Skip to content

CHANGELOG.md

Latest commit

 

History

History
1062 lines (830 loc) · 50.9 KB

CHANGELOG.md

File metadata and controls

1062 lines (830 loc) · 50.9 KB
Raw

Changelog

All notable changes to VIGICam are documented here. Format follows Keep a Changelog. Versions follow Semantic Versioning: MAJOR.MINOR.PATCH.

  • PATCH — bug fixes, no new entities or config changes
  • MINOR — new entities, new features, backwards-compatible
  • MAJOR — breaking changes that require re-adding the integration

[Unreleased]

[0.7.2] - 2026-07-06

Fixed

  • Driveway camera RTSP connection storm — async_camera_image() opened a brand-new ffmpeg/RTSP connection to stream2 on every single snapshot request (dashboard thumbnails, notification snapshots), with no caching or lock against overlapping calls. On a camera that only tolerates a handful of concurrent RTSP sessions, this routinely piled up alongside the live-view stream and pushed it past its limit, causing the camera to start rejecting connections (429 Too Many Requests) and the live view to fail entirely, including from the camera's own local web UI. Snapshot requests are now cached for 10 seconds and serialized behind a lock, so overlapping requests share one ffmpeg call instead of each opening a new connection.

[0.7.1] - 2026-06-19

Added

  • Stream quality option — new toggle in Configure: "Use main stream (HD, stream1)". Off by default, which uses the sub-stream (stream2) for fast startup and smooth playback on Raspberry Pi and low-powered hardware. Enable for full HD resolution on more powerful hosts.
  • Dashboard camera cards now show a live feed (not just a snapshot) when Camera View is set to Live in the card editor — documented in USAGE.md.

Changed

  • Default live stream is now stream2 (sub-stream) instead of stream1. Existing installs are unaffected — the option defaults to off, preserving the previous sub-stream behaviour. Users who were relying on stream1 can re-enable it via Configure.

[0.7.0] - 2026-06-18

Added

  • Target Tracking switch (PTZ cameras only) — enables/disables the camera's auto-tracking feature, which pans and tilts to follow a detected subject.

[0.6.2] - 2026-06-18

Fixed

  • Removed deprecated homeassistant key from manifest.json (rejected by hassfest validator).
  • Sorted manifest.json keys to match required order: domain, name, then alphabetical.

Added

  • Hassfest validation workflow (.github/workflows/hassfest.yaml) for HA integration compliance checks.

[0.6.1] - 2026-06-18

Changed

  • smart_frame.py renamed to event_image.py; internal function names updated to match (fetch_latest_smart_frame → fetch_latest_event_image, get_smart_frames → get_event_images). No user-visible entity changes.
  • Configure dialog capability summary now shows "Event Image Capture" instead of "Smart Frame".

[0.6.0] - 2026-06-18

Added

  • Feature groups — Options flow (Settings → Devices & Services → VIGICam → Configure) lets you enable or disable Camera Stream, Detection Events, and Image Controls per camera. Disabling a group removes its entities from the registry immediately.
  • Last Detection image entity — updates on every detection event. Downloads a still from the SD card at the moment of the event if Event Image Capture is configured; otherwise falls back to a live RTSP grab. Exposes source, event_label, and detection_zone attributes.
  • Event Image Capture capability sensor — binary sensor indicating whether the camera has an SD card partition formatted for image storage. Probed via picture_total_space_accurate in hd_info (verified against live cameras; cameras without image storage always return 0B).
  • PTZ capability, OpenAPI capability, SD Card, and ONVIF Events capability sensors — all five capability sensors appear in the Diagnostics section of each camera device page.
  • Spotlight light entity — proper HA light entity with brightness control (0–255 mapped to camera's 1–4 scale). Replaces the old Spotlight Intensity number entity.
  • Privacy Mask switch — toggle the camera's lens mask on/off.
  • Image controls — optional tuning entities (brightness, contrast, saturation, WDR, HLC, flip, rotate, etc.) available when the Image Controls feature group is enabled.
  • Frigate integration awareness — detects Frigate at the same IP and suggests which feature groups to disable.
  • Entity removal confirmation step — options flow lists active entities that will be removed when a feature group is disabled, so changes aren't accidental.
  • detection_zone attribute on detection binary sensors and the Last Detection image entity — identifies which configured zone fired (e.g. Area1, Line1).
  • OpenAPI detection sensors (require OpenAPI enabled in camera settings): Vehicle Detected, Audio Anomaly, Loitering, Scene Change, Object Left or Taken, Area Entry, Area Exit.
  • SD card sensors (require OpenAPI): Recording Duration, Oldest Recording, Record Capacity Remaining, Video Space Free.
  • Uptime sensor (require OpenAPI) — hours since last camera reboot.
  • PTZ services: vigicam.ptz_move_to (absolute position), vigicam.ptz_save_preset, vigicam.ptz_delete_preset.
  • HACS Action workflow (.github/workflows/validate.yaml) — validates the repository against HACS requirements daily and on every push/PR.

Changed

  • source attribute on the Last Detection image entity: "smart_frame" → "event_capture". Update any automations that check this value.
  • smart_frame_label attribute renamed to event_label. Update any automations that reference this attribute.
  • "Smart Frame Capture" capability sensor renamed to "Event Image Capture" (cap_smart_frame → cap_event_capture). Existing installations: delete the old entity from the entity registry.
  • PTZ preset select entity tracks last-selected preset in HA (camera has no position API).
  • Entity naming uses alphabetical prefixes — Alarm, Detection, PTZ clusters in device view.

Fixed

  • Camera dashboard thumbnails work on HA 2026.x+ (replaced removed FFmpegManager.get_image).
  • SSL certificate errors on cameras using TP-Link internal CA (e.g. VIGI C320I).
  • Smart Detection and Line Crossing sensors now show "Clear / Detected" not "Off / On".
  • RTSP credentials redacted in log output if a snapshot grab fails.

[0.6.0b13] - 2026-06-18

Fixed

  • Event Image Capture sensor still false positive on VIGI C540V — get_picture_list also returns error_code 0 on both cameras (like get_media_list), so the previous fix did not help. The probe now reads picture_total_space_accurate from harddisk_manage hd_info instead. Cameras with a split SD card partition return a non-zero byte count (e.g. 3221225472B on the InSight S245); cameras without image storage return 0B (VIGI C540V). Verified against live cameras before release.

[0.6.0b12] - 2026-06-18

Fixed

  • Event Image Capture sensor false positive on VIGI C540V — the startup probe used get_media_list(media_type=2), which both cameras accept regardless of whether SD card capture is supported. Switched to get_picture_list(media_type=2), which backs the "Capture Playback" submenu in the VIGI app. Cameras without a split SD card partition (e.g. VIGI C540V) do not have this submenu and return an error on this call, so the sensor now correctly shows Off on those cameras.

[0.6.0b11] - 2026-06-18

Added

  • Capability diagnostic sensors — full section in USAGE.md documenting all five sensors (PTZ, OpenAPI, Event Image Capture, SD Card, ONVIF Events) with automation examples.
  • Event image capture setup guide in USAGE.md — step-by-step instructions: Storage → Capture Management → Event-triggered capture; Storage → Storage Management → Capture Capacity slider + disk format; Event → Smart Event → Upload Capture. Includes a note clarifying that the VIGI app's "Smart Frame" setting is a separate feature (live-view bounding box overlay) and does not need to be enabled for SD card image capture to work.
  • detection_zone attribute documentation added to the Binary Sensors and Last Detection sections in USAGE.md.
  • Image history guidance in USAGE.md — explains VIGI app Capture Playback as the right tool for reviewing past events; covers Gallery Card + automation approach for users who want a HA-based snapshot archive.

Changed

  • "Smart Frame Capture" capability sensor renamed to "Event Image Capture" (cap_smart_frame → cap_event_capture). "Smart Frame" is a VIGI app display feature (bounding box overlay on the live feed) unrelated to saving images to SD card. The sensor now accurately describes what it detects: whether the camera supports saving a full-frame still to SD card at the moment of each detection event. Existing installations will see a new entity; the old cap_smart_frame entity can be removed from the entity registry.
  • source attribute on the Last Detection image entity renamed from "smart_frame" to "event_capture". Update any automations that check this value.
  • smart_frame_label attribute on the Last Detection image entity renamed to event_label. Update any automations that reference this attribute.

[0.6.0b10] - 2026-06-18

Added

  • Capability diagnostic sensors — four binary sensor entities now appear in the Diagnostics section of every camera device page:

    • PTZ — on if pan/tilt/zoom was detected at startup
    • OpenAPI — on if the OpenAPI endpoint is reachable
    • Event Image Capture — on if the camera supports saving event images to SD card
    • SD Card — on/off updated live as the card is inserted or removed
    • ONVIF Events — on while the ONVIF pull-point subscription is active; goes off if the subscription drops or detection events are disabled

    These are automation-friendly: use them as conditions to write automations that adapt to whichever cameras have PTZ, or to check at a glance what the integration detected on a given camera.

[0.6.0b9] - 2026-06-18

Added

  • Last Detection image entity — when a detection event fires, the integration grabs an image and stores it as a Last Detection entity on the device. On cameras with Smart Frame enabled (and an SD card), it downloads the AI-cropped Smart Frame. On cameras without Smart Frame (e.g. VIGI C540V), it falls back to capturing a still from the live RTSP stream two seconds after the event arrives.
  • detection_zone attribute on detection binary sensors — when the camera's ONVIF event identifies which detection zone triggered (e.g. Area1, Line1), that name is now exposed as a detection_zone attribute on the binary sensor and the Last Detection image entity. Use this in automations to distinguish Area1 from Area2 on the same camera.

Fixed

  • DETECTION_EVENT_SUFFIXES mismatch — event-driven binary sensors use unique-ID suffixes like event_person, event_motion, etc., but the cleanup set contained the bare keys (person, motion). The mismatch meant disabling Detection Events did not remove stale entities from the registry. Corrected to use the actual suffixes.

[0.6.0b8] - 2026-06-18

Fixed

  • Capability summary now appears as hint text below the Camera Stream toggle instead of in the step description. The step description is not rendered by all HA frontend versions; data_description (toggle hint text) always renders.
  • Frigate note moved into the hint text below the Detection Events toggle.
  • Capability line is now compact and single-line: SD card ✓ · PTZ ✗ · OpenAPI ✓ · Smart Frame ✓ · Frigate ✗

[0.6.0b7] - 2026-06-18

Added

  • Entity removal confirmation step — when disabling a feature group would remove entities that are currently active in HA, the options flow now shows a confirmation screen listing those entities before saving. Closing the dialog without submitting cancels the change. Entities that are disabled (never individually enabled) are removed silently without a prompt.

[0.6.0b6] - 2026-06-18

Added

  • Capability summary in Configure dialog — the options flow now shows a ✓/✗ list of what was detected at startup: SD card, PTZ controls, OpenAPI, Smart Frame capture, and Frigate integration. Useful for confirming the integration found a feature, or spotting why an expected feature is missing.
  • Frigate auto-defaults — when Frigate is detected and the camera has never been configured via the options flow before, Camera Stream and Detection Events are pre-set to off in the dialog to avoid duplicate entities. The user can override before saving.

[0.6.0b5] - 2026-06-18

Added

  • docs/FRIGATE_SETUP.md — step-by-step guide for running VIGICam alongside Frigate: recommended architecture, which feature groups to disable, automation examples combining both integrations.
  • docs/MOVING_TO_FRIGATE.md — migration checklist for existing standalone VIGICam users adding Frigate: what changes, how to update automations, how to reverse the change.
  • docs/USAGE.md additions:
    • Feature Groups section — explains the three feature groups, how to configure them, and the Repairs notifications.
    • Spotlight light entity — replaces the removed Spotlight Intensity number entity.
    • Privacy Mask switch — blanks the camera feed on demand; includes automation example.
    • Image Controls section — full reference for all image tuning numbers, switches, and selects added in v0.6.0b3.

[0.6.0b4] - 2026-06-18

Added

  • Frigate integration awareness — VIGICam now detects whether Frigate has a camera configured at the same IP address. When detected:
    • The Options flow shows an advisory note suggesting you disable Camera Stream and Detection Events to avoid duplicate entities.
    • A has_frigate flag is stored in entry data for future features.
  • Frigate gone repair notice — if VIGICam was previously detected alongside Frigate but Frigate is no longer present (removed or reconfigured), a Repairs notification appears in HA's Problems section, guiding you to re-enable Camera Stream and Detection Events.

Changed

  • Frigate detection logic extracted into its own frigate.py module (was an inline function in config_flow.py). Covers both modern Frigate config entries and older builds that embed the camera URL in entity unique IDs.

[0.6.0b3] - 2026-06-18

Added

  • Image controls — optional camera tuning entities available when the Image Controls feature group is enabled in the integration options. All are in the Configuration entity category and hidden by default (enable individually from the entity settings gear):
    • Numbers: Image Brightness, Contrast, Saturation, Chroma, Sharpness, WDR Gain, Exposure Gain (all 0–100 sliders)
    • Selects: Flip, Rotate, Flicker (50/60 Hz), White Balance, Exposure Type
    • Switches: WDR, HLC, Dehaze, EIS, Auto Exposure Anti-flicker, Backlight Compensation, Lens Distortion Correction, Full Colour People Enhance, Full Colour Vehicle Enhance
  • Privacy Mask switch — toggle the camera's lens mask (screen blackout) on/off. Appears automatically when the camera supports it; always visible regardless of the Image Controls feature group setting.
  • image_common and lens_mask are now fetched every coordinator poll cycle so all new entities stay up to date.

[0.6.0b2] - 2026-06-18

Added

  • Spotlight light entity — the camera spotlight is now a proper HA light entity with brightness control, replacing the separate Spotlight Intensity number entity. Use it in automations with light.turn_on(brightness=200), include it in scenes, and control it from the default light card on dashboards. Brightness maps the camera's 1–4 internal scale to HA's 0–255 range.

Removed

  • Spotlight Intensity number entity — superseded by the spotlight light entity's brightness control. Existing automations using the number entity should be updated to use light.turn_on with brightness: instead. The entity is removed from the registry automatically on first load.

[0.6.0b1] - 2026-06-18

Added

  • Feature groups — new Options flow (Settings → Devices & Services → VIGICam → Configure) lets you enable or disable three feature groups per camera:
    • Camera stream (on by default) — the RTSP live-feed entity. Disable when Frigate provides the stream.
    • Detection events (on by default) — motion/person/intrusion binary sensors and ONVIF/OpenAPI event subscriptions. Disable when Frigate handles detection.
    • Image controls (off by default) — camera tuning entities added in a later release; toggle on here when ready.
  • Automatic entity cleanup — when a feature group is turned off, any entities that belonged to it are removed from the entity registry immediately on reload. No stale unavailable stubs left behind.
  • SD card removal notification — if the SD card is removed while HA is running, a notice appears in Settings → System → Repairs explaining what happened. Dismiss it to keep the entities (e.g. card temporarily removed); it clears automatically when the card is reinserted and the integration reloads.
  • Frigate awareness — the Options flow detects whether Frigate has a camera at the same IP and shows a contextual note recommending which feature groups to disable.
  • light platform stub added (Phase 2 will introduce the spotlight light entity).

Changed

  • README rewritten — new intro selling VIGI/InSight hardware, two-path setup explanation (standalone vs Frigate), and a "Why VIGI" section covering PoE, business-spec build quality, InSight line, and on-device AI.
  • OpenAPI event stream now logs the full raw event payload at DEBUG level, making it easier to inspect what the camera sends when zone-based detection fires.

[0.5.0b8] - 2026-06-17

Fixed

  • Firmware Version sensor now consistently appears in the Diagnostics section when enabled. Previously it had no entity category, so it appeared under Information/Sensors on one camera and was invisible on another. Still hidden by default — enable via the entity settings gear.

[0.5.0b7] - 2026-06-17

Fixed

  • Detection Person and Detection Vehicle switches showed wrong state on cameras where smart detection is configured via the camera web UI or OpenAPI. The camera web UI writes smart detection state through OpenAPI; the JSON API people_enabled / vehicle_enabled fields are legacy and not kept in sync. When OpenAPI is available, these switches now read from getPeopleDetectionSwitch / getVehicleDetectionSwitch and write via setPeopleDetectionSwitch / setVehicleDetectionSwitch.
  • OpenAPI is now probed before the first coordinator refresh so smart-detection state is already correct when HA entities are created on startup.

[0.5.0b6] - 2026-06-17

Fixed

  • PTZ Preset dropdown always showed "unknown" — the camera has no API to report its current position, so the dropdown now tracks the last preset selected from HA. Selecting from the dropdown, calling vigicam.goto_preset, or using button jogs / vigicam.ptz / vigicam.ptz_move_to all update the dropdown correctly. State resets to unknown on HA restart (camera may have been moved externally).

[0.5.0b5] - 2026-06-17

Fixed

  • OpenAPI SD card sensors (Recording Duration, Oldest Recording, Record Capacity Remaining, Video Space Free) and Uptime sensor all reported "unknown" — the coordinator was storing the full raw API response instead of unwrapping the inner result dict. Sensors now report correct values.
  • With loop recording enabled and the video partition full, Record Capacity Remaining and Video Space Free correctly report 0.

Changed

  • USAGE.md fully updated for v0.5.0: new "OpenAPI — unlocking additional sensors" setup section, automation examples for every binary sensor (all ONVIF and all 7 OpenAPI sensors), examples for every service, SD card sensor descriptions updated with NVR note and loop-recording behaviour.

[0.5.0b4] - 2026-06-17

Added

  • vigicam.ptz_move_to service — move the camera to an absolute pan/tilt/zoom position (pan and tilt: -1.0 to 1.0, zoom: -1.0 to 1.0). Requires PTZ camera + OpenAPI.
  • vigicam.ptz_save_preset service — save the camera's current position as a named preset (slot 1–8). If no slot id is given, the first unused slot is chosen automatically. Requires PTZ camera + OpenAPI.
  • vigicam.ptz_delete_preset service — delete a preset by name. Requires PTZ camera + OpenAPI.
  • PTZ Preset select entity refreshes automatically after save/delete (on next 30s poll).

Notes

  • The camera does not expose a current position API — no position sensor is possible. Use presets to track meaningful positions via the PTZ Preset select entity.

[0.5.0b3] - 2026-06-17

Added

  • SD Card Recording Duration sensor — total hours of video stored on the SD card (requires OpenAPI + SD card present).
  • SD Card Oldest Recording sensor — datetime of the earliest stored clip (requires OpenAPI + SD card present).
  • SD Card Record Capacity Remaining sensor — hours of recording space left (diagnostic, requires OpenAPI + SD card present).
  • SD Card Video Space Free sensor — GB of video storage free (diagnostic, requires OpenAPI + SD card present).
  • Uptime sensor — hours since last camera reboot (diagnostic, hidden by default, requires OpenAPI).
  • Coordinator now polls getSdCardStatus and getDeviceStatus via OpenAPI on every 30s refresh cycle when OpenAPI is available.

[0.5.0b2] - 2026-06-17

Added

  • Vehicle Detected binary sensor — fires on vehicle detection events via OpenAPI subscribeMsg (previously only available as part of the "Smart Detection" catch-all).
  • Audio Anomaly binary sensor — fires on sound anomaly detection events.
  • Loitering binary sensor — fires when loitering is detected (InSight S245 only).
  • Scene Change binary sensor — fires when the camera scene changes unexpectedly (InSight S245 only).
  • Object Left or Taken binary sensor — fires on abandoned/removed object events.
  • Area Entry / Area Exit binary sensors — fire on configured area crossing events.
  • All 7 sensors require OpenAPI to be enabled in camera settings; they appear automatically when has_openapi=True. Smart Detection (ONVIF) continues as a catch-all fallback.
  • The integration enables msg_push_enabled for each detection type on first connect so events are guaranteed to fire via subscribeMsg.

[0.5.0b1] - 2026-06-17

Added

  • OpenAPI client (openapi.py) — connects to the TP-Link IPC OpenAPI on port 20443 using the two-step SHA-256 doAuth flow, with a 25-minute stok cache and automatic re-authentication. Lays the groundwork for Vehicle Detection, Audio Anomaly, and other split detection sensors (coming in v0.5.0b2).
  • OpenAPI feature detection — at startup the integration probes port 20443; if available, has_openapi=True is set and the OpenAPI client is activated. If not available, an INFO log message explains how to enable it in the camera settings. The coordinator re-checks every 5 minutes so the flag activates without a restart if the user enables OpenAPI later.
  • SD card detection (has_sd_card) — at startup the integration checks whether a usable SD card is present (status normal or full). SD card sensors (Used %, Total, Free, Status) are now only registered when an SD card is detected, keeping the device page clean for cameras on NVR-only storage.

[0.4.0] - 2026-06-17

Added

  • Last Detection image entity (image.<camera>_last_detection) — updates automatically on every detection event (intrusion, line crossing, motion, person detected, smart detection). On cameras with Smart Frame support, downloads the AI-cropped Smart Frame from the SD card via WebSocket; on cameras without Smart Frame support (e.g. VIGI C540V), falls back to grabbing a still from the live RTSP stream. The source attribute indicates which method was used (smart_frame / rtsp_snapshot). Smart Frame images also expose smart_frame_label and file_id attributes.
  • Smart Frame support is now detected automatically at startup — no manual configuration needed.

Fixed

  • Camera dashboard thumbnails were broken in recent HA versions — FFmpegManager.get_image no longer exists. Replaced with asyncio.create_subprocess_exec throughout.
  • Setup failed with SSL certificate errors on cameras using the TP-Link internal CA (e.g. VIGI C320I). Two-stage fix: SSL errors now propagate unwrapped from api.py so the config flow's no-verify fallback session can run correctly.
  • Config flow had no logger — all setup failures were silently swallowed with no debug output. Errors are now logged at debug level with full exception detail.

[0.4.0b7] - 2026-06-17

Added

  • Last Detection image entity now works on cameras without Smart Frame support (e.g. VIGI C540V). When Smart Frame capture is unavailable, the entity falls back to grabbing a still from the live RTSP stream (stream1) via ffmpeg on any detection event — intrusion, line crossing, motion, person detected, or smart detection. The source attribute on the entity indicates "smart_frame" or "rtsp_snapshot" so you can tell which method was used.

[0.4.0b6] - 2026-06-17

Fixed

  • Setup still fails with SSL error on cameras using the TP-Link internal CA (e.g. VIGI C320I). Root cause identified: api.py's _post method was catching ClientConnectorCertificateError as a generic aiohttp.ClientError and wrapping it in VIGIError, so the config flow's SSL-fallback branch never ran. SSL errors now propagate unwrapped from _post, allowing Attempt 2 (no-verify session) to execute correctly.

[0.4.0b5] - 2026-06-17

Fixed

  • Setup fails with SSL certificate error on cameras using the TP-Link internal CA (CN=TPRI-CA / CN=TPRI-DEVICE), including the VIGI C320I. The SSL fallback in the config flow was using async_create_clientsession(verify_ssl=False) which is unreliable across HA/aiohttp versions. Attempt 2 now uses session=None so VIGICamera creates its own session with an explicit no-verify SSL context — the same approach used successfully in the rest of the integration.

[0.4.0b4] - 2026-06-17

Fixed

  • Config flow had no logger — connection failures during setup were silently swallowed with no debug output. Errors during cannot_connect and invalid_auth are now logged at debug level with full exception detail, making it possible to diagnose setup failures on unsupported or different-firmware cameras.

[0.4.0b3] - 2026-06-17

Added

  • Smart Frame support is now detected automatically at startup. Cameras that do not support Smart Frame image capture (e.g. VIGI C540V) will not show the Last Detection image entity. Mirrors the existing PTZ capability detection pattern.

Changed

  • Documented Smart Frame model limitation in docs/USAGE.md — the entity will simply be absent on unsupported cameras rather than showing as unavailable.

[0.4.0b2] - 2026-06-16

Fixed

  • Camera dashboard thumbnail was broken — FFmpegManager.get_image no longer exists in current HA versions. Replaced with asyncio.create_subprocess_exec (same pattern as Smart Frame downloads). Restores stream2 for low-bitrate dashboard thumbnails; stream1 remains the live view source.

Added

  • Documented the Last Detection image entity requirements in docs/USAGE.md (Smart Frame capture must be enabled; SD card must be formatted for image storage).

[0.4.0b1] - 2026-06-16

Added

  • Last Detection image entity (image.<camera>_last_detection) — downloads the most recent AI-cropped Smart Frame from the camera's SD card via WebSocket (wss://<ip>:8443/stream) each time an ONVIF detection event fires. Shows the detected object as cropped by the camera's on-device AI, not a raw stream grab. Exposes detection_type (ONVIF trigger: "person", "smart_detection", etc.) and smart_frame_label ("Person", "Smart Detection", etc.) as state attributes. Requires Smart Frame capture enabled in camera settings (Event → Smart Frame) and an SD card inserted; image is unchanged if either is unavailable.
  • Added docs/camera_api_research.md — local API research notes covering the JSON API, ONVIF, RTSP replay, Smart Frame WebSocket protocol, and SD recording behaviour.

[0.3.26] - 2026-06-16

Fixed

  • Smart Detection and Line Crossing sensors showed "Off" / "On" instead of "Clear" / "Detected". Added device_class=BinarySensorDeviceClass.MOTION to both — consistent with Motion, Person, Tamper, and Intrusion which already had the correct device class.

Docs

  • Added a note to Smart Detection in USAGE.md clarifying that vehicle detection fires this sensor (not a separate entity) — the Detection Vehicle switch only enables the feature on camera; the ONVIF event topic is shared across all smart detection types.

[0.3.25] - 2026-06-15

Fixed

  • RTSP credentials redacted in logs — if a snapshot grab fails, the error is now logged with the password replaced by *** so credentials cannot appear in the HA log. The RTSP URL with credentials is only held in memory and never written to any log output.

[0.3.24] - 2026-06-15

Performance

  • Sub-stream thumbnails — dashboard card previews and history snapshots now grab a frame from stream2 (sub-stream, low bitrate) instead of stream1 (HD). The live view when tapping the card still uses the full HD stream1. Significantly reduces CPU load on the HA host for thumbnail requests, especially on Raspberry Pi.

[0.3.23] - 2026-06-15

Added

  • times and pause inputs in the announce blueprint (camera_announce.yaml) — Repeat count (default 1, max 50) and Pause between repeats (default 1 s) added to the VIGI Camera — Speak on Trigger blueprint.
  • New blueprint: VIGI Camera — Play Audio File on Trigger (camera_play_file.yaml) — trigger-based automation to play a pre-recorded file from the HA media browser, a www/ URL, a file path, or an external URL. Includes slot, repeat count, and pause inputs with guidance on getting the media browser URL.

[0.3.22] - 2026-06-15

Added

  • Device name suggestion during setup — config flow now has a second step after credentials are validated. The camera's configured name (dev_name / alias, URL-decoded) is pre-filled so users can confirm or change it before the entry is created. Existing entries are unaffected.
  • MAC Address diagnostic sensor — shows the camera's MAC address. Hidden by default; enable from entity settings. No extra API calls — pulled from the existing device_info poll.

[0.3.21] - 2026-06-15

Fixed

  • times parameter ignored in play_file and speak — play_audio was sleeping only pause seconds (default 1 s) between plays. If the clip was longer than pause, the next test_audio fired while the camera was still playing, and the camera dropped it. Fix: play_audio now accepts audio_duration (calculated from WAV size) and sleeps for audio_duration + pause between plays, ensuring the clip always finishes first. vigicam.play_audio (existing slot, unknown duration) falls back to pause-only.

[0.3.20] - 2026-06-15

Fixed

  • vigicam.play_file "No such file" on HA OS — HA OS mounts the media directory at /media/ (a separate partition), not /config/media/. Path resolution now uses hass.config.media_dirs.get("local") which returns the correct base path per installation type. Fallback to config_dir/media/ for Container/Core/Supervised.

[0.3.19] - 2026-06-15

Fixed

  • vigicam.play_file 401 Unauthorized on HA media browser URLs — HA's /media/local/ endpoint requires authentication that the shared aiohttp session does not carry. URLs matching /media/local/ or /local/ are now automatically resolved to their equivalent file paths on the HA host before fetching, avoiding the auth requirement entirely. External URLs are unaffected.

Added

  • HACS brand icon — added custom_components/vigicam/brand/icon.png (wall-mounted CCTV camera). HA 2026.3+ changed icon delivery for custom integrations to a brand/ subdirectory inside the integration folder.

[0.3.18] - 2026-06-15

Added

  • vigicam.play_file service — fetch any audio file (HTTP URL or absolute path on the HA host), convert to 8 kHz mono WAV via ffmpeg, upload to a camera slot, and play it. Accepts any format ffmpeg supports. Same 15 s / 256 KB limit as vigicam.speak. Files in /config/www/ are reachable as http://<ha-ip>:8123/local/filename.
  • times and pause parameters added to vigicam.speak and vigicam.play_file — repeat the audio N times with a configurable gap. Range 1–50 (matches camera UI range). Note: "Alarm Sound Repetitions" entity controls the alarm trigger button only; use times in the service call to repeat speak/play_file announcements.

[0.3.17] - 2026-06-15

Fixed

  • vigicam.speak ffmpeg conversion failing — pipe:1 output path was accidentally dropped when switching to -f s16le in 0.3.16. ffmpeg had no output destination and exited with an error before producing any PCM data.

[0.3.16] - 2026-06-15

Fixed

  • vigicam.speak upload rejected (-67306) — root cause fixed. ffmpeg writing WAV to stdout (pipe) cannot seek back to fill in the correct RIFF/data chunk sizes, leaving a 0x7FFFFFFF placeholder that the camera's strict header parser rejects. Fix: output raw PCM (-f s16le, no header) from ffmpeg and construct the WAV header in Python with the exact data size. The codec reverts to pcm_s16le 8 kHz mono which is confirmed to work.

[0.3.15] - 2026-06-15

Fixed

  • vigicam.speak upload rejected (-67306) — switched ffmpeg codec from pcm_s16le (16-bit linear PCM) to pcm_mulaw (G.711 μ-law). IP cameras typically use μ-law as their native audio format for alert sounds; 16-bit PCM is rejected at the codec level.

[0.3.14] - 2026-06-15

Fixed

  • vigicam.speak upload rejected (-67306) — ffmpeg now passes -map_metadata -1 and -fflags +bitexact to produce a minimal WAV without extra metadata chunks that the camera firmware rejects.
  • Speak: delete slot before uploading — clears the slot first to avoid rejection when the camera treats an already-occupied slot as immutable.

[0.3.13] - 2026-06-15

Fixed

  • hass.config.version does not exist in HA 2026.x — replaced with homeassistant.__version__. This crash was preventing the TTS diagnostic log from printing, masking the root cause.
  • Added approach 4: entity_components lookup (HA 2025.x+ entity-based TTS architecture).
  • Added approach 5: brute-force scan of all hass.data values for any object with async_get_tts_audio, as a last resort that works regardless of storage key.
  • Diagnostic error log now reliably prints the TTS-related hass.data keys so the correct access method can be identified if all else fails.

[0.3.12] - 2026-06-15

Fixed

  • vigicam.speak TTS lookup — import DATA_TTS_MANAGER directly from HA so the correct hass.data key is used regardless of which version renamed it. Added a module-level async_get_tts_audio fallback (HA 2024.10+ public API). If all three approaches fail, the error log now includes the actual TTS-related hass.data keys present so the correct approach can be identified.

[0.3.11] - 2026-06-15

Fixed

  • vigicam.speak not working — tts.get_tts_audio action is not registered in HA 2025.x+. The service now uses the internal TTS manager (hass.data["tts_manager"]) directly, which is available in all HA 2024.6+ versions and works with any configured TTS engine (Piper, Nabu Casa Cloud, etc.). tts.get_tts_audio is kept as a fallback for older builds.
  • Alarm Sound Repetitions max value corrected from 10 → 50 (matches camera web UI range).

[0.3.10] - 2026-06-15

Fixed

  • Blueprint (camera_announce.yaml) rejected by HA with "extra keys not allowed" — removed invalid example: key from the message input (not a valid blueprint input field).

[0.3.9] - 2026-06-15

Added

  • vigicam.speak service — full TTS-to-camera pipeline in one service call. Accepts a text message, generates TTS audio via tts.get_tts_audio, converts to 8 kHz mono WAV via ffmpeg (guaranteeing camera compatibility regardless of TTS engine output format/bitrate), uploads to the camera, and plays it. Requires HA 2024.6+.
  • Blueprint updated (blueprints/automation/vigicam/camera_announce.yaml) — now uses vigicam.speak directly. Configure: trigger + camera + message template + TTS engine + language. The message field is a template selector so trigger data can be included.

Notes

  • vigicam.speak uses tts.get_tts_audio (HA 2024.6+, return_response=True). On older HA the service will log a clear error pointing to the version requirement.
  • ffmpeg conversion uses the binary from HA's ffmpeg manager (ffmpeg dependency already declared in manifest). All TTS engine outputs (MP3, OGG, WAV at any rate) are resampled to 8 kHz mono PCM WAV, which is always within the camera's format limits.
  • If the converted WAV exceeds 256 KB (about 15 seconds of speech at 8 kHz), the service logs an error asking you to shorten the message.

[0.3.8] - 2026-06-15

Added

  • vigicam.delete_audio service — delete a custom audio slot (101, 102, or 103) from the camera. Safe to call on an already-empty slot (camera returns success).
  • Alarm Sound Repetitions number entity — controls how many times the alarm sound plays per trigger (was only configurable via the camera web UI). Range 1–10. This is what controls the "10 second loop" behaviour of the Alarm Trigger button.
  • vigicam.play_audio now supports times and pause parameters — repeat the audio N times with a configurable gap between plays (default 1s). Useful for announcements that should be heard twice.
  • Blueprint: VIGI Camera — Announce on Trigger (blueprints/automation/vigicam/camera_announce.yaml) — importable automation blueprint. Configure a trigger + camera + slot; the automation plays the pre-uploaded audio when the trigger fires. Repeat count configurable.

Notes

  • Custom audio behaviour summary: play_audio / test_audio plays a slot once (or N times with pause). The Alarm Trigger button uses manual_msg_alarm which loops the configured alarm type based on sound_alarm_times — that is now the Alarm Sound Repetitions entity. These are two separate playback paths with different use cases.
  • 3 custom slots (101–103). Listing slots is not supported on tested firmware (-40106) but upload, play, and delete all work. Overwriting is safe — just re-upload to the same slot.
  • Blueprint uses tts.speak trigger selector requiring HA 2024.6+.

[0.3.7] - 2026-06-15

Added

  • vigicam.upload_audio service — uploads an audio file (via URL) to a custom slot on the camera (slots 101, 102, or 103). Set play: true to play immediately after upload. Supported formats: WAV mono 8 kHz ≤15 s ≤256 KB; MP3 mono ≤15 s ≤128 KB ≤64 kbps.
  • vigicam.play_audio service — plays any audio slot through the camera speaker. Built-in: slot 0 = Alarm Tone, slot 1 = Ring Tone. Custom: slots 101–103 (uploaded first).

Notes

  • Upload is a two-step API sequence discovered from the camera's web UI JavaScript: DO system/upload_usr_def_audio to acquire an upload URL, then multipart POST to it.
  • TTS workflow: generate TTS audio in HA → use the TTS proxy URL as url in vigicam.upload_audio → camera plays the spoken message through its speaker.
  • The camera converts uploaded audio internally; download size may differ from upload size.
  • get usr_def_audio_alarm/usr_def_audio returns -40106 on tested firmware (listing not supported), but upload and playback both work fine.

[0.3.6] - 2026-06-15

Added

  • Alarm Trigger button — press to fire the camera's configured alarm sound immediately (plays for 10 seconds then auto-stops; works regardless of whether the master Alarm switch is on — this is a manual test/trigger, not detection-linked).
  • Alarm Stop button — press to cancel an in-progress alarm trigger early.

Notes

  • Trigger uses msg_alarm.manual_msg_alarm {action: start/stop}, discovered in the camera's web UI JavaScript source. The 10-second countdown matches the camera's built-in behaviour; calling stop cancels it at any point.
  • Sound type (Alarm Tone / Ring Tone / Custom) is whatever is configured in the camera's web UI under Active Defence → Sound Alarm. Changing that setting changes what plays when the button is pressed.
  • Both buttons appear for any camera that reports alarm support (not PTZ-only).

[0.3.5] - 2026-06-15

Changed

  • Entity name grouping — related entities now sort together alphabetically in the HA device view:
    • Detection switches renamed with Detection prefix: Detection Motion, Detection Person, Detection Tamper, Detection Vehicle (were Motion Detection etc.)
    • Alarm switches renamed with Alarm prefix: Alarm Light, Alarm Sound (were Light Alarm, Sound Alarm); Alarm unchanged.
    • PTZ jog buttons renamed with PTZ prefix: PTZ Pan Left, PTZ Pan Right, PTZ Tilt Up, PTZ Tilt Down, PTZ Zoom In, PTZ Zoom Out (were Pan Left etc.)
    • Result: Controls section shows three clear clusters — Alarm (A), Detection (D), and PTZ (PT) — instead of being scattered across the alphabet.

Notes

  • Display names change; existing entity IDs in the HA registry are preserved for existing installations (HA stores entity IDs by unique_id, not by name). New installations will get entity IDs matching the new names.

[0.3.4] - 2026-06-15

Added

  • ONVIF PTZ services — three new HA services for PTZ cameras:
    • vigicam.ptz — continuous move in a direction (left/right/up/down/zoom_in/zoom_out) with configurable speed (0.0–1.0) and optional duration (auto-stops after N seconds). If no duration is given, call vigicam.ptz_stop to stop.
    • vigicam.ptz_stop — stop all camera movement immediately.
    • vigicam.goto_preset — move to a named preset by name (e.g. "Full Stable Yard"); cleaner alternative to select.select_option for automations with many presets.
  • PTZ direction buttons — six button entities per PTZ camera (Pan Left/Right, Tilt Up/Down, Zoom In/Out) that each jog the camera for 1 second then stop. Useful for dashboard control cards; for custom durations use the vigicam.ptz service.

Changed

  • Per-preset "Go to {name}" buttons removed — they never worked (bug fixed in 0.3.3) and don't scale to cameras with many presets. Use the PTZ Preset select entity or the new vigicam.goto_preset service for preset navigation.

Notes

  • ONVIF PTZ uses profile_1 (mainStream). All tested VIGI firmware versions use this token; it is not configurable per-camera currently.
  • Continuous move does not time out on the camera side — always call ptz_stop (or use the duration parameter) to prevent runaway movement.

[0.3.3] - 2026-06-15

Fixed

  • PTZ preset buttons and select never created — get_presets() read field names pan/tilt/zoom but the camera API returns position_pan/position_tilt/position_zoom. The IndexError was swallowed by the except clause, so get_presets() always returned [], causing has_ptz = False at startup. Preset buttons and the PTZ Preset select entity were therefore never registered for any camera. Now only reads id and name (neither buttons nor select use the position values).
  • Preset names shown URL-encoded — camera returns "Zoomed%20Stable%20View"; now decoded with urllib.parse.unquote() so entity names and select options show the correct human-readable name.

[0.3.2] - 2026-06-15

Added

  • IP Address diagnostic sensor — shows the camera's current IP address (from network.wan).
  • Connection Type diagnostic sensor — shows DHCP or Static, indicating whether the camera is using a dynamic or fixed IP address.

[0.3.1] - 2026-06-15

Added

  • Intrusion binary sensor (ONVIF IsIntrusion) — fires when intrusion zones are triggered (configured via VIGI app or web UI)
  • Line Crossing binary sensor (ONVIF IsLineCross) — fires when line crossing rules are triggered
  • Smart Detection binary sensor (ONVIF IsTPSmartEvent) — TP-Link catch-all topic covering vehicle, sound, loitering, abandoned object and scene change detection; these cannot be distinguished at the ONVIF level
  • Light Alarm switch — controls light_alarm_enabled in msg_alarm (the camera's flashing light response to detection events)
  • Sound Alarm switch — controls sound_alarm_enabled in msg_alarm

Fixed

  • ONVIF event dispatch was checking Value/IsMotion for all event types. VIGI cameras use per-topic boolean fields (IsTamper, IsPeople, IsIntrusion, IsLineCross, IsTPSmartEvent). Dispatch now scans the data dict for any Is* field.
  • ONVIF topic keyword map updated with verified topic names from GetEventProperties on both cameras; more specific matches (TPSmartEvent, CellMotion) now take priority over generic ones.

Notes

  • Smart detection zone/area configuration (line crossing paths, intrusion regions, etc.) is not exposed by the local JSON API (-40106). Use the VIGI app or camera web UI to configure rules; this integration reports when they fire.
  • Vehicle, sound, loitering, abandoned object, scene change detection all share the IsTPSmartEvent ONVIF topic — they cannot be separated at this level.

[0.3.0] - 2026-06-15

Added

  • Real-time motion/person/vehicle/tamper binary sensors via ONVIF pull-point subscription. On startup, a background task subscribes to the camera's ONVIF event service and long-polls for events every 8 seconds. Events fire immediately rather than waiting for the 30-second coordinator cycle.
  • Detected state auto-clears after 15 seconds if the camera does not send an explicit "active=false" event.
  • Subscription auto-renews 5 minutes before the 1-hour expiry. Re-subscribes on error after a 15-second backoff.

Notes

  • Auth: VIGI cameras require WS-Security PasswordDigest = SHA1(nonce + created + raw_password). Using SHA1(password) first — as the onvif-zeep-async library does — returns NotAuthorized on these cameras. No external ONVIF library needed.
  • Subscription address is returned on port 1024; pull calls go to that address, not the main service URL on port 80.
  • ONVIF topic → entity mapping uses keyword matching. Unknown topics are logged at DEBUG level — add to TOPIC_KEYWORD_MAP in onvif_events.py as needed.

[0.2.2] - 2026-06-15

Added

  • Loop Recording binary sensor — shows whether loop recording is active on the SD card. Read from loop_record_status in harddisk_manage response; cannot be set via the local API (all write attempts return -40101). Only created if the camera reports the field.

Fixed

  • SD Card Used % was reporting the camera's percent field which is inconsistent across firmware versions (camera 1 reports 0%, camera 2 reports 100%, both with full disks). Now calculated from total_space_accurate and free_space_accurate (byte values) for reliable results across all cameras.

[0.2.1] - 2026-06-15

Fixed

  • SD card sensors (used %, total, free, status) all showing unknown — the camera wraps disk data one level deeper than expected: hd_info[0] returns {"hd_info_1": {...}}, not the disk dict directly. get_storage() now unwraps that extra nesting level.
  • _parse_gb() failed to parse storage values — camera returns "116.8GB" and "0B" but the old parser only stripped G/M/K suffixes, not B, so float("116.8GB") raised ValueError and returned None. Now handles GB, MB, KB, B, and G suffixes.

[0.2.0] - 2026-06-15

Added

  • HACS custom repository support (hacs.json)
  • Tamper detection switch (InSight S245 and other cameras that support it)
  • Dynamic entity creation — entities are only registered if the camera reports support for that feature in its first API response; works across all camera models
  • SSL auto-detection in config flow — tries verified connection first (proper certs), falls back to unverified for self-signed certs; stores preference per camera
  • Camera thumbnail on device page via ffmpeg frame grab from RTSP stream
  • "Visit" link on device info page (configuration_url → camera web UI)

Fixed

  • All sensors showing unknown — ssl.create_default_context() made blocking load_default_certs() calls inside the async event loop (detected by HA's loop guard on Python 3.14+); refactored to use HA's async_get/create_clientsession helpers which handle SSL context creation correctly
  • Binary sensors were showing config-enabled state as live events — motion/person/ vehicle/tamper all read enabled: on/off and HA displayed this as "Detected", which was misleading; replaced with switches (which is the correct entity type for enable/disable controls)
  • Tamper was a binary sensor showing "Tampering: detected" at all times because tamper detection was enabled on the camera; moved to a switch
  • Firmware version showed full build string ("2.2.0 Build 250904 Rel.60109n") — now cleaned to the version number only ("2.2.0")
  • Model name included hardware revision ("VIGI C540V 1.0") — now stripped ("VIGI C540V")
  • Storage sensors (SD card) showing unknown — coordinator only caught VIGIError; broader exception handling added; hd_info handled as both list and dict (firmware version difference between camera models)

Notes

  • Binary sensor platform is now a placeholder — real-time motion/person/vehicle/tamper events require ONVIF event subscriptions, planned for a future release

[0.1.0] - 2026-06-15

Added

  • Initial release
  • Local HTTPS API client (api.py) with VIGI-specific auth (stok at top level, not nested under result as in Tapo cameras)
  • DataUpdateCoordinator polling all camera state every 30 seconds
  • Config flow — IP, username, password with live credential validation
  • camera — RTSP HD stream (stream1) with CameraEntityFeature.STREAM
  • switch — motion detection, person detection, vehicle detection, status LED, alarm, speaker mute, microphone mute
  • select — night vision mode (5 options), PTZ preset (PTZ cameras only)
  • sensor — SD card used %, total GB, free GB, status, firmware version
  • button — one per named PTZ preset (PTZ cameras only, auto-detected)
  • number — speaker volume, motion sensitivity, spotlight intensity
  • binary_sensor — placeholder for future ONVIF event support
  • MIT licence with attribution to Tapo Control, pytapo, and vigi_camera_lighting