dckiller51/bodymiscale 2026.5.0

on GitHub

2026.5.0

🙏 This release is dedicated to @mano3m, whose PR #355 was the inspiration and technical foundation for this version — state restoration, Dutch translations, profile ID filtering and weight range filtering. Thank you for this outstanding contribution.

🆕 New features

Profile system and interactive notifications

• New profile.py — complete profile management architecture:
  • NotificationCoordinator: central coordinator for sending interactive push notifications to mobile devices
  • Notification translations loaded dynamically based on HA server language (_get_notify_translations)
  • Added optional weight pre-filter for notification profiles (notify_weight_min / notify_weight_max) — when configured, a notification is only sent to a user if the measured weight falls within their range, allowing up to 5+ users to share the same scale without exceeding the 3-action limit on mobile notifications
  • ProfileFilter (ABC) + implementations NoFilterProfile, WeightRangeFilter, ProfileIdFilter: automatic profile selection based on measured weight, a sensor identifier, or no filter
  • build_profile_filter(): factory that instantiates the correct filter based on configuration
  • Support for 4 profile detection methods: none, weight_range, notification, profile_id
• New sensor sensor.profile_id — active profile identifier during a weigh-in

Dutch translations

• Added translations/nl.json

Impedance sensors created by the integration

• bodymiscale now creates its own impedance sensors (no more external input_number helpers needed for storage)
• Automatic state restoration on startup via RestoreSensor — no more unknown state after HA restart

Improved cold start

• Current state of scale sensors is read at startup in BodyScaleMetricsHandler.__init__ — calculations are available immediately without waiting for a new weigh-in, including after migration from a previous version

⚙️ Changes

Restructured config flow (VERSION 1 → 4)

• Added async_migrate_entry — automatic migration of existing v1, v2 and v3 config entries:
  • v1 → v2: move name/birthday/gender to data, move height/sensors to options
  • v2 → v3: add impedance_mode and calculation_mode, purge last_measurement_time
  • v3 → v4: add profile_method
• Refactored form schemas into 4 distinct functions:
  • _get_user_schema — identity (name, birthday, gender)
  • _get_modes_schema — height, impedance mode, calculation mode, profile method
  • _get_sensors_schema — scale sensors
  • _get_profile_schema — profile method and notification
• Removed _get_impedance_schema and _get_main_options_schema (replaced)
• Added _purge_impedance_keys, _purge_other_method_keys, _validate_weight_range
• Removed manual last_measurement_time sensor configuration — bodymiscale now manages the timestamp automatically from state.last_changed after each valid measurement, simplifying setup and avoiding stale timestamps

Internal metrics (metrics/__init__.py)

• New _MetricsStore — internal store with differentiated TTL between source metrics (weight, impedance) and derived metrics, replaces the simple dict

New constants (const.py)

• CONF_NOTIFY_DEVICE_ID, CONF_NOTIFY_SERVICE — notification configuration
• CONF_NOTIFY_WEIGHT_MIN, CONF_NOTIFY_WEIGHT_MAX - weight range for the notification
• CONF_PROFILE_ID, CONF_SENSOR_PROFILE_ID — profile identification
• CONF_PROFILE_METHOD, PROFILE_METHOD_NONE/NOTIFY/WEIGHT/ID, PROFILE_METHOD_OPTIONS
• CONF_WEIGHT_MIN, CONF_WEIGHT_MAX — weight range for profile filter
• EVENT_MOBILE_APP_NOTIFICATION_ACTION — HA event for notification response
• NOTIFICATION_COORDINATOR, NOTIFICATION_TAG, PENDING_MEASUREMENT_TIMEOUT
• MAIN_ENTITIES, CONSTRAINT_PROFILE_ID_MIN/MAX

Developer tooling

• Migration from setup.cfg + mypy.ini + requirements.txt + bandit.yaml → unified pyproject.toml
• Replaced flake8/black/isort with Ruff (lint + format)
• Modernized .devcontainer.json — astral-sh.ruff extension, removed black/flake8/autopep8
• Added .codespell-ignore for scientific terms in docstrings
• pre-commit: added codespell, yamllint, bandit hooks

🗑️ Removals

• cachetools removed from requirements in manifest.json — no longer used
• bandit.yaml removed — bandit config migrated to pyproject.toml
• setup.cfg, mypy.ini, requirements.txt removed — consolidated into pyproject.toml

🔧 Bug fixes

• Fixed [#362] — Metric.SKELETAL_MUSCLE_MASS was incorrectly listed as a dependency of Metric.BODY_SCORE but is only available in dual frequency mode; removed from dependencies so body score is correctly calculated in single impedance mode
• Fixed [#276], [#342], [#343] — the S400 is now supported via profile ID filtering (the S400 sends a profile ID per user via BLE); note that the S400 uses its own calculation engine.
• Fixed [#320] — profile ID filtering is now natively supported; configure a profile ID sensor per user and bodymiscale will automatically route measurements to the correct profile
• Fixed [#11], [#38], [#238], [#252], [#282] — profile filtering (weight range, profile ID, notification) allows multiple users to share the same scale sensors; combined with automatic state restoration, dedicated per-user input_number helpers are no longer needed
• Fixed [#171], [#235], [#330] — TTLCache expiry during reload caused metrics to be cleared for all but the first user to trigger a callback; replaced by _MetricsStore which preserves source metrics independently per user across restarts and reloads
• Possibly fixed [#117] — the new _MetricsStore with differentiated TTL ensures derived metrics are recalculated even when source values are unchanged
• Fixed [#37] — automatic state restoration via RestoreSensor ensures all metrics persist across restarts and reloads
• YAML → UI migration: birthday, height and impedance sensors correctly preserved when migrating from an old YAML config to UI configuration
• Fixed metric dependency graph being shared across all profile instances — depended_by was mutated on global _METRIC_DEPS objects, causing each metric recalculation to fire N times (where N = number of configured users). Each BodyScaleMetricsHandler now owns an isolated copy of the dependency graph.

📐 Blueprint

• Blueprint: Updated interactive_notification_user_selection_weight_data_update.yaml. (thank you @DJaeger)
  • Added dual-impedance support (Low & High Frequency) for Xiaomi S400 scales.
  • Integrated smart Min/Max sorting for LF/HF values.
  • Added 30s data freshness check for improved accuracy.
  • Added 2-decimal rounding for assistance mode.