sensor.sun_next_rising flips to tomorrow at sunrise; the old anchor only
pulled it back when >12h away. With short June nights tomorrow's sunrise
is within 12h by mid-afternoon, so from ~17:30 local until sunset the
pair described tomorrow and the curve returned minimum.
Replace the clock-distance heuristic with ordering-based anchoring in a
new pure helper anchor_sun_events() (color_and_brightness.py):
- daytime (sunrise > sunset): pull sunrise back one day, unconditionally
- stale "today's sunset" still on yesterday: push sunset forward one day
- post-sunset ramp tail (both flipped, still within half_width): pull
both back so the down-ramp completes instead of snapping to min
8 regression tests incl. the June-afternoon repro and a winter check.
Config:
- .ruff.toml target-version py310 -> py312, matching pyproject's
requires-python (>=3.12) and the 3.12 'type' aliases switch.py
already ships. Unblocks two false invalid-syntax errors.
- Ignore D102 in tests (test-class methods), consistent with the
existing D100/D103 test ignores.
Code (no behavior changes):
- COM812 trailing commas + UP017 dt.UTC via ruff --fix, black re-wrap
- RUF002: unicode minus/en-dash -> ASCII hyphen in docstrings
- SIM105: contextlib.suppress for the lux-reading parse in config_flow
- TRY300: move return out of try in _read_lux_sensor
- PLC0415: hoist inline asyncio/logging imports to module top
- E741: rename ambiguous 'l' comprehension variable to light_id
- PT006/PT018: parametrize tuple + split compound assertion
- D102: docstring for OptionsFlow.async_step_init
- ARG001/ARG002: noqa with justification on HA-required signatures
- PLR0912/PLR0915: noqa on prepare_adaptation_data and
_update_attrs_and_maybe_adapt_lights (upstream-inherited complexity;
splitting would hurt readability more than it helps)
./scripts/lint passes clean; 148 tests pass.
Device page sorts entities alphabetically by name, so 'Min/Max …' always
listed max above min ('a' < 'i'). Rename the four range numbers to
'Brightness lower/upper' and 'Color temp lower/upper' so each quantity's
lower bound sorts first. Pin suggested_object_id to the field key so new
profiles keep slugging number.<profile>_min_brightness etc., identical to
profiles created before the rename; existing entity_ids are registry-
preserved via unchanged unique_ids.
New-profile defaults: brightness 5-100% -> 10-90% (5% proved too dim on
gloomy days; 90% leaves manual-boost headroom), min color temp 2200K ->
2000K (deeper warm tone won in practice). Existing profiles unaffected —
saved options and restored slider values take precedence.
Specs synced: runtime-range-controls naming table + scenarios,
output-sensors collision references. Two new tests cover the slug pin
and the lower/upper friendly-name composition.
Thread 2 (bug): runtime range number entities seeded from native_min on a
fresh profile (only a name stored at setup), so new groups ran at
max_brightness=1% and a 1000-1000K color-temp range until the options
dialog was saved once. Each RANGE_ENTITIES row now carries a sensible
DEFAULT_* and _options_value() falls back to it. Adds a regression test.
Thread 3: collapse 'Sun schedule' and 'Light control' by default
(now 2 open / 5 closed): Targets + Daytime curve expanded, the rest
collapsed.
Existing profiles unaffected (their options are already saved); only
newly-created groups get the corrected seed. Bump 2.3.0 -> 2.3.1-cdit.1.
The 'Lux reduction' sensor published the retained brightness factor, so it
read 100% when nothing was being reduced (0% actual reduction) — the label
and value were inverses. Now it publishes the actual reduction:
100 - round(factor*100), so 0% = no dimming, higher = more cut, None when
the gate is inactive.
Updates the output-sensors spec scenarios to match (71 -> 29, 100 -> 0,
adds the gate-inactive None case) and adds a publish-level value test
(none existed — the prior scenarios were never enforced in code).
Two config-flow bugs in the Ambient lux / Advanced sections:
- Conditional fields (target_lux, send_split_delay) only appeared on a
fresh re-open of the dialog, never in the same session — violating the
options-flow spec ('SHALL re-render ... in the same session'). Now, when
a driver is enabled but its dependent field was hidden from the rendered
schema, the form re-renders (carrying the user's edits) instead of saving.
- The options form could not be saved at all without selecting a lux
sensor: vol.Optional(CONF_LUX_SENSOR, default='') fed '' into the
illuminance EntitySelector, which rejects empty strings. Omit the
default (vol.UNDEFINED) when no sensor is configured.
Extracted _has_pending_reveal() and _overlay_range_values() helpers to
keep async_step_init under the branch limit. Added 3 flow-level tests
exercising the real re-render path (prior tests only hit the schema
builder, which is how these slipped through). 141 tests pass.
Optional per-profile lux sensor binding that dims lights when daylight
alone exceeds a user-set target (factor = target/current). Lights turn
off entirely below min_brightness. Two config fields, live lux reading
in the options flow, two conditional output sensors, sensor state
subscription with 5pp significance guard. 26/26 tasks, 138 tests green.
The CDiT fork is single-household and was carrying an entire upstream
docs/ mkdocs site documenting features the fork has removed (sleep mode,
take-over-control, brightness_mode selector, YAML config). The README
mostly auto-generated from those pages via markdown-code-runner.
What's gone:
- docs/ — all 11 pages + assets + run_markdown_code_runner.py
- .github/workflows/docs.yml — zensical+shinylive GitHub Pages build
- .github/workflows/markdown-code-runner.yml — auto-edit of README from docs/
- .github/update-services.py / update-strings.py — would clobber the
hand-edited services.yaml / strings.json
- scripts/update-generated-content — orchestrator for the above
- custom_components/adaptive_lighting/_docs_helpers.py + docs_gen.py —
only used by the deleted markdown-code-runner
- zensical.toml — docs-site config
- "docs" dependency group in pyproject.toml (markdown-code-runner,
shinylive, zensical, etc.) + lock regenerated
What's new:
- README.md rewritten to describe the fork's actual surface — 3 switches,
4 number entities, 3 sensor entities per profile, 2 services (apply +
change_switch_settings), entry-only setup, UI tuning, sun source.
Preserves the existing "What's new in 2.1" / "2.2" block quotes.
- webapp/README.md annotated to flag the Shiny simulator models the
upstream curve (with brightness_mode, sleep mode, etc.) and does not
reflect the fork.
- CLAUDE.md command table no longer references update-generated-content.
Sensor.py + test_sensor_platform.py: incidental black reformat from the
lint pass (multi-arg function calls split per line).
Tests still pass (123/123). Lint clean. Sensors implementation
unchanged.
Three sensor entities per AL profile expose the curve's current outputs
plus actual sun elevation as graphable numerics with
SensorStateClass.MEASUREMENT, so HA's recorder + apexcharts-card chart
them natively:
- sensor.<profile>_output_brightness (% — from self._settings)
- sensor.<profile>_output_color_temp (K — from self._settings)
- sensor.<profile>_sun_elevation (° — from sun.sun.attributes.elevation)
Architecture: master switch publishes computed outputs to
hass.data[DOMAIN][entry_id]["outputs"] after each curve tick, then fires
a per-entry dispatcher signal. Sensors subscribe and read from the cache.
Single computation path, push-based updates, no polling.
The existing master switch attributes (brightness_pct, color_temp_kelvin,
synthetic sun_position in [-1,+1]) are unchanged — sensors are purely
additive. Manifest bumped to 2.2.0-cdit.1 (minor; no breaking changes).
Artifact realignment along the way: Q2 answer assumed sun_position was
Sun2 elevation degrees, but reading color_and_brightness.py revealed
it's a synthetic float in [-1,+1] derived from the brightness curve.
Switched to pulling actual elevation from sun.sun (Decision 8 added),
renamed sensor from sun_position to sun_elevation. Friendly names use
asymmetric "Output" prefix to dodge collision with the adapt-brightness
switch and Min/Max color-temp numbers.
Remaining: group 6 (manual live-HA verification on
homeassistant.onca-blenny.ts.net) — requires HACS deploy.
15 new sensor tests pass; full suite green (123/123). openspec
validate --strict green. ruff clean on new files.
Adds four live-tunable `number` entities per AL profile (min/max
brightness, min/max color temp) that own the runtime curve values.
Slider changes take effect on the next curve tick — no integration
reload. State persists across HA restart via `RestoreNumber`.
Curve math now reads `min_brightness`, `max_brightness`,
`min_color_temp`, `max_color_temp` from the four runtime entities
via the entity registry, falling back to `entry.options` when an
entity is unavailable. The options flow seeds its four range fields
from the current entity state so the dialog matches reality.
Also fixes entity friendly names via HA's `has_entity_name`
composition: profile "Dining MVP" now reads as "Dining MVP",
"Dining MVP Brightness", "Dining MVP Color" — short enough for
HA's tightest cards. `unique_id`s are unchanged so existing
entity_ids stay stable.
Manifest bumped to 2.1.0-cdit.1 (minor, no breaking changes).
14 new tests in `tests/test_number_platform.py`; 108 passing
overall. OpenSpec change archived once 9.x live-HA verification
completes.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Production HA caught this on the first save attempt: HA rejects
`config_entry.add_update_listener()` when the integration also uses
`OptionsFlowWithReload`, raising:
ValueError: Config entry update listeners should not be used with
OptionsFlowWithReload
The listener was a vestigial bit of the pre-redesign reload plumbing
that should have been deleted alongside the manual reload path (design
decision D6). Removed:
- `config_entry.add_update_listener(async_update_options)` registration
in `async_setup_entry`
- `async_update_options` helper (only existed to call `async_reload`)
- `UNDO_UPDATE_LISTENER` tracking in `async_unload_entry`
- The `UNDO_UPDATE_LISTENER` import (still defined in const.py for any
external code that imports it; pruning that is a separate cleanup)
Updated `tests/test_init.py::test_successful_setup_on_current_version`
to assert `entry.entry_id in hass.data[DOMAIN]` instead of the listener
key. All 94 tests still pass locally.
This is the first real-HA-only bug we've shipped — props to the
production deploy for catching it before merge.
Test infrastructure:
- Add `pytest-homeassistant-custom-component` (PHACC) as a `test`
dependency group. PHACC ships `hass`, `enable_custom_integrations`,
`MockConfigEntry`, and friends without needing to clone HA core as a
sibling directory. Modernizes the test setup from upstream's
`setup-symlinks` pattern.
- conftest.py: add `auto_enable_custom_integrations` autouse fixture
so HA discovers the integration under `custom_components/` during
tests. Keep the upstream template-deprecation no-op.
- Import paths: all tests now import from `custom_components.adaptive_lighting`
(not `homeassistant.components.adaptive_lighting`) and from
`pytest_homeassistant_custom_component.common` (not `tests.common`).
New tests (groups 7.2-7.16):
- tests/test_color_and_brightness.py (16 tests): TestBrightnessCurve
asserts min-before-sunrise, midpoint-at-event, max-during-day, sunset
ramp symmetry. TestColorTempCurve verifies the same shape applies to
K. TestSunPosition checks the synthetic +1/-1/0 derivation.
TestTanhDayCurveDirect exercises the helper directly.
- tests/test_config_flow.py (12 tests): six sections in order, each
section contains only its specified fields, conditional visibility
of send_split_delay, default sun entities, strict-typed entity
selectors, NumberSelector slider/box configs, BooleanSelector for
every flag, full user→create-entry flow, YAML-managed entry aborts
with `yaml_managed` reason, options flow renders the sectioned schema.
- tests/test_init.py (6 tests): successful setup on current version,
stale version raises ConfigEntryError via async_migrate_entry,
unload is clean, tombstone removes orphan sleep entity + logs INFO,
tombstone is idempotent, tombstone respects config_entry ownership.
Source modernizations driven by the tests:
- __init__.py: add `async_migrate_entry` that surfaces the
"incompatible — delete and recreate" message and sets the entry to
MIGRATION_ERROR. Removed direct `ConfigEntryError` from
`async_setup_entry` (HA routes version mismatches through the
migration handler now).
- switch.py: removed upstream's YAML-managed-entry auto-remove hack.
YAML profiles now load normally and the options flow handles the
"you must edit configuration.yaml" message (spec R7).
Existing tests updated to PHACC paths: test_adaptation_utils.py
(38 tests), test_hass_utils.py (22 tests). Both pass without
modification beyond the import fix.
Deleted: tests/test_switch.py (2,999 LOC of upstream tests, most
covering sleep mode / take-over-control / manual-control state
machines that no longer exist; CDiT-specific switch tests deferred
to a follow-up change).
Result: `uv run --group test pytest tests/` → 94 passed in 0.57s.
Deferred from this change:
- 4.3: manual UI test that toggling a field and saving reloads cleanly
(requires a real HA instance, can't be done from CLI).
- 9.4: GitHub repo description / topics update (do via `gh repo edit`
outside the change scope).
openspec status: 4/4 artifacts complete; strict-validate green.
## Summary
- fix hassfest validation by replacing raw options-description URLs with Home Assistant translation placeholders
- extend the pytest CI matrix to cover the latest patch release for each Home Assistant month, plus dev
- align CI/dev container Python versions and tests with newer Home Assistant behavior
## Validation
- GitHub Actions pytest matrix passed for 2024.12.5 through 2026.4.3 plus dev
- Docker passed for linux/amd64 and linux/arm64
- hassfest, HACS validation, pre-commit, pre-commit.ci, markdown-code-runner, docs build, and Release Drafter passed
* test: regression test — AL must not override manual brightness with separate_turn_on_commands
End-to-end scenario: user adjusts brightness via a directly-bound Zigbee
switch (e.g. IKEA RODRET). No HA service call is made; ZHA reports the new
brightness via async_update_entity. On the next adaptation interval AL must
detect the change and stop overriding the user's brightness.
The test verifies the user-visible symptom: after two adaptation cycles
following a simulated direct-Zigbee brightness change, the light's brightness
must still be the manually set value — not AL's own target.
NOTE: this test FAILS on the current code. It is committed here to document
the bug before the fix is applied in the next commit.
* fix: merge last_service_data across split calls to fix detect_non_ha_changes with separate_turn_on_commands
When separate_turn_on_commands=True, each adaptation cycle makes two
light.turn_on calls (brightness, then color_temp). Previously each call
overwrote last_service_data[light], so after the cycle only the color_temp
key remained. _attributes_have_changed() then saw old_brightness=None and
silently skipped the brightness comparison, so a manually-set brightness was
never detected and AL kept overriding it.
Fix: merge instead of overwrite so all split-call attributes accumulate:
self.manager.last_service_data[light] = {
**self.manager.last_service_data.get(light, {}),
**service_data,
}
* test: add intermediate assertions to regression test
Two assertions were promised in the PR description but missing:
1. After the force-adapt, assert that last_service_data contains BOTH
brightness AND color — directly proving the merge fix works.
2. After the first non-forced update, assert that BRIGHTNESS is in
manual_control — proving detection fired, not just that the final
state is right.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* [pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
* refactor: remove spurious comments, trim test docstring and assertions
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* refactor: strip verbose comments from test, trim assert messages
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* [pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
* test: add message to bare assert
## Summary
- Fixes regression in v1.30.0 where lights turned on by automations were incorrectly marked as "manually controlled"
- Makes `adapt_only_on_bare_turn_on` respect individual attribute tracking from #1356
## Root Cause
PR #1356 added a call to `update_manually_controlled_from_event()` in the `turn_on_off_event_listener.on()` handler for ALL `light.turn_on` events, including when turning a light on from OFF state.
When an automation turns on a light with brightness/color attributes, this incorrectly marked the light as "manually controlled", preventing Adaptive Lighting from adapting it.
## Fix
1. Only call `update_manually_controlled_from_event()` when the light was **already ON** before the turn_on event. Turning on from OFF is handled by `_respond_to_off_to_on_event()`.
2. Make `adapt_only_on_bare_turn_on` respect `take_over_control_mode`:
- With `PAUSE_CHANGED`: Only pause adaptation of specified attributes, continue adapting unspecified ones
- With `PAUSE_ALL`: Pause all adaptation (existing behavior)
## Expected Behavior After Fix
| Scenario | `adapt_only_on_bare_turn_on` | `take_over_control_mode` | Result |
|----------|------------------------------|--------------------------|--------|
| Turn on from OFF with brightness | `false` | Either | NOT manually controlled |
| Turn on from OFF with brightness | `true` | `PAUSE_ALL` | All adaptation paused |
| Turn on from OFF with brightness | `true` | `PAUSE_CHANGED` | Only brightness paused, color adapts |
| Turn on from OFF without attributes | Either | Either | NOT manually controlled |
| Change brightness while ON | Either | Either | Brightness manually controlled |
## Test plan
- [x] Turn on light via automation with brightness/color (`adapt_only_on_bare_turn_on=false`) - should adapt
- [x] Turn on light via scene (`adapt_only_on_bare_turn_on=true`, `PAUSE_ALL`) - should pause all adaptation
- [x] Turn on light with brightness only (`adapt_only_on_bare_turn_on=true`, `PAUSE_CHANGED`) - should adapt color
- [x] Both intercept=True and intercept=False paths tested for consistency
- [x] CI tests pass
Fixes#1378
Co-authored-by: Mario Guggenberger <mg@protyposis.net>
* refactor: introduce light control parameter enum
* refactor: replace manual control flag with parameter enum
* test: update deprecated color temp attribute
* build: set execution bits on task scripts
* feat: individual manual control of brightness and color
* test: add tests for individual manual control evaluation
* fix: sequential manual changes not always detected
If multiple attributes of a light were changed within an interval, only the last change was detected because the check in the interval only used the latest event. For example, if there was a brightness change and a following color change, only the color attribute was detected as manually controlled. To fix this, the manual control attribute flags are now set directly from the event handler so that all events are processed.
* fix: invalid service description
* docs: fix missing space in config description
* refactor: pluralize multivalued bitmask enum name
* feat: add option to duplicate existing lighting instance
A menu step in the config flow was added that allows a user to create a
new instance or duplicate the options of an existing one.
* [pre-commit.ci] auto fixes from pre-commit.com hooks
for more information, see https://pre-commit.ci
* Add type annotation, tests, and translations for duplicate feature
- Add type annotation for source_options class attribute
- Add menu step translations to en.json
- Add tests for menu display, new instance creation, and duplication
* Simplify source_options access with class-level default
---------
Co-authored-by: Bas Nijholt <bas@nijho.lt>
* Add regression tests for SimpleSwitch initial state bug
Adds tests that verify SimpleSwitch._state is set immediately in __init__
rather than waiting for async_added_to_hass(). These tests currently FAIL
because _state is None after __init__, which causes an infinite loop in
_setup_listeners when the entity is disabled (since async_added_to_hass
is never called for disabled entities).
Regression tests for: https://github.com/basnijholt/adaptive-lighting/issues/1264
* Fix infinite loop when disabling SimpleSwitch entities
The issue was that SimpleSwitch._state was initialized to None in __init__,
but only set to a boolean value in async_added_to_hass(). When an entity
is disabled, async_added_to_hass() is never called, so _state stayed None.
The _setup_listeners() method has a while loop that waits for
_state is not None for all SimpleSwitch children (sleep_mode_switch,
adapt_brightness_switch, adapt_color_switch). With _state stuck at None,
this created an infinite loop.
The fix sets _state to initial_state directly in __init__ instead of
waiting for async_added_to_hass() to set it. The async_added_to_hass()
will still properly restore state from the last session or set based
on initial_state as before.
Fixes: https://github.com/basnijholt/adaptive-lighting/issues/1264
* Update .devcontainer
* Drop support for ≤2023.6
* Add .vscode/settings.json
* Use async_process_ha_core_config
* Fix for HA ≤2023.10
* pop normalized_name
* Add comments
* use . instead of source
* set pytest args
* set country
* use py3.13 for dev branch of core
* set country in test_adaptive_lighting_time_zones_with_default_settings
* Add comment
* Implement call intercept for multiple lights
* remove comment
* skip if no eids
* add comment
* fix type
* Fix skipped
* indentation
* Add logging and fix error
* Fix for HA ≤2023.04
* simplify
* remove unused ignores
* Debug mode
* Add test
* Make test failing
* rename switch
* rename lights
* Fix tests
* Rename lights in tests
* Remove unused dependencies
* Improve tests
* More tests
* Remove the DEBUG_MODE
* Add doc-string
* Extra test
* assert
* extra test
* Comments
* fix
* fix
* expand light groups
* more logging
* sort
* Revert is_proactively_adapting checks
This reverts commit 39fd8f2be0.
* simplify the mapping
* Revert "Revert is_proactively_adapting checks"
This reverts commit 18803e8e50.
* test
* no light groups
* do not expand
* Do not expand_light_groups in intercept
* more logging
* Fix
* add comment
* Add multi_light_intercept config option
* Update README.md, strings.json, and services.yaml
* add light group
* fix platform
* add simple test
* turn off again
* Test without take over control
* improve test and fix it in one way
* Fixes
* add cleanup fixture
* format
* Update test_switch.py
* add __str__
* remove unneeded call
* simplify service_data construction
* Generalize is_our_context
* Fix multi_light_intercept: false
* add comments
* add docs
* Update README.md, strings.json, and services.yaml
* Add feature line
* move function
---------
Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
* Before scheduling turn_on do a last-minute check if lights are off
* Pass force
* add comment
* fix for proactive
* commetn
* No default for force
* rm newline
* no force
* Return bool
