2.27.0 brings forecast/runtime parity, opt-in movement minimization for sun-tracking covers, and sun-state classification exposed through diagnostics and the `decision_trace` sensor. A false manual override that fired when a cover returned from unavailable is fixed. The release tooling also gains a ZIP-structure guard to prevent a double-nested HACS install path.

- **Forecast/runtime parity** ([#556]): the position forecast now calls the same snapshot-free primitives as the live pipeline, so min/max limits, floor-at-1%, rounding, and movement minimization all match runtime by construction.

- **Opt-in movement minimization** ([#555]): a new `CONF_MINIMIZE_MOVEMENTS` option quantizes sun-tracking positions to a configurable number of coverage steps, reducing small cover movements without changing average coverage.

- **Sun-state classification** ([#552], [#553], [#554]): a `SunState` enum (`HITTING`, `IN_FOV_NOT_VALID`, `OUTSIDE_FOV`) is computed per cycle, surfaced in diagnostics, and exposed as `sun_state` on the `decision_trace` sensor attributes.

- **False manual override suppressed** ([#546], [#551]): covers returning from unavailable no longer trigger a spurious manual override when no prior command target exists.

The position forecast in `forecast.py` previously re-implemented a subset of the live position math independently. It never applied min/max position limits, did not floor solar samples at 1% or round them, and used the raw default height rather than the sunset position during the sunset window. The result was a forecast that visually diverged from what the live pipeline would produce.

Three snapshot-free primitives now own the shared math in `pipeline/helpers.py`: `solar_position_from_geometry` computes the sun-tracking position from geometry alone, `default_position_with_limits` computes the default (sunset-aware) position with min/max limits applied, and `apply_config_limits` applies the configured position bounds. The live pipeline helpers in `pipeline/helpers.py` are thin adapters over these primitives. The forecast calls the same functions, so limits, floor-at-1, rounding, and movement minimization match runtime by construction.

`compute_effective_default` gained an `eval_time` keyword argument so the forecast can compute the sunset-aware default at each sample's projected time rather than at the current moment. A parity test locks `compute_solar_position` and `compute_default_position` against future drift between the two paths.

- `forecast.py` — now calls shared primitives instead of duplicating position math

- `pipeline/helpers.py` — `solar_position_from_geometry`, `default_position_with_limits`, `apply_config_limits` extracted here

Small positional corrections as the sun moves across the sky cause covers to jog frequently. The new `CONF_MINIMIZE_MOVEMENTS` option (default off) quantizes the computed sun-tracking position to a configurable number of evenly-spaced coverage steps via `quantize_to_coverage_steps()` in `position_utils.py`. Step count is controlled by `CONF_MAX_COVERAGE_STEPS` (default `DEFAULT_MAX_COVERAGE_STEPS`). Quantization applies after the raw solar position is computed, so it interacts correctly with min/max limits and the floor-at-1% rounding already in place.

The feature applies to tilt, blind, and awning cover types. It is surfaced in the config flow options UI, in `services.yaml` for service calls, and in all three translation files.

- `position_utils.py` — `quantize_to_coverage_steps()`

- `pipeline/handlers/solar.py` — applies quantization when `CONF_MINIMIZE_MOVEMENTS` is enabled

- `services.yaml` — `CONF_MINIMIZE_MOVEMENTS`, `CONF_MAX_COVERAGE_STEPS` exposed

- `translations/en.json`, `translations/de.json`, `translations/fr.json` — labels and descriptions

Each update cycle now classifies the sun's relationship to the cover window using the new `SunState` StrEnum in `engine/sun_geometry.py`, with three values: `HITTING` (sun is directly hitting the cover), `IN_FOV_NOT_VALID` (sun is within the field of view but not at a valid elevation or angle), and `OUTSIDE_FOV` (sun is outside the configured field of view entirely). The `in_fov()` method on sun geometry classes drives the classification.

`diagnostics/builder.py` reads the classification from the pipeline result and includes it in the diagnostics payload, making it visible in the HA diagnostics download. `sensor.py` surfaces `sun_state` as an attribute on the `decision_trace` sensor, so dashboards and automations can read the classification directly without downloading diagnostics. The attribute degrades gracefully when the value is absent.

- `engine/sun_geometry.py` — `SunState` enum, `in_fov()` method

- `diagnostics/builder.py` — classification included in diagnostics output

- `sensor.py` — `sun_state` attribute on `decision_trace` sensor

**False manual override on cover returning from unavailable ([#546], [#551]):** The override detector in `managers/manual_override/` treated any position delta as a manual override, even when the integration had never recorded a command target — for example, when a cover came back online after being unavailable. The fix threads `has_recorded_target` through the detection path: when no target has been recorded, the position-delta branch does not mark an override. A user-context change (explicit user action) still triggers the override regardless. The new constant `_NON_POSITION_COVER_STATES` centralizes the set of cover states excluded from position-delta detection.

The release script in `scripts/release` now verifies the built ZIP structure before publishing, guarding against a double-nested directory layout that causes HACS to install the integration at the wrong path ([#544], [#545]). This is a distribution-only change with no runtime effect.

- 4214 tests passing

- Home Assistant 2026.3.0+

[#544]: https://github.com/jrhubott/adaptive-cover-pro/issues/544

[#545]: https://github.com/jrhubott/adaptive-cover-pro/issues/545

[#546]: https://github.com/jrhubott/adaptive-cover-pro/issues/546

[#551]: https://github.com/jrhubott/adaptive-cover-pro/issues/551

[#552]: https://github.com/jrhubott/adaptive-cover-pro/issues/552

[#553]: https://github.com/jrhubott/adaptive-cover-pro/issues/553

[#554]: https://github.com/jrhubott/adaptive-cover-pro/issues/554

[#555]: https://github.com/jrhubott/adaptive-cover-pro/issues/555

[#556]: https://github.com/jrhubott/adaptive-cover-pro/issues/556