You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
This commit was created on GitHub.com and signed with GitHub’s verified signature.
Changelog
1.3.1 - 2026-04-12
Documentation sync and quality scale
Updated README.md, INSTALLATION.md, IMPLEMENTATION_SUMMARY.md, and validate.py to reflect the event-driven subscription model (ring detection is exclusively event-driven with no polling fallback, backup polling is a low-frequency safety net at 60 s default)
quality_scale.yaml: strict-typing marked done (mypy strict clean, pyright clean with framework-level suppressions via pyrightconfig.json)
Added pyrightconfig.json to suppress HA-framework-level pyright false positives (cached_property override, stub fallback typing) that affect all HA integrations equally
Stripped vestigial polling ring-detection code from _async_update_data — baseline state capture retained for event handlers
Adopted HA _attr_* pattern for all static entity properties, fixing pyright reportIncompatibleVariableOverride errors
1.3.0 - 2026-04-12
Motion detection binary sensor
New binary_sensor.<intercom>_motion entity with device_class=motion, driven by the device's MotionDetected log events (state: "in" = motion started, "out" = motion ended)
Motion detection capability is auto-detected via /api/system/caps — the sensor is only created when motionDetection is "active,licensed" on the device
The log subscription (/api/log/subscribe) now includes MotionDetected events alongside CallStateChanged / CallSessionStateChanged when motion detection is available
last_motion timestamp exposed as extra state attribute
System capabilities (/api/system/caps) fetched once during coordinator static cache initialisation
Diagnostics output includes system_caps and motion_detection sections
RTSP probe skipped entirely when /api/system/caps reports rtspServer as not "active" — avoids a TCP+Digest handshake timeout on devices without the RTSP license
Translations updated (EN, CS) for the new entity
1.2.0 - 2026-04-12
RTSP credentials as separate configuration
The 2N RTSP server has its own user database, independent of the HTTP API accounts. RTSP credentials are now configured separately in the camera options step (rtsp_username, rtsp_password) with no fallback to HTTP API credentials
RTSP probe upgraded from port-reachability check to full Digest authentication handshake (unauthenticated OPTIONS → 401 challenge → authenticated OPTIONS) — the integration no longer marks RTSP as "available" when credentials are wrong
Special characters in RTSP credentials are URL-encoded in the stream URL
Without RTSP credentials configured, the integration skips RTSP entirely and falls back to MJPEG
RTSP credentials are redacted in diagnostics output
Translations updated (EN, CS) with labels and descriptions for the new fields
1.1.0 - 2026-04-11
HA 2026.4+ compliance
Imported homeassistant.components.persistent_notification API (hass.components.X accessor was removed in HA 2025.1)
OptionsFlow no longer stores config_entry; uses self.config_entry from the framework
DataUpdateCoordinator constructed with the config_entry kwarg so HA tags traces with the entry
Switched from Camera + stream_source + ffmpeg to homeassistant.components.mjpeg.MjpegCamera
Credentials are passed to MjpegCamera separately and never appear in the URL exposed to logs, diagnostics, or dashboards
Camera transport (RTSP / MJPEG / public-MJPEG) is resolved once at coordinator setup; the entity reads coordinator.camera_transport_info instead of probing on every property access
Event handling and call lifecycle
Push-driven log subscription via /api/log/subscribe + /api/log/pull + /api/log/unsubscribe with re-subscribe and exponential backoff (capped at ~60 s)
Polling fallback through /api/call/status keeps ringing detection alive when the push channel is degraded
New 2n_intercom.answer_call and 2n_intercom.hangup_call services with target.config_entry and a reason selector (normal / rejected / busy)
sensor.<intercom>_call_state exposes an active_session_id attribute so automations can hang up the exact session they answered
Hangup logic: only the firmware's idempotent code 14 + "session not found" response is treated as success — every other rejection (including code 14 + "Unsupported Content-Type") is logged as a real failure
Centralised device-error parsing across the entire client: any non-success response surfaces as a WARNING with code / description / param; only the idempotent hangup case stays at DEBUG
Diagnostics and real-state entities
sensor.<intercom>_sip_registration — derived from /api/phone/status
binary_sensor.<intercom>_input_1 — real /api/io/status input state
binary_sensor.<intercom>_relay_1_active — real cached /api/switch/status relay state
Lock fallback is_locked prefers cached switch/status for relay 1 and only falls back to optimistic state when switch/caps confirms relay 1 doesn't exist
Configuration flows
Reauth flow (async_step_reauth) — credential rejection raises ConfigEntryAuthFailed instead of looping on ConfigEntryNotReady
Reconfigure flow (async_step_reconfigure) — change host/port/protocol/credentials/SSL without removing the entry (HA 2024.10+)
Performance / KISS
Static caps (switch/caps, io/caps, camera transport) fetched once at setup, not refetched per poll — only switch/status, io/status, phone/status, and the call/status fallback poll on the 5-second interval
Shared TwoNIntercomEntity base class — device_info, available, and _attr_has_entity_name deduplicated across every platform
Snapshot caching collapsed into a single layer at the coordinator
Removed dead PLATFORMS constant; _get_platforms() now returns Platform enum members
