode 3 left the target temperature stale (climate: trailing '3' on WHO 4 dimension 12 is read as protection/off, so _target_temperature goes stale #454 → fix(climate): a dimension 12 mode 3 of a running zone is its new setpoint (#454) #552), and that 11 zones that never answer cost about 70 s of startup polling (perf(climate): stop polling heating zones that never answer #555, test: pin the #553 debug log where the gateway NACKs the dead climate zones #557). None of these had been reported.
• @manfredgittmaier-afk (MH200N): your alarm probe showed that an MH200/MH200N without an alarm central unit still answers *#5*0## with 8 zone frames. Without it, those plants would have got phantom alarm entities (test(466): add MH200N burglar alarm, diagnostic and timed turn-on trace (#466) #488 → fix(alarm): ignore phantom zone telemetry and purge orphaned partitions on MH200/MH200N (#466, #488) #490). Your F500N report fixed the tuner reports (fix(media_player): parse real OWNSoundEvent dimensions and handle malformed RDS frames #497), and your clock observation fixed myhome.sync_time on the MH200N (MH200N: myhome.sync_time sets date and timezone but not the time (DIMENSION 22 vs 0) #520 → fix(services): send DIMENSION 0 alongside DIMENSION 22 during sync_time (#520) #522).
• @anotherjulien (F454 + MH202): the first real lock (WHO 14) and F520 energy frames, which gave us F520 discovery in the bus sweep (test(466): add F454 sweep and MH202 traces for lock, energy, dimmer and covers (#466) #491, feat(sensor): proactive bus sweep and discovery for F520 energy meters (Dimension 1200 / 113) (#494) #506). You also compared the same F418U2 dimmer through two gateways, which settled how dimension 1 and 4 behave (test(golden): add BTicino F418U2 modern dimmer samples on F454 and MH200 (#466, #501) #502, test(golden): add BTicino F418U2 modern dimmer samples on MH202 (#466, #501) #505), and your two-gateway plant was one of the test beds for multi-gateway support (feat(gateway): multiple gateway and shared bus support (#453) #459).
• @nicolacavallo84 and @xtimmy86x: we were about to drop the 4890 touch-screen family as "not a gateway". Your H4890 traces and wiring diagram proved us wrong, and the family is back with real traces (feat(gateway): reinstate 4890 family with authentic H4890 traces (#466) #484). The MyHomeServer1 + H4890 plant was the other multi-gateway test bed (feat(gateway): multiple gateway and shared bus support (#453) #459).
• @TheDarkWizard (MyHomeServer1): your KW8011 captures taught us how CEN+ wire addresses (21, 22, …) map to scenario objects (test(traces): ingest physical MyHomeServer1 CEN+, cover, and diagnostic golden samples (#466) #512 → feat(cenplus): support wire WHERE address resolution and KW8011 lifecycle (#466) #515). They also gave us the first WHO 1001 diagnostics in the corpus, and your "1-second delay scenario" workaround for Living Now controls.
• @f18m (MyHomeServer1 + LN-4660M2): your centralized shutter captures showed that these buttons send advanced automation frames to the general address. They are now device triggers you can automate on (test(golden): add MyHomeServer1 and LN-4660M2 centralized cover samples (#445, #466) #499 → feat(device_trigger): add centralized shutter controls to gateway device triggers (#445, #466) #503).
• @lyubomirtraykov (F461): your sweep confirmed the F461's WHO 1013 identity on real hardware, so it is identified correctly instead of by guesswork (test(466): add BTicino F461 gateway trace closing WHO 13/1013 diagnostics gap #526, test(466): confirm F461 WHO1013 identity 134 on physical hardware #527). Your fan-coil report fixed the fan modes ([BUG] Climat Zones do not support Fan OFF #410 → fix(climate): omit unsupported 'off' from fancoil fan modes (#410) #412).
• @caiosweet (MH200N + MH201): the first full MH200N sweep in the corpus, the MH201 trace and old scenario-module frames (test(466): add MH200N bus sweep trace replay and golden fixtures (#466) #469, test(traces): add issue #466 traces and golden samples #477). Your duplicate-probe report fixed a probe that came back as a second entity (Discovered temperature probe 105 is recreated and polled with *#4*105*15##, causing retry warnings #441 → fix(sensor): one unique id per temperature probe, however it was created (#441) #452).
• @wave68runner: CEN short presses from an MH200N and a MyHomeServer1 on one bus, the first WHO 15 frames for the MyHomeServer1 (test: add classic CEN (WHO 15) short-press traces from #541 #548). Your latency report led to the command-session cap (high latency #425 → fix(gateway): cap command workers at the gateway profile's session limit #450).
• @fedem95: your MH201 CEN+ hold trace showed one 2-second hold firing long_press five times (CEN / CEN+ buttons don't show up in the new automation UI #418 → fix(cen+): long_press once per hold, repeats as long_press_repeat #439).
• @marcob79, @xduss, @Interstellar0verdrive and @pakob81: general covers that never stopped (Mh201 and General cover button behaviour #433 → fix(cover): a general / area / group cover follows its shutters (#433) #436), a pump read as a heating zone (Pump actuator incorrectly interpreted as zone actuator #431 → fix(climate): a pump status (WHERE 0#N) is not zone N (#431) #432), the cover calibration field test (docs(covers): the remaining points from the b13 field test #417), and YAML names lost on secondary buses (Since updating MyHOME to 0.9.4 #408 → fix(yaml): normalize single-digit MAC octets and harden secondary bus key indexing (#408) #414).

The Hardware Trace Availability Matrix now has at least one trace for 8 gateway models (F454, F461, 4890 family, MH200, MH200N, MH201, MH202, MyHomeServer1). A bug in code that a trace covers fails CI before it can reach a release. The matrix is on the docs site and in #466, and CI keeps it up to date.

Thank you, all of you. Most of this release came out of your traces. 🙏

---

📖 Documentation on GitHub Pages

Everything is at openwebnet-ha.github.io/MyHOME/dev:

• Versioned docs: /dev/ follows the development branch and matches b14 right now; /beta/ and a frozen per-beta snapshot (/2.0.0b14/) will follow shortly. The v0.9.4 docs stay at /0.9.4/.
• Beta differences: what changed between the betas, side by side.
• Trace availability: the matrix above, generated from the test fixtures.
• Checked against the code in CI: the gateway profile table, the options reference and the supported-domains table are generated from the code, and the docs build fails if a page no longer matches the code.
• New and updated pages: multiple gateways, sound system and multi-room, Lovelace recipes (speaker tiles, active-speakers card), cover travel-time calibration, and repair issues. Notes like [!NOTE] and [!WARNING] now show as boxes on the site instead of as raw text.

---

🌟 Most important new features and fixes

🔀 Multiple gateways and shared buses (#459, #525, #531)

• Gateways on separate buses, several gateways sharing one SCS bus, and a warm-standby gateway that takes over while the primary is offline.
• When you add a second gateway, the config flow now asks for the bus topology first, so it no longer creates duplicates of every device the first gateway already has.

🔊 Sound system (WHO 16) and Music Assistant

• Source selection on the F441/F441M matrix works (fix(media_player): make F441M source selection work — routing address used the wrong digits #422). The routing address used the wrong digits; switching over IP is clean.
• Multi-room grouping with Music Assistant and Spotify Connect (feat(media_player): multi-room audio grouping and Music Assistant integration #460). Groups survive a restart (fix(audio): keep the decoder books across restarts and close the #532 audit leftovers #535). The group is kept when the anti-hiss timer switches its rooms off (fix(audio): keep the group when the anti-hiss timer switches its rooms off (live-tested) #538). Handover when the leading room leaves or is switched off (feat(media_player): add group handover, 100% pre-gain, and F441 multiroom trace #521, fix(media_player): grace period before a member's group-leave OFF #528, fix(media_player): turning off a group's leading room no longer silences the whole group #554). Joining a room adds it to the group instead of replacing the members (fix(media_player): async_join_players is additive, not a snapshot #530). Parked groups wake faster (perf(media_player): wake a parked group faster #544).
• Tuner entity for F500/F500N sources: station, frequency and RDS text (feat(media_player): tuner entity for WHO=16 sound sources (F500) #427, fix(media_player): parse real OWNSoundEvent dimensions and handle malformed RDS frames #497).
• Streaming companion discovery: a streamer whose own integration cannot play a URL (e.g. Cambridge Audio) is matched automatically to its DLNA or Cast endpoint (feat(media_player): automatic companion decoder discovery & transparent DLNA streaming bridge #487, feat(media_player): add cast platform and host/name companion resolution #519).
• Amplifiers are discovered at startup with the correct status request (fix(gateway): discover WHO 16 with the dimension-5 status request *#16*0*5## #438).

🌡️ Climate (WHO 4)

• Home+Control / MyHomeServer1 zone state (dimension 7): mode and setpoint follow what the plant reports (feat(climate): follow WHO 4 dimension 7 zone state on Home+Control plants (#429) #457).
• Setpoint changes from the plant update the target temperature (fix(climate): a dimension 12 mode 3 of a running zone is its new setpoint (#454) #552). Probes and pumps are no longer mistaken for zones (fix(climate): a pump status (WHERE 0#N) is not zone N (#431) #432, fix(climate): a temperature probe is not zone WHERE-100 #550), and one WHERE grammar is shared by climate and sensors, with a CI guard (refactor: one WHO 4 WHERE grammar and a CI guard against raw WHERE reads #556).
• Zones that never answer are skipped at startup after two silent restarts and asked again after a week. A Repairs issue tells you which zone (perf(climate): stop polling heating zones that never answer #555).

🪟 Covers

• General, area and group covers follow their shutters instead of waiting for a stop that never comes (fix(cover): a general / area / group cover follows its shutters (#433) #436).
• Centralized shutter buttons (LN-4660M2 and similar) as device triggers (feat(device_trigger): add centralized shutter controls to gateway device triggers (#445, #466) #503).

🔘 CEN / CEN+

• A CEN+ hold fires long_press once; the repeats fire the new long_press_repeat trigger (fix(cen+): long_press once per hold, repeats as long_press_repeat #439).
• CEN+ wire addresses resolve to their scenario object (KW8011) (feat(cenplus): support wire WHERE address resolution and KW8011 lifecycle (#466) #515). CEN devices no longer offer triggers they can never fire (fix: cover level 255, per-family CEN triggers, corrected CEN golden frames #430).

💡 Lights

• Adaptive Lighting compatibility without flooding the bus (feat(light): adaptive lighting compatibility and bus traffic optimization (#461) #462).
• An unknown lighting WHAT (the MH200 "WHAT 19" fault) no longer switches a light on or off, and is shown as an unknown_state attribute (fix(light): keep the last state when a light reports an unknown WHAT #456, fix(light): an unknown WHAT does not keep a state restored at startup #458, fix(gateway): skip lighting scope events when is_on is None #465).

🔌 Gateway and reliability

• Fixed: a general "all lights off" (*1*0*0##) stopped the event listener and made every entity unavailable until a reload (fix(gateway): a general "all lights off" (*1*0*0##) killed the event listener — entities unavailable until reload #455).
• Gateway identity from WHO 1013 diagnostics ends the false "identity mismatch" on modern gateways, and the device page shows the model data (Fix false gateway identity mismatch (#370) by using WHO=1013 diagnostics #420, Show the gateway's own model data in HA (device page, diagnostics) #423).
• Command workers capped at what the gateway accepts (MH200N) (fix(gateway): cap command workers at the gateway profile's session limit #450). The event session is closed on unload (fix(gateway): close the event session when the listener is cancelled #444). sync_time sets the clock on the MH200N (fix(services): send DIMENSION 0 alongside DIMENSION 22 during sync_time (#520) #522).
• The options flow was changed ahead of Home Assistant 2026.12, which would otherwise have broken the integration (fix(config_flow): migrate to OptionsFlowWithReload and eliminate entry update listener (#510) #514).
• F520 energy meters are found by the bus sweep (feat(sensor): proactive bus sweep and discovery for F520 energy meters (Dimension 1200 / 113) (#494) #506). The sweep now also queries alarm and sound (feat(sweep): add WHO 5 alarm and WHO 16 sound to bus sweep with MH200 golden samples #470). The bus card backfills and exports its whole buffer (fix(bus-monitor): backfill the card's whole buffer and export the full ring #449). Phantom alarm zones are gone on MH200/MH200N (fix(alarm): ignore phantom zone telemetry and purge orphaned partitions on MH200/MH200N (#466, #488) #490).

🌍 Translations

• Translations now go through Crowdin (chore(translations): set up Crowdin localization pipeline and anti-drift tooling #463). If you would like to help with Italian, French, Dutch or another language, let us know.

---

🛡️ What we improved in CI

• 2,834 tests (b13: about 1,720), with 100% coverage kept as a gate.
• Trace replay grew with every contribution: more trace fixtures and 240+ golden frames, all parsed and replayed on every PR. The trace matrix is regenerated from the fixtures and synced to Community Call for Real-World Gateway Traces: Help Us Eliminate Hardware Blind Spots! 📡 #466 automatically (docs: add dynamic hardware trace availability matrix #464, feat(ci): auto-sync hardware trace matrix to issue #466 and normalize 4890 family naming #486, fix(ci): restore issue #466 sync and write permissions for hardware trace matrix #517).
• The strict-typing gate can fail now. A shell pipe was hiding the ratchet's exit code, so the job stayed green with 11 real mypy --strict errors. The gate is fixed and the errors are cleared (fix(typing): make the strict-typing gate fail, and clear the 11 errors behind it #560).
• New guards in the standards check: no update listeners or direct reloads, which break on HA 2026.12 (ci: forbid update listeners and direct flow reloads (#510) #543); every file compiles and from __future__ stays in the right place (test(standards): enforce valid Python syntax and strict __future__ positioning #518); no raw WHERE reads outside reviewed code (refactor: one WHO 4 WHERE grammar and a CI guard against raw WHERE reads #556).
• Docs drift check in CI and pre-commit (feat(docs): native anti-drift sentinel and proactive PR documentation synchronization #495, feat(docs): automate per-beta documentation parity, options sentinels, and delta tracking matrix #507, feat(ci): dynamically calibrate supported entity domains table in README #483).
• Cross-repo testing with OWNd: MyHOME runs against the pinned, latest and development OWNd, and OWNd runs the MyHOME suite, so a library change cannot quietly break the integration (test(profiles): accept a separate OWNd MH200 profile with WHO 16 #440, test(ownd): accept one or two sends for a NACKed status request #451, test(profiles): adapt MH200N tests to support WHO 16 forward-compatibility (OWNd#63) #473).
• Least-privilege workflow permissions for every workflow, which clears the CodeQL alerts (ci(security): add explicit least-privilege workflow permissions #500).
• Flaky and noisy tests fixed: event-bus races, unmocked sockets, unawaited coroutines (test(config_flow): prevent unmocked socket connects on reauth and options flow reload #413, test: fix unawaited coroutine warning in test_platforms #478, test(config_flow): prevent unmocked socket connect on options flow reload #481, test(reauth,options): mock async_reload in reauth and options flows to eliminate socket warnings #516, fix(tests): eliminate dispatcher/event-bus race in CEN failover tests #529, fix(tests): wait for the event bus in the delegation filtering test #546).
• Under the hood, with no change in behaviour: __init__, cover, gateway sessions, light and the media player were split into smaller modules, so the code is easier to review and test (refactor(init): decompose __init__.py into legacy_yaml and migrate sibling modules (Milestone 1) #467, refactor(cover): encapsulate calibration hub, motion math, and isolate multi-gateway state (Milestone 2) #468, refactor: decompose init, cover, and gateway session/event runners (Milestones 1–3) #472, refactor(light): extract SoftwareFadeEngine and DaliFeatureLock subsystems (Milestone 4) #485, refactor(media_player): split the audio zone entity into layered modules #559).

---

⚠️ Behaviour changes to check after updating

• CEN+ hold automations: long_press now fires once per hold. If you relied on it repeating, use long_press_repeat.
• Sound source selection now really switches the matrix (it was a no-op before). Volume 0 is no longer a mute.
• Fan coils no longer offer an off fan mode; the thermostats do not support it.
• Heating zones that never answer are skipped at startup (see Repairs).

---

🧪 How to test

Update to v2.0.0b14 via HACS (enable Show beta versions). Home Assistant 2026.3 or newer is required.

We would especially like feedback on:

• Multiple gateways / shared buses: does the topology step make sense for your setup?
• Multi-room audio with Music Assistant on F441/F441M plants.
• Climate on MyHomeServer1 / Home+Control plants.
• Centralized cover buttons and CEN+ triggers in automations.

And please keep the traces coming in #466. A sweep takes a minute from the bus card (Export Trace) or with myhome.sweep_bus, and it goes straight into the test suite. The gaps still open are in the matrix. 📡

Found a bug? Open an issue and link this discussion.

Thank you all for helping build the next MyHOME! 🚀

[manfredgittmaier-afk]

b14 feedback – MH200N (model 44) + F441/F500N audio, F411 covers, F428 dry contact

Upgraded from the v2-phase1-architecture branch to 2.0.0b14 (OWNd 2.0.0b9). Everything I use works:

Volume down on the amplifiers now works (the 161000*… issue is gone) 🎉
F500N tuner: presets, next/previous, seek, frequency, RDS title
Source switching on the F441 matrix
sync_time now also sets the time on the MH200N (DIMENSION 0 fix): clock is within ~3 s
Dry contact motion detector (WHO 25, inverted: true), turn_on_timed, general "all lights off" event
Phantom alarm partitions are gone

One small issue: button.mh200n_gateway_calibrate_all_covers is permanently unavailable, although this gateway has many discovered F411 covers (each has its own working "Calibrate travel time" button). It looks like the "no covers on this gateway" check from d2e2fa3 doesn't find the covers here. No functional impact – I've disabled the entity for now.

Side note for the calibration docs: my F411 actuators report the stop only after their 60 s run-time limit (last_run_seconds: 60.34), so auto-calibration can't measure them. Setting the travel times manually with set_cover_travel_time works fine.

Thanks for the great work!

[GreenGrassBlueOcean]

@Interstellar0verdrive tagging you hare as subject matter expert regarding the cover issue. I also opened a new issue #565

[Interstellar0verdrive]

Answered on #565, with what I found in button.py and what I measured on my own installation today.

[TheDarkWizard]

@GreenGrassBlueOcean

I have two quick questions: the side panel is still not available—is this add-on still dedicated to you developers?
I assume the submenu to calibrate the covers is only intended for servers that do not "receive" the opening/closing percentage from the actuators, is that correct?

P.s
The credits on the V2-architecure page should be updated :P

[Interstellar0verdrive]

@TheDarkWizard on your second question, since the calibration is the part I am working on: yes. The calibration only exists for covers whose position is estimated from their travel times, which is why the calibration buttons are created "per timed cover" in button.py. An actuator that reports its own position does not need it, and the calibration view lists only the time-based covers. The side panel and the credits are for @GreenGrassBlueOcean to answer.

[GreenGrassBlueOcean]

@TheDarkWizard

Regarding the side panel: @xtimmy86x is doing the side panel, while I am doing the backend.

UX design is the hardest part of software design. It’s like a first impression—you only get one chance to make a perfect impression.

As a backend engineer, we can make all the mistakes in the world and fix them with a new PR. So essentially we have a really good team: @Interstellar0verdrive knowing all those nitty-gritty details about the deepest thoughts of Italian electronic cover designers, @anotherjulien as the OpenWebNet documentation oracle, @xtimmy86x crafting the frontend, and myself harnessing myself between traces and continuous integration.

So yes, I am confident that @xtimmy86x will give us a side panel during beta testing. But see this as a beautiful woman preparing for a gala—don't disturb her! 😉

P.S. Noted on the credits on the V2 architecture page, we'll make sure they get updated!

[TheDarkWizard]

Thanks for the update on the role division! I'll take note so I know who to tag next time!

So I don't even have to ask how long it "she" takes to get ready?