Frigate

A complete and local NVR

Get it on umbrelOS

Version0.17.1-2

CategoryHome & Automation

Source codePublic

Developed byBlake Blackshear 1 app

Submitted by~dibref-labter 2 apps

Compatible withumbrelOS 0.5 or later

A complete and local NVR designed for Home Assistant with AI object detection. Uses OpenCV and Tensorflow to perform realtime object detection locally for IP cameras.

🛠️ SET-UP

You can configure your cameras directly within the Frigate app, there is no need to manually edit the configuration files.
Sign in to Frigate with username admin and the "default app password" displayed on Frigate's page in the app store (shown after install). The same default app password is also used as your RTSP password if needed.

What's new

Version 0.17.1-2+4 days ago

Fixed an issue that could prevent Home Assistant and other apps on your local network from connecting to Frigate.

If Frigate asks you to sign in after updating, use your existing Frigate username and password. If you have not set one, use the default credentials shown by Umbrel before Frigate opens or on Frigate's page in the App Store.

Existing installs are updated automatically. Your cameras and Frigate settings are preserved, and you can change your Frigate password in Frigate after signing in.

Version 0.17.1-1+this month

This package revision keeps Frigate on upstream 0.17.1 and protects the web UI behind Umbrel app authentication.

Frigate now opens through Umbrel's app-authenticated proxy instead of exposing Frigate's internal unauthenticated UI directly. The default package uses Frigate's documented reverse-proxy mode, so Umbrel remains the login layer and no separate Frigate admin password is required for the default setup.

Full upstream release notes for Frigate 0.17.1 are available at https://github.com/blakeblackshear/frigate/releases/tag/v0.17.1

Version 0.17.1+3 months ago

⚠️ As usual, Frigate will attempt to update its configuration automatically. It is still recommended to back up your current config and database before upgrading.

🚨 This is a maintenance release with security fixes and bugfixes:

Restricted raw config endpoint to admin role only
Fixed cross-camera authentication in timeline and media endpoints
Fixed RKNN ArcFace input format
Fixed storage cleanup being attempted when Frigate is in safe mode

Other improvements in this update include:

Improved playback of videos in Tracking Details
Fixed preview retrieval to handle missing previews gracefully
Fixed thumbnail encoding logic
Fixed go2rtc config handling
Fixed GenAI issues
Fixed audio event maintainer
Push notifications now filtered by user role camera access
Various UI fixes and improvements

Full release notes are available at https://github.com/blakeblackshear/frigate/releases

Version 0.17.0+3 months ago

⚠️ As usual, Frigate will attempt to update its configuration automatically. It is still recommended to back up your current config and database before upgrading.

🚨 This is a major release with several breaking changes:

GenAI config restructured: global genai now only configures the provider; other fields moved under objects -> genai
Recordings retention is now fully tiered with separate record -> continuous and record -> motion fields
LPR models updated: most users should manually switch to the small model
strftime_fmt has been fully removed; date/time formatting is now based on UI language selection
Camera auto-detection resolution logic has changed; you may need to explicitly set detect width and height if Frigate hangs on startup
The exec, expr, and echo sources for go2rtc are now removed by default to reduce security risk
Nvidia GTX 900 series GPUs are no longer supported

New features and improvements in this update include:

New classification model training: train local models for state classification (e.g. gate open/closed) and object classification (e.g. identify specific dogs or people)
Custom viewer roles to limit user access to specific cameras
GenAI can now summarize review items with a title, description, and activity classification
Semantic Search Triggers to automate actions when a tracked object matches an image or description
Nvidia GPU performance improved with CUDA Graphs support
Intel OpenVINO now supports NPU inference with reduced memory and CPU usage
RKNN improvements including automatic ONNX model conversion and accelerated enrichment support
Apple Silicon NPU object detection support
YOLOv9 support on Google Coral devices
New community-supported detectors: MemryX, Degirum SDK, and Synaptics
Add Camera Wizard in the UI to add cameras without editing config files
Many more settings can now be saved dynamically without restarting Frigate
Configuration Safe Mode highlights invalid config issues in the UI
Audio transcription support using local models
LPR normalization configuration and updated PaddleOCR models with Chinese character support
IPv6 can now be toggled via config
Revamped Settings menu with sidebar navigation

Full release notes are available at https://github.com/blakeblackshear/frigate/releases

Version 0.16.4+4 months ago

⚠️ As usual, Frigate will attempt to update its configuration automatically. It is still recommended to back up your current config and database before upgrading.

🚨 This release includes a critical security fix for authenticated users: