Repository navigation
Fix availability false-unavailable + add curb calibration endpoint-trim retry - #44
Merged
Pascal-ZeGerman merged 11 commits intoSep 11, 2026
Merged
Conversation
Fixes the root Pages URL 404 (only docs/demo and docs/explorer had index.html). Splits the two pages by audience: demo for everyone, explorer for maintainers/technical users. README's Live Demo section now leads with the hosted URLs instead of local-only instructions. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01VQxcp5EeadKNjM5u7WAR2r
…e behavior The device tracker picker only lives under Reconfigure (three-dot menu), not Configure (gear icon) — easy to miss and led to confusion after the 3.3.0-rc4 reconfigure flow shipped. Also clarify that VIN/Car Name are generic device_tracker attribute reads, not tied to any specific vehicle brand's integration. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01NVTaYxJNE3ANndgwg7LJPt
…backstop + tracker health) The sensor flipped to unavailable after 8h of device_tracker silence even though the resolved curb position was still correct. `last_gps_update` only advances on a real state_changed event, and the periodic heartbeat re-resolves from cached coordinates without touching it -- so a parked car (telematics asleep on ignition-off, or a Shortcut-posted tracker that only fires on phone movement) tripped the gate on every overnight park. Silence and failure are now treated as different things: - stale_timeout becomes a 168h (7-day) backstop for "no location event at all" rather than a freshness SLA. 7 days matches the NYC ASP recurrence cycle, so a car untouched for a week is ordinary rather than exceptional. - New coordinator.tracker_unavailable reads the tracker's live HA state; an explicit unavailable/unknown is a real integration failure (dead API, expired token) and marks entities unavailable immediately. Read live rather than cached because a tracker already unavailable at HA start never fires an event. A MISSING entity is deliberately not a failure -- that would flap the sensor on every restart for manually-posted trackers. - async_migrate_entry rewrites the shadowing legacy stale_timeout=8 stored in existing entries; without it the new default would reach nobody. Only the exact legacy value is touched. - _async_on_gps_update pushes an entity refresh on a tracker-health transition so availability re-evaluates at once instead of at the next pipeline run. The GPS-pipeline-health binary sensor also consults tracker_unavailable, since it shares stale_timeout and would otherwise stay ON for a week after the source integration died. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
The prior table-only version made users read the whole Configure/Reconfigure comparison before finding out which menu to click. Lead with a literal numbered click path for the common case (swapping the GPS source device tracker), then keep the table as a reference for what Configure covers. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01NVTaYxJNE3ANndgwg7LJPt
…rb calibration - ENDPOINT_TRIM_FT = 25.0 fixed module constant (spike 008 backtested value) - derive_segment_calibration() gains endpoint_trim_ft (default 0.0 = no-op) - derive_segment_calibration_with_endpoint_retry(): untrimmed-first, retry once with the 25 ft trim only on failure, return the untrimmed baseline if both fail
…antics - Synthetic corner-flare fixture reproduces the spike 008 mechanism offline: spread_n ~16.98 / spread_s ~17.70 untrimmed (rejected), c=+1.0 width=34.0 spreads ~0 with the 25 ft trim (calibrated) - TestEndpointTrim: constant, 0.0 default is a no-op, corner rejection, trim rescue, degenerate short-block collapses safely - TestEndpointRetry: passes-untrimmed / rescued-by-trim / fails-both branches plus the zero-trim opt-out
- _derive_segment_fields calls derive_segment_calibration_with_endpoint_retry; the roadbed cross-check still runs on whatever comes back - docstrings + inline comment name the retry and note the two-gate pipeline - TestCurbCalibrationEndpointTrim: corner-contaminated segment (~17 ft spreads untrimmed) now reaches segments.json calibrated with c=+1.0, width=34.0
…fast path The prior fix (e5a9a93) added a tracker-health fast path but only wired it into 2 of 15 entities and left the GPS-stale watchdog notification on the old full-stale_timeout wait, so a genuinely broken tracker still took up to 7 days to alert the user instead of firing immediately. - Watchdog notification now shares the same tracker-health fast path as entity availability, firing immediately instead of waiting stale_timeout. - Detect a deleted/renamed device_tracker via the entity registry (tracker_registered/tracker_failed), distinct from "not started yet", so a removed entity also triggers the immediate alert. - Consolidate the duplicated three-gate availability policy into coordinator.gps_data_available, shared by ASPNextMoveTimeSensor, ASPGpsPipelineHealthBinarySensor, ASPActiveNowBinarySensor, and the pipeline-derived diagnostic sensors (latitude/longitude/resolved street/resolution status/confidence/SODA level/last resolved); tracker name/VIN sensors gate on tracker health alone via a lighter mixin. - Coalesce the v1->v2 config-entry migration into a single async_update_entry call instead of two. - Extract tracker-health-transition handling out of _async_on_gps_update into its own method. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Follow-up cleanup pass on 8ac5637 (reuse/simplification/efficiency/altitude): - Extract _gps_watchdog_arm_normal from _gps_watchdog_rearm so a real GPS location fix (which already proves the tracker isn't unavailable) arms the backstop directly instead of re-deriving tracker_failed, including an unnecessary entity-registry lookup on every location update. - Merge the two near-duplicate coordinator test stubs in test_sensor_availability_backstop.py into one _StubCoordinator. - Replace the hand-rolled gps_data_available reimplementation in test_gps_pipeline_health_binary_sensor.py's stub with a rebind of the real property, matching the sibling test file's pattern. - Document why gps_data_available intentionally omits tracker_failed (entity availability stays lenient on a missing entity to avoid reboot-flapping; the one-shot watchdog notification does not need to) and the three-tier availability-gating convention for diagnostic sensors, so both read as deliberate design rather than oversights. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
![github-advanced-security[bot]](https://avatars.githubusercontent.com/in/57789?s=60&v=4){kind=small}
|
|
||
| def _patch_downloads(self, monkeypatch): | ||
| monkeypatch.setattr( | ||
| build_index, "_download_cscl_geojson", lambda: self._make_cscl_gdf() |
CI's ruff-format-check job was failing on these 3 files.
| monkeypatch.setattr( | ||
| build_index, "_download_cscl_geojson", lambda: self._make_cscl_gdf() | ||
| ) | ||
| monkeypatch.setattr(build_index, "_fetch_asp_signs", lambda: set()) |
new_state's None-ness isn't narrowed by has_gps_location alone since has_location() isn't a TypeGuard.
|
|
||
| if not has_gps_location: | ||
| return | ||
| assert new_state is not None # has_gps_location implies this |
Pascal-ZeGerman
deleted the
ship/availability-backstop-and-curb-calibration
branch
September 11, 2026 16:59
This was referenced Sep 12, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
endpoint_trim_ftand a fallback-retry to curb calibration, and wire the endpoint-trim retry into the index buildTest plan
pytest tests/test_curb_calibration.py tests/test_sensor_availability_backstop.py tests/test_gps_pipeline_health_binary_sensor.py tests/test_build_index.py tests/test_ha_integration.py(192 passed, run during code review)🤖 Generated with Claude Code