Release notes straight from the repository. Installers and update scripts are attached to each release on GitHub →
Atmovio 5.3.20
latestmusic instrument and performer filters
Add visible multi-select instrument and ensemble filters, including piano and orchestra; combine them with a song title or search by instruments alone.
Search titles, mood tags and creators through Openverse's field queries so performer filters are honored; require every selected instrument and retain CC0 / CC BY restrictions.
Preserve the typed title when selecting instruments, clear outdated results after filter changes, and explain that instrument matches depend on the track's tags.
Atmovio 5.3.19
finished video library and visible progress
Show only available, completed main event videos and manual clips in the AI video gallery; keep intermediate safety parts out of the library and export the complete event as one file.
Add a prominent live panel for recording/export, Studio rendering and YouTube upload, including confirmed progress, estimated completion and recoverable errors.
Delete all saved AI videos and hidden parts in the background with confirmation and a visible result; preserve active event parts, videos used by processing/upload, failed deletions, raw recordings and YouTube outputs.
Generate thumbnails lazily and reject empty files as finished videos.
Atmovio 5.3.18
distinguish missing Studio videos from queued work
Show a creation queue only for queued Studio jobs, in both the YouTube gallery and video detail.
Explain unavailable finished files and offer disk checks or explicit recreation from the original clip.
Clarify that the creation queue runs on the Raspberry Pi and avoid promising upload readiness for unavailable files.
Atmovio 5.3.17
AI films belong to detections
Show evaluated film evidence within its detection instead of duplicating failed films in a separate gallery.
Keep only compact links to films without an evaluation above history, including incomplete films, with camera filtering.
Open the full detection evidence from error thumbnails too.
Atmovio 5.3.16
AI films scroll with the page
Remove the height limit and nested scrollbar from AI film cards on every screen size.
Atmovio 5.3.15
mobile layouts and trim track
Use one aligned trim track with independent accessible handles and concise time values.
Stack Studio and detection editor columns on narrow screens; keep the mobile save controls in normal flow.
Remove nested scrolling from mobile AI film cards and align video actions in two columns.
Atmovio 5.3.14
non-destructive Studio trimming
Select a source range with two draggable handles or precise offsets, preview it without creating a file, and see the selected and accelerated duration.
Persist the trim with the Studio job and apply it before speed, intro, overlays and music during rendering; original AI clips remain intact.
Validate ranges against the source file and warn that existing AI metadata describes the original full clip.
Atmovio 5.3.13
descriptive timelapse titles
Preserve the observed phenomenon in timelapse-only exports and Studio evidence instead of replacing it with a generic label.
Ask AI metadata generation for a concrete title based on observed clouds and development.
Atmovio 5.3.12
notify after the complete AI video
Detection notifications with automatic video enabled wait for the playable final main export.
Intermediate safety parts and manually created clips do not release the notification.
Export failure alerts remain separate from successful video notifications.
Atmovio 5.3.11
local system history and charts
Add System → Graphs for HDD usage, CPU temperature/utilization, RAM, system load and AI activity, with shared date ranges and Today/Yesterday/3/7/30-day presets.
Collect once a minute in a bounded subprocess; store history alongside snapshots on the external disk and skip collection when storage is unavailable.
Keep 30 days by default, configurable from 1 to 365 days; downsample long ranges and leave missing measurements blank.
Bundle the history collector with installers and preserve it in update backups.
Atmovio 5.3.10
grouped event videos
Show linked YouTube video creation, queue, readiness, upload and failure states on each source AI video, with direct Studio links and automatic overview refresh during active work.
Keep intermediate event parts together in a separate detail instead of individual AI video gallery cards; link them from the final video or an unfinished event group.
Persist the final export's event association, including migration of existing exact export/job matches.
Deleting the final export also deletes its parts, with a group-wide check for active Studio/YouTube work and retained records when Frigate does not acknowledge deletion.
Atmovio 5.3.9
music preview and Openverse filters
Preview selected library music directly in the video form; pause other music previews and add Openverse license, creator and file format filters.
Atmovio 5.3.8
storage stability, export recovery and log copying
Preserve managed recording mode during upgrades so restarting the guard does not unnecessarily recreate Frigate. Allow up to 30 seconds for a bounded write probe, while still reacting to explicit write failures and mount loss.
Offer recovery of missing exports using their original full range; check recorder/source availability before retrying and protect ready or actively creating files.
Keep full and intermediate exports pending through recorder outages, retaining their original time ranges; recover legacy connectivity failures on upgrade.
Treat failed recording API requests as unavailable, not proof of missing recordings.
Keep automatic file creation and interrupted exports visible in the AI video overview until a saved file is confirmed.
Copy hidden logs and input values using a selected textarea on HTTP; offer manual copying when the browser rejects clipboard access.
Atmovio 5.3.7
compact dashboard for any camera count
Show all dashboard cameras, removing the previous four-camera limit.
Adapt the camera grid to the available width and cap individual cards at 420 px, including single-camera setups.
Reduce card spacing and keep AI summaries compact, with descriptions available in detection details.
Verify dashboard rendering for 1, 2, 3, 4, 5 and 8 cameras.
Atmovio 5.3.6
dashboard and recording workflow
AI videa rozlišují průběžné části, pevnou délku a souvislé video události; u typu je výrazně uvedena délka zdrojového záznamu.
Give camera previews the full dashboard width with two spacious 16:9 cards per row and up to four cameras.
Move recent AI detections below the cameras in a compact two-column overview; stack both sections on small screens.
Add camera OD–DO selection, full-file preparation, visible preview speed controls and a two-hour preview limit consistent with the form.
Optionally sample nine frames from the actual exported video in a background queue and save editable titles and descriptions with the AI video; Studio is a separate optional next step.
Refresh YouTube upload progress independently every three seconds, confirm smaller chunks, and remove unmeasured countdown estimates.
Save intermediate event parts without triggering automatic Studio/YouTube jobs; keep the full-event export and send detection notifications only after a playable part is confirmed.
Keep publication explicit; persist metadata failures/retries and do not overwrite existing Studio edits.
Atmovio 5.3.5
visible AI video progress
Show recording wait, export progress, saved files and failures directly above detection details and in detection history.
Highlight pending videos with dated deadlines; distinguish export scheduling from completion.
Preserve links from continuing AI detections to their shared video, including fixed-length clips.
Refresh detail status without reloading the page or resetting edited video ranges.
Atmovio 5.3.4
grouped sky and timelapse phenomena
Add 23 cloud, optical, night-sky and temporal categories in compact groups, including custom camera rules.
Preserve existing selections; new installations include the extended catalog by default.
Require time-series evidence for motion categories and reject those IDs from single-image responses.
Ask for timelapse reasoning and observed frame bounds; distinguish camera artifacts from weather.
Atmovio 5.3.3
timelapse rule in AI detection
Move the timelapse trigger and score threshold into AI detection → Cameras and phenomena.
Preserve existing values and prevent video-settings saves from changing the detection rule.
Atmovio 5.3.2
consistent AI navigation and timelapse triggers
Keep AI detection and video settings connected with shared navigation; strengthen primary and settings menus.
Add an opt-in, independent timelapse score threshold for automatic clips from completed films.
Ask AI to assess observed cloud motion and development separately from single-scene interest.
compact administration and redesigned system navigation
Shorten administration labels and page introductions; move optional explanations into accessible disclosures across cameras, AI, storage, live view, videos, studio, system and diagnostics. Keep critical warnings visible.
Replace widely spaced internal tabs with a compact sidebar, clear selected page and dedicated content heading.
Share the navigation across overview, access, integration, maintenance and update pages; use a compact mobile grid.
Put the primary Atmovio password form first and constrain its width for readable, consistent settings.
Simplify AI timing into interval, frame capacity and editable auto-calculated duration, alongside a live contact-sheet layout, pixel dimensions and effective evaluation time. Recommended values fill only the draft.
Atmovio 5.3
clearer administration and bounded AI evidence
Separate system overview, passwords, API integration and maintenance into dedicated pages; preserve existing actions.
Correct mounted-disk messaging during recorder transitions. Never offer formatting merely because usage statistics are unavailable for a mounted recording disk. Log storage transitions and expose actual probe failures and timeouts.
Poll recorder readiness and restart acknowledgement; refresh recovered status pages without discarding edited forms. Failed camera images recover automatically and show a single illustrated placeholder. Failed video cards explain causes.
Show diagnostic entries newest first while preserving multiline details. Improve file inputs and laptop action layouts.
Review missing optional music, intro, title, description and enabled-but-empty overlay text before studio creation.
Configurable maximum frames per completed AI film. The 5-minute / 40-minute / 9-frame preset keeps evidence compact; films end at the earlier of the duration or frame limit, with the boundary snapshot starting the next film. Contact sheets with up to 9 frames use 720×405 photo areas; larger sets retain 480×270. Existing films finish unchanged. The compatibility default is 36 frames; long, dense future films can now close earlier to respect that limit.
Escape version and log JSON in the updater's HTML attribute so progress tracking initializes correctly.
Show completion and the running version at the top without a manual refresh. Confirm success only after the installer exits with its HTTP health-check success marker; expose failure and unconfirmed outcomes honestly.
Recover polling across restarts with request timeouts, a prolonged-outage message and an explicit login link when the session expires. Recover the last result after reopening the page.
Atmovio 5.2
completed AI films and continuous event clips
Separate snapshot collection from AI evaluation: completed-film mode is the new default for enabled filmstrips. Existing sample intervals and film durations remain; new installations default to 5-minute samples / 60-minute films. The previous rolling-window mode remains selectable in Settings → AI → Timing and AI film.
Persist per-camera films and samples across restarts. Collect the next film while background inference processes the previous one. Daily AI limits pause inference, not collection; provider failures retry twice and remain visible.
Show ongoing films in AI detections, chronological snapshot playback, exact timestamps, selected AI frames and the actual contact sheet sent to the provider. At most 36 evenly spaced frames are submitted; all collected frames remain viewable.
Evaluate the whole film, including events that ended before its last frame. Validate model frame indices and pad event boundaries with neighbouring samples. Missing boundary evidence falls back to the full observed film.
Create interesting-film clips independently of email delivery and notification cooldown. Join continuing matching events across adjacent films; finalize when the event ends, on continuation timeout, or in bounded six-hour parts. Keep accumulated AI evidence with the exported source for subsequent title/description generation.
Optional fixed-length source clips (1–120 minutes, including the before-margin) for both AI modes. Exact device-local date/time OD–DO controls preview or create a corrected clip; a future end queues creation. Corrections preserve existing videos. Invalid/reversed/overlong and ambiguous DST ranges are rejected.
Manual clip playback/export uses the observed event range for completed films. Legacy detection and clip behavior remains available. Night/outage partial films are explicit; historical unsaved frames cannot be reconstructed.
Plain-language illustrated guide inside AI/video settings, with a non-saving timing simulator and a live explanation beside the real timing fields. Covers overlapping events, fixed clip length, OD–DO and YouTube.
Compact desktop header with a single horizontal navigation row, quieter service status and a separate account/tools menu. Original logo retained; mobile navigation and account dropdown work independently.
Consistent form controls, labels, spacing, focus states, tabs and save bars across the administration. Camera, AI-provider, AI-video and YouTube-default forms now use explanatory sections with related fields aligned together.
Accessible label associations for existing forms, labelled navigation controls, light/dark themes and responsive layouts. Field names, save actions and recording/AI/video processing behavior remain unchanged.
Atmovio 5.0
clearer administration and video metadata from AI filmstrips
Horizontal navigation: Dashboard · Živý přenos · AI detekce · AI videa · YouTube videa · Nastavení. The dashboard summarizes cameras, recent evaluations, source clips, YouTube outputs, storage and system health. Quieter light and dark themes, responsive layout, original Atmovio logo retained.
Separate source AI clips and processed YouTube videos, including processing/upload states and failed automatic exports. Settings have a persistent submenu; YouTube settings are grouped into video defaults, intro/music and channel connection. AI-video settings save independently of the detector configuration, camera selection and custom rules.
AI detections show all evaluations by default. Combine camera, phenomenon, score, date and alert/error filters. Skipped dark/unchanged frames remain aggregate counters, not fictitious image history.
Filmstrip evaluations retain their time range and an observed evolution description. Both title and description can be proposed from stored visual evidence and the clip interval. A preview requires explicit application before saving; automatic studio videos use both generated fields, falling back to templates together if the AI request fails. Evolution outside the exported clip is excluded from the metadata prompt. This is analysis of sampled frames, not the MP4.
Exported clips retain their AI evidence after the original detection is deleted. Existing 4.7 data is migrated additively. Camera display-name edits preserve stable camera IDs and references. Deleting a detection removes its filmstrip too. Retention skips videos being processed/uploaded; active source clips cannot be deleted underneath a worker. Intentionally deleted automatic studio outputs are not silently recreated; explicit recreation remains available.
Offline regression checks cover settings isolation, authentication/CSRF, all page states, filmstrip evidence, video metadata, retention, camera rename and migration. Generated install/update scripts and tag-based releases retained.
Atmovio 4.7
the AI sees the film, not a photo
Filmstrip evaluation: instead of one snapshot the AI now gets one picture made of the current frame (large, "teď") plus a strip of older frames from the same camera with time offsets (default 6 frames over the last 60 minutes, Nastavení → AI → Kdy se dívat → "Filmový pás"). It still is one API call per check, so the daily limit is unchanged. Atmovio collects a small thumbnail from every check (also skipped ones) to build the strip; after a restart the strip fills up within an hour.
Atmovio 4.6.6
delete everything about a video, any time
"Smazat z RPi" moved to the "Co dál" box and works whether or not the video went to YouTube: the studio video (always), the source clip and the AI detection with its snapshot are each a switch – tick all three and nothing about the video remains on the Pi. Blocked only while an upload is running.
Atmovio 4.6.5
the countdown is actually visible
The "vytváří se… / zbývá asi … · hotovo asi v …" text in the rendering box was dark-on-dark in the light theme. It is now light, large and shows a ☕ when the wait is over three minutes.
Atmovio 4.6.4
capital letter at the start of titles
Templates for title, description, intro and overlay text: the first letter of every line is capitalised, so a template starting with {jev} ("výrazné červánky – Březolupy …") no longer starts lowercase.
Atmovio 4.6.3
description always reaches YouTube
If a video has an empty description at upload time (videos touched by the 4.6.0 bug), it is rebuilt from the description template, the AI text and the music credit before uploading. Angle brackets, which YouTube rejects, are replaced. The log now records the title and description length of every upload.
Atmovio 4.6.2
titles that attract viewers
✨ Navrhnout titulek (AI) next to the title field (new video and finished video): the configured AI provider proposes five catchy Czech YouTube titles (≤ 70 characters, no clickbait) from the camera, date, phenomenon and the AI description – click one to use it. Automatic videos get an AI title too (switch in Nastavení → Video studio; the template stays as fallback). Default template changed to {jev} nad obcí {kamera} · časosběr {datum}.
Atmovio 4.6.1
upload keeps title, time estimates, clean-up after upload
Fix: uploading to YouTube wiped the video's title and description (the upload form re-sent empty values). The upload now uses the saved title/description; an empty title falls back to the clip name.
When will it be done: rendering and uploading show a live countdown ("zbývá asi 2 min") and the clock time ("hotovo asi v 17:05" / "na YouTube asi v 17:07") – also while a video waits in the queue (including videos ahead of it) and in the Videa list.
Clean up after upload: once a video is on YouTube, the video page offers "Uklidit z RPi" – delete the studio video, the source clip and optionally the AI detection with its snapshot, freeing disk space.
Atmovio 4.6
YouTube upload
YouTube: a finished studio video is uploaded to your channel from the video page (visibility, playlist or a new playlist, title and description taken from the page) with a progress bar and the link when done; or automatically for alert videos (Nastavení → Video studio → YouTube). Videos in the Videa list show a ▶ YouTube badge with the link.
Linking for humans: a step-by-step guide inside Atmovio (Google Cloud project → enable YouTube Data API → consent screen → "TVs and Limited Input devices" client), then *Propojit YouTube* shows a short code you type at google.com/device – no redirect URLs, works from any browser, brand accounts supported. Playlists are loaded automatically; "Propojit znovu / jiný kanál" switches channels. Full guide in docs/youtube.md.
Resumable upload in 8 MB chunks with retries; quota and token errors are explained in plain words (about 6 uploads a day; publish the consent screen so the link does not expire after 7 days).
Atmovio 4.5.5
readable overlay text, Openverse check
Text in the picture sits on a translucent dark strip, so it is readable over a bright sky as well as over the fade-out at the end.
Nastavení → Video studio: "Vyzkoušet hledání" button next to the Openverse registration shows whether the music search works with the current registration.
Atmovio 4.5.4
fades and finer speed steps
Fade from black / fade to black: the sped-up recording now brightens from black at the start (1 s) and darkens at the end (2 s); the music fade-out is never shorter than the picture fade. Both lengths are in Nastavení → Video studio → "Obraz – rozjasnění a ztmavení". The intro is not affected.
Speed in steps of 10×: a slider with −10 / +10 buttons from 10× to 240× (was a fixed set of six values), on the new-video page and for the default speed in settings.
Atmovio 4.5.3
music for every video, intro sound, time estimate
Find music per video: in Studio → Hudba, "Najít hudbu k tomuto videu" searches Openverse (Creative Commons music, only CC0 / CC BY so YouTube never complains), with quick moods (klidná, klavír, západ slunce, epická…), a player for every result and Použít – the track is downloaded to the Pi and selected for this video. The author credit is added to the video description automatically. Optional one-click registration at Openverse (e-mail + confirmation link) in Nastavení → Video studio lifts the small anonymous search limit.
Intro keeps its sound: an intro video with audio is no longer silent; the music starts after the intro and fades in (intro sound fades out over its last 0.8 s). Intro without music → sound during the intro, silence after.
How long will it take: the new-video page estimates the rendering time for the chosen speed, and the progress page shows "zbývá asi …" computed from the real encoding pace.
Atmovio 4.5.2
videos no longer lost on updates
Root cause of "zaseklo se – Frigate export nedokončil": every Atmovio update (and every restart of the storage guard) re-created the Frigate container, which killed exports that were being cut at that moment – with 1.5-hour clips that happened a lot. The guard now adopts the running Frigate instead of recreating it when the disk mode has not changed, so updates no longer interrupt recording or exports.
Automatic recovery: an export that Frigate did start but was restarted underneath (power cut, manual restart) is detected (Frigate's uptime is shorter than the export's age) and re-submitted automatically, up to twice, while the recording still exists. Only then does the video show "Vytvořit znovu".
Atmovio 4.5.1
see the speed before rendering
Studio: "Ukázat, jak bude video rychlé" plays the source clip right on the page at the chosen speed (10–240×, switch speeds while it plays, loops until closed) so you can pick the right speed by eye before anything renders. Above 16× the preview steps through the clip, the finished video is smooth 30 fps.
Atmovio 4.5
Video studio
Video studio (Videa → ⏩ Studio, Nastavení → Video studio): a finished clip becomes a shareable time-lapse. Speed 10× / 20× / 30× / 60× / 120× / 240× (the page tells you how long the result will be), optional intro (an MP4 up to 15 s, or just a logo image that gets the camera name and date), text in the picture from a template ({kamera} · {datum} · {rychlost}×), background music with a smooth fade in and fade out (looped when shorter than the video), and an editable title and description pre-filled from the AI text.
Preview first: the video renders in the background (progress shown, one at a time so recording is not disturbed); you watch it, edit the title/description, download the MP4. Nothing is uploaded anywhere yet – YouTube upload with playlists arrives in 4.6, the "Nahrát na YouTube" button already shows where it will be.
Automatic mode: with "Automatika" on, every automatic alert video is turned into a studio video with the default settings as soon as Frigate finishes cutting it.
Studio videos live on the recordings disk (atmovio/studio), are listed at the top of Videa and are deleted with the same retention as clips. Uploads of intro/music are stored in the Atmovio data folder.
Installer/updater add fonts-dejavu-core (needed for text in the picture).
Atmovio 4.4.4
AI outages no longer swallow interesting skies
Retries on provider hiccups: a 503/529 "overloaded", a timeout or a dropped connection from the AI provider is retried twice (after 5 s and 20 s) with the same snapshot before giving up; the request timeout is 60 s instead of 90 s.
Quick re-check after a failure: when the AI still fails, the camera is checked again after 2 minutes (up to 3×) instead of waiting for the full interval – a shower or a sunset is not lost to a minute-long Google outage.
Logs explain both cases in plain words; the history record says "zkusím znovu za 2 min".
Atmovio 4.4.3
last AI evaluation in the web heartbeat
Webhook heartbeat carries last_eval for every AI-watched camera – the most recent evaluation (score, phenomenon, description, time, whether it alerted), including evaluations below the threshold. A website receiving the heartbeat can show "what the AI just said" under the current snapshot, exactly like the Atmovio dashboard.
Atmovio 4.4.2
readable logs
Logs made readable: every Atmovio log line gets a level and a plain-language explanation – ✖ needs action, ⚠ worth a look, ◌ transient (hosting or AI provider hiccup that resolved itself), · normal. The summary counts only real problems; a hosting timeout is no longer an "error". "Surový text" toggle keeps the raw log.
Report a bug: when the log contains errors, a "Nahlásit chybu na GitHub" button opens a pre-filled issue with the error lines and version (nothing personal).
Logs: "Vymazat tento log" moved to the page header instead of dangling under the toolbar.
Release workflow updates an existing release instead of failing when a tag is force-pushed.
Atmovio 4.4.1
camera page cleanup, stuck exports
Camera page is a dashboard only – no forms; "Upravit kameru" opens the edit form on Nastavení → Kamery (/cameras/edit/<id>), AI is configured on one place in Nastavení → AI.
Stuck video exports (Frigate restarted mid-export, "vytváří se…" for days) are flagged after 2 hours with "Vytvořit znovu" (re-submits the same range) and delete buttons, on Videa and on the camera page. When the recording for that range is already gone (retention), the export says so and offers delete only.
Atmovio 4.4
AI settings for humans
AI settings rebuilt into five tabs: Who evaluates · Cameras and phenomena · When to look · Automatic video · Test and statistics. One "Save" bar saves everything; the page stays on the tab you were on.
Per-camera rules: every camera has its own "AI watches" switch, and either the default settings or its own threshold and own list of phenomena (e.g. a good west-facing camera alerts from 7/10, a poorer north one from 5/10). Automatic video is switched per camera too.
"Anything photogenic" – a separate switch (default on, also per camera): alert even when none of the selected phenomena matched but the AI scored the view above the threshold. "Jiná zajímavá obloha" renamed to "Fotogenická / zajímavá obloha".
Custom phenomena: add your own (name + description for the AI, e.g. "contrails"); the AI looks for them on every snapshot and they can be selected like the built-in ones.
History page has a button to the AI settings; score badges use the camera's own threshold.
Atmovio 4.3
history, disk and log polish
History page rebuilt: button filters (alerts / all / errors · camera · min. score), one row per evaluation grouped by day with the full AI text and a colour score badge, 30 rows per page with paging; delete actions moved to the page header.
Dashboard shows an available update in the Raspberry Pi tile and in the System box (also in /api/v1/status → update).
Sunrise/sunset are highlighted in the header strip, dawn/dusk muted.
S.M.A.R.T.: a disk is no longer listed twice after /dev/sdX names swap on reboot (keyed by serial/model, one row per role).
S.M.A.R.T.: a disk with ≥ 200 reallocated sectors is flagged as worn ("plan a replacement") even when the count is stable.
Webhook heartbeat: a hosting hiccup shorter than 3 minutes is no longer logged; longer outages log start and end with duration.
systemd unit files get mode 644 before daemon-reload – no more "world-inaccessible" warnings in the system log.
Atmovio 4.2
AI in focus
AI in focus: every camera card on the dashboard shows the latest AI evaluation of that camera with score, phenomenon and the model's description (also below the threshold), with links to the detail and the camera's history. The events feed is gone; "Recent sky alerts" (with the AI text) sits above the storage / AI / system boxes.
AI provider chooser: Gemini, Groq/OpenRouter, Claude, OpenAI and Ollama as cards with price, free limits, accuracy, where to get the key and recommended models (still editable by hand).
Detection page plays the exported clip when one exists (smooth, seekable); the raw Frigate segment is used only for a custom range, with an explanation why its time may jump.
Camera fps badge explains that 5 fps is the preview/AI stream; recording keeps the camera's full quality.
Atmovio 4.1
new admin design
Redesigned web UI. Horizontal menu in the header (Přehled · Kamery · Záznamy · Detekce a AI · Videa · Nastavení ▾ · Nástroje ▾), quick actions (live view, add camera), account menu with update notice and logout. No more sidebar.
Dark and light mode with a switch in the header (remembered in the browser; dark is the default).
New dashboard: hero strip with date/time and sun times, KPI tiles with coloured icons (cameras, storage, AI detections, alerts, Raspberry Pi), camera cards with live/settings buttons and an "add camera" card, recent events feed, storage ring, AI today with 7-day statistics, system tiles.
Unified cards, tables, forms and buttons across all pages; camera cards on the Kamery page got action buttons.
Help section links to the REST API / Home Assistant guide on atmovio.com.
Atmovio 4.0
renamed to Atmovio
SkyWatch is now Atmovio (the old name collides with existing products). Everything is renamed: repository VladimirVecera/atmovio, install dir /opt/nvr/atmovio, service atmovio.service, update-atmovio.sh, env ATMOVIO_DIR, new API keys start with at_ (old sw_ keys keep working).
The updater migrates an existing SkyWatch 3.x install in place (moves the directory, renames the service, database, log and snapshot folder, rewrites config.json); configuration, cameras, detections and videos are kept. Releases also ship update-skywatch.sh (same file, old name) so 3.2–3.4 installs can update from the web UI.
Webhook heartbeat sends atmovio_url and, for older receivers, skywatch_url with the same value.
New logo.
Atmovio 3.4
Systém page: wider cards; the API card explains where the outgoing webhook (data + alerts to your website) is configured.
Docs: hardware page (supported Pi, parts, disk sizing), README hardware and AI sections, full REST API field reference and the REST-vs-webhook overview.
Atmovio 3.3
Webhook heartbeat carries api: a snapshot of everything the REST API returns (status, cameras, notified detections, videos, events – identical fields) plus thumbnails of new detections/videos, so a website that cannot reach the Pi can show the same data. status.ai.threshold added to the API.
Atmovio 3.2
Update from the web UI: Atmovio checks GitHub Releases once a day (can be turned off), shows a banner when a newer version exists, and installs it on one click (Nastavení → Systém → Aktualizace) – downloads update-atmovio.sh from the release, verifies SHA-256 against SHA256SUMS, runs it as a separate systemd unit and shows progress/result; the existing automatic rollback applies.
GET /api/v1/status reports update.latest / update.available.
Release workflow: pushing a v* tag builds a GitHub release with install.sh, update-atmovio.sh and SHA256SUMS.
Atmovio 3.1
Open-source release: generic webhook (any website), example PHP receiver, read-only REST API with API keys, documentation in English and Czech.
Per-camera page: live view, status, AI toggles, detections, videos and stream settings in one place.
Cameras overview with large auto-refreshing snapshots; live MJPEG on demand (single camera or all).
Automatic video export after an alert (per camera, −N/+M minutes, optional 25× timelapse).
In-browser video player with speeds up to 120×; range requests for seeking.
AI watching from dawn to dusk (civil/nautical/astronomical twilight), dark-frame skipping, independent daily statistics (7-day table), skipped frames no longer stored.
History shows notified detections by default and marks exported videos.
Updater keeps only the last two backups and removes orphaned virtualenvs.
New logo, favicon fix, nicer date formats everywhere.
Atmovio 3.0
New UI (Pico CSS + Alpine.js, dark mode, mobile bottom bar), settings grouped under "Nastavení".
Detection detail with clip playback, neighbouring alerts and clip export; videos page.
SMART disk monitoring with trend-based warnings; storage guard; VPN split-tunnel normalisation.
Camera discovery verifies real frames from every stream; Frigate reads cameras directly.
Internal versions before the public release.
Installed Atmovio checks for new releases daily and installs them on one click (Nastavení → Systém → Aktualizace).
