Agent DVR

VersionUpdate InformationDate
8.1.7.0New

Security update warning: admins now get a warning when they connect, and a “Security update required” row at the top of the server menu, if their version of Agent has a known security issue. Users without a subscription who click Update are sent to the download page.

Changed

Remember me on the local web UI no longer stores your password in the browser. Agent now gives the browser a sign-in token instead:
Tokens expire after 30 days without use.
They stop working if the password is changed or the user is renamed or deleted.
Signing out revokes the token.
Passwords that were already saved are swapped for a token the next time the page loads.
The show-password button has been removed from the login fields.
Signed-out screen: after you sign out, Agent stays signed out instead of reopening the login form. The product name shows top left, with a login button top right.
“Login Failed” now only shows when saved credentials were actually rejected.
Account permissions are now shown as badges.
Lower memory use: about 58% fewer memory allocations overall. Overlay labels (AI boxes, zones, line counts) are no longer laid out again every frame, and many video, audio, AI and streaming paths now reuse their buffers.
WebRTC:
Playback no longer resamples every frame for each viewer, so it uses less CPU and looks sharper.
Encode buffers are sized to the frame (previously 4 MB at 1080p and 16 MB at 4K).
Audio is now encoded as mono.
Faster image resizing and motion tracking.
Dropped network connections are now logged as one line instead of a full stack trace.

Fixes

Fixed alerts going missing when the AI description provider failed (for example a bad API key or model, or a timeout). The push notification still went out, but the alert was not saved, didn’t show in the UI, and its alert actions and MQTT messages didn’t run. A keywords-only AI reply is also no longer billed twice.
Fixed “Object found” actions with a repeat timeout of 0 sending several push notifications a second. AI-triggered pushes are now limited to one every 5 seconds per action and device.
Fixed audio recognition being fed leftover audio from the previous chunk.
Fixed a possible crash when a camera stopped while it was writing pre-recorded video.
macOS: normal encoder activity is no longer logged as “Error receiving packet”.

Packages

SQLite 3.39.2 ? 3.54.0 (fixes CVE-2025-6965).
ModelContextProtocol 2.3.0, Microsoft.Extensions 10.0.13, AWSSDK.S3 4.0.104.6, FluentFTP 55.1.0, NAudio 3.3.0, PuppeteerSharp 25.15.0, SharpCompress 0.50.7, System.ServiceModel.Primitives 10.1.0 and DnsClient 1.8.0.
OpenCV native libraries rebuilt for all platforms.
10/23/2026
8.1.6.0New

Three new screensavers: Eye (the logo iris behind a camera lens, in your palette’s accent colour), Globe (a dotted Earth with day and night, city lights and connection arcs) and Scan (a flight over contour terrain with a LiDAR sweep and detections). They replace the previous six screensavers.

Changed

Screensavers now follow palette changes straight away, instead of keeping the previous colours.
If you had picked one of the old screensavers, it switches to Random.

Fixes

Fixed Agent closing when you used talk to play audio through the computer’s speakers with the output device set to Auto.
Talk errors on the server are now logged instead of stopping Agent.
10/21/2026
8.1.5.0Analytics

Much lower CPU use when nothing is moving. On cameras with the Simple or Tracking motion detector, Analytics now drops to one check a second when the detector sees no motion and everything it’s following is standing still, such as a parked car or people waiting in a queue. It keeps counting them, and goes back to full speed the moment anything moves. Previously a parked car in view kept a camera at full rate around the clock.
Warning when this can’t work: a camera’s Analytics settings now say when it has no Simple or Tracking motion detector, so Analytics can’t slow down and uses more CPU.
Edit Camera button in a camera’s Analytics settings, replacing the “Device enabled” switch, opens that camera’s full settings.
Deleting a camera from Analytics settings ? Edit Camera no longer drops you back into the settings of the camera you just deleted.
With the UI minimised, Analytics no longer leaves a gap at the top for the hidden menus.

Video decoding

With Limit Decoding, the camera selected in a layout’s main panel now gets full-rate decoding, as maximised and single cameras already did.

Edit Devices

Redesigned list: devices are grouped by location, with a filter box. Each row shows a status dot, a camera or mic icon, the paired mic underneath, and labels for what’s switched on (Schedule, Alerts, Detector, Analytics). On phones the labels move to their own line.

Fixes

Detection feedback now asks “did we get this right?” for small, distant detections too. Previously anything under 48 pixels was skipped, so far-off false alarms never got the prompt.
Clicking OK in Views (Server Menu) while on a screen other than Live, such as Analytics, threw an error; fixed. The view-cycling setting also no longer overwrites the floor plans’ own setting.
Danish: “Analytics” now reads “Analyse”.
10/21/2026
8.1.4.08.1.4.0   New

Age and gender demographics. Analytics can now report visitor age bands and gender from face detections. The overlay shows age and gender on detected faces in live view.

Limit Decoding. New option in video source settings - Advanced. Limit Decoding will decode only keyframes for the live view to lower CPU usage. Agent can now decode only the streams each camera actually needs at any moment. A new server setting in playback - Advanced ("Full Decode in Live View") forces full decode while a camera is open in live view. Limit Decoding will be bypassed if the camera is set to Encode when recording.

Changed

Age and gender model updated to v2.
The Analytics Lines editor now shows live video while you draw lines.
Emailed lines crossing reports are grouped by location.
Limit Decoding now warns when live view would require full decoding, rather than silently overriding the limit.

Fixes

Fixed a crash on Linux with older NVIDIA GPUs (Pascal and earlier, such as the GTX 10-series) where closing the first GPU encode session while other encode sessions were still running could crash the process. This is a known NVIDIA driver bug; Agent now works around it.

Fix for some MJPEG sources using the internal decoder switching off permanently after a server timeout instead of reconnecting.
10/20/2026
8.1.3.0New

Serbian (Latin script) language.
AI blur (objects, faces, number plates, loitering) now also covers the full-screen HD view and “Main and Max”. Before, it was only applied to the low resolution view.
Recording settings warn that raw recordings don’t include AI blur.
The Detector tab warns when a camera uses the old People or HAAR Objects detectors, and links to AI Object Recognition.

Changed

Updated to OpenCV 5.0 on all platforms. ARM builds get faster image processing.
Flip and rotate are much faster.
Detection feedback now counts every rejection. An object you keep rejecting (for example a stump detected as “dog”) can now block itself as intended.
If WebRTC is switched off in the browser (usually by a VPN or privacy extension), the web client now says so. It used to say the browser was too old.
Translation updates.

Fixes

Fixed a crash on Linux with older NVIDIA GPUs (Pascal and Maxwell, such as the GTX 10-series and Tesla P100) using NVENC. On newer drivers, recording could crash about 15 minutes after startup while other GPU encodes were running.
Fixed colour fringing on horizontal edges when a camera is flipped horizontally.
Fixed raw recordings playing back with the wrong rotation after the camera’s rotation was changed.
Fixed flip and rotation being lost in raw recordings when video wasn’t the first stream.
Fixed occasional alert frame loading, which failed with the previous OpenCV build.
Fixed the “Name” field label, which showed the language’s own name in some languages (for example “Deutsch” in German).
10/19/2026
8.1.2.0

(Beta):

8.1.2.0 New

Detector tab: a warning now shows when the People or HAAR Objects detector is selected. These detectors use older, less accurate technology, and the warning links to AI Object Recognition instead.
MCP: a new get_analytics_occupancy tool tells you how full a place got over a period. It gives the peak count and when it happened, the average, the count at the start and end, and how long and when the count was at or above capacity (or another number). It can break this down hour by hour or day by day, so it can answer questions like “how full did we get on Saturday?”.
Trip Wires: the live overlay now draws one arrow across each wire, the same as the editor. Named wires show their name beside the wire.

Changed

Alerts: a detection later in the same alert interval now adds its labels to the alert. For example, a person seen after a car now appears when you filter alerts by Person or search for it, not only on the recording.
Recordings: the label that started a recording is now marked at the moment it was detected, so playback can jump to it like any other tag.
Web UI: when WebRTC is turned off in the browser, the error now says so and suggests that a VPN or privacy extension may be blocking it. It used to tell you to use a newer browser.
MCP: every tool parameter now has a defined type. query_api and control_api only accept their arguments as a JSON object; the query-string form is gone. Tools that change rules or Analytics now name the API command they run and link to the API reference.

Fixes

Alerts raised from a snapshot (for example, a Snapshot Taken action set to Trigger Alert) used the current live frame instead of the snapshot image.
Plugins: alerts raised by a plugin lost their message and image, and dropped the plugin’s AI results. All three are now kept.

Packages

Emgu CV 4.12 to 5.0 (OpenCV 5.0) on every platform. This adds libjpeg-turbo 3.1.2, and KleidiCV on ARM64. 32-bit ARM Linux no longer needs the system’s libz.so.1.
.NET libraries 10.0.12 to 10.0.13. Also AWSSDK.S3 4.0.104.4, LibVLCSharp 3.10.2, Google.Apis.Drive.v3 1.77.0, NAudio 3.2.0, PuppeteerSharp 25.14.0 and SharpCompress 0.50.5.
10/16/2026
8.1.1.08.1.1.0 New

Analytics: people and vehicle counting, for Standard and Professional licenses. Open it from the main menu (shortcut 8). It runs its own detector and tracker on each camera, separate from object recognition, with its own counting zones and lines. The first time, these are copied from the motion detector's.
Analytics dashboards: shared by everyone on the server. Reports include line crossings, zones, occupancy, time in zone (average and median), busy times, queue length, live view, speed, heatmaps and object paths over live video, and comparison with up to three earlier periods. Each report can have its own period, and can be shown full screen, printed or saved as CSV.
Analytics counters: occupancy counts across any number of cameras' lines, with an optional capacity, a full-screen door display, and rules that call a URL when the count goes above or below a number.
Analytics speed: two counting lines a known distance apart measure each object's speed.
Analytics alerts: objects or a queue in a zone, time in a zone, a line crossed, following closely, occupancy and speed. An alert can notify admins, send an email or call a URL. Each alert has hours when it is active and a Test button.
Analytics emailed reports: a dashboard's figures daily or weekly, with heatmaps and paths drawn over each camera's reference image.
Analytics sales and conversion: enter or import (CSV / JSON) sales figures by the day or hour. The Conversion report sets them against visitors to give sales per visitor and the conversion rate.
Analytics is for admin users only and has its own retention setting. It can be hidden in Server Settings - User Interface - Sections. Other licenses see a page describing it.
MCP: assistant tools for analytics counts, live view, time in zone, speeds, alerts, conversion and recording sales figures, plus creating, changing and deleting counters, speed measurements and alerts. New counting_report prompt.
Zones and trip wires can be named. The names show in the zone and line editors, and on the zone pickers in actions and detector dialogs.
API: setconfidence sets the minimum confidence on a camera's object, face or LPR tab. Profiles now keep each camera's object detection confidence.

Changed

Trip Wires detector: each wire has one arrow across it, and the direction options read Inward / Outward instead of Left / Right. Saved settings, MQTT and API field names are unchanged.
Dialogs with up to four sections show them as tabs. The section menu has previous / next arrows.
Copy Settings: a new Push Alerts section (push on/off, template, video and audio options). Alerts also copies the AI description, video preview and reset settings.
The alert dialog remembers whether the photo or the video is shown large.
Logs: camera logins, passwords, RTMP stream keys and alert action URL tokens (Telegram, Discord, Slack, IFTTT) are no longer written to the log.

Fixes

Storage: the daily archive clean could delete a whole drive's files when the archive path looked like E:\{NAME}. It could also delete live and locked recordings when the archive folder contained a storage location. These paths are now skipped, and a storage location whose archive overlaps a media folder can't be saved. A missing archive folder no longer stops the other locations being cleaned.
HTTP / MJPEG cameras read through FFmpeg froze on the last frame when the camera closed the connection. They now reconnect.
Deleting a device that was failing to connect kept it reconnecting until restart.
VLC sources: stopping a stalled stream could freeze enable, disable and reconnect for every device.
ONVIF imaging: settings the camera doesn't report show "Not Supported" instead of a list of blank entries. Saving no longer fails when the camera is disconnected, and IR cut filter changes are now sent.
SMTP: the embedded snapshot has a valid Content-Id.
Copy Settings: copying Timelapse now includes the filename template. LPR is listed once.
Photos: {CAMERA} in a photo filename dropped the photo and reset the template. It now works the same as {NAME}.
Microphone level: a silent or disabled microphone broke the level display.
10/15/2026
8.1.0.0New

Assistant Rules: quiet rules. An assistant can now set up a rule for something NOT happening: “let me know if there is no movement on the kitchen camera in the next 2 minutes”, “tell me if nobody has come to the door by 9am”. A rule fires once per quiet spell. It is not watching while its time window is closed, the rule is paused or the device is offline.
MCP and API: create_rule / createrule accept quiet_minutes and quiet_since.

Changed

SMTP: Embed Images is on by default again for new installs. Existing installs keep their saved setting.
Push-to-talk: a change to the Push To Talk setting reaches connected clients straight away, without a page reload.

Fixes

ONVIF detector: recording now follows motion until the camera reports it has stopped, or until the Event Timeout expires. Cameras that send a single on/off pair (Axis) stopped recording at the minimum record time.
SMTP: with Embed Images on, the snapshot is sent as a proper inline image. Some mail clients showed it twice, and strict ones (such as Evolution) showed nothing inline.
Push-to-talk: a quick tap no longer briefly switches the microphone on. It shows “Hold the button to talk” instead.
File Upload: uploads from the File Upload dialog were not saved, and the file picker only offered .wav. Audio uploads now offer .wav, .mp3 and .ogg.
Faces: photos uploaded for face recognition were sent as PNG instead of JPEG, several times larger than intended.
Web UI: a timestamped tag stays on the recording after reopening the dialog.
10/9/2026
8.0.9.0

(Beta):

- AI models: new installs get new built-in models: a Standard object detector, Standard
(SFace) face recognition with a much faster face finder, and a new licence plate finder
that reads tilted plates better. Existing installs keep the models they already have.
The plate model can be chosen in Server Settings - AI Models - License Plate Recognition.
- Assistant Rules: assistants can now set a camera's alert mode, motion detector and when
AI runs, so "only alert when someone is there" works end to end. They can also change
sensitivity (low/medium/high, lower/higher), switch alerts or the detector on and off on a
timetable (quiet hours), speak or play a sound through a camera straight away, rename a
device or set its location, and create rules that run every N minutes.
- RTMP: new "Audio from all cameras" option on an RTMP server. A streamed view can now
carry the mixed sound of every camera in it instead of the first camera only. Off by
default.
- Audit: delete and archive requests are written to sessionlog.txt with the user and
session. Requests through the API are logged as API.
- Licences: every package now includes LICENSE.txt and a Licenses folder with third-party
notices.

Changed
- Performance: less CPU for AI inference on the CPU, live view, hardware decoding,
digital zoom and grid layouts, with sharper tiles.
- Text overlays: new text renderer. Characters the chosen font lacks come from other
installed fonts, and marks above or below letters are no longer clipped. Plugins that use
ImageSharp must now bundle it.
- Plugins: only signed packages from files.ispyconnect.com are installed. A failed
download no longer removes the installed plugin.
- Audio: the WASAPI and MP3 source types are retired. Existing devices migrate to Local
Device (re-select the device) and Network Microphone.
- Source dialogs: the video and audio source dialogs now show the source type as the
dialog title.
- Face recognition: faces smaller than 40 pixels are never named.
- TURN server: updated to fix a remote crash (CVE-2026-54909). It now restarts
automatically if it exits.
- Codec: a camera's Default codec option shows which server default it uses.
- SMTP: Embed Images is off by default on new installs.
- Scripts: output on stderr is logged as an error only when the script fails.

Fixes
- Windows: GPU AI on new installs has been broken since 7.5.0 because DirectML.dll was
missing from the package. Some models ran on CPU and others failed.
- Intel QSV: fell back to software encoding on recent Intel drivers (N150, UHD 630).
- TLS: rtmps streaming and https/rtsps/HLS sources failed with "I/O error". Connection
errors now include FFmpeg's reason.
- Codec: a camera set to Default always recorded H264/AAC .mp4 and ignored a software
H265, VP8 or VP9 server default.
- Video files over HTTP: an .mp4 or on-demand HLS playlist on a web server was treated as
a live stream and raced to the end in seconds. It now plays at normal speed, like a local
file.
- All audio: VLC, NDI, WAV and iSpy Server sources sounded wrong in an all-audio view.
The mixer also kept running after the view closed, and a change to the system clock could
silence or repeat the mix.
- Live view: tool buttons that were on (All audio, Trace, PTZ, mute) looked off after
visiting another screen.
- Faces: adding several photos at once could store green or purple thumbnails and learn
from them. Switching face models no longer compares incompatible face data.
- Colour: desktop capture had a slight green tint, and full-range sources are now
converted correctly.
- TURN: the server was not restarted after a public IP change.
- Storage: moving a storage location or device folder inside itself copied it into
itself repeatedly.
- Motion tracking: false detections after the tracker was full or PTZ had moved, and a
memory leak on every detector settings change.
- Object detection: colour models were padded with the wrong colour.
- RTMP: false "not keeping up" warnings while reconnecting.
- VLC source: a possible crash on streams with an odd width.
- Timestamp: font changes now apply to static text straight away.
- Windows: sound playback follows a change of Audio Out Device.

Packages
- Removed: SixLabors.ImageSharp 3.1.12 and ImageSharp.Drawing 2.1.7. Overlay text now uses
AgentDVR.Fonts (our fork of SixLabors.Fonts 1.0.1, Apache-2.0) with our own rasteriser.
- Removed: NAudio.WinMM.
- NAudio.Core and NAudio.Wasapi 2.3.0 to 3.1.0.
- mDNS: Makaretu.Dns 2.0.1 and Makaretu.Dns.Multicast 0.27.0 replaced by the maintained
Makaretu.Dns.New 3.1.2 and Makaretu.Dns.Multicast.New 0.38.0.
- TURN server: pion/turn 4.1.4, pion/stun 3.1.7, pion/dtls 3.1.10, x/crypto 0.57.0, built
with Go 1.27.1.
- Added: DirectML.dll in the Windows package.
10/6/2026
8.0.8.08.0.8.0 New
- Video errors banner: Agent now keeps a count of video corruption on each camera (lost network
packets, smeared or broken frames) over the last 30 minutes. If the errors are frequent or
severe enough to see, a banner at the top of the camera's General settings tab shows the
count and what to do. On UDP it recommends Prefer TCP, because the usual cause is packet loss
over Wi-Fi. On TCP it points at the camera's Wi-Fi signal, bitrate or firmware. A summary is
written to the log every 5 minutes while errors are occurring.

Changed
- RTSP transport: once a camera or microphone has connected and delivered video or audio,
Agent uses only its saved transport (TCP, UDP or HTTP tunnelling). A camera that was slow to
answer once can no longer turn Prefer TCP into UDP. Auto Switch Transport now searches only
for a new source, or after you change the source or transport settings. It saves a different
transport only after that transport works on two connections, so cameras that are still
booting after a power cut are not switched to UDP permanently. With Auto Switch off, the
behaviour is unchanged.
- FFmpeg: if the GPL FFmpeg build you selected is not installed yet (for example after an
update to a new FFmpeg version) or cannot be loaded, Agent starts straight away on the
standard build that ships with it. It downloads the GPL build in the background and
restarts to use it when the download finishes. There is no longer an "Installing FFmpeg"
page at startup.
- Windows: every executable and DLL in the package is now code-signed, which helps
enterprise app-control policies.

Fixes
- Auto Switch Transport: a failed connection flipped the saved Prefer TCP setting every
time. This happened on microphones even after hours of working. After an outage the setting
could be left on either value.
- Installer: the Agent service is now stopped only after the new package has downloaded and
its signature has been checked. Before, a failed download or signature check left the
existing install stopped and not recording. If extraction fails, the service is started
again.
- Installer: the port check now opens a socket instead of sending an HTTP request. A port
used by another program (a web app returning 404, or a service that isn't a web server)
used to count as free, and Agent then moved to a random port.
- Windows: after an automatic update, Apps & Features (and winget) showed the old version
number.
- Windows: the Start menu shortcut opens the port Agent is actually using, even when Agent
had to move off a busy port.
- Network discovery: with no network connection, ONVIF discovery no longer logs an error on
every poll, and cameras found earlier are no longer removed from the list while offline.
9/24/2026
8.0.7.08.0.7.0 New
- Assistant Rules: tell a connected AI assistant (Claude, ChatGPT or any MCP client) what you
want in a sentence and it sets up a rule that runs on your server. "Tell me if a red car turns
up on the driveway in the next 6 hours", "if someone is at the side gate tonight, play a dog
barking", "if the bins aren't out on Sunday at 6pm, text me", "arm the system at 10pm every
night". Rules trigger on anything your cameras detect, can ask your AI provider a yes or no
question about the picture (a red car, someone carrying a parcel), follow a timetable including
sunrise and sunset, and can notify you, speak or play a sound through the camera, record, move
a PTZ camera or call a webhook. They can cover one camera, a group, a location or the server
itself. Manage them in Server Settings - Assistant Rules. Examples are in the user guide under
AI Assistants.

- MCP and API: new tools and commands for creating, listing, pausing and deleting rules, plus
setgroups to put a device in a group. See the API docs.

Fixes
- Schedule: settings no longer revert after a restart. Startup now applies only the schedule
entries that were missed while the server was down, instead of replaying the whole day.
- Digital PTZ: presets keep their zoom point across restarts, and now work on Virtual Cameras
dewarp.
- Scripts: scripts run as actions on Windows now log their output, errors and exit code.
- OneDrive: a refused upload now logs the reason instead of "Value cannot be null".
- Network discovery: the mDNS advertiser no longer restarts whenever an unrelated network
adapter changes state.
9/21/2026
8.0.6.0New
- Alerts: Merge Delay. New per-camera setting on the Alerts tab (seconds, 0 = off). With a Merge
Delay set, the first trigger opens a pending alert and every later trigger within the window
folds into it. When the window closes you get one alert, one push notification and one run of
your alert actions. It uses the best frame and carries all the tags, with recognised names and
plates first (e.g. "face,person,bicycle"). Recording still starts immediately and never waits
on the delay.

- Trusted Proxies: entries now accept CIDR ranges (e.g. 10.0.0.0/16) as well as single IPs, and
X-Forwarded-For hops with a :port suffix are handled. This fixes setups behind Azure
Application Gateway and similar proxies with dynamic backend IPs. Any parsing error now fails
closed (the proxy is treated as untrusted).

Fixes
- Recording: migrated iSpy configs could carry a stale frame rate value that forced a 1-frame
GOP. That produced all-intra recordings around 5x the normal size and stopped NVENC from
opening ("GOP Length should be greater than number of B frames + 1"). The recorder now sizes
the GOP from the Max Record Framerate only, and the GOP can never be shorter than 2 frames.
- AI: "Trigger Face Recognition", "Trigger LPR" and "Trigger Object Recognition" actions were
silently lost if they arrived inside the controller's interval or while a run was already in
progress. The trigger now stays queued until a run actually starts.
- API: /command/alert is no longer rejected with "Feature in use" while a Merge Delay window is
open. Its tag merges into the pending alert.
- Web UI: Escape closes the alerts list.
9/17/2026
8.0.5.0Fixed

HD Recordings on Windows hardware/MF encoders started sharp and turned progressively blocky.
Playback: "Resize File Playback" is now on by default for new installs and decodes at the size of your live view rather than the tile size. This fixes corrupted playback of high-resolution recordings on servers without a GPU and improves quality when a recording is maximised from a grid.
Local network discovery went silent after sleep/wake or a network change until Agent was restarted.

Changed

License and account details move out of config.json into a separate license.json that is never exported or included in backups. Existing installs migrate automatically.
Failed login log entries include the client IP and username so fail2ban/CrowdSec can key off Agent's log.
Web UI: new password requirements with a breach warning, and a faster user account page.
9/16/2026
8.0.4.0Fixed:
- RTSP cameras stuck on "Connecting" after a disconnect until deleted and re-added:
- Crash in the VLC decoder from a stop/dispose race on the media player.
- Windows installer: the port field showed junk characters and the "port in use" check failed against the retyped value, caused by a byte order mark in the port file Agent writes for the installer.
- Web UI: selecting recordings by dragging lost the selection when new recordings arrived mid-drag.
Changed
- applySchedule and applyScheduleToDevice now accept the same group and location targeting as the other device commands (group takes a name or a comma-separated list, case insensitive), and skip devices whose schedule is switched off, matching the automatic per-minute schedule run.
AI assistants (MCP)
- New get_workflow tool returns the built-in workflows (morning briefing, door report, wildlife report, camera health, day in review) as step-by-step text for assistants that do not support MCP prompts, such as ChatGPT.
- get_alerts now points assistants at the AI-written alert description (Describe Alerts) for cameras that alert without recording.
- get_activity_summary description clarified so assistants go on to fetch images for "show me" questions instead of stopping at the counts.
9/10/2026
8.0.3.0MCP server built in. AI assistants such as Claude Code and Claude Desktop can connect directly to your Agent at /mcp (LAN, no cloud). Query cameras, view live snapshots, browse recordings and alerts, get activity summaries and timeline sheets, arm/disarm, and control devices. Claude Desktop extension bundle available. Toggle under Server Settings > Local Server.
New API commands for integrations: getCapabilities (feature flags and a full command catalogue), recording frame sampling, contact sheets, activity summaries, and time-windowed alert queries.
WebRTC H.264 fix: profile and level now apply to all software encoders, including Media Foundation on Windows. Should fix iOS and iPad live video failing to play.
Plugins: fixed MissingMethodException crashes in plugin code on trimmed builds.
Trip wires: crossing counters drawn as plain numbers, fixing garbled labels in non-Latin locales.
Fixes: motion graph "Collection was modified" errors during recording, VLC source stop/dispose race, object tracker crash on frame size change, OneDrive token errors now report the real cause, start page selector showing "Live Video" twice.
9/9/2026
8.0.2.0## Agent DVR 8.0.2

- Trip wires: Left/Right follow the draw direction, arrowhead shown on the overlay, existing wires migrated
- Trip wires: path-based crossing detection catches fast objects and wires near the frame edge
- Trip wires: per-wire and total counters via {TRIPCOUNT}, {TRIPCOUNTLEFT}, {TRIPCOUNTRIGHT}, {TRIPJSON}, MQTT event and retained topics, Home Assistant sensors, resettripwires command and Reset Count button
- Trip wires: counts drawn as numbers on the side they count; updated editor UI
- AI: object, face, LPR and Ask AI share one frame snapshot and JPEG per frame
- AI: batched feedback embeddings, pooled tensor buffers, cached model lookups, keep-alive HTTP clients
- AI: LPR keeps per-plate confidence
- AI: tracked objects re-acquire on frame size change instead of failing until timeout
- Recording: raw writer on its own thread, slow disks no longer stall the camera pipeline
- Encoding: x264/OpenH264/x265 thread count from frame size
- Streaming: MJPEG skips encoding when nothing changed; pre-roll only kept where raw record is possible
- Performance: fewer per-frame copies, inline processing for small frames, scaler contexts rented per caller, idle frame pools trimmed
- Fixed AI attachments getting the alert image without detection boxes
- Storage: bulk delete and archive batched, one transaction per batch, seconds instead of minutes on slow disks
- OneDrive: fixed "unauthorized" on upload after successful authorisation
- OneDrive: token refresh errors show the real reason; fixed trimmed-build deserialisation failure
- Web UI: restore button hidden on pointer layouts
- Cleaner shutdown of devices
9/8/2026
8.0.1.0Workaround bug in firefox on Android preventing video playback
Ensure resize isn't defaulted-on when adding cameras
Add connection failure messaging
Fix wait race condition on ios app
9/4/2026
8.0.0.0Update translations
Fix rendering issues with long strings in some translations
Fix bugs with set/ append timestamp API
Fix bug with firing face recognized on unknown faces
Ensure changes to config made via API survive object reloads


9/3/2026
7.9.9.0Add Belarusian translation set
Show warning on AI panels if AI not supported due to old OS
MQTT watchdog to ensure MQTT stays alive
Shutdown watchdog to ensure restarts complete on locked files
Clone request bugfix for some ONVIF cameras
Sources update
Updates to licensing system (old licenses auto-update when online)
Strip pasted quotes from camera source URLs and paths.

Revert glibc bump for older linux support
Store timestamp append setting to survive restarts


9/2/2026
7.9.8.0Fix issue with saving AI Audio tags
Add option to display all detected AI audio tags in overlay
Add mask overlay to HD live video streams
Remove some obsolete code
Add mask warning on recording tab if saving raw
Add AutoSwitchTransport to audio handler


8/25/2026
7.9.7.0Fix issue with loading ffmpeg on ARM8/23/2026
7.9.6.0Fix issue with doubletake authentication
Improve error logging for platerecognizer
Fix arrow navigation toggling watched
Fix opendrive reset issue
Fix onedrive personal issue
Fix issue with discovery drowning NAT on podman
Add API command integration for plugins
8/23/2026
7.9.5.0

(Beta):

fix some ai translation issues
de-duplicate labels in alerts
add video preview toggle to alerts
add 400ms debounce to alert navigation
fix issue with mp4 preview generation using openH264
fix issue with overlay not drawing on alert images
8/21/2026
7.9.4.0

(Beta):

Embed ffmpeg 9.0.1 LGPL
Fix issues with VLC decoder
Support multi touch touchpad drag on timelines
Optimise recordings lookup

8/20/2026
7.9.3.0

(Beta):

Update to ffmpeg 9
Patch ffmpeg to handle cameras that don't list transports properly
ONVIF ws-auth fallbacks for bad camera clocks,
Allow getStatus for non-admin app users
Package bumps
Add auto reconnect to initial load failure on local
Persist zoom on restart
Override turn server address on localhost to loopback address
Log user activity to sessionlog.txt
8/18/2026
7.9.2.0Update download UI
Fix some missing icons
8/13/2026
7.9.1.0Add v380 detection and native talk setup
Update gemini AI support
Add notes field to addcamera API
Default Gemini model to v1beta
Show address in ONVIF device list
8/11/2026
7.9.0.0Move Describe option from Ask AI to Alerts tab
Send series of grabs to ask AI to improve description
Integrate AI description with mobile push
Harden AI prompt
Update minimum token spend for Gemini to improve responses
Notify about free Gemini quota
Track 429 responses from AI integration
Fix issues with Mac OS updater
Fix bug with filtered unread alerts causing the alert viewer to become unresponsive
8/9/2026
7.8.9.0Added video previews to alerts so you can see what triggered them at a glance in the alert viewer
Fixed remote viewing failing on routers that block loopback connections when using port forwarding
Add "all power" to live controls (need to enable in account menu - controls) - toggle power to all devices in the current view

8/7/2026
7.8.8.0Fix initial delay when editing devices on first load
Add HD on-demand mode (for low bandwidth installations - only connects to the HD endpoint when the camera is maximised or recording). This is in video source settings.
Minor UI updates


8/6/2026
7.8.7.0Discord webhooks, Home Assistant discovery
- Call URL action: Discord webhook support (payload_json/files[n] multipart
naming so image attachments work) and automatic application/json
content-type for JSON bodies (Slack, Mattermost, Google Chat, Teams)
- Home Assistant MQTT auto-discovery (HADiscovery flag, retained states,
alerton/off + detectoron/off commands)
- Move alert action processing off the camera frame-processing thread
8/5/2026
7.8.6.0Fix UI hang when GPU not responding
Updates to time machine UI
8/5/2026
7.8.5.0fix rendering issue with German
fix issue passing oauth groups to the web ui
standardise fill mode in playback settings
require admin for local API with protect API checked
various security related updates
fix issue with CPU usage under docker
8/4/2026
7.8.4.0Fix issue with bulk delete of files and photos when devices are minimized in the UI
Fix issue with high CPU detection on Docker
8/1/2026
7.8.3.0Dont allocate GPU decoder for HD streams unless decoding (memory optimisation)
Recycle request pool on bad ONVIF reconnects
Sanitize control characters to console
7/31/2026
7.8.2.0Fix issue with IPv6
Persist access key for reconnecting clients
7/30/2026
7.8.1.0UI updates (timeline)
Add ffmpeg system logging to startup
7/29/2026
7.8.0.0Fix issue with streamfile.cgi
Add range support to streamfile
Work around issue with some ONVIF event streams sending initialized in every notification packet.
7/28/2026
7.7.9.0V4L2 M2M hardware decoder support for ARM7/28/2026
7.7.8.0Fix reconnect exit on hung/ terminated connect attempt
UI theming updates
Add multiple drives to stats report
Update stats display in server info panel
Add ptz commands to API
Add delete files/ photos to API

7/28/2026
7.7.7.0

Fix issue with timelapse display on timeline
Show current profile in hover on lock icon
Fix issue with some (non-conforming) ONVIF device connections
Update audio visualizations
Fix issue with displaying alert via id in the url
Add 2 more palettes to UI settings
Update camera connecting animation




7/25/2026
7.7.5.0

(Beta):

Major UI refresh
Rewrite ONVIF support
Add command support for new apps
Cap WAL file size (database)
Webrtc updates for pacing and streaming
Add default font
Add global timestamp config to settings
Fix bug with string/ numeric ONVIF PTZ presets
Fix bug with cloud auth (google drive)
Zoomable and Pannable floorplans
Move file management serverside (no limit on number of files in the browser/ management screens)
Add save-over to ONVIF PTZ controller
Theme updates (moved to palettes in account menu - 4 new themes)

7/24/2026
7.7.4.0

UI updates
Fix some bugs with PTZ Presets and Task editors
Menu filter now searches all fields and help text
Improve device discovery

7/11/2026
7.7.3.0Retain HD camera on window resize
Support High DPI video (new setting in server settings - playback - default is disabled)
Update to web client for browser support for high DPI for scaling
Support dynamic resize in file playback mode (fill available area)
Improve pan after zoom accuracy
Improve resolution and encoding in live video view.
Fix rare remote reconnect wedge
7/10/2026
7.7.2.0Improve generation of timestamps in raw recordings (when enabled)
Major update of translations
7/8/2026
7.7.1.0

Add audio ducking to reduce echo on talk
Make UI more responsive for offline devices
Optimise paired device handling and state management
Add tapo, reolink, sip and rtp talk modules
Auto detect talk on add camera
Add talk error reporting back to session
Auto detect talk model when adding new devices
Improve PTZ Auto Tracking

7/6/2026
7.7.0.0Fix a rare condition where webservice reconnect can drop out
UI updates (time display in cut tool)
7/3/2026
7.6.9.0Fix issues with some ONVIF cameras
Updates for ONVIF talk
Fixes for rendering issues on older web browsers
Talk now follows camera selection and logs what it's doing
Fix NMS issue on local AI recognizer
7/2/2026
7.6.8.0Fix issue with streaming network audio via mp3 cuasing stuttering in the web ui
Add sentinel tracking to disable GPU devices that hard-error on init
Ignore some incompatible GPU devices
6/29/2026
7.6.7.0Fix issue with resolution detection on some ONVIF HikVision cameras.6/28/2026
7.6.6.0Fix issue with resetting retry counter in onvif event stream
Fix issue with detector zone save and apply (and associated actions)
Update tracking overlays to use color-blind safe colors
6/25/2026
7.6.5.0Fix issue with occasional audio drift n live view
Fix issue with QSV encoder (on Windows, if you have ffmpeg 8.1 folder in Agent delete and restart that folder to get the latest version)
Update onvif event stream retry and back-off logic.
Fix issue with writing tail of encoded recordings
6/24/2026
7.6.4.0Fix issue with opencv on old processors
Add automatic device detection/ add system for ONVIF, NDI and USB cameras
Add default camera login data to server settings - User Interface
Fix race issue playing files from live view
6/23/2026
7.6.3.0Fix issue with save config on close
Dynamically add threads to file playback based on resolution
6/21/2026
7.6.2.0Fix issue with keyframe detection on ffmpeg 8.1 for playback
Fix issue with copy button over insecure contexts
Fix issue with mjpeg streamer not closing on disconnect (CPU leak)
Change database scan to quick to improve startup time

6/20/2026
7.6.1.0
Major Update (with new FFmpeg 8.1)

See Beta release notes below for a summary of the recent changes - please backup your existing install before upgrading. Only upgrade if you have local access to roll-back.

Highlights:

  • FFmpeg 8.1 update
  • Learning AI update via alert feedback and local object recognition
  • VR re-write
  • Time machine merged into Live View
  • UI refresh
  • New Shortcut Key Manager
  • Lots of performance tweaks
  • Security fixes and package updates
  • New webserver
6/19/2026
7.6.0.0

(Beta):

More code cleanup
Move to Kestrel web server (remove old http.sys dependency)
Add cross platform SSL support (Need to reconfigure SSL if on Windows)
Resolve issue with interval reconnect and broken ONVIF streams
Update packages
6/18/2026
7.5.9.0

(Beta):

Add Tasks to PTZ buttons
General code cleanup
Add last-frame watchdog for broken camera streams
Security updates / server tightening
Default SSL validation on for SMTP, MQTT, LDAP - turn it off in settings if you have problems connecting.
Add Validate SSL toggle into device source settings.
Add signature verification to downloads and updates
Memory and CPU usage improvements
Add lock/unlock/investigate into time machine
Rewrite key shortcut manager - add collision detection.



6/17/2026
7.5.7.0

(Beta):

Fix XML parsing issue
6/12/2026
7.5.6.0

(Beta):

Feature Update: Local AI models can now learn from your alerts - alerts triggered or filtered by object recogntion now feed-back with a "did we get this right?" panel in alert viewer and learn to filter false positives over time. Enable by checking "Learn from feedback" in Object Recognition settings.
GPU accelerated screensavers
Updated + cleaned IP camera database
Bug fix for picture-in-picture issue
Other bug fixes and improvements.
FFmpeg library updated for QSV support.
6/12/2026
7.5.5.0

(Beta):

Fix bug with USB device parsing on windows (ffmpeg 8.1)
Fix bug with USB device access on linux
Fix bug with overlays on maximised streams
6/6/2026
7.5.4.0

(Beta):

UI updates (shift-click and drag/ scroll select on files and photos)
Add second CDN as backup for downloads of resources
Add debug output for docker integration (volume mapping)
Add date to session management (if not today)
Windows Tray App updated (with live video player)

6/5/2026
7.5.3.0

(Beta):

UI bug fixes
Fix issue qith qmax on VP8
Enumerate all dev/dri devices on linux
6/4/2026
7.5.2.0

(Beta):

UI bug fixes
6/3/2026
7.5.1.0

(Beta):

Resolve some transparency issues
Add Track to server settings to set update stream
Remove size limit on configuration upload
6/2/2026
7.5.0.0

(Beta):

UI rewrite (mostly technical but also some changes - main menu is at top left)
Remove time machine (replaced by new tab on live view)
Rewrite VR viewer
Update to custom built ffmpeg 8.1 (installers for all platforms available)
Update to latest version of openCV
Add hardware detection and setup wizard
Fix issues with nVidia GPU support for inbuilt AI on linux
Fix bug with email storage reports
Update tamper detection
Add sharding to ONNX parallel processing
Fix overlay bug on AI detected objects
Fix issue with jpeg set-range ffmpeg messages
Fix incompatibility issue with ONNX and opencv both using CUDA
Add FTP to UI
Major performance improvements

6/1/2026
7.4.8.0

(Beta):

Speed up connections on Windows
UI updates (removed jquery)
Fix some more JSON issues from migration
Updated Korean translations
Support timestamp merging within json packets using string interpolation.
5/5/2026
7.4.7.0

(Beta):

Fix issue with opencv / CUDA
Security audit updates - multiple path traversal fixes
Webrtc update
Various error message in-fills
Add multiple API field validation sections
Escape AIJSON in shell script calls
4/30/2026
7.4.6.0Fix bug with sped-up audio
Update LDAP to support more login methods
Disable cloud provider if authorization code cleared
Other minor bug fixes from code review
4/28/2026
7.4.5.0Fix bug with PiP serialization
Update logs UI
Merge in some webrtc/ sipsorcery updates
Fix issue with EOF detection in playback
Fix issue with update script on MacOS
Remove ONVIF debug logging
4/27/2026
7.4.4.0Fix another issue with json serializer
Fix issue with OSX FFmpeg discovery
4/23/2026
7.4.3.0Fix issue serializing ICE candidates
Play recording from start if start seek finds EOF
4/22/2026
7.4.2.0Fix potential memory issue
Use timestamp font size on overlays
Fix a couple of import issues from XML config for new json parser.
4/21/2026
7.4.0.0

(Beta):

Support runtime bitrate changes on GPU when driver supported
Update AI models
Remove AI enhancer options (not useful)
Add filter to menu system
Update JSON parsing library
Fix issue with pass-through AIJSON in merge tags
Add AIJSON for internal AI
Remove dynamic code
Add train option to photos for facial recognition

4/20/2026
7.3.9.0Fix bug clearing alerts on install4/12/2026
7.3.8.0Fix a potential issue with GPU driver installation for AI models
Fix issue with zombie alerts left over from deleted cameras.
Close alert panel on deletion of all alerts.
4/12/2026
7.3.7.0Updates for webrtc GPU code
Fix issue with default tags in actions (push and email alerts)
4/9/2026
7.3.6.0webrtc gpu update
fix issue with face uploads on linux
4/7/2026
7.3.5.0QSV webrtc update
Doubletake null reference guards
4/6/2026
7.3.4.0Add inference timing to system information
Fix issue with runaway onvif cameras
allow turns: in turn config
add nvidia hardware detection to docker/ linux
Fix issue with dstalign on photos
Fix bug with QSV encode in webrtc
Fix bug with timelapse
add config options to ai overlay
drawing optimizations
update ui, fix issue with mini flag
add alternate face processor
fix bugs with object not found in case
pixellate instead of blur
fix bug with motion passthrough
fix bug with onvif authentication on some cameras
Add diagnostic logging to ONVIF HTTP auth handler
Add inference timing to listen AI
Update server info panel with GPU statistics
Add GPU column to devices view in server info panel
Fix issue with mini mode
Update wireframe screensaver
Add support for larger local facial recognition model (in server settings - AI)
Add support for specifying GPU device to use for local AI
fix filename display in server summary
Update wireframe screensaver





4/4/2026
7.3.3.0Fix issue with resizing detection areas from codeproject AI
Add in system tab to system settings information panel with CPU, drive space and memory
Add in inference timing summaries to controls tab of system information panel.
3/20/2026
7.3.2.0Add support for up to 2 users in basic/ standard
Update webrtc QSV H264 encoder
3/19/2026
7.3.1.0Fix issue with some streams reconnecting on transient errors3/18/2026
7.3.0.0Fix issue passing options to stream analyzer3/17/2026
7.2.9.0Fix issue with AI photos
Fix issue with reading some non-blocking decoders (mjpeg)
3/17/2026
7.2.8.0Fix issue with stream GPU fallback to CPU.3/16/2026
7.2.7.0

Fix issue with local devices

Updates to Mac OS installers
If you are on Mac OS make sure you backup your configuration before updating. Run the installer on the download page if you have issues.

3/16/2026
7.2.5.0UI update (break view controls into new tab)
Don't set CPU flags on hardware encoder
Fix issue with GPU encode on 4K streams
Enable AI on 32 bit ARM
3/11/2026
7.2.4.0

(Beta):

Fix issue with some GPU encoders at high resolution
Minor UI updates
3/6/2026
7.2.3.0

(Beta):

Switch webrtc H264 mode off by default
Save alert filter history
Fix issues with some webrtc H264 hardware encoders
3/4/2026
7.2.2.0

(Beta):

Fix text rendering issue
Fix reading metadata from raw streams for adjustments in playback
3/3/2026
7.2.1.0

(Beta):

Fix webrtc GPU encoding issue on some platforms.
If you have issues uncheck Use GPU in server settings - playback and let us know.
3/3/2026
7.2.0.0

(Beta):

Update UI
Update packages to latest versions
Add built-in AI support for object recognition, LPR and facial recognition (requires license)
Add custom model support to built-in AI.
Add built in facial recognition database
Add face capture for adding faces
Add AI support for 32 bit ARM.
Code optimisation
Fix issue with running bash scripts in an action on MacOS
Add H264 webrtc encoder.
Add GPU support for H264 webrtc encoder
Add toggles to settings / playback for forcing H264 (default is on)
Updates to ONVIF code.
Add shortcut keys for toggling the UI, tools and commands. See Account menu - Key Manager. Reset it to see the new settings.

3/2/2026
7.1.9.0Add XML fixes for some reolink ONVIF cameras1/29/2026
7.1.8.0

Add name option to API actionRun, actionOff and actionOn calls (to trigger multiple actions in one API call)
Fix bug when copying both schedule and actions between devices
Add webrtc logging level integration
Remove error log when TURNS connection is closed
Add default support for doubletake as a face recognizer
Fix bug in calling openbrowser from deployed app
Fix bug with doubletake not working when no AI servers have been configured

1/28/2026
7.1.7.0Add clear option to lists of actions/ schedules/ ptz schedules
Update default encoder settings
Add auto mode to H264 encoder (changes based on resolution)
Add support for secure stun and turn endpoints for remote access.
Add WebRTC logging option to logging tab
Move TURN server configuration to it's own section in server settings
1/26/2026
7.1.6.0Add H264 profile to recording settings
Add support for AV1 encoding (CPU+GPU)
Add support for time machine scrubbing (enable it in server settings - user interface)
1/21/2026
7.1.5.0Fix bug in file transfer1/19/2026
7.1.4.0Fix bug with mjpeg server
Improve digital zooming
1/19/2026
7.1.3.0Add global audio feature (you might need to enable the ALL button in account menu - controls - live video). Listen to all active devices simultaneously in live view (experimental).
Fix bug with loop file option in file source
Fix bug with adding additional views with floorplans in the layout
Fix rare bug where cameras fail to reconnect
Fix bug with audio rendering + AI audio overlay
Add None option to audio visualizer
Fix (potential) bug with vulkan GPU integration
Fix bug with digest auth in some onvif PTZ devices
Add import option to server menu - devices to import content (video/ audio/ photos) to cameras (from server-local paths).
Fix bug with cloud uploads when multiple providers configured
Fix bug with opendrive uploader
Add decoder device detection to video source advanced settings dialog
Fix bug with appending a timestamp when the timestamp is empty
1/18/2026
7.1.2.0UI tweaks
Set default theme to cosmo dark
Fix bug with clicking on floorplan when camera is maximised in live view
1/9/2026
7.1.1.0Fix issue with streaming video to Alexa Show
Add tags to archive merger
Add ONVIF support for switchbot
1/8/2026
7.1.0.0Fix bug with storage management spawning too many tasks
Add support for doubletake facial recognition
Automatic scale mode when resizing images
Add support for switchbot connection via onvif
1/2/2026
7.0.9.0Fix issue with disconnecting cameras12/27/2025
7.0.8.0Potential fix for memory leak on some GPU devices when recording
Fix blocky display when using high performance resizer on mobile devices
Add {ALERTID} merge tag to tasks
Fix an issue where FTP servers create nested folders
12/24/2025
7.0.7.0Fix issue with RKMPP hardware ID
Add alertid to url parameters
Svae virtual camera layouts between restarts
Add roll control to virtual cameras via middle mouse button or shift key
Fix potential issue getting group membership in LDAP
Fix issue with image streamer/ timelapse.
12/22/2025
7.0.6.0Fix record issue when recording raw with h265 and restarted stream
Add debug logging
Add alerts and detector state to MQTT stats
12/16/2025
7.0.5.0Add xml fixing patches for more onvif cameras
Add support for virtual cameras with 360 dewarp (including digital ptz controls)
Dont connect to network storage unless needed when saving
12/14/2025
7.0.4.0Pass through tags to photos from actions
Add support for custom TURN servers (to server settings - local server)
12/10/2025
7.0.3.0Fix bug with camera authentication12/9/2025
7.0.2.0UI Bug fixes
Fix file source on non-windows platform issue
Support mjpeg with no content-type via internal decoder
Fix oauth sign-in issue
12/9/2025
7.0.0.0UI view bug fix
Add filter by time only option
Add filter by all tags option
Add protection around closing writer
UI updates
11/27/2025
6.9.9.0Fix bug with initial seek on clicking alerts
Add camera names to tag filtering
Add info text for onboarding
Remove Filter badge (use highlighted icons instead)
Highlight active icons in toolbar
11/26/2025
6.9.8.0Update themes (+ add 2 more)
UI update
Add Grid toggle to UI settings
Add Server Summary to server settings
Fix photo rotation issue for AI images
Optimise high res jpeg generation
11/25/2025
6.9.7.0Fix UI issues11/17/2025
6.9.6.0Playback UI bugfixes11/17/2025
6.9.5.0Restore High Performance Resizing option
Fix issues with audio visualizations
Fix issues with some UI commands
11/17/2025
6.9.4.0Force update of css file
Smoother fast playback
11/14/2025
6.9.3.0Merge in HEVC raw recording fix11/13/2025
6.9.2.0Add audio visualization
Add Norwegian translations
Fix crashing bug with NDI
Fix bug with FTP
11/13/2025
6.9.1.0

(Beta):

Update net 10 binaries
11/12/2025
6.9.0.0

(Beta):

Major: Upgrade to .Net 10 platform
UI update (for new multi server view)
Fix audio visualizer rendering issues
Add new audio visualizer
Fix bug with buffered encoder timestamps.
Move talk to worklet browser module
Performance updates

11/12/2025
6.8.9.0Fix issue with some GPU devices
Simplify audio processing
11/7/2025
6.8.8.0Fix bug on playback11/6/2025
6.8.7.0Fix connecting to some non-standard streams11/6/2025
6.8.6.0Fix issue with merged web script11/6/2025
6.8.5.0Fix issue with GPU/ CPU fallback
Update for upcoming multi-server web viewer
11/6/2025
6.8.4.0Fix resource leak in ONVIF events.
Update Gemini integration
Fix bug with downloading files via time machine
11/2/2025
6.8.3.0Fix issue with local device resolution setting
Fix issue with auto switching between TCP and UDP
Fix compacting database issue
Add options to set Gemini Ask AI video length and add audio
10/30/2025
6.8.2.0Fix crashing bug in speed detector
Add toggle to local server - playback to control file playback rendering size (toggle off to save full frame resolution photos during playback)
Add debug output to show ffmpeg location on startup
10/29/2025
6.8.1.0Switch scaling algorithms for image quality issues10/27/2025
6.8.0.0Fix memory leak in detector rewrite
Fix crashing bug with object tracker
Add retry/ persistance to cloud uploads
Add support for WasabiS3
Keep mini mode between reloads
10/27/2025
6.7.9.0Setup incremental vacuum on database
Optimise motion detector
Use opencv to resize video with adaptive scaling
10/26/2025
6.7.8.0Fix issue with oauth logins
10/25/2025
6.7.7.0Fix bug with motion detection
Fix crashing bug with facial recognition
Fix potential deadlock on writer close
Fix issue with timestamp location
10/25/2025
6.7.6.0Remove some codec flags (possible corrupted video fix)
Fix for timeout issues on ONVIF event streams.
10/24/2025
6.7.5.0Add buffer size calculation back into new media player10/24/2025
6.7.3.0

Drop corrupted packets (reduce false motion triggers)

10/24/2025
6.7.3.0Fixes for VAAPI HEVC decoder
Fix other GPU issues
Add support for client-secret based oauth logins
10/23/2025
6.7.2.0Fix issue with fallback from GPU to CPU decoding10/22/2025
6.7.1.0Fix some GPU driver issues10/22/2025
6.7.0.0Fix bug with settings not saving
Fix bug with resize
10/22/2025
6.6.9.0Bug fixes
Performance improvements
Move LDAP to Users tab in server settings
Add OAUTH to users tab in server setrtings
10/21/2025
6.6.8.0

(Beta):

Fix issues with buffer and timelapse
Fix crashing issue on reconnecting HD streams
10/11/2025
6.6.7.0

(Beta):

Major update/ code rewrite
Fully async media pipeline
Rewritten webrtc library
Optimised and streamlined sources
Fix issue with IOS playback
Massively reduce memory allocations
Updated libraries
Security patches
Fix issue with Gemini



10/10/2025
6.6.6.0Fix issue with ONVIF reconnecting event streams
Fix bugs in webrtc data channel code (faster downloads!)
9/10/2025
6.6.5.0Fix issue with constant reconnects to onvif event stream
dont connect onvif event stream if not needed
Updates to webrtc library
9/8/2025
6.6.4.0Fix resource leak in timelapse recorder
9/6/2025
6.6.3.0Fix potential race condition when adding files to database
Use net9 version of webrtc library
Fix UI issue with floorplans
Fix connections between macOS and linux hosts
9/5/2025
6.6.2.0

Refactor file player
HttpClient factory fixes (reuse http clients)
Update layout optimizer
Update libraries
Add open browser scripts for all platforms
Updates for launcher apps
Fix onedrive and onedrive business upload issue
Smoother file playback progress bar
Refactor ONVIF
Support ONVIF event stream when not using ONVIF motion detector
Add FTP Filename field to photos tab
Prefer IPv4 when connecting to remote server
Add repair database option if database corruption detected (to server menu)
Optimise JSON storage in database
Fix ffmpeg executable reference in docker
LDAP rewrite

9/4/2025
6.6.1.0Add back in {MSG} tag
Use prebuilt binaries for macOS
Ignore renew requests in onvif event streams that dont support them
Exit if port not available instead of use new port
8/5/2025
6.6.0.0MacOS compat update for upcoming installer
8/1/2025
6.5.9.0Fix OBJECTNAME merge tag7/30/2025
6.5.8.0Fix issue with onvif event receiver
Auto convert messaging tags to new format
7/30/2025
6.5.7.0Fix issue with multiple model object recogniser7/28/2025
6.5.6.0Support paths in FTP server config
Riollback some updates to sipsorcery code
Add support for multiple simultaneous object recognition models
Fix bug with call URL action (header parsing)
Normalise action tags
Improve keyframe detection code for raw recording
Raw recording buffer can now overlap previous recordings
Add support for UUID based onvif event receivers

7/28/2025
6.5.5.0Fix issue with FTP update
Fix issue with Call URL action
Allow negative schedule offset via UI
Add object recognition model to profiles
Normalise paths in cloud uploads
Add setinterval api endpoints for various intervals
7/23/2025
6.5.4.0Updates to google drive uploaded (duplicate folder fix)
General code cleanup for deprecated code
Library updates
Add {ALERTCOUNT} tag to merge fields (resets when camera is enabled).
Pass through detected tag from ONVIF detector (in {TAGS} parameter)
7/9/2025
6.5.3.0Fix issue with wrap-around PTZ scheduler
Add Gemini as LPR recogniser
Add Ollama, vLLM and LM Studio integrations for LPR and Ask AI
Update translations
Fix bug with PTZ Permissions
7/1/2025
6.5.2.0Resolve issue with moving folders between filesystems6/25/2025
6.5.1.0Add srihome put support for ptz
Fix bug with api tagging of recordings
Add re-order to actions
Fix issue with digital nav/ selected device
Add console logging to docker
6/25/2025
6.5.0.0Fix issue with log color coding
Fix issue with Agent calling commands in original commands folder
Fix issue with tag filtering on ask ai recognized event
Fix issue with tagging recordings on audio recognition
Optimise dummy image feed
Installer audio recognition engine file to app data folder
6/23/2025
6.4.9.0Fix bug with merging tags
Add support for uploading complete zip files with full config
Add overlays, sounds files, commands and floorplan images to config downloads
Move overlays, sound files, commands and floorplan images to data folder.
Prepopulate action parameters based on config.
6/19/2025
6.4.8.0Add template support for SMS, Push and Email actions
Fix issue with filtering events for no object recognized and no plate recognized.
Consolidate merge code to use one merge method
Detect and update TURN IP address
Add back in local access to 4K resolution for basic subscription+
6/18/2025
6.4.7.0

(Beta):

Fix issue with choppy audio
Add template support for push notifications
Detect and use public IP address in TURN server (if not specified)
Auto update TURN server public IP address
VISCA updates
NDI PTZ updates
6/12/2025
6.4.6.0

(Beta):

Add PTZ saved preview and edit support to all PTZ providers
Add tags to push notifications
6/11/2025
6.4.5.0

(Beta):

Add VISCA PTZ provider
Fix bug with PTZ images
Add support for AAC talk encoding
6/10/2025
6.4.4.0

(Beta):

Updates to UI
Add device selection drop-down to editor
Rewrite file player to use double buffer
Translation updates
6/9/2025
6.4.3.0

(Beta):

UI updates
Fix for issue with Safari on Mac OS
Add OBSBOT PTZ provider
Add OPUS webrtc audio support
Add abs-sendtime to webrtc channels
Update to webrtc SSL library
6/1/2025
6.4.2.0

(Beta):

Add OSBOT and VISCA PTZ providers
UI updates
Security updates to various libraries
5/27/2025
6.4.1.0

(Beta):

Add digital PTZ preset support
UI updates
Update to google drive folder discovery
5/23/2025
6.4.0.0

(Beta):

Fix issue where trying to connect to local USB devices that aren't plugged in could hang.
More UI updates
Add motion graph display to timeline
Remove camera live thumbnails from timeline
Rework time machine display to show motion graph and alert events and display current device and "others" instead of all.
Make UI more responsive
Fix issue with fast-forward audio.
Remove theme dependant styling
5/22/2025
6.3.9.0

(Beta):

Don't zoom out on edit
Add back in mute control
Change controls defaults to hide some rarely used options
Add drag and drop to rearrange controls (in account menu - controls)
Updates to volume control
5/15/2025
6.3.8.0

(Beta):

UI bug fixes
Fix issue with stalling downloads on mobiles
Add Change View action to actions and scheduler (changes the view on all connected browsers)
Add hungarian language pack
Fix potential issue when saving config during power outage
Revert S3 stream uploader changes

5/14/2025
6.3.7.0

(Beta):

Save 10 days worth of backup configurations
Updates to UI (auto move playback controls, fix bug with alert images, add clear button to alerts filter)
Use file stream for S3 upload.
5/12/2025
6.3.6.0

(Beta):

Fix image rotation issue
5/8/2025
6.3.5.0

(Beta):

Fixes for streaming to ipads
Major overhaul of UI
Enhanced local video device support
Fixes for download speed of files through UI
Bug fixes
Browser compatibility updates
Updates for upcoming mobile app upgrades
Add inbuilt STUN server for when TURN server is disabled in server settings.
5/7/2025
6.3.4.0Fix issue with missing file (Windows)4/17/2025
6.3.3.0Fix memory leak in image sources (image/ MJPEG)
Ffix rendering issues on timeline and timemachine timeline.
4/16/2025
6.3.2.0Fix for crash bug on some iphones when viewing advanced detector settings
Update UI for zone selection
4/15/2025
6.3.1.0Fix for detecting local devices when drivers stream back to stdOut4/13/2025
6.3.0.0Updates to TURN server
Add debug output to device parsing
4/13/2025
6.2.9.0

Add dewarper for quad rectilinear feeds
Add option to dewarp high resolution streams (to adjustments)
Fix issue with local USB audio devices on linux (use pulse instead of alsa)
Fix issue with listing local devices on first load of Agent on linux and OSX
Add schedule summary viewer to server menu

4/11/2025
6.2.8.0Add support for talking through other cameras
Update dropbox for large file support
Fix issue with Google Drive on linux and OSX
Add volume setting for talk
Fix OSX binaries (certificate issue)
Fix issues with frame timing
Fix M3U8 streams
Fix for image encoding (grabThumb)
Updates to LDAP (ActiveDirectory+UseBasic support)
Migration to new licensing system
Add Gemini video option for Ask AI
Fix bugs with Image source endpoint
4/8/2025
6.2.7.0Fix bug with google drive paths
Add Tag Recording action
Treat 400 errors from ONVIF cameras as digest challenges
3/10/2025
6.2.6.0Add Syslog integration (see server settings - logging)
Add Hikvision motion detection type with Hikvision AI support
Fix issue with google drive uploads
Rotate and flip photos according to camera rules
3/8/2025
6.2.5.0Fix bug with volume control
Add audio back to webm/ mp4 live streamer
3/6/2025
6.2.4.0Fix issue with TURN server detection
Fix issue with ARM build for MacOS
3/6/2025
6.2.3.0Fix bug with detector view on raspberry pi3/4/2025
6.2.2.0Remove B-frames from RTMP streamer (AWS compat)
Add account menu - controls UI for enabling and disabling control visibility throughout the UI
Fix bug with rotation of video
Add TURN_IPADDRESS to environment variables for docker setup
Fix bug with audio level recording in meta data when recording raw
Prevent update in UI in docker
3/4/2025
6.2.1.0update updater
add debug output to blocked alerts
dont log failed stun requests
pass zonelist to child actions
ignore ssl errors on reolink AI calls
add start menu shortcut option to installer (all users)
2/25/2025
6.2.0.0

(Beta):

Add TURN servers for all platforms
2/21/2025
6.1.9.0

(Beta):

Add dynamic bitrate adjustment for webrtc streams
Add TWCC support to webrtc library
2/20/2025
6.1.8.0Add auto mute toggle to UI settings
Add notes field to devices
Update html renderer
Tray app checks if it's already running
2/16/2025
6.1.7.0

(Beta):

Add extra detection for stun/ turn server
2/14/2025
6.1.6.0Add TCP support to TURN server integration for NAT traversal2/13/2025
6.1.5.0Add support for coturn
Fix blink bug at start of streaming
Tweak streaming settings
2/13/2025
6.1.4.0

(Beta):

Add workaround for missing libva driver on linux
Fix stuttering issue when connecting streams
Add MQTT protocol selection to MQTT settings
Fix frame-forward bug on playback
2/10/2025
6.1.3.0

Optimise javascript

2/6/2025
6.1.2.0Fix name encoding issue when setting ptz preset name
Fix issue with image selection when using base64image tags with AI events
2/6/2025