Daikin Cycle ML

Home Assistant integration that turns your ESPAltherma sensor into a complete heat pump intelligence platform — cycle detection, per-mode COP, seasonal SPF, weather-normalized degradation, runtime + energy accounting, installer-grade curve advice, and 19 ML-powered alerts.

Local-only. No cloud. No accounts. Everything runs inside your Home Assistant box.

IoT class: calculated · Integration type: hub · 50 sensors · 15 binary sensors · 10 services · 19 alerts

Showcase

Simple Card	Heating Curve	Daily Summary

One-glance overview	LWT vs outdoor	8-day cycle history

Lovelace dashboard cards — see dashboard/ for installation.

What you get

Detect and fix short cycling

The integration watches every compressor start→run→stop transition and scores each cycle 0-100 based on runtime, water delta-T, off-time, backup-heater use, and defrost events. 8 pendulum patterns across Heating / Cooling / DHW detect short cycling hourly and daily. When cycling is excessive, the integration tells you exactly what to change — raise minimum run time, lower LWT setpoint, widen thermostat hysteresis.

Real COP intelligence, not just a number

9 per-mode COP sensors (heating / DHW / cooling × day / week / month) plus SPF seasonal, SPF year-to-date, and rolling 365-day SCOP. Weather-normalized 7-day-vs-7-day degradation with outdoor bin matching isolates hardware wear from weather variation. 30-day regression trend catches slow drift before it becomes a problem.

Datasheet-backed COP normalization

15 bundled Daikin Altherma models (8 EPRA + 7 ERLA) with full EN 14511 reference data. Your live COP is normalized to A7/W35 and compared to spec. cop_vs_datasheet_pct tells you if your pump is underperforming its design point — critical for detecting refrigerant loss, dirty filters, or sizing issues. Import your own datasheet for any non-bundled model.

Installer-grade LWT advice

Dynamic heat-curve advice in Daikin installer language: slope delta and offset delta, not just "raise/lower LWT". The suggested step is 1–3 °C scaled to the delta between measured and optimal LWT, with comfort guardrails (floor + ceiling on projected indoor temp) and tracking-error dampening so you never over-correct. Works with your LW setpoint (main) attribute for real accuracy.

Runtime, energy, and HA Energy Dashboard

12 sensors for compressor runtime, BUH runtime, defrost count/duration, duty cycle, and electrical/thermal kWh per mode. All energy sensors are HA Energy Dashboard compatible (state_class: total_increasing). Accumulator persistence across HA restarts — no more lost daily totals after a reboot.

19 alerts, fully tunable

19 alert types across 7 domains — pendulum, short cycle, ML, setpoint, COP/heat-curve, component health, data quality. Every alert has its own dedup window and — new in v1.9 — 11 per-alert toggles on top of 5 group toggles. Quiet hours with critical bypass, bilingual EN + NL templates, delivery to any notify.* service plus persistent notifications.

Per-installation machine learning

12-dimensional feature vector per cycle. Per-mode EWMA baseline with outlier skip. Adaptive thresholds (opt-in) that learn what "normal" looks like for your house. Weekly k-means clustering into 3 archetypes (pendulum / normal / dhw_like). Z-score anomaly detection flags unusual cycles before they become alerts. Energy anomaly detection with Welford baseline catches stuck valves and runaway demand.

Zero config complexity

8-step setup wizard with smart auto-import (v1.8) that detects non-canonical attribute names via alias table + fuzzy matching. Options menu with grouped sections and a dedicated notifications sub-menu. Reconfigure flows to change source sensor or model without re-adding. Diagnostics download for one-click troubleshooting support.

Production-grade quality

6724 tests, 100.00% coverage, mypy strict, Pylint 10/10, IQS Bronze + Silver + Gold. 8 GitHub Actions workflows enforced on every PR. No cloud. No telemetry. No accounts. Your heat pump data stays on your HA instance.

What's new

v1.9.0 — Alerts v2

3 new alerts targeting database health, ML retraining, and energy anomalies:

db_retention_warning — DB not being pruned, catches stuck maintenance
kmeans_retrain_stale — clustering hasn't succeeded in 14 days, catches ML degradation
energy_daily_anomaly — daily electrical consumption more than 2σ above baseline

Per-alert toggles — 11 new individual switches on top of the existing 5 group toggles. Silence specific alerts without touching their whole family.

Dedup windows — 19 new numeric fields, one per alert type. Fine-grained control over how often each alert can repeat.

Q4 fix — setpoint_osc no longer fires on monotonic drift (only real reversals).

I7/I8 fixes — retention fallbacks, defrost duration, energy accumulators decoupled from runtime lifecycle (daily energy totals now survive HA restarts even if runtime restore fails).

v1.8.0 — Smart auto-import

Onboarding that works even when your ESPAltherma firmware uses non-standard attribute names.

Entity discovery — scans sensor.* entities for overlap with the 13 canonical attributes
Alias table + fuzzy matching — 14+ community variants (benthouse, akaQ, Videopac, ...) resolve automatically
Anti-alias guard — Leaving water temp. after BUH (R2T) never wrongly resolves to R1T
Diagnose screen — new wizard step showing per-canonical match kind (EXACT/ALIAS/FUZZY/MISSING) and sanity warnings

v1.6.0 — HVAC Parity

Turns the integration from a cycle-detector into a full HVAC performance platform.

9 per-mode COP sensors — heating / DHW / cooling × day / week / month
SPF / SCOP — seasonal, year-to-date, rolling 365-day
15 bundled Daikin datasheets (8 EPRA + 7 ERLA) — normalized COP at A7/W35 + comparison vs spec
Weather-normalized degradation — outdoor-binned 7d-vs-7d, BUH/defrost excluded
30-day regression trend — isolates hardware wear from weather
Runtime + energy accounting — 12 new sensors, HA Energy Dashboard compatible
OptionsFlow overhaul — grouped sections + notifications sub-menu
Slope / offset advice in Daikin installer language

v1.5.0 — Dynamic LWT step

Dynamic LWT step 1-3 °C — scaled to measured vs optimal delta, not hardcoded
Comfort dual-loop — floor + ceiling on projected indoor temp
Setpoint comparison — uses real LW setpoint (main) instead of last bucket sample
Tracking-error dampening — smaller steps when unit lags setpoint
8 new attributes on heating_curve_advice

v1.4.x — Hardening

v1.4.0 — entity slug renames to sensor.daikin_cycle_ml_<key>
v1.4.1-1.4.5 — cycle/quality/notify fixes, hydrate path, rollup extension to 720 h

v1.3.0 — COP analytics foundation

COP hourly rollup + KPI sensors (cop_mean_day/_week/_month)
REST endpoint GET /api/daikin_cycle_ml/cop_hourly
sensor.cop_curve_recent (48 h recorder-safe window)
export_cop_hourly service

See CHANGELOG.md for the complete history.

Version history

Version	Date	Headline
v1.9.0	2026-10-20	Alerts v2: 3 new alerts, per-alert toggles, dedup windows, `recompute_kmeans` service
v1.8.1	2026-10-08	R331 fix — alert advice bleed
v1.8.0	2026-10-08	Smart auto-import wizard
v1.7.2	2026-09	Hotfix for issue #49
v1.7.1	2026-09	Security updates
v1.7.0	2026-09	Defrost history helper sensors
v1.6.2	2026-10-06	Onboarding fix — Automatic/Manual attribute mode (issue #25)
v1.6.1	2026-10-06	Energy tick fix — standby power not booked as heating
v1.6.0	2026-10-06	HVAC Parity — 3× features, breaking `cop_today`
v1.5.0	2026-10-03	Dynamic LWT step + comfort dual-loop
v1.4.0	2026-09-29	Entity slug renames (breaking)
v1.3.0	2026-09-26	COP hourly rollup + REST endpoint
v1.2.x	2026-09-20	IQS Silver + Gold
v1.1.0	2026-09-15	HACS-compliant layout
v1.0.x	2026-09-10	Initial release

Why this exists

Short cycling is the #1 hidden cause of reduced COP and premature compressor wear on Daikin Altherma heat pumps. Existing HA integrations miss it because they look at temperature or power — not at the actual compressor runtime pattern.

Daikin Cycle ML is built for the ESPAltherma community: users who already have rich local data but no tool to interpret it. It detects every cycle, learns what "normal" looks like for your installation per mode, and warns you when patterns drift — long before you'd notice from your energy bill.

No cloud. No external API. No dependency on Daikin's infrastructure. Just your data, analysed locally.

What it answers

Is my heat pump short cycling? How often? Is it getting worse?
Why is my COP lower this week than last week — weather or hardware?
Is my heat curve too high? What should I change it to?
Is my backup heater running more than it should?
Is my defrost behaviour normal for the current weather?
Is the pump degrading over time, or is it weather?

What it does NOT do

Doesn't write to your heat pump (read-only)
Doesn't call any cloud API (fully local)
Doesn't replace your thermostat (diagnostic layer, not control)

Documentation

Complete, per-topic guides are in docs/:

Doc	What it covers
Configuration	Source sensor requirements, setup wizard, options menu, reconfigure flows, cookbook
Entities	All 50 sensors, 15 binary sensors, 10 services, 8 repair issues, automations reference
Alerts & Notifications	All 19 alert types, delivery channels, dedup, quiet hours, per-alert toggles, templates
Machine Learning	Feature vector, EWMA baseline, adaptive thresholds, k-means clustering, anomaly detection
Database	Schema reference, `model_state` keys, retention, migrations, backup/recovery
Development	Dev environment, test suite, CI gates, IQS status, scope boundaries, release procedure
Troubleshooting	Setup issues, cycle detection failures, alert issues, DB problems, log reference
Upgrade	Version-by-version breaking changes, migration steps, downgrade recovery

Installation

Requirements

Home Assistant 2026.9.0 or newer
Python 3.14 (bundled with HA 2026.9)
ESPAltherma publishing attributes on a sensor (typically sensor.althermasensors)
Recommended: outdoor temp + leaving-water temp attributes (needed for dT and COP)

Install via HACS

Install HACS if you haven't already
HACS → Integrations → search "Daikin Cycle ML"
Select Daikin Cycle ML → Download
Restart Home Assistant (Settings → System → Restart)
Settings → Devices & Services → Add Integration → "Daikin Cycle ML"
Follow the wizard

Integration not appearing in the HACS search?

The integration may not yet be listed in the HACS default catalog (submission pending). If the HACS search finds nothing, add it as a custom repository:

HACS → Integrations → three dots (top right) → Custom repositories
Add https://github.com/elRadix/daikin_cycle_ml with category Integration
Click Add → find Daikin Cycle ML in the list → Download
Restart Home Assistant (Settings → System → Restart)
Settings → Devices & Services → Add Integration → "Daikin Cycle ML"
Follow the wizard

Manual install

Copy custom_components/daikin_cycle_ml/ to /config/custom_components/daikin_cycle_ml/
Restart Home Assistant (full restart, not just config reload)
Settings → Devices & Services → Add Integration → "Daikin Cycle ML"
Follow the wizard

Upgrading from an older version

See docs/UPGRADE.md for version-by-version notes. Only two breaking changes in the entire history:

v1.4.0 — entity slug renames
v1.6.0 — cop_today changed to heating-only

Everything else is backward compatible.

First hour

Before you start the wizard

Verify your source sensor exposes the required attributes.

Developer Tools → States
Search sensor.althermasensors
Expand "Attributes"
Look for these 13 keys (exact names):

INV frequency (rps)
Operation Mode
I/U operation mode
3way valve
Defrost Operation
Leaving water temp after BUH (R2T)
Inlet water temp (R4T)
Flow sensor (l/min)
Water pump operation
Outdoor air temp (R1T)
DHW tank temp (R5T)
BUH Step1
BUH Step2

If one or more are missing: enable the register in ESPAltherma's .h file, restart the ESP, wait for republish. Or use Manual mode in the wizard to map them yourself.

Optional but recommended:

LW setpoint (main) — enables modern LWT advice + setpoint_oscillating alert
indoor_temp_sensor (a separate room sensor) — enables comfort guardrails

Setup wizard (8 steps)

The wizard adapts to your setup:

Your situation	Wizard path
Standard ESPAltherma, bundled model	7 steps (user → attributes → cycle → pendulum → quality → notifications → finalize)
Standard, non-canonical attribute names	7 steps + map_attributes
Custom model	9 steps (adds HP spec + attribute mapping)
Custom with alias/fuzzy detection	10 steps (adds diagnose screen)

Key decisions:

Source sensor — usually sensor.althermasensors
Model — 17 bundled Daikin models (8 EPRA + 7 ERLA) + Custom. Pick your exact model if listed. Bundled = datasheet-backed COP comparison.
Attribute mode — auto for canonical names, manual if your firmware differs
RPS threshold — default 3. Lower = catches brief modulation dips, higher = misses short cycles.
Thresholds — defaults are sensible; tune later once you have data

Full wizard walkthrough in docs/CONFIGURATION.md §3.

Immediate post-install verification

After the wizard completes, check in this order:

Within 1 minute:

sensor.daikin_cycle_ml_source_health — shows < 30 s
binary_sensor.daikin_cycle_ml_source_stale — should be off
binary_sensor.daikin_cycle_ml_missing_attributes — should be off

Within 1 hour:

sensor.daikin_cycle_ml_cycle_state — should transition idle ↔ running when compressor starts/stops
sensor.daikin_cycle_ml_current_cycle — shows live cycle data during a running cycle
sensor.daikin_cycle_ml_cycles_today — should increment after first cycle completes

Within 24 hours:

First entries in daily_summary (check diagnostics)
sensor.daikin_cycle_ml_last_cycle — should show the last cycle's quality score

Within 7 days:

K-means clustering runs Sunday 04:00 local → clusters assigned
Energy baseline accumulates → energy_daily_anomaly can fire (needs ≥ 7 days)
ml_anomaly starts firing (needs ≥ 20 cycles per mode)

If nothing happens in the first hour

Check in this order:

Is the compressor actually running? Check sensor.daikin_cycle_ml_thermal_power_live — should be > 0 when heating
Is RPS > threshold? Options → Device → compressor_rps_threshold. Try lowering to 2.
Is the source sensor fresh? sensor.daikin_cycle_ml_source_health < 30 s
Are required attributes present? binary_sensor.daikin_cycle_ml_missing_attributes off
Wait longer — the first cycle takes ~1 h because it needs a full start→run→stop

Still stuck? See docs/TROUBLESHOOTING.md §3.

First day — what to look at

After 24 h of data, these entities start being useful:

Cycle health

sensor.daikin_cycle_ml_cycles_today — how many cycles? Typical: 6-20. Over 40 indicates excessive cycling.
sensor.daikin_cycle_ml_quality_today — average quality 0-100. Above 70 is good. Below 50 means something to fix.
sensor.daikin_cycle_ml_last_cycle — hover for attributes: duration, dT, RPS, COP, cluster

Alerts

binary_sensor.daikin_cycle_ml_pendulum_hourly — going on/off during active heating? Check setpoint / hysteresis
binary_sensor.daikin_cycle_ml_short_run — cycles under 20 min? Raise minimum run time or lower LWT setpoint
binary_sensor.daikin_cycle_ml_short_off — off-times under 5 min? Check thermostat hysteresis

COP

sensor.daikin_cycle_ml_cop_today — heating-only COP. Typical: 2.5-5.0 depending on outdoor temp and LWT
sensor.daikin_cycle_ml_cop_vs_datasheet_pct — if you selected a bundled model, this shows your COP as % of spec (should be near 100% in mild weather)

Advice

sensor.daikin_cycle_ml_heating_curve_advice — the main advice sensor. States:
- keep — no change needed
- lower_lwt — heat curve is too high, lower by step_c degrees
- raise_lwt — heat curve is too low
- no_data — no recent heating samples, or DHW active
- unknown — check the reason attribute

If reason: within_deadband, no action needed. If reason: no_recent_heating, wait for heating activity.

First week — what to tune

Tune cycle thresholds to your installation

Default thresholds work for most installs, but every installation is different:

If you get too many pendulum_hourly alerts:

Options → Pendulum → raise pendulum_cycles_per_hour from 4 to 6 or 8
Or Options → Notifications → Content → disable alert_group_pendulum

If you get too many short_run alerts:

Options → Pendulum → raise short_run_threshold_min from 20 to 30 or 45
Or Options → Notifications → Dedup → raise short_run window to 120 min

If you get too many short_off alerts:

Options → Pendulum → raise short_off_threshold_min from 5 to 10
Check if your thermostat has a minimum off-time setting

Set up the comfort guardrails

The heat curve advice uses projected indoor temperature to decide how aggressive to be:

comfort_min_c (default 20°C) — never advise lowering LWT below what would project this indoor temp
comfort_max_c (default 24°C) — never advise raising LWT above this

Adjust to your comfort preferences:

If advice is too conservative (never suggests lowering): raise comfort_min_c to 21 or 22
If advice is too aggressive (suggests lowering too much): lower comfort_min_c to 19

Configure notifications

Default notification behavior:

Persistent notification in HA sidebar (always)
No phone push (needs a notify_service)

To enable phone notifications:

Options → Notifications → Delivery → Notify service → select your notify.* entity
Options → Notifications → Delivery → Emoji → keep on (phone notifications have emoji)
Options → Notifications → Quiet hours → Enabled → set 22:00 to 07:00 to avoid night alerts

Review the first week's data

Check diagnostics:

Settings → Devices & Services → Daikin Cycle ML → menu → Download diagnostics
Look at database.extra.daily_summary — cycles per day, mode breakdown
Look at database.extra.kmeans_state — cluster labels and sizes
Look at database.extra.baseline_state — per-mode sample counts

Compare with your own observation:

Do cycles/day match what you see in real life?
Are cluster labels meaningful? (or all "normal"?)
Is the energy baseline growing?

Recipes & how-to's

Send LWT advice to your phone

Automation that notifies you when the pump advises a heat-curve change:

alias: "Daikin: LWT advice → phone"
mode: single
trigger:
  - platform: state
    entity_id: sensor.daikin_cycle_ml_heating_curve_advice
    to:
      - "lower_lwt"
      - "raise_lwt"
condition:
  - condition: template
    value_template: >-
      {{ state_attr('sensor.daikin_cycle_ml_heating_curve_advice', 'betrouwbaarheid') | float(0) >= 0.6 }}
action:
  - service: notify.mobile_app_your_phone
    data:
      title: "Daikin LWT advice"
      message: >-
        {{ state_attr('sensor.daikin_cycle_ml_heating_curve_advice', 'state_label') }}
        · setpoint {{ state_attr('sensor.daikin_cycle_ml_heating_curve_advice', 'huidige_setpoint') }}
        → {{ state_attr('sensor.daikin_cycle_ml_heating_curve_advice', 'doel_setpoint') }} °C

Log every cycle to HA Logbook

See cycle history in the HA logbook:

alias: "Daikin: log cycles"
mode: queued
max_exceeded: silent
trigger:
  - platform: state
    entity_id: sensor.daikin_cycle_ml_last_cycle
action:
  - service: logbook.log
    data:
      name: "Daikin cycle"
      message: >-
        {{ state_attr('sensor.daikin_cycle_ml_last_cycle', 'duration_min') }} min
        · quality {{ states('sensor.daikin_cycle_ml_last_cycle') }}
        · dT {{ state_attr('sensor.daikin_cycle_ml_last_cycle', 'dt_avg_k') }} K
        · {{ state_attr('sensor.daikin_cycle_ml_last_cycle', 'mode') }}

Daily summary to your phone

alias: "Daikin: daily summary"
mode: single
trigger:
  - platform: time
    at: "22:00:00"
action:
  - service: notify.mobile_app_your_phone
    data:
      title: "Daikin today"
      message: >-
        {{ states('sensor.daikin_cycle_ml_cycles_today') }} cycles
        · COP {{ states('sensor.daikin_cycle_ml_cop_today') }}
        · heating {{ states('sensor.daikin_cycle_ml_electrical_energy_heating_today') }} kWh

Alert on data source death

Triggers when the ESPAltherma link is dead for 2+ minutes:

alias: "Daikin: source stale"
mode: single
trigger:
  - platform: state
    entity_id: binary_sensor.daikin_cycle_ml_source_stale
    to: "on"
    for: "00:02:00"
action:
  - service: notify.mobile_app_your_phone
    data:
      title: "Daikin source stale"
      message: "ESPAltherma sensor not updating — check ESP device"

Energy Dashboard integration

The 6 energy sensors are directly compatible:

Settings → Dashboards → Energy
Click Add consumption under Grid or Solar
Select sensor.daikin_cycle_ml_electrical_energy_heating_today
All 6 energy sensors have state_class: total_increasing — the dashboard handles them automatically

More recipes in docs/ENTITIES.md §9 and docs/CONFIGURATION.md §6.

Frequently asked questions

Does this work without ESPAltherma?

No. The integration reads attributes from an ESPAltherma sensor. Without those attributes, cycle detection and COP analysis cannot work. It's not a cloud API client.

Does it write to my heat pump?

No. The integration is read-only toward the heat pump. All advice requires user action or user automation.

Can I use it alongside the official Daikin integration?

Yes. They don't conflict — different entities, different data sources.

How much space does it use?

Typically 10-40 MB after a full year, depending on retention settings. See docs/DATABASE.md §10.3.

Can I export the data?

Yes. Use daikin_cycle_ml.export_cycles and daikin_cycle_ml.export_cop_hourly services to export to JSON or CSV. See docs/ENTITIES.md §5.

Is my data sent anywhere?

No. All data stays in /config/.storage/daikin_cycle_ml.db on your HA instance. No cloud, no telemetry, no accounts.

What if my model isn't in the bundled list?

Use Model = Custom in the wizard. You can optionally provide a datasheet JSON for full functionality, or leave it empty for Basic profile defaults.

Why does `ml_anomaly` fire but I don't see anything wrong?

ml_anomaly flags statistical deviations, not necessarily problems. Check top_dim attribute — it tells you which dimension triggered. top_dim: v5 (outdoor temp) is usually a weather outlier, not a bug.

Why does my LWT advice stay on "no_data"?

Check the reason attribute. reason: dhw_active = DHW is running, advice pauses. reason: no_recent_heating = no heating activity. reason: within_deadband = advice is "keep", no action needed.

Where do I tune the sensitivity of cycling alerts?

Options → Pendulum → pendulum_cycles_per_hour (default 4) and pendulum_cycles_per_day (default 40). Raise both if you get too many alerts; lower if you want earlier warning.

Why doesn't Test-all show all 19 alerts?

Known HA quirk — kmeans_retrain_stale may not appear in the panel under bulk notification. The alert is functional, just cosmetically missing in this one case. See docs/ALERTS.md §10.3.

Can I downgrade if I don't like a new version?

Downgrading is not officially supported but is technically possible. See docs/UPGRADE.md §8. All historical data will be lost unless you export it first.

Quality

Metric	Value
Tests	6724 passed / 4 skipped
Coverage	100.00% (6724 stmts / 1922 branches)
mypy strict	35 files, 0 errors
Pylint	10.00/10
CI workflows	8, all green
IQS	Bronze + Silver + Gold
Entities	61 registered
Services	10
Repairs	8

Run the full local quality gate before every push:

scripts/ci-local.sh

Full development guide in docs/DEVELOPMENT.md.

Where the data lives

Daikin Cycle ML persists to:

/config/.storage/daikin_cycle_ml.db — SQLite: cycles, features, COP samples, model state
/config/.storage/daikin_cycle_ml.alerts — HA Store: dedup timestamps
/config/.storage/daikin_cycle_ml.user_datasheets — HA Store: user-imported model specs

All data stays on your HA instance. No telemetry, no cloud, no accounts.

Backup procedure in docs/DATABASE.md §8.

Scope

In scope:

Cycle detection and quality scoring
Per-mode COP / SPF / SCOP / degradation analysis
Daikin datasheet integration
Runtime + energy accounting
LWT heat-curve advice
ML (baseline, adaptive thresholds, clustering, anomaly)
19 alert types
Setup wizard + OptionsFlow

Out of scope:

Daikin cloud API
Setpoint writes (read-only)
Cost tracking / € calculations
Consumption forecasting
InfluxDB export (use HA recorder)
Web UI cycle explorer

Full list in docs/DEVELOPMENT.md §1.

Known limitations

kmeans_retrain_stale may not appear in Test-all — HA quirk under bulk notification burst, not a bug. See docs/ALERTS.md §10.3.
Energy baseline is not weather-normalized — cold snaps may trigger false energy_daily_anomaly alerts. See docs/ML.md §7.5.
defrost_used / buh_used variance may cause false ml_anomaly at shoulder seasons. See docs/ML.md §11.7.
No supervised ML — label_cycle collects labels but they are not used for training yet.

See docs/ML.md §11 for the full ML limitation list.

Support

Daikin Cycle ML is free and open-source, built and maintained in my own time. No cloud, no accounts, no telemetry — your heat pump data stays on your Home Assistant instance.

If this integration has helped you reduce short cycling, understand your cycles, or improve your COP, a coffee is always appreciated and helps keep the project going.

Get help

Issues: https://github.com/elRadix/daikin_cycle_ml/issues
Discussions: https://github.com/elRadix/daikin_cycle_ml/discussions
Tweakers ESPAltherma topic — largest user base, mixed NL/EN
Before opening an issue: see docs/TROUBLESHOOTING.md and download diagnostics (Settings → Devices & Services → Daikin Cycle ML → menu → Download diagnostics)

Contributing

See docs/DEVELOPMENT.md for:

Dev environment setup (scripts/bootstrap.sh)
Test suite (pytest -q — 100% coverage required)
Code quality gates (ruff, mypy strict, markdown guard)
Standard workflow (branch, commit, push, PR)
IQS status and how to maintain it

Every PR must:

Pass all 4 local gates (scripts/ci-local.sh)
Have 100% test coverage for new code
Update docs/ for any user-facing change
Reference the tracker issue (Refs #<num> on PR body line 1)

License

MIT — see LICENSE.

Acknowledgements

ESPAltherma community — the source-sensor format that makes this possible
Daikin Altherma users on Tweakers and HA forums — for the feedback loop that shaped every version
Home Assistant core team — for the framework that makes custom components this easy
IQS reviewers — for pushing toward Bronze + Silver + Gold compliance

Name		Name	Last commit message	Last commit date
.github		.github
assets		assets
custom_components		custom_components
dashboard		dashboard
docs		docs
scripts		scripts
tests		tests
.gitignore		.gitignore
.pre-commit-config.yaml		.pre-commit-config.yaml
.pylintrc		.pylintrc
AGENTS.md		AGENTS.md
ALERTS_V2.md		ALERTS_V2.md
CHANGELOG.md		CHANGELOG.md
DOCUMENTATION.md		DOCUMENTATION.md
HANDOFF.md		HANDOFF.md
HANDOFF_SOP.md		HANDOFF_SOP.md
LICENSE		LICENSE
README.md		README.md
ROADMAP.md		ROADMAP.md
conftest.py		conftest.py
hacs.json		hacs.json
pyproject.toml		pyproject.toml
requirements_test.txt		requirements_test.txt

Latest commit

History

Folders and files

Repository files navigation