Merge branch 'dev' into AXERA-axcl

.cursor/rules/frontend-always-use-translation-files.mdc

@@ -0,0 +1,6 @@
+---
+globs: ["**/*.ts", "**/*.tsx"]
+alwaysApply: false
+---
+
+Never write strings in the frontend directly, always write to and reference the relevant translations file.

.gitignore

@@ -15,6 +15,7 @@ frigate/version.py
 web/build
 web/node_modules
 web/coverage
+web/.env
 core
 !/web/**/*.ts
 .idea/*

LICENSE

@@ -1,6 +1,6 @@
 The MIT License
 
-Copyright (c) 2020 Blake Blackshear
+Copyright (c) 2025 Frigate LLC (Frigate™)
 
 Permission is hereby granted, free of charge, to any person obtaining a copy
 of this software and associated documentation files (the "Software"), to deal

Makefile

@@ -14,6 +14,7 @@ push-boards: $(BOARDS:%=push-%)
 
 version:
 	echo 'VERSION = "$(VERSION)-$(COMMIT_HASH)"' > frigate/version.py
+	echo 'VITE_GIT_COMMIT_HASH=$(COMMIT_HASH)' > web/.env
 
 local: version
 	docker buildx build --target=frigate --file docker/main/Dockerfile . \

README.md

@@ -1,8 +1,10 @@
 <p align="center">
-  <img align="center" alt="logo" src="docs/static/img/frigate.png">
+  <img align="center" alt="logo" src="docs/static/img/branding/frigate.png">
 </p>
 
-# Frigate - NVR With Realtime Object Detection for IP Cameras
+# Frigate NVR™ - Realtime Object Detection for IP Cameras
+
+[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)
 
 <a href="https://hosted.weblate.org/engage/frigate-nvr/">
 <img src="https://hosted.weblate.org/widget/frigate-nvr/language-badge.svg" alt="Translation status" />
@@ -12,7 +14,7 @@
 
 A complete and local NVR designed for [Home Assistant](https://www.home-assistant.io) with AI object detection. Uses OpenCV and Tensorflow to perform realtime object detection locally for IP cameras.
 
-Use of a GPU or AI accelerator such as a [Google Coral](https://coral.ai/products/) or [Hailo](https://hailo.ai/) is highly recommended. AI accelerators will outperform even the best CPUs with very little overhead.
+Use of a GPU or AI accelerator is highly recommended. AI accelerators will outperform even the best CPUs with very little overhead. See Frigate's supported [object detectors](https://docs.frigate.video/configuration/object_detectors/).
 
 - Tight integration with Home Assistant via a [custom component](https://github.com/blakeblackshear/frigate-hass-integration)
 - Designed to minimize resource use and maximize performance by only looking for objects when and where it is necessary
@@ -33,6 +35,15 @@ View the documentation at https://docs.frigate.video
 
 If you would like to make a donation to support development, please use [Github Sponsors](https://github.com/sponsors/blakeblackshear).
 
+## License
+
+This project is licensed under the **MIT License**.
+
+- **Code:** The source code, configuration files, and documentation in this repository are available under the [MIT License](LICENSE). You are free to use, modify, and distribute the code as long as you include the original copyright notice.
+- **Trademarks:** The "Frigate" name, the "Frigate NVR" brand, and the Frigate logo are **trademarks of Frigate LLC** and are **not** covered by the MIT License.
+
+Please see our [Trademark Policy](TRADEMARK.md) for details on acceptable use of our brand assets.
+
 ## Screenshots
 
 ### Live dashboard
@@ -66,3 +77,7 @@ We use [Weblate](https://hosted.weblate.org/projects/frigate-nvr/) to support la
 <a href="https://hosted.weblate.org/engage/frigate-nvr/">
 <img src="https://hosted.weblate.org/widget/frigate-nvr/multi-auto.svg" alt="Translation status" />
 </a>
+
+---
+
+**Copyright © 2025 Frigate LLC.**

TRADEMARK.md

@@ -0,0 +1,58 @@
+# Trademark Policy
+
+**Last Updated:** November 2025
+
+This document outlines the policy regarding the use of the trademarks associated with the Frigate NVR project.
+
+## 1. Our Trademarks
+
+The following terms and visual assets are trademarks (the "Marks") of **Frigate LLC**:
+
+- **Frigate™**
+- **Frigate NVR™**
+- **Frigate+™**
+- **The Frigate Logo**
+
+**Note on Common Law Rights:**
+Frigate LLC asserts all common law rights in these Marks. The absence of a federal registration symbol (®) does not constitute a waiver of our intellectual property rights.
+
+## 2. Interaction with the MIT License
+
+The software in this repository is licensed under the [MIT License](LICENSE).
+
+**Crucial Distinction:**
+
+- The **Code** is free to use, modify, and distribute under the MIT terms.
+- The **Brand (Trademarks)** is **NOT** licensed under MIT.
+
+You may not use the Marks in any way that is not explicitly permitted by this policy or by written agreement with Frigate LLC.
+
+## 3. Acceptable Use
+
+You may use the Marks without prior written permission in the following specific contexts:
+
+- **Referential Use:** To truthfully refer to the software (e.g., _"I use Frigate NVR for my home security"_).
+- **Compatibility:** To indicate that your product or project works with the software (e.g., _"MyPlugin for Frigate NVR"_ or _"Compatible with Frigate"_).
+- **Commentary:** In news articles, blog posts, or tutorials discussing the software.
+
+## 4. Prohibited Use
+
+You may **NOT** use the Marks in the following ways:
+
+- **Commercial Products:** You may not use "Frigate" in the name of a commercial product, service, or app (e.g., selling an app named _"Frigate Viewer"_ is prohibited).
+- **Implying Affiliation:** You may not use the Marks in a way that suggests your project is official, sponsored by, or endorsed by Frigate LLC.
+- **Confusing Forks:** If you fork this repository to create a derivative work, you **must** remove the Frigate logo and rename your project to avoid user confusion. You cannot distribute a modified version of the software under the name "Frigate".
+- **Domain Names:** You may not register domain names containing "Frigate" that are likely to confuse users (e.g., `frigate-official-support.com`).
+
+## 5. The Logo
+
+The Frigate logo (the bird icon) is a visual trademark.
+
+- You generally **cannot** use the logo on your own website or product packaging without permission.
+- If you are building a dashboard or integration that interfaces with Frigate, you may use the logo only to represent the Frigate node/service, provided it does not imply you _are_ Frigate.
+
+## 6. Questions & Permissions
+
+If you are unsure if your intended use violates this policy, or if you wish to request a specific license to use the Marks (e.g., for a partnership), please contact us at:
+
+**help@frigate.video**

docker/main/build_pysqlite3.sh

@@ -5,6 +5,12 @@ set -euxo pipefail
 SQLITE3_VERSION="3.46.1"
 PYSQLITE3_VERSION="0.5.3"
 
+# Install libsqlite3-dev if not present (needed for some base images like NVIDIA TensorRT)
+if ! dpkg -l | grep -q libsqlite3-dev; then
+  echo "Installing libsqlite3-dev for compilation..."
+  apt-get update && apt-get install -y libsqlite3-dev && rm -rf /var/lib/apt/lists/*
+fi
+
 # Fetch the pre-built sqlite amalgamation instead of building from source
 if [[ ! -d "sqlite" ]]; then
   mkdir sqlite

docker/main/install_deps.sh

@@ -20,6 +20,7 @@ apt-get -qq install --no-install-recommends -y \
     libgl1 \
     libglib2.0-0 \
     libusb-1.0.0 \
+    python3-h2 \
     libgomp1  # memryx detector
 
 update-alternatives --install /usr/bin/python3 python3 /usr/bin/python3.11 1
@@ -95,6 +96,9 @@ if [[ "${TARGETARCH}" == "amd64" ]]; then
 
     apt-get -qq install -y ocl-icd-libopencl1
 
+    # install libtbb12 for NPU support
+    apt-get -qq install -y libtbb12
+
     rm -f /usr/share/keyrings/intel-graphics.gpg
     rm -f /etc/apt/sources.list.d/intel-gpu-jammy.list
 
@@ -115,6 +119,11 @@ if [[ "${TARGETARCH}" == "amd64" ]]; then
     wget https://github.com/intel/compute-runtime/releases/download/24.52.32224.5/intel-level-zero-gpu_1.6.32224.5_amd64.deb
     wget https://github.com/intel/intel-graphics-compiler/releases/download/v2.5.6/intel-igc-opencl-2_2.5.6+18417_amd64.deb
     wget https://github.com/intel/intel-graphics-compiler/releases/download/v2.5.6/intel-igc-core-2_2.5.6+18417_amd64.deb
+    # npu packages
+    wget https://github.com/oneapi-src/level-zero/releases/download/v1.21.9/level-zero_1.21.9+u22.04_amd64.deb
+    wget https://github.com/intel/linux-npu-driver/releases/download/v1.17.0/intel-driver-compiler-npu_1.17.0.20250508-14912879441_ubuntu22.04_amd64.deb
+    wget https://github.com/intel/linux-npu-driver/releases/download/v1.17.0/intel-fw-npu_1.17.0.20250508-14912879441_ubuntu22.04_amd64.deb
+    wget https://github.com/intel/linux-npu-driver/releases/download/v1.17.0/intel-level-zero-npu_1.17.0.20250508-14912879441_ubuntu22.04_amd64.deb
 
     dpkg -i *.deb
     rm *.deb
@@ -136,6 +145,6 @@ rm -rf /var/lib/apt/lists/*
 
 # Install yq, for frigate-prepare and go2rtc echo source
 curl -fsSL \
-    "https://github.com/mikefarah/yq/releases/download/v4.33.3/yq_linux_$(dpkg --print-architecture)" \
+    "https://github.com/mikefarah/yq/releases/download/v4.48.2/yq_linux_$(dpkg --print-architecture)" \
     --output /usr/local/bin/yq
 chmod +x /usr/local/bin/yq

docker/main/install_memryx.sh

@@ -2,9 +2,9 @@
 set -e
 
 # Download the MxAccl for Frigate github release
-wget https://github.com/memryx/mx_accl_frigate/archive/refs/heads/main.zip -O /tmp/mxaccl.zip
+wget https://github.com/memryx/mx_accl_frigate/archive/refs/tags/v2.1.0.zip -O /tmp/mxaccl.zip
 unzip /tmp/mxaccl.zip -d /tmp
-mv /tmp/mx_accl_frigate-main /opt/mx_accl_frigate
+mv /tmp/mx_accl_frigate-2.1.0 /opt/mx_accl_frigate
 rm /tmp/mxaccl.zip
 
 # Install Python dependencies

docker/main/requirements-wheels.txt

@@ -56,7 +56,7 @@ pywebpush == 2.0.*
 # alpr
 pyclipper == 1.3.*
 shapely == 2.0.*
-Levenshtein==0.26.*
+rapidfuzz==3.12.*
 # HailoRT Wheels
 appdirs==1.4.*
 argcomplete==2.0.*

docker/main/requirements.txt

@@ -1,2 +1 @@
 scikit-build == 0.18.*
-nvidia-pyindex

docker/main/rootfs/usr/local/nginx/conf/nginx.conf

@@ -73,6 +73,8 @@ http {
         vod_manifest_segment_durations_mode accurate;
         vod_ignore_edit_list on;
         vod_segment_duration 10000;
+
+        # MPEG-TS settings (not used when fMP4 is enabled, kept for reference)
         vod_hls_mpegts_align_frames off;
         vod_hls_mpegts_interleave_frames on;
 
@@ -105,6 +107,10 @@ http {
             aio threads;
             vod hls;
 
+            # Use fMP4 (fragmented MP4) instead of MPEG-TS for better performance
+            # Smaller segments, faster generation, better browser compatibility
+            vod_hls_container_format fmp4;
+
             secure_token $args;
             secure_token_types application/vnd.apple.mpegurl;
 
@@ -274,6 +280,18 @@ http {
                 include proxy.conf;
             }
 
+            # Allow unauthenticated access to the first_time_login endpoint
+            # so the login page can load help text before authentication.
+            location /api/auth/first_time_login {
+                auth_request off;
+                limit_except GET {
+                    deny all;
+                }
+                rewrite ^/api(/.*)$ $1 break;
+                proxy_pass http://frigate_api;
+                include proxy.conf;
+            }
+
             location /api/stats {
                 include auth_request.conf;
                 access_log off;
@@ -302,6 +320,12 @@ http {
                 add_header Cache-Control "public";
             }
 
+            location /fonts/ {
+                access_log off;
+                expires 1y;
+                add_header Cache-Control "public";
+            }
+
             location /locales/ {
                 access_log off;
                 add_header Cache-Control "public";

docker/memryx/user_installation.sh

@@ -24,10 +24,13 @@ echo "Adding MemryX GPG key and repository..."
 wget -qO- https://developer.memryx.com/deb/memryx.asc | sudo tee /etc/apt/trusted.gpg.d/memryx.asc >/dev/null
 echo 'deb https://developer.memryx.com/deb stable main' | sudo tee /etc/apt/sources.list.d/memryx.list >/dev/null
 
-# Update and install memx-drivers
+# Update and install specific SDK 2.1 packages
-echo "Installing memx-drivers..."
+echo "Installing MemryX SDK 2.1 packages..."
 sudo apt update
-sudo apt install -y memx-drivers
+sudo apt install -y memx-drivers=2.1.* memx-accl=2.1.* mxa-manager=2.1.*
+
+# Hold packages to prevent automatic upgrades
+sudo apt-mark hold memx-drivers memx-accl mxa-manager
 
 # ARM-specific board setup
 if [[ "$arch" == "aarch64" || "$arch" == "arm64" ]]; then
@@ -37,11 +40,5 @@ fi
 
 echo -e "\n\n\033[1;31mYOU MUST RESTART YOUR COMPUTER NOW\033[0m\n\n"
 
-# Install other runtime packages
+echo "MemryX SDK 2.1 installation complete!"
-packages=("memx-accl" "mxa-manager")
-for pkg in "${packages[@]}"; do
-    echo "Installing $pkg..."
-    sudo apt install -y "$pkg"
-done
 
-echo "MemryX installation complete!"

docker/tensorrt/Dockerfile.amd64

@@ -21,7 +21,7 @@ FROM deps AS frigate-tensorrt
 ARG PIP_BREAK_SYSTEM_PACKAGES
 
 RUN --mount=type=bind,from=trt-wheels,source=/trt-wheels,target=/deps/trt-wheels \
-    pip3 uninstall -y onnxruntime tensorflow-cpu \
+    pip3 uninstall -y onnxruntime \
     && pip3 install -U /deps/trt-wheels/*.whl
 
 COPY --from=rootfs / /

docker/tensorrt/Dockerfile.arm64

@@ -112,7 +112,7 @@ RUN apt-get update \
     && apt-get install -y protobuf-compiler libprotobuf-dev \
     && rm -rf /var/lib/apt/lists/*
 RUN --mount=type=bind,source=docker/tensorrt/requirements-models-arm64.txt,target=/requirements-tensorrt-models.txt \
-    pip3 wheel --wheel-dir=/trt-model-wheels -r /requirements-tensorrt-models.txt
+    pip3 wheel --wheel-dir=/trt-model-wheels --no-deps -r /requirements-tensorrt-models.txt
 
 FROM wget AS jetson-ffmpeg
 ARG DEBIAN_FRONTEND
@@ -145,7 +145,8 @@ COPY --from=trt-wheels /etc/TENSORRT_VER /etc/TENSORRT_VER
 RUN --mount=type=bind,from=trt-wheels,source=/trt-wheels,target=/deps/trt-wheels \
     --mount=type=bind,from=trt-model-wheels,source=/trt-model-wheels,target=/deps/trt-model-wheels \
     pip3 uninstall -y onnxruntime \
-    && pip3 install -U /deps/trt-wheels/*.whl /deps/trt-model-wheels/*.whl \
+    && pip3 install -U /deps/trt-wheels/*.whl \
+    && pip3 install -U /deps/trt-model-wheels/*.whl \
     && ldconfig
 
 WORKDIR /opt/frigate/

docker/tensorrt/requirements-amd64.txt

@@ -13,7 +13,6 @@ nvidia_cusolver_cu12==11.6.3.*; platform_machine == 'x86_64'
 nvidia_cusparse_cu12==12.5.1.*; platform_machine == 'x86_64'
 nvidia_nccl_cu12==2.23.4; platform_machine == 'x86_64'
 nvidia_nvjitlink_cu12==12.5.82; platform_machine == 'x86_64'
-tensorflow==2.19.*; platform_machine == 'x86_64'
 onnx==1.16.*; platform_machine == 'x86_64'
 onnxruntime-gpu==1.22.*; platform_machine == 'x86_64'
 protobuf==3.20.3; platform_machine == 'x86_64'

docker/tensorrt/requirements-arm64.txt

@@ -1 +1,2 @@
 cuda-python == 12.6.*; platform_machine == 'aarch64'
+numpy == 1.26.*; platform_machine == 'aarch64'

docs/docs/configuration/advanced.md

@@ -25,7 +25,7 @@ Examples of available modules are:
 
 - `frigate.app`
 - `frigate.mqtt`
-- `frigate.object_detection`
+- `frigate.object_detection.base`
 - `detector.<detector_name>`
 - `watchdog.<camera_name>`
 - `ffmpeg.<camera_name>.<sorted_roles>` NOTE: All FFmpeg logs are sent as `error` level.
@@ -53,6 +53,17 @@ environment_vars:
   VARIABLE_NAME: variable_value
 ```
 
+#### TensorFlow Thread Configuration
+
+If you encounter thread creation errors during classification model training, you can limit TensorFlow's thread usage:
+
+```yaml
+environment_vars:
+  TF_INTRA_OP_PARALLELISM_THREADS: "2" # Threads within operations (0 = use default)
+  TF_INTER_OP_PARALLELISM_THREADS: "2" # Threads between operations (0 = use default)
+  TF_DATASET_THREAD_POOL_SIZE: "2" # Data pipeline threads (0 = use default)
+```
+
 ### `database`
 
 Tracked object and recording information is managed in a sqlite database at `/config/frigate.db`. If that database is deleted, recordings will be orphaned and will need to be cleaned up manually. They also won't show up in the Media Browser within Home Assistant.
@@ -247,7 +258,7 @@ curl -X POST http://frigate_host:5000/api/config/save -d @config.json
 if you'd like you can use your yaml config directly by using [`yq`](https://github.com/mikefarah/yq) to convert it to json:
 
 ```bash
-yq r -j config.yml | curl -X POST http://frigate_host:5000/api/config/save -d @-
+yq -o=json '.' config.yaml | curl -X POST 'http://frigate_host:5000/api/config/save?save_option=saveonly' --data-binary @-
 ```
 
 ### Via Command Line

docs/docs/configuration/camera_specific.md

@@ -164,13 +164,35 @@ According to [this discussion](https://github.com/blakeblackshear/frigate/issues
 Cameras connected via a Reolink NVR can be connected with the http stream, use `channel[0..15]` in the stream url for the additional channels.
 The setup of main stream can be also done via RTSP, but isn't always reliable on all hardware versions. The example configuration is working with the oldest HW version RLN16-410 device with multiple types of cameras.
 
+<details>
+  <summary>Example Config</summary>
+
+:::tip
+
+Reolink's latest cameras support two way audio via go2rtc and other applications. It is important that the http-flv stream is still used for stability, a secondary rtsp stream can be added that will be using for the two way audio only.
+
+NOTE: The RTSP stream can not be prefixed with `ffmpeg:`, as go2rtc needs to handle the stream to support two way audio.
+
+Ensure HTTP is enabled in the camera's advanced network settings. To use two way talk with Frigate, see the [Live view documentation](/configuration/live#two-way-talk).
+
+:::
+
 ```yaml
 go2rtc:
   streams:
+    # example for connecting to a standard Reolink camera
     your_reolink_camera:
       - "ffmpeg:http://reolink_ip/flv?port=1935&app=bcs&stream=channel0_main.bcs&user=username&password=password#video=copy#audio=copy#audio=opus"
     your_reolink_camera_sub:
       - "ffmpeg:http://reolink_ip/flv?port=1935&app=bcs&stream=channel0_ext.bcs&user=username&password=password"
+    # example for connectin to a Reolink camera that supports two way talk
+    your_reolink_camera_twt:
+      - "ffmpeg:http://reolink_ip/flv?port=1935&app=bcs&stream=channel0_main.bcs&user=username&password=password#video=copy#audio=copy#audio=opus"
+      - "rtsp://username:password@reolink_ip/Preview_01_sub
+    your_reolink_camera_twt_sub:
+      - "ffmpeg:http://reolink_ip/flv?port=1935&app=bcs&stream=channel0_ext.bcs&user=username&password=password"
+      - "rtsp://username:password@reolink_ip/Preview_01_sub
+    # example for connecting to a Reolink NVR
     your_reolink_camera_via_nvr:
       - "ffmpeg:http://reolink_nvr_ip/flv?port=1935&app=bcs&stream=channel3_main.bcs&user=username&password=password" # channel numbers are 0-15
       - "ffmpeg:your_reolink_camera_via_nvr#audio=aac"
@@ -201,22 +223,7 @@ cameras:
           roles:
             - detect
 ```
-
+</details>
-#### Reolink Doorbell
-
-The reolink doorbell supports two way audio via go2rtc and other applications. It is important that the http-flv stream is still used for stability, a secondary rtsp stream can be added that will be using for the two way audio only.
-
-Ensure HTTP is enabled in the camera's advanced network settings. To use two way talk with Frigate, see the [Live view documentation](/configuration/live#two-way-talk).
-
-```yaml
-go2rtc:
-  streams:
-    your_reolink_doorbell:
-      - "ffmpeg:http://reolink_ip/flv?port=1935&app=bcs&stream=channel0_main.bcs&user=username&password=password#video=copy#audio=copy#audio=opus"
-      - rtsp://reolink_ip/Preview_01_sub
-    your_reolink_doorbell_sub:
-      - "ffmpeg:http://reolink_ip/flv?port=1935&app=bcs&stream=channel0_ext.bcs&user=username&password=password"
-```
 
 ### Unifi Protect Cameras
 

docs/docs/configuration/custom_classification/object_classification.md

@@ -10,9 +10,19 @@ Object classification allows you to train a custom MobileNetV2 classification mo
 Object classification models are lightweight and run very fast on CPU. Inference should be usable on virtually any machine that can run Frigate.
 
 Training the model does briefly use a high amount of system resources for about 1–3 minutes per training run. On lower-power devices, training may take longer.
-When running the `-tensorrt` image, Nvidia GPUs will automatically be used to accelerate training.
 
-### Sub label vs Attribute
+## Classes
+
+Classes are the categories your model will learn to distinguish between. Each class represents a distinct visual category that the model will predict.
+
+For object classification:
+
+- Define classes that represent different types or attributes of the detected object
+- Examples: For `person` objects, classes might be `delivery_person`, `resident`, `stranger`
+- Include a `none` class for objects that don't fit any specific category
+- Keep classes visually distinct to improve accuracy
+
+### Classification Type
 
 - **Sub label**:
 
@@ -25,6 +35,15 @@ When running the `-tensorrt` image, Nvidia GPUs will automatically be used to ac
   - Ideal when multiple attributes can coexist independently.
   - Example: Detecting if a `person` in a construction yard is wearing a helmet or not.
 
+## Assignment Requirements
+
+Sub labels and attributes are only assigned when both conditions are met:
+
+1. **Threshold**: Each classification attempt must have a confidence score that meets or exceeds the configured `threshold` (default: `0.8`).
+2. **Class Consensus**: After at least 3 classification attempts, 60% of attempts must agree on the same class label. If the consensus class is `none`, no assignment is made.
+
+This two-step verification prevents false positives by requiring consistent predictions across multiple frames before assigning a sub label or attribute.
+
 ## Example use cases
 
 ### Sub label
@@ -56,18 +75,22 @@ classification:
 
 ## Training the model
 
-Creating and training the model is done within the Frigate UI using the `Classification` page.
+Creating and training the model is done within the Frigate UI using the `Classification` page. The process consists of two steps:
 
-### Getting Started
+### Step 1: Name and Define
+
+Enter a name for your model, select the object label to classify (e.g., `person`, `dog`, `car`), choose the classification type (sub label or attribute), and define your classes. Include a `none` class for objects that don't fit any specific category.
+
+### Step 2: Assign Training Examples
+
+The system will automatically generate example images from detected objects matching your selected label. You'll be guided through each class one at a time to select which images represent that class. Any images not assigned to a specific class will automatically be assigned to `none` when you complete the last class. Once all images are processed, training will begin automatically.
 
 When choosing which objects to classify, start with a small number of visually distinct classes and ensure your training samples match camera viewpoints and distances typical for those objects.
 
-// TODO add this section once UI is implemented. Explain process of selecting objects and curating training examples.
-
 ### Improving the Model
 
 - **Problem framing**: Keep classes visually distinct and relevant to the chosen object types.
-- **Data collection**: Use the model’s Train tab to gather balanced examples across times of day, weather, and distances.
+- **Data collection**: Use the model’s Recent Classification tab to gather balanced examples across times of day, weather, and distances.
 - **Preprocessing**: Ensure examples reflect object crops similar to Frigate’s boxes; keep the subject centered.
 - **Labels**: Keep label names short and consistent; include a `none` class if you plan to ignore uncertain predictions for sub labels.
 - **Threshold**: Tune `threshold` per model to reduce false assignments. Start at `0.8` and adjust based on validation.

docs/docs/configuration/custom_classification/state_classification.md

@@ -10,7 +10,17 @@ State classification allows you to train a custom MobileNetV2 classification mod
 State classification models are lightweight and run very fast on CPU. Inference should be usable on virtually any machine that can run Frigate.
 
 Training the model does briefly use a high amount of system resources for about 1–3 minutes per training run. On lower-power devices, training may take longer.
-When running the `-tensorrt` image, Nvidia GPUs will automatically be used to accelerate training.
+
+## Classes
+
+Classes are the different states an area on your camera can be in. Each class represents a distinct visual state that the model will learn to recognize.
+
+For state classification:
+
+- Define classes that represent mutually exclusive states
+- Examples: `open` and `closed` for a garage door, `on` and `off` for lights
+- Use at least 2 classes (typically binary states work best)
+- Keep class names clear and descriptive
 
 ## Example use cases
 
@@ -38,15 +48,25 @@ classification:
 
 ## Training the model
 
-Creating and training the model is done within the Frigate UI using the `Classification` page.
+Creating and training the model is done within the Frigate UI using the `Classification` page. The process consists of three steps:
 
-### Getting Started
+### Step 1: Name and Define
 
-When choosing a portion of the camera frame for state classification, it is important to make the crop tight around the area of interest to avoid extra signals unrelated to what is being classified.
+Enter a name for your model and define at least 2 classes (states) that represent mutually exclusive states. For example, `open` and `closed` for a door, or `on` and `off` for lights.
 
-// TODO add this section once UI is implemented. Explain process of selecting a crop.
+### Step 2: Select the Crop Area
+
+Choose one or more cameras and draw a rectangle over the area of interest for each camera. The crop should be tight around the region you want to classify to avoid extra signals unrelated to what is being classified. You can drag and resize the rectangle to adjust the crop area.
+
+### Step 3: Assign Training Examples
+
+The system will automatically generate example images from your camera feeds. You'll be guided through each class one at a time to select which images represent that state.
+
+**Important**: All images must be assigned to a state before training can begin. This includes images that may not be optimal, such as when people temporarily block the view, sun glare is present, or other distractions occur. Assign these images to the state that is actually present (based on what you know the state to be), not based on the distraction. This training helps the model correctly identify the state even when such conditions occur during inference.
+
+Once all images are assigned, training will begin automatically.
 
 ### Improving the Model
 
 - **Problem framing**: Keep classes visually distinct and state-focused (e.g., `open`, `closed`, `unknown`). Avoid combining object identity with state in a single model unless necessary.
-- **Data collection**: Use the model’s Train tab to gather balanced examples across times of day and weather.
+- **Data collection**: Use the model’s Recent Classifications tab to gather balanced examples across times of day and weather.

docs/docs/configuration/face_recognition.md

@@ -70,7 +70,7 @@ Fine-tune face recognition with these optional parameters at the global level of
 - `min_faces`: Min face recognitions for the sub label to be applied to the person object.
   - Default: `1`
 - `save_attempts`: Number of images of recognized faces to save for training.
-  - Default: `100`.
+  - Default: `200`.
 - `blur_confidence_filter`: Enables a filter that calculates how blurry the face is and adjusts the confidence based on this.
   - Default: `True`.
 - `device`: Target a specific device to run the face recognition model on (multi-GPU installation).
@@ -114,9 +114,9 @@ When choosing images to include in the face training set it is recommended to al
 
 :::
 
-### Understanding the Train Tab
+### Understanding the Recent Recognitions Tab
 
-The Train tab in the face library displays recent face recognition attempts. Detected face images are grouped according to the person they were identified as potentially matching.
+The Recent Recognitions tab in the face library displays recent face recognition attempts. Detected face images are grouped according to the person they were identified as potentially matching.
 
 Each face image is labeled with a name (or `Unknown`) along with the confidence score of the recognition attempt. While each image can be used to train the system for a specific person, not all images are suitable for training.
 
@@ -140,7 +140,7 @@ Once front-facing images are performing well, start choosing slightly off-angle
 
 Start with the [Usage](#usage) section and re-read the [Model Requirements](#model-requirements) above.
 
-1. Ensure `person` is being _detected_. A `person` will automatically be scanned by Frigate for a face. Any detected faces will appear in the Train tab in the Frigate UI's Face Library.
+1. Ensure `person` is being _detected_. A `person` will automatically be scanned by Frigate for a face. Any detected faces will appear in the Recent Recognitions tab in the Frigate UI's Face Library.
 
    If you are using a Frigate+ or `face` detecting model:
 
@@ -161,6 +161,8 @@ Start with the [Usage](#usage) section and re-read the [Model Requirements](#mod
 
 Accuracy is definitely a going to be improved with higher quality cameras / streams. It is important to look at the DORI (Detection Observation Recognition Identification) range of your camera, if that specification is posted. This specification explains the distance from the camera that a person can be detected, observed, recognized, and identified. The identification range is the most relevant here, and the distance listed by the camera is the furthest that face recognition will realistically work.
 
+Some users have also noted that setting the stream in camera firmware to a constant bit rate (CBR) leads to better image clarity than with a variable bit rate (VBR).
+
 ### Why can't I bulk upload photos?
 
 It is important to methodically add photos to the library, bulk importing photos (especially from a general photo library) will lead to over-fitting in that particular scenario and hurt recognition performance.
@@ -186,7 +188,7 @@ Avoid training on images that already score highly, as this can lead to over-fit
 No, face recognition does not support negative training (i.e., explicitly telling it who someone is _not_). Instead, the best approach is to improve the training data by using a more diverse and representative set of images for each person.
 For more guidance, refer to the section above on improving recognition accuracy.
 
-### I see scores above the threshold in the train tab, but a sub label wasn't assigned?
+### I see scores above the threshold in the Recent Recognitions tab, but a sub label wasn't assigned?
 
 The Frigate considers the recognition scores across all recognition attempts for each person object. The scores are continually weighted based on the area of the face, and a sub label will only be assigned to person if a person is confidently recognized consistently. This avoids cases where a single high confidence recognition would throw off the results.
 

docs/docs/configuration/genai.md

@@ -17,18 +17,17 @@ To use Generative AI, you must define a single provider at the global level of y
 genai:
   provider: gemini
   api_key: "{FRIGATE_GEMINI_API_KEY}"
-  model: gemini-1.5-flash
+  model: gemini-2.0-flash
 
 cameras:
   front_camera:
-    objects:
     genai:
-        enabled: True # <- enable GenAI for your front camera
+      enabled: True # <- enable GenAI for your front camera
-        use_snapshot: True
+      use_snapshot: True
-        objects:
+      objects:
-          - person
+        - person
-        required_zones:
+      required_zones:
-          - steps
+        - steps
   indoor_camera:
     objects:
       genai:
@@ -71,7 +70,7 @@ You should have at least 8 GB of RAM available (or VRAM if running on GPU) to ru
 genai:
   provider: ollama
   base_url: http://localhost:11434
-  model: llava:7b
+  model: qwen3-vl:4b
 ```
 
 ## Google Gemini
@@ -80,7 +79,7 @@ Google Gemini has a free tier allowing [15 queries per minute](https://ai.google
 
 ### Supported Models
 
-You must use a vision capable model with Frigate. Current model variants can be found [in their documentation](https://ai.google.dev/gemini-api/docs/models/gemini). At the time of writing, this includes `gemini-1.5-pro` and `gemini-1.5-flash`.
+You must use a vision capable model with Frigate. Current model variants can be found [in their documentation](https://ai.google.dev/gemini-api/docs/models/gemini).
 
 ### Get API Key
 
@@ -97,7 +96,7 @@ To start using Gemini, you must first get an API key from [Google AI Studio](htt
 genai:
   provider: gemini
   api_key: "{FRIGATE_GEMINI_API_KEY}"
-  model: gemini-1.5-flash
+  model: gemini-2.0-flash
 ```
 
 :::note
@@ -112,7 +111,7 @@ OpenAI does not have a free tier for their API. With the release of gpt-4o, pric
 
 ### Supported Models
 
-You must use a vision capable model with Frigate. Current model variants can be found [in their documentation](https://platform.openai.com/docs/models). At the time of writing, this includes `gpt-4o` and `gpt-4-turbo`.
+You must use a vision capable model with Frigate. Current model variants can be found [in their documentation](https://platform.openai.com/docs/models).
 
 ### Get API Key
 
@@ -139,18 +138,19 @@ Microsoft offers several vision models through Azure OpenAI. A subscription is r
 
 ### Supported Models
 
-You must use a vision capable model with Frigate. Current model variants can be found [in their documentation](https://learn.microsoft.com/en-us/azure/ai-services/openai/concepts/models). At the time of writing, this includes `gpt-4o` and `gpt-4-turbo`.
+You must use a vision capable model with Frigate. Current model variants can be found [in their documentation](https://learn.microsoft.com/en-us/azure/ai-services/openai/concepts/models).
 
 ### Create Resource and Get API Key
 
-To start using Azure OpenAI, you must first [create a resource](https://learn.microsoft.com/azure/cognitive-services/openai/how-to/create-resource?pivots=web-portal#create-a-resource). You'll need your API key and resource URL, which must include the `api-version` parameter (see the example below). The model field is not required in your configuration as the model is part of the deployment name you chose when deploying the resource.
+To start using Azure OpenAI, you must first [create a resource](https://learn.microsoft.com/azure/cognitive-services/openai/how-to/create-resource?pivots=web-portal#create-a-resource). You'll need your API key, model name, and resource URL, which must include the `api-version` parameter (see the example below).
 
 ### Configuration
 
 ```yaml
 genai:
   provider: azure_openai
-  base_url: https://example-endpoint.openai.azure.com/openai/deployments/gpt-4o/chat/completions?api-version=2023-03-15-preview
+  base_url: https://instance.cognitiveservices.azure.com/openai/responses?api-version=2025-04-01-preview
+  model: gpt-5-mini
   api_key: "{FRIGATE_OPENAI_API_KEY}"
 ```
 
@@ -196,10 +196,10 @@ genai:
   model: llava
 
 objects:
-    prompt: "Analyze the {label} in these images from the {camera} security camera. Focus on the actions, behavior, and potential intent of the {label}, rather than just describing its appearance."
+  prompt: "Analyze the {label} in these images from the {camera} security camera. Focus on the actions, behavior, and potential intent of the {label}, rather than just describing its appearance."
-    object_prompts:
+  object_prompts:
-      person: "Examine the main person in these images. What are they doing and what might their actions suggest about their intent (e.g., approaching a door, leaving an area, standing still)? Do not describe the surroundings or static details."
+    person: "Examine the main person in these images. What are they doing and what might their actions suggest about their intent (e.g., approaching a door, leaving an area, standing still)? Do not describe the surroundings or static details."
-      car: "Observe the primary vehicle in these images. Focus on its movement, direction, or purpose (e.g., parking, approaching, circling). If it's a delivery vehicle, mention the company."
+    car: "Observe the primary vehicle in these images. Focus on its movement, direction, or purpose (e.g., parking, approaching, circling). If it's a delivery vehicle, mention the company."
 ```
 
 Prompts can also be overridden at the camera level to provide a more detailed prompt to the model about your specific camera, if you desire.

docs/docs/configuration/genai/config.md

@@ -35,18 +35,18 @@ Each model is available in multiple parameter sizes (3b, 4b, 8b, etc.). Larger s
 
 :::tip
 
-If you are trying to use a single model for Frigate and HomeAssistant, it will need to support vision and tools calling. https://github.com/skye-harris/ollama-modelfiles contains optimized model configs for this task.
+If you are trying to use a single model for Frigate and HomeAssistant, it will need to support vision and tools calling. qwen3-VL supports vision and tools simultaneously in Ollama.
 
 :::
 
 The following models are recommended:
 
-| Model             | Notes                                                       |
+| Model             | Notes                                                                |
-| ----------------- | ----------------------------------------------------------- |
+| ----------------- | -------------------------------------------------------------------- |
-| `Intern3.5VL`     | Relatively fast with good vision comprehension
+| `qwen3-vl`        | Strong visual and situational understanding, higher vram requirement |
-| `gemma3`          | Strong frame-to-frame understanding, slower inference times |
+| `Intern3.5VL`     | Relatively fast with good vision comprehension                       |
-| `qwen2.5vl`       | Fast but capable model with good vision comprehension       |
+| `gemma3`          | Strong frame-to-frame understanding, slower inference times          |
-| `llava-phi3`      | Lightweight and fast model with vision comprehension        |
+| `qwen2.5-vl`      | Fast but capable model with good vision comprehension                |
 
 :::note
 

docs/docs/configuration/genai/review_summaries.md

@@ -7,38 +7,95 @@ Generative AI can be used to automatically generate structured summaries of revi
 
 Requests for a summary are requested automatically to your AI provider for alert review items when the activity has ended, they can also be optionally enabled for detections as well.
 
-Generative AI review summaries can also be toggled dynamically for a camera via MQTT with the topic `frigate/<camera_name>/review_descriptions/set`. See the [MQTT documentation](/integrations/mqtt/#frigatecamera_namereviewdescriptionsset).
+Generative AI review summaries can also be toggled dynamically for a [camera via MQTT](/integrations/mqtt/#frigatecamera_namereviewdescriptionsset).
 
 ## Review Summary Usage and Best Practices
 
 Review summaries provide structured JSON responses that are saved for each review item:
 
 ```
-- `scene` (string): A full description including setting, entities, actions, and any plausible supported inferences.
+- `title` (string): A concise, direct title that describes the purpose or overall action (e.g., "Person taking out trash", "Joe walking dog").
-- `confidence` (float): 0-1 confidence in the analysis.
+- `scene` (string): A narrative description of what happens across the sequence from start to finish, including setting, detected objects, and their observable actions.
+- `confidence` (float): 0-1 confidence in the analysis. Higher confidence when objects/actions are clearly visible and context is unambiguous.
 - `other_concerns` (list): List of user-defined concerns that may need additional investigation.
 - `potential_threat_level` (integer): 0, 1, or 2 as defined below.
-
-Threat-level definitions:
-- 0 — Typical or expected activity for this location/time (includes residents, guests, or known animals engaged in normal activities, even if they glance around or scan surroundings).
-- 1 — Unusual or suspicious activity: At least one security-relevant behavior is present **and not explainable by a normal residential activity**.
-- 2 — Active or immediate threat: Breaking in, vandalism, aggression, weapon display.
 ```
 
-This will show in the UI as a list of concerns that each review item has along with the general description.
+This will show in multiple places in the UI to give additional context about each activity, and allow viewing more details when extra attention is required. Frigate's built in notifications will also automatically show the title and description when the data is available.
 
 ### Defining Typical Activity
 
-Each installation and even camera can have different parameters for what is considered suspicious activity. Frigate allows the `activity_context_prompt` to be defined globally and at the camera level, which allows you to define more specifically what should be considered normal activity. It is important that this is not overly specific as it can sway the output of the response. The default `activity_context_prompt` is below:
+Each installation and even camera can have different parameters for what is considered suspicious activity. Frigate allows the `activity_context_prompt` to be defined globally and at the camera level, which allows you to define more specifically what should be considered normal activity. It is important that this is not overly specific as it can sway the output of the response.
+
+<details>
+  <summary>Default Activity Context Prompt</summary>
 
 ```
-- **Zone context is critical**: Private enclosed spaces (back yards, back decks, fenced areas, inside garages) are resident territory where brief transient activity, routine tasks, and pet care are expected and normal. Front yards, driveways, and porches are semi-public but still resident spaces where deliveries, parking, and coming/going are routine. Consider whether the zone and activity align with normal residential use.
+### Normal Activity Indicators (Level 0)
-- **Person + Pet = Normal Activity**: When both "Person" and "Dog" (or "Cat") are detected together in residential zones, this is routine pet care activity (walking, letting out, playing, supervising). Assign Level 0 unless there are OTHER strong suspicious behaviors present (like testing doors, taking items, etc.). A person with their pet in a residential zone is baseline normal activity.
+- Known/verified people in any zone at any time
-- Brief appearances in private zones (back yards, garages) are normal residential patterns.
+- People with pets in residential areas
-- Normal residential activity includes: residents, family members, guests, deliveries, services, maintenance workers, routine property use (parking, unloading, mail pickup, trash removal).
+- Deliveries or services during daytime/evening (6 AM - 10 PM): carrying packages to doors/porches, placing items, leaving
-- Brief movement with legitimate items (bags, packages, tools, equipment) in appropriate zones is routine.
+- Services/maintenance workers with visible tools, uniforms, or service vehicles during daytime
+- Activity confined to public areas only (sidewalks, streets) without entering property at any time
+
+### Suspicious Activity Indicators (Level 1)
+- **Testing or attempting to open doors/windows/handles on vehicles or buildings** — ALWAYS Level 1 regardless of time or duration
+- **Unidentified person in private areas (driveways, near vehicles/buildings) during late night/early morning (11 PM - 5 AM)** — ALWAYS Level 1 regardless of activity or duration
+- Taking items that don't belong to them (packages, objects from porches/driveways)
+- Climbing or jumping fences/barriers to access property
+- Attempting to conceal actions or items from view
+- Prolonged loitering: remaining in same area without visible purpose throughout most of the sequence
+
+### Critical Threat Indicators (Level 2)
+- Holding break-in tools (crowbars, pry bars, bolt cutters)
+- Weapons visible (guns, knives, bats used aggressively)
+- Forced entry in progress
+- Physical aggression or violence
+- Active property damage or theft in progress
+
+### Assessment Guidance
+Evaluate in this order:
+
+1. **If person is verified/known** → Level 0 regardless of time or activity
+2. **If person is unidentified:**
+   - Check time: If late night/early morning (11 PM - 5 AM) AND in private areas (driveways, near vehicles/buildings) → Level 1
+   - Check actions: If testing doors/handles, taking items, climbing → Level 1
+   - Otherwise, if daytime/evening (6 AM - 10 PM) with clear legitimate purpose (delivery, service worker) → Level 0
+3. **Escalate to Level 2 if:** Weapons, break-in tools, forced entry in progress, violence, or active property damage visible (escalates from Level 0 or 1)
+
+The mere presence of an unidentified person in private areas during late night hours is inherently suspicious and warrants human review, regardless of what activity they appear to be doing or how brief the sequence is.
 ```
 
+</details>
+
+### Image Source
+
+By default, review summaries use preview images (cached preview frames) which have a lower resolution but use fewer tokens per image. For better image quality and more detailed analysis, you can configure Frigate to extract frames directly from recordings at a higher resolution:
+
+```yaml
+review:
+  genai:
+    enabled: true
+    image_source: recordings # Options: "preview" (default) or "recordings"
+```
+
+When using `recordings`, frames are extracted at 480px height while maintaining the camera's original aspect ratio, providing better detail for the LLM while being mindful of context window size. This is particularly useful for scenarios where fine details matter, such as identifying license plates, reading text, or analyzing distant objects.
+
+The number of frames sent to the LLM is dynamically calculated based on:
+
+- Your LLM provider's context window size
+- The camera's resolution and aspect ratio (ultrawide cameras like 32:9 use more tokens per image)
+- The image source (recordings use more tokens than preview images)
+
+Frame counts are automatically optimized to use ~98% of the available context window while capping at 20 frames maximum to ensure reasonable inference times. Note that using recordings will:
+
+- Provide higher quality images to the LLM (480p vs 180p preview images)
+- Use more tokens per image due to higher resolution
+- Result in fewer frames being sent for ultrawide cameras due to larger image size
+- Require that recordings are enabled for the camera
+
+If recordings are not available for a given time period, the system will automatically fall back to using preview frames.
+
 ### Additional Concerns
 
 Along with the concern of suspicious activity or immediate threat, you may have concerns such as animals in your garden or a gate being left open. These concerns can be configured so that the review summaries will make note of them if the activity requires additional review. For example:

docs/docs/configuration/hardware_acceleration_enrichments.md

@@ -5,7 +5,7 @@ title: Enrichments
 
 # Enrichments
 
-Some of Frigate's enrichments can use a discrete GPU / NPU for accelerated processing.
+Some of Frigate's enrichments can use a discrete GPU or integrated GPU for accelerated processing.
 
 ## Requirements
 
@@ -18,8 +18,10 @@ Object detection and enrichments (like Semantic Search, Face Recognition, and Li
 - **Intel**
 
   - OpenVINO will automatically be detected and used for enrichments in the default Frigate image.
+  - **Note:** Intel NPUs have limited model support for enrichments. GPU is recommended for enrichments when available.
 
 - **Nvidia**
+
   - Nvidia GPUs will automatically be detected and used for enrichments in the `-tensorrt` Frigate image.
   - Jetson devices will automatically be detected and used for enrichments in the `-tensorrt-jp6` Frigate image.
 

docs/docs/configuration/license_plate_recognition.md

@@ -3,18 +3,18 @@ id: license_plate_recognition
 title: License Plate Recognition (LPR)
 ---
 
-Frigate can recognize license plates on vehicles and automatically add the detected characters to the `recognized_license_plate` field or a known name as a `sub_label` to tracked objects of type `car` or `motorcycle`. A common use case may be to read the license plates of cars pulling into a driveway or cars passing by on a street.
+Frigate can recognize license plates on vehicles and automatically add the detected characters to the `recognized_license_plate` field or a [known](#matching) name as a `sub_label` to tracked objects of type `car` or `motorcycle`. A common use case may be to read the license plates of cars pulling into a driveway or cars passing by on a street.
 
 LPR works best when the license plate is clearly visible to the camera. For moving vehicles, Frigate continuously refines the recognition process, keeping the most confident result. When a vehicle becomes stationary, LPR continues to run for a short time after to attempt recognition.
 
 When a plate is recognized, the details are:
 
-- Added as a `sub_label` (if known) or the `recognized_license_plate` field (if unknown) to a tracked object.
+- Added as a `sub_label` (if [known](#matching)) or the `recognized_license_plate` field (if unknown) to a tracked object.
-- Viewable in the Review Item Details pane in Review (sub labels).
+- Viewable in the Details pane in Review/History.
 - Viewable in the Tracked Object Details pane in Explore (sub labels and recognized license plates).
 - Filterable through the More Filters menu in Explore.
-- Published via the `frigate/events` MQTT topic as a `sub_label` (known) or `recognized_license_plate` (unknown) for the `car` or `motorcycle` tracked object.
+- Published via the `frigate/events` MQTT topic as a `sub_label` ([known](#matching)) or `recognized_license_plate` (unknown) for the `car` or `motorcycle` tracked object.
-- Published via the `frigate/tracked_object_update` MQTT topic with `name` (if known) and `plate`.
+- Published via the `frigate/tracked_object_update` MQTT topic with `name` (if [known](#matching)) and `plate`.
 
 ## Model Requirements
 
@@ -30,7 +30,7 @@ In the default mode, Frigate's LPR needs to first detect a `car` or `motorcycle`
 
 ## Minimum System Requirements
 
-License plate recognition works by running AI models locally on your system. The models are relatively lightweight and can run on your CPU or GPU, depending on your configuration. At least 4GB of RAM is required.
+License plate recognition works by running AI models locally on your system. The YOLOv9 plate detector model and the OCR models ([PaddleOCR](https://github.com/PaddlePaddle/PaddleOCR)) are relatively lightweight and can run on your CPU or GPU, depending on your configuration. At least 4GB of RAM is required.
 
 ## Configuration
 
@@ -74,8 +74,8 @@ Fine-tune the LPR feature using these optional parameters at the global level of
   - Default: `small`
   - This can be `small` or `large`.
   - The `small` model is fast and identifies groups of Latin and Chinese characters.
-  - The `large` model identifies Latin characters only, but uses an enhanced text detector and is more capable at finding characters on multi-line plates. It is significantly slower than the `small` model. Note that using the `large` model does not improve _text recognition_, but it may improve _text detection_.
+  - The `large` model identifies Latin characters only, and uses an enhanced text detector to find characters on multi-line plates. It is significantly slower than the `small` model.
-  - For most users, the `small` model is recommended.
+  - If your country or region does not use multi-line plates, you should use the `small` model as performance is much better for single-line plates.
 
 ### Recognition
 
@@ -178,7 +178,7 @@ lpr:
 
 :::note
 
-If you want to detect cars on cameras but don't want to use resources to run LPR on those cars, you should disable LPR for those specific cameras.
+If a camera is configured to detect `car` or `motorcycle` but you don't want Frigate to run LPR for that camera, disable LPR at the camera level:
 
 ```yaml
 cameras:
@@ -306,7 +306,7 @@ With this setup:
 - Review items will always be classified as a `detection`.
 - Snapshots will always be saved.
 - Zones and object masks are **not** used.
-- The `frigate/events` MQTT topic will **not** publish tracked object updates with the license plate bounding box and score, though `frigate/reviews` will publish if recordings are enabled. If a plate is recognized as a known plate, publishing will occur with an updated `sub_label` field. If characters are recognized, publishing will occur with an updated `recognized_license_plate` field.
+- The `frigate/events` MQTT topic will **not** publish tracked object updates with the license plate bounding box and score, though `frigate/reviews` will publish if recordings are enabled. If a plate is recognized as a [known](#matching) plate, publishing will occur with an updated `sub_label` field. If characters are recognized, publishing will occur with an updated `recognized_license_plate` field.
 - License plate snapshots are saved at the highest-scoring moment and appear in Explore.
 - Debug view will not show `license_plate` bounding boxes.
 

docs/docs/configuration/live.md

@@ -174,7 +174,7 @@ For devices that support two way talk, Frigate can be configured to use the feat
 - Ensure you access Frigate via https (may require [opening port 8971](/frigate/installation/#ports)).
 - For the Home Assistant Frigate card, [follow the docs](http://card.camera/#/usage/2-way-audio) for the correct source.
 
-To use the Reolink Doorbell with two way talk, you should use the [recommended Reolink configuration](/configuration/camera_specific#reolink-doorbell)
+To use the Reolink Doorbell with two way talk, you should use the [recommended Reolink configuration](/configuration/camera_specific#reolink-cameras)
 
 As a starting point to check compatibility for your camera, view the list of cameras supported for two-way talk on the [go2rtc repository](https://github.com/AlexxIT/go2rtc?tab=readme-ov-file#two-way-audio). For cameras in the category `ONVIF Profile T`, you can use the [ONVIF Conformant Products Database](https://www.onvif.org/conformant-products/)'s FeatureList to check for the presence of `AudioOutput`. A camera that supports `ONVIF Profile T` _usually_ supports this, but due to inconsistent support, a camera that explicitly lists this feature may still not work. If no entry for your camera exists on the database, it is recommended not to buy it or to consult with the manufacturer's support on the feature availability.
 
@@ -214,6 +214,42 @@ For restreamed cameras, go2rtc remains active but does not use system resources
 
 Note that disabling a camera through the config file (`enabled: False`) removes all related UI elements, including historical footage access. To retain access while disabling the camera, keep it enabled in the config and use the UI or MQTT to disable it temporarily.
 
+### Live player error messages
+
+When your browser runs into problems playing back your camera streams, it will log short error messages to the browser console. They indicate playback, codec, or network issues on the client/browser side, not something server side with Frigate itself. Below are the common messages you may see and simple actions you can take to try to resolve them.
+
+- **startup**
+
+  - What it means: The player failed to initialize or connect to the live stream (network or startup error).
+  - What to try: Reload the Live view or click _Reset_. Verify `go2rtc` is running and the camera stream is reachable. Try switching to a different stream from the Live UI dropdown (if available) or use a different browser.
+
+  - Possible console messages from the player code:
+
+    - `Error opening MediaSource.`
+    - `Browser reported a network error.`
+    - `Max error count ${errorCount} exceeded.` (the numeric value will vary)
+
+- **mse-decode**
+
+  - What it means: The browser reported a decoding error while trying to play the stream, which usually is a result of a codec incompatibility or corrupted frames.
+  - What to try: Ensure your camera/restream is using H.264 video and AAC audio (these are the most compatible). If your camera uses a non-standard audio codec, configure `go2rtc` to transcode the stream to AAC. Try another browser (some browsers have stricter MSE/codec support) and, for iPhone, ensure you're on iOS 17.1 or newer.
+
+  - Possible console messages from the player code:
+
+    - `Safari cannot open MediaSource.`
+    - `Safari reported InvalidStateError.`
+    - `Safari reported decoding errors.`
+
+- **stalled**
+
+  - What it means: Playback has stalled because the player has fallen too far behind live (extended buffering or no data arriving).
+  - What to try: This is usually indicative of the browser struggling to decode too many high-resolution streams at once. Try selecting a lower-bandwidth stream (substream), reduce the number of live streams open, improve the network connection, or lower the camera resolution. Also check your camera's keyframe (I-frame) interval — shorter intervals make playback start and recover faster. You can also try increasing the timeout value in the UI pane of Frigate's settings.
+
+  - Possible console messages from the player code:
+
+    - `Buffer time (10 seconds) exceeded, browser may not be playing media correctly.`
+    - `Media playback has stalled after <n> seconds due to insufficient buffering or a network interruption.` (the seconds value will vary)
+
 ## Live view FAQ
 
 1. **Why don't I have audio in my Live view?**
@@ -277,3 +313,38 @@ Note that disabling a camera through the config file (`enabled: False`) removes
 7. **My camera streams have lots of visual artifacts / distortion.**
 
    Some cameras don't include the hardware to support multiple connections to the high resolution stream, and this can cause unexpected behavior. In this case it is recommended to [restream](./restream.md) the high resolution stream so that it can be used for live view and recordings.
+
+8. **Why does my camera stream switch aspect ratios on the Live dashboard?**
+
+   Your camera may change aspect ratios on the dashboard because Frigate uses different streams for different purposes. With go2rtc and Smart Streaming, Frigate shows a static image from the `detect` stream when no activity is present, and switches to the live stream when motion is detected. The camera image will change size if your streams use different aspect ratios.
+
+   To prevent this, make the `detect` stream match the go2rtc live stream's aspect ratio (resolution does not need to match, just the aspect ratio). You can either adjust the camera's output resolution or set the `width` and `height` values in your config's `detect` section to a resolution with an aspect ratio that matches.
+
+   Example: Resolutions from two streams
+
+   - Mismatched (may cause aspect ratio switching on the dashboard):
+
+     - Live/go2rtc stream: 1920x1080 (16:9)
+     - Detect stream: 640x352 (~1.82:1, not 16:9)
+
+   - Matched (prevents switching):
+     - Live/go2rtc stream: 1920x1080 (16:9)
+     - Detect stream: 640x360 (16:9)
+
+   You can update the detect settings in your camera config to match the aspect ratio of your go2rtc live stream. For example:
+
+   ```yaml
+   cameras:
+     front_door:
+       detect:
+         width: 640
+         height: 360 # set this to 360 instead of 352
+       ffmpeg:
+         inputs:
+           - path: rtsp://127.0.0.1:8554/front_door # main stream 1920x1080
+             roles:
+               - record
+           - path: rtsp://127.0.0.1:8554/front_door_sub # sub stream 640x352
+             roles:
+               - detect
+   ```

docs/docs/configuration/object_detectors.md

@@ -3,6 +3,8 @@ id: object_detectors
 title: Object Detectors
 ---
 
+import CommunityBadge from '@site/src/components/CommunityBadge';
+
 # Supported Hardware
 
 :::info
@@ -13,8 +15,8 @@ Frigate supports multiple different detectors that work on different types of ha
 
 - [Coral EdgeTPU](#edge-tpu-detector): The Google Coral EdgeTPU is available in USB and m.2 format allowing for a wide range of compatibility with devices.
 - [Hailo](#hailo-8): The Hailo8 and Hailo8L AI Acceleration module is available in m.2 format with a HAT for RPi devices, offering a wide range of compatibility with devices.
-- [MemryX](#memryx-mx3): The MX3 Acceleration module is available in m.2 format, offering broad compatibility across various platforms.
+- <CommunityBadge /> [MemryX](#memryx-mx3): The MX3 Acceleration module is available in m.2 format, offering broad compatibility across various platforms.
-- [DeGirum](#degirum): Service for using hardware devices in the cloud or locally. Hardware and models provided on the cloud on [their website](https://hub.degirum.com).
+- <CommunityBadge /> [DeGirum](#degirum): Service for using hardware devices in the cloud or locally. Hardware and models provided on the cloud on [their website](https://hub.degirum.com).
 
 **AMD**
 
@@ -34,16 +36,16 @@ Frigate supports multiple different detectors that work on different types of ha
 
 - [ONNX](#onnx): TensorRT will automatically be detected and used as a detector in the `-tensorrt` Frigate image when a supported ONNX model is configured.
 
-**Nvidia Jetson**
+**Nvidia Jetson** <CommunityBadge />
 
 - [TensortRT](#nvidia-tensorrt-detector): TensorRT can run on Jetson devices, using one of many default models.
 - [ONNX](#onnx): TensorRT will automatically be detected and used as a detector in the `-tensorrt-jp6` Frigate image when a supported ONNX model is configured.
 
-**Rockchip**
+**Rockchip** <CommunityBadge />
 
 - [RKNN](#rockchip-platform): RKNN models can run on Rockchip devices with included NPUs.
 
-**Synaptics**
+**Synaptics** <CommunityBadge />
 
 - [Synaptics](#synaptics): synap models can run on Synaptics devices(e.g astra machina) with included NPUs.
 
@@ -258,41 +260,55 @@ Hailo8 supports all models in the Hailo Model Zoo that include HailoRT post-proc
 
 ## OpenVINO Detector
 
-The OpenVINO detector type runs an OpenVINO IR model on AMD and Intel CPUs, Intel GPUs and Intel VPU hardware. To configure an OpenVINO detector, set the `"type"` attribute to `"openvino"`.
+The OpenVINO detector type runs an OpenVINO IR model on AMD and Intel CPUs, Intel GPUs and Intel NPUs. To configure an OpenVINO detector, set the `"type"` attribute to `"openvino"`.
 
-The OpenVINO device to be used is specified using the `"device"` attribute according to the naming conventions in the [Device Documentation](https://docs.openvino.ai/2024/openvino-workflow/running-inference/inference-devices-and-modes.html). The most common devices are `CPU` and `GPU`. Currently, there is a known issue with using `AUTO`. For backwards compatibility, Frigate will attempt to use `GPU` if `AUTO` is set in your configuration.
+The OpenVINO device to be used is specified using the `"device"` attribute according to the naming conventions in the [Device Documentation](https://docs.openvino.ai/2025/openvino-workflow/running-inference/inference-devices-and-modes.html). The most common devices are `CPU`, `GPU`, or `NPU`.
 
-OpenVINO is supported on 6th Gen Intel platforms (Skylake) and newer. It will also run on AMD CPUs despite having no official support for it. A supported Intel platform is required to use the `GPU` device with OpenVINO. For detailed system requirements, see [OpenVINO System Requirements](https://docs.openvino.ai/2024/about-openvino/release-notes-openvino/system-requirements.html)
+OpenVINO is supported on 6th Gen Intel platforms (Skylake) and newer. It will also run on AMD CPUs despite having no official support for it. A supported Intel platform is required to use the `GPU` or `NPU` device with OpenVINO. For detailed system requirements, see [OpenVINO System Requirements](https://docs.openvino.ai/2025/about-openvino/release-notes-openvino/system-requirements.html)
 
 :::tip
 
+**NPU + GPU Systems:** If you have both NPU and GPU available (Intel Core Ultra processors), use NPU for object detection and GPU for enrichments (semantic search, face recognition, etc.) for best performance and compatibility.
+
 When using many cameras one detector may not be enough to keep up. Multiple detectors can be defined assuming GPU resources are available. An example configuration would be:
 
 ```yaml
 detectors:
   ov_0:
     type: openvino
-    device: GPU
+    device: GPU # or NPU
   ov_1:
     type: openvino
-    device: GPU
+    device: GPU # or NPU
 ```
 
 :::
 
 ### OpenVINO Supported Models
 
+| Model                                 | GPU | NPU | Notes                                                        |
+| ------------------------------------- | --- | --- | ------------------------------------------------------------ |
+| [YOLOv9](#yolo-v3-v4-v7-v9)           | ✅  | ✅  | Recommended for GPU & NPU                                    |
+| [RF-DETR](#rf-detr)                   | ✅  | ✅  | Requires XE iGPU or Arc                                      |
+| [YOLO-NAS](#yolo-nas)                 | ✅  | ✅  |                                                              |
+| [MobileNet v2](#ssdlite-mobilenet-v2) | ✅  | ✅  | Fast and lightweight model, less accurate than larger models |
+| [YOLOX](#yolox)                       | ✅  | ?   |                                                              |
+| [D-FINE](#d-fine)                     | ❌  | ❌  |                                                              |
+
 #### SSDLite MobileNet v2
 
 An OpenVINO model is provided in the container at `/openvino-model/ssdlite_mobilenet_v2.xml` and is used by this detector type by default. The model comes from Intel's Open Model Zoo [SSDLite MobileNet V2](https://github.com/openvinotoolkit/open_model_zoo/tree/master/models/public/ssdlite_mobilenet_v2) and is converted to an FP16 precision IR model.
 
+<details>
+  <summary>MobileNet v2 Config</summary>
+
 Use the model configuration shown below when using the OpenVINO detector with the default OpenVINO model:
 
 ```yaml
 detectors:
   ov:
     type: openvino
-    device: GPU
+    device: GPU # Or NPU
 
 model:
   width: 300
@@ -303,6 +319,8 @@ model:
   labelmap_path: /openvino-model/coco_91cl_bkgr.txt
 ```
 
+</details>
+
 #### YOLOX
 
 This detector also supports YOLOX. Frigate does not come with any YOLOX models preloaded, so you will need to supply your own models.
@@ -311,6 +329,9 @@ This detector also supports YOLOX. Frigate does not come with any YOLOX models p
 
 [YOLO-NAS](https://github.com/Deci-AI/super-gradients/blob/master/YOLONAS.md) models are supported, but not included by default. See [the models section](#downloading-yolo-nas-model) for more information on downloading the YOLO-NAS model for use in Frigate.
 
+<details>
+  <summary>YOLO-NAS Setup & Config</summary>
+
 After placing the downloaded onnx model in your config folder, you can use the following configuration:
 
 ```yaml
@@ -331,6 +352,8 @@ model:
 
 Note that the labelmap uses a subset of the complete COCO label set that has only 80 objects.
 
+</details>
+
 #### YOLO (v3, v4, v7, v9)
 
 YOLOv3, YOLOv4, YOLOv7, and [YOLOv9](https://github.com/WongKinYiu/yolov9) models are supported, but not included by default.
@@ -341,6 +364,9 @@ The YOLO detector has been designed to support YOLOv3, YOLOv4, YOLOv7, and YOLOv
 
 :::
 
+<details>
+  <summary>YOLOv Setup & Config</summary>
+
 :::warning
 
 If you are using a Frigate+ YOLOv9 model, you should not define any of the below `model` parameters in your config except for `path`. See [the Frigate+ model docs](/plus/first_model#step-3-set-your-model-id-in-the-config) for more information on setting up your model.
@@ -353,7 +379,7 @@ After placing the downloaded onnx model in your config folder, you can use the f
 detectors:
   ov:
     type: openvino
-    device: GPU
+    device: GPU # or NPU
 
 model:
   model_type: yolo-generic
@@ -367,6 +393,8 @@ model:
 
 Note that the labelmap uses a subset of the complete COCO label set that has only 80 objects.
 
+</details>
+
 #### RF-DETR
 
 [RF-DETR](https://github.com/roboflow/rf-detr) is a DETR based model. The ONNX exported models are supported, but not included by default. See [the models section](#downloading-rf-detr-model) for more informatoin on downloading the RF-DETR model for use in Frigate.
@@ -377,6 +405,9 @@ Due to the size and complexity of the RF-DETR model, it is only recommended to b
 
 :::
 
+<details>
+  <summary>RF-DETR Setup & Config</summary>
+
 After placing the downloaded onnx model in your `config/model_cache` folder, you can use the following configuration:
 
 ```yaml
@@ -394,6 +425,8 @@ model:
   path: /config/model_cache/rfdetr.onnx
 ```
 
+</details>
+
 #### D-FINE
 
 [D-FINE](https://github.com/Peterande/D-FINE) is a DETR based model. The ONNX exported models are supported, but not included by default. See [the models section](#downloading-d-fine-model) for more information on downloading the D-FINE model for use in Frigate.
@@ -404,6 +437,9 @@ Currently D-FINE models only run on OpenVINO in CPU mode, GPUs currently fail to
 
 :::
 
+<details>
+  <summary>D-FINE Setup & Config</summary>
+
 After placing the downloaded onnx model in your config/model_cache folder, you can use the following configuration:
 
 ```yaml
@@ -418,15 +454,17 @@ model:
   height: 640
   input_tensor: nchw
   input_dtype: float
-  path: /config/model_cache/dfine_s_obj2coco.onnx
+  path: /config/model_cache/dfine-s.onnx
   labelmap_path: /labelmap/coco-80.txt
 ```
 
 Note that the labelmap uses a subset of the complete COCO label set that has only 80 objects.
 
+</details>
+
 ## Apple Silicon detector
 
-The NPU in Apple Silicon can't be accessed from within a container, so the [Apple Silicon detector client](https://github.com/frigate-nvr/apple-silicon-detector) must first be setup. It is recommended to use the Frigate docker image with `-standard-arm64` suffix, for example  `ghcr.io/blakeblackshear/frigate:stable-standard-arm64`.
+The NPU in Apple Silicon can't be accessed from within a container, so the [Apple Silicon detector client](https://github.com/frigate-nvr/apple-silicon-detector) must first be setup. It is recommended to use the Frigate docker image with `-standard-arm64` suffix, for example `ghcr.io/blakeblackshear/frigate:stable-standard-arm64`.
 
 ### Setup
 
@@ -614,12 +652,23 @@ detectors:
 
 ### ONNX Supported Models
 
+| Model                         | Nvidia GPU | AMD GPU | Notes                                               |
+| ----------------------------- | ---------- | ------- | --------------------------------------------------- |
+| [YOLOv9](#yolo-v3-v4-v7-v9-2) | ✅         | ✅      | Supports CUDA Graphs for optimal Nvidia performance |
+| [RF-DETR](#rf-detr)           | ✅         | ❌      | Supports CUDA Graphs for optimal Nvidia performance |
+| [YOLO-NAS](#yolo-nas-1)       | ⚠️         | ⚠️      | Not supported by CUDA Graphs                        |
+| [YOLOX](#yolox-1)             | ✅         | ✅      | Supports CUDA Graphs for optimal Nvidia performance |
+| [D-FINE](#d-fine)             | ⚠️         | ❌      | Not supported by CUDA Graphs                        |
+
 There is no default model provided, the following formats are supported:
 
 #### YOLO-NAS
 
 [YOLO-NAS](https://github.com/Deci-AI/super-gradients/blob/master/YOLONAS.md) models are supported, but not included by default. See [the models section](#downloading-yolo-nas-model) for more information on downloading the YOLO-NAS model for use in Frigate.
 
+<details>
+  <summary>YOLO-NAS Setup & Config</summary>
+
 :::warning
 
 If you are using a Frigate+ YOLO-NAS model, you should not define any of the below `model` parameters in your config except for `path`. See [the Frigate+ model docs](/plus/first_model#step-3-set-your-model-id-in-the-config) for more information on setting up your model.
@@ -643,6 +692,8 @@ model:
   labelmap_path: /labelmap/coco-80.txt
 ```
 
+</details>
+
 #### YOLO (v3, v4, v7, v9)
 
 YOLOv3, YOLOv4, YOLOv7, and [YOLOv9](https://github.com/WongKinYiu/yolov9) models are supported, but not included by default.
@@ -653,6 +704,9 @@ The YOLO detector has been designed to support YOLOv3, YOLOv4, YOLOv7, and YOLOv
 
 :::
 
+<details>
+  <summary>YOLOv Setup & Config</summary>
+
 :::warning
 
 If you are using a Frigate+ YOLOv9 model, you should not define any of the below `model` parameters in your config except for `path`. See [the Frigate+ model docs](/plus/first_model#step-3-set-your-model-id-in-the-config) for more information on setting up your model.
@@ -676,12 +730,17 @@ model:
   labelmap_path: /labelmap/coco-80.txt
 ```
 
+</details>
+
 Note that the labelmap uses a subset of the complete COCO label set that has only 80 objects.
 
 #### YOLOx
 
 [YOLOx](https://github.com/Megvii-BaseDetection/YOLOX) models are supported, but not included by default. See [the models section](#downloading-yolo-models) for more information on downloading the YOLOx model for use in Frigate.
 
+<details>
+  <summary>YOLOx Setup & Config</summary>
+
 After placing the downloaded onnx model in your config folder, you can use the following configuration:
 
 ```yaml
@@ -701,10 +760,15 @@ model:
 
 Note that the labelmap uses a subset of the complete COCO label set that has only 80 objects.
 
+</details>
+
 #### RF-DETR
 
 [RF-DETR](https://github.com/roboflow/rf-detr) is a DETR based model. The ONNX exported models are supported, but not included by default. See [the models section](#downloading-rf-detr-model) for more information on downloading the RF-DETR model for use in Frigate.
 
+<details>
+  <summary>RF-DETR Setup & Config</summary>
+
 After placing the downloaded onnx model in your `config/model_cache` folder, you can use the following configuration:
 
 ```yaml
@@ -721,10 +785,15 @@ model:
   path: /config/model_cache/rfdetr.onnx
 ```
 
+</details>
+
 #### D-FINE
 
 [D-FINE](https://github.com/Peterande/D-FINE) is a DETR based model. The ONNX exported models are supported, but not included by default. See [the models section](#downloading-d-fine-model) for more information on downloading the D-FINE model for use in Frigate.
 
+<details>
+  <summary>D-FINE Setup & Config</summary>
+
 After placing the downloaded onnx model in your `config/model_cache` folder, you can use the following configuration:
 
 ```yaml
@@ -742,6 +811,8 @@ model:
   labelmap_path: /labelmap/coco-80.txt
 ```
 
+</details>
+
 Note that the labelmap uses a subset of the complete COCO label set that has only 80 objects.
 
 ## CPU Detector (not recommended)
@@ -861,16 +932,16 @@ detectors:
 
 model:
   model_type: yolonas
-  width: 320   # (Can be set to 640 for higher resolution)
+  width: 320 # (Can be set to 640 for higher resolution)
-  height: 320  # (Can be set to 640 for higher resolution)
+  height: 320 # (Can be set to 640 for higher resolution)
   input_tensor: nchw
   input_dtype: float
   labelmap_path: /labelmap/coco-80.txt
   # Optional: The model is normally fetched through the runtime, so 'path' can be omitted unless you want to use a custom or local model.
   # path: /config/yolonas.zip
-          # The .zip file must contain:
+  # The .zip file must contain:
-          # ├── yolonas.dfp          (a file ending with .dfp)
+  # ├── yolonas.dfp          (a file ending with .dfp)
-          # └── yolonas_post.onnx    (optional; only if the model includes a cropped post-processing network)
+  # └── yolonas_post.onnx    (optional; only if the model includes a cropped post-processing network)
 ```
 
 #### YOLOv9
@@ -889,16 +960,15 @@ detectors:
 
 model:
   model_type: yolo-generic
-  width: 320   # (Can be set to 640 for higher resolution)
+  width: 320 # (Can be set to 640 for higher resolution)
-  height: 320  # (Can be set to 640 for higher resolution)
+  height: 320 # (Can be set to 640 for higher resolution)
   input_tensor: nchw
   input_dtype: float
   labelmap_path: /labelmap/coco-80.txt
   # Optional: The model is normally fetched through the runtime, so 'path' can be omitted unless you want to use a custom or local model.
   # path: /config/yolov9.zip
-          # The .zip file must contain:
+  # The .zip file must contain:
-          # ├── yolov9.dfp          (a file ending with .dfp)
+  # ├── yolov9.dfp          (a file ending with .dfp)
-          # └── yolov9_post.onnx    (optional; only if the model includes a cropped post-processing network)
 ```
 
 #### YOLOX
@@ -924,8 +994,8 @@ model:
   labelmap_path: /labelmap/coco-80.txt
   # Optional: The model is normally fetched through the runtime, so 'path' can be omitted unless you want to use a custom or local model.
   # path: /config/yolox.zip
-          # The .zip file must contain:
+  # The .zip file must contain:
-          # ├── yolox.dfp          (a file ending with .dfp)
+  # ├── yolox.dfp          (a file ending with .dfp)
 ```
 
 #### SSDLite MobileNet v2
@@ -951,9 +1021,9 @@ model:
   labelmap_path: /labelmap/coco-80.txt
   # Optional: The model is normally fetched through the runtime, so 'path' can be omitted unless you want to use a custom or local model.
   # path: /config/ssdlite_mobilenet.zip
-          # The .zip file must contain:
+  # The .zip file must contain:
-          # ├── ssdlite_mobilenet.dfp          (a file ending with .dfp)
+  # ├── ssdlite_mobilenet.dfp          (a file ending with .dfp)
-          # └── ssdlite_mobilenet_post.onnx    (optional; only if the model includes a cropped post-processing network)
+  # └── ssdlite_mobilenet_post.onnx    (optional; only if the model includes a cropped post-processing network)
 ```
 
 #### Using a Custom Model
@@ -973,18 +1043,19 @@ To use your own model:
 For detailed instructions on compiling models, refer to the [MemryX Compiler](https://developer.memryx.com/tools/neural_compiler.html#usage) docs and [Tutorials](https://developer.memryx.com/tutorials/tutorials.html).
 
 ```yaml
-  # The detector automatically selects the default model if nothing is provided in the config.
+# The detector automatically selects the default model if nothing is provided in the config.
-  #
+#
-  # Optionally, you can specify a local model path as a .zip file to override the default.
+# Optionally, you can specify a local model path as a .zip file to override the default.
-  # If a local path is provided and the file exists, it will be used instead of downloading.
+# If a local path is provided and the file exists, it will be used instead of downloading.
-  #
+#
-  # Example:
+# Example:
-  # path: /config/yolonas.zip
+# path: /config/yolonas.zip
-  #
+#
-  # The .zip file must contain:
+# The .zip file must contain:
-  # ├── yolonas.dfp          (a file ending with .dfp)
+# ├── yolonas.dfp          (a file ending with .dfp)
-  # └── yolonas_post.onnx    (optional; only if the model includes a cropped post-processing network)
+# └── yolonas_post.onnx    (optional; only if the model includes a cropped post-processing network)
 ```
+
 ---
 
 ## NVidia TensorRT Detector
@@ -1092,16 +1163,16 @@ A synap model is provided in the container at /mobilenet.synap and is used by th
 Use the model configuration shown below when using the synaptics detector with the default synap model:
 
 ```yaml
-detectors:  # required
+detectors: # required
-  synap_npu:  # required
+  synap_npu: # required
-    type: synaptics  # required
+    type: synaptics # required
 
-model:  # required
+model: # required
-  path: /synaptics/mobilenet.synap  # required
+  path: /synaptics/mobilenet.synap # required
-  width: 224  # required
+  width: 224 # required
-  height: 224  # required
+  height: 224 # required
-  tensor_format: nhwc  # default value (optional. If you change the model, it is required)
+  tensor_format: nhwc # default value (optional. If you change the model, it is required)
-  labelmap_path: /labelmap/coco-80.txt  # required
+  labelmap_path: /labelmap/coco-80.txt # required
 ```
 
 ## Rockchip platform
@@ -1275,97 +1346,101 @@ Explanation of the paramters:
 
 ## DeGirum
 
-DeGirum is a detector that can use any type of hardware listed on [their website](https://hub.degirum.com). DeGirum can be used with local hardware through a DeGirum AI Server, or through the use of `@local`. You can also connect directly to DeGirum's AI Hub to run inferences. **Please Note:** This detector *cannot* be used for commercial purposes.
+DeGirum is a detector that can use any type of hardware listed on [their website](https://hub.degirum.com). DeGirum can be used with local hardware through a DeGirum AI Server, or through the use of `@local`. You can also connect directly to DeGirum's AI Hub to run inferences. **Please Note:** This detector _cannot_ be used for commercial purposes.
 
 ### Configuration
 
 #### AI Server Inference
 
 Before starting with the config file for this section, you must first launch an AI server. DeGirum has an AI server ready to use as a docker container. Add this to your `docker-compose.yml` to get started:
+
 ```yaml
 degirum_detector:
-    container_name: degirum
+  container_name: degirum
-    image: degirum/aiserver:latest
+  image: degirum/aiserver:latest
-    privileged: true
+  privileged: true
-    ports:
+  ports:
-      - "8778:8778"
+    - "8778:8778"
 ```
+
 All supported hardware will automatically be found on your AI server host as long as relevant runtimes and drivers are properly installed on your machine. Refer to [DeGirum's docs site](https://docs.degirum.com/pysdk/runtimes-and-drivers) if you have any trouble.
 
 Once completed, changing the `config.yml` file is simple.
+
 ```yaml
 degirum_detector:
-    type: degirum
+  type: degirum
-    location: degirum # Set to service name (degirum_detector), container_name (degirum), or a host:port (192.168.29.4:8778)
+  location: degirum # Set to service name (degirum_detector), container_name (degirum), or a host:port (192.168.29.4:8778)
-    zoo: degirum/public # DeGirum's public model zoo. Zoo name should be in format "workspace/zoo_name". degirum/public is available to everyone, so feel free to use it if you don't know where to start. If you aren't pulling a model from the AI Hub, leave this and 'token' blank.
+  zoo: degirum/public # DeGirum's public model zoo. Zoo name should be in format "workspace/zoo_name". degirum/public is available to everyone, so feel free to use it if you don't know where to start. If you aren't pulling a model from the AI Hub, leave this and 'token' blank.
-    token: dg_example_token # For authentication with the AI Hub. Get this token through the "tokens" section on the main page of the [AI Hub](https://hub.degirum.com). This can be left blank if you're pulling a model from the public zoo and running inferences on your local hardware using @local or a local DeGirum AI Server
+  token: dg_example_token # For authentication with the AI Hub. Get this token through the "tokens" section on the main page of the [AI Hub](https://hub.degirum.com). This can be left blank if you're pulling a model from the public zoo and running inferences on your local hardware using @local or a local DeGirum AI Server
-```
-Setting up a model in the `config.yml` is similar to setting up an AI server.
-You can set it to:
-- A model listed on the [AI Hub](https://hub.degirum.com), given that the correct zoo name is listed in your detector
-    - If this is what you choose to do, the correct model will be downloaded onto your machine before running.
-- A local directory acting as a zoo. See DeGirum's docs site [for more information](https://docs.degirum.com/pysdk/user-guide-pysdk/organizing-models#model-zoo-directory-structure).
-- A path to some model.json.
-```yaml
-model:
-    path: ./mobilenet_v2_ssd_coco--300x300_quant_n2x_orca1_1 # directory to model .json and file
-    width: 300 # width is in the model name as the first number in the "int"x"int" section
-    height: 300 # height is in the model name as the second number in the "int"x"int" section
-    input_pixel_format: rgb/bgr # look at the model.json to figure out which to put here
 ```
 
+Setting up a model in the `config.yml` is similar to setting up an AI server.
+You can set it to:
+
+- A model listed on the [AI Hub](https://hub.degirum.com), given that the correct zoo name is listed in your detector
+  - If this is what you choose to do, the correct model will be downloaded onto your machine before running.
+- A local directory acting as a zoo. See DeGirum's docs site [for more information](https://docs.degirum.com/pysdk/user-guide-pysdk/organizing-models#model-zoo-directory-structure).
+- A path to some model.json.
+
+```yaml
+model:
+  path: ./mobilenet_v2_ssd_coco--300x300_quant_n2x_orca1_1 # directory to model .json and file
+  width: 300 # width is in the model name as the first number in the "int"x"int" section
+  height: 300 # height is in the model name as the second number in the "int"x"int" section
+  input_pixel_format: rgb/bgr # look at the model.json to figure out which to put here
+```
 
 #### Local Inference
 
 It is also possible to eliminate the need for an AI server and run the hardware directly. The benefit of this approach is that you eliminate any bottlenecks that occur when transferring prediction results from the AI server docker container to the frigate one. However, the method of implementing local inference is different for every device and hardware combination, so it's usually more trouble than it's worth. A general guideline to achieve this would be:
+
 1. Ensuring that the frigate docker container has the runtime you want to use. So for instance, running `@local` for Hailo means making sure the container you're using has the Hailo runtime installed.
 2. To double check the runtime is detected by the DeGirum detector, make sure the `degirum sys-info` command properly shows whatever runtimes you mean to install.
 3. Create a DeGirum detector in your `config.yml` file.
 
 ```yaml
 degirum_detector:
-    type: degirum
+  type: degirum
-    location: "@local" # For accessing AI Hub devices and models
+  location: "@local" # For accessing AI Hub devices and models
-    zoo: degirum/public # DeGirum's public model zoo. Zoo name should be in format "workspace/zoo_name". degirum/public is available to everyone, so feel free to use it if you don't know where to start.
+  zoo: degirum/public # DeGirum's public model zoo. Zoo name should be in format "workspace/zoo_name". degirum/public is available to everyone, so feel free to use it if you don't know where to start.
-    token: dg_example_token # For authentication with the AI Hub. Get this token through the "tokens" section on the main page of the [AI Hub](https://hub.degirum.com). This can be left blank if you're pulling a model from the public zoo and running inferences on your local hardware using @local or a local DeGirum AI Server
+  token: dg_example_token # For authentication with the AI Hub. Get this token through the "tokens" section on the main page of the [AI Hub](https://hub.degirum.com). This can be left blank if you're pulling a model from the public zoo and running inferences on your local hardware using @local or a local DeGirum AI Server
-
 ```
 
 Once `degirum_detector` is setup, you can choose a model through 'model' section in the `config.yml` file.
 
 ```yaml
 model:
-    path: mobilenet_v2_ssd_coco--300x300_quant_n2x_orca1_1
+  path: mobilenet_v2_ssd_coco--300x300_quant_n2x_orca1_1
-    width: 300 # width is in the model name as the first number in the "int"x"int" section
+  width: 300 # width is in the model name as the first number in the "int"x"int" section
-    height: 300 # height is in the model name as the second number in the "int"x"int" section
+  height: 300 # height is in the model name as the second number in the "int"x"int" section
-    input_pixel_format: rgb/bgr # look at the model.json to figure out which to put here
+  input_pixel_format: rgb/bgr # look at the model.json to figure out which to put here
 ```
 
-
 #### AI Hub Cloud Inference
 
 If you do not possess whatever hardware you want to run, there's also the option to run cloud inferences. Do note that your detection fps might need to be lowered as network latency does significantly slow down this method of detection. For use with Frigate, we highly recommend using a local AI server as described above. To set up cloud inferences,
+
 1. Sign up at [DeGirum's AI Hub](https://hub.degirum.com).
 2. Get an access token.
 3. Create a DeGirum detector in your `config.yml` file.
 
 ```yaml
 degirum_detector:
-    type: degirum
+  type: degirum
-    location: "@cloud" # For accessing AI Hub devices and models
+  location: "@cloud" # For accessing AI Hub devices and models
-    zoo: degirum/public # DeGirum's public model zoo. Zoo name should be in format "workspace/zoo_name". degirum/public is available to everyone, so feel free to use it if you don't know where to start.
+  zoo: degirum/public # DeGirum's public model zoo. Zoo name should be in format "workspace/zoo_name". degirum/public is available to everyone, so feel free to use it if you don't know where to start.
-    token: dg_example_token # For authentication with the AI Hub. Get this token through the "tokens" section on the main page of the (AI Hub)[https://hub.degirum.com).
+  token: dg_example_token # For authentication with the AI Hub. Get this token through the "tokens" section on the main page of the (AI Hub)[https://hub.degirum.com).
-
 ```
 
 Once `degirum_detector` is setup, you can choose a model through 'model' section in the `config.yml` file.
 
 ```yaml
 model:
-    path: mobilenet_v2_ssd_coco--300x300_quant_n2x_orca1_1
+  path: mobilenet_v2_ssd_coco--300x300_quant_n2x_orca1_1
-    width: 300 # width is in the model name as the first number in the "int"x"int" section
+  width: 300 # width is in the model name as the first number in the "int"x"int" section
-    height: 300 # height is in the model name as the second number in the "int"x"int" section
+  height: 300 # height is in the model name as the second number in the "int"x"int" section
-    input_pixel_format: rgb/bgr # look at the model.json to figure out which to put here
+  input_pixel_format: rgb/bgr # look at the model.json to figure out which to put here
 ```
 
 ## AXERA
@@ -1423,7 +1498,7 @@ COPY --from=ghcr.io/astral-sh/uv:0.8.0 /uv /bin/
 WORKDIR /dfine
 RUN git clone https://github.com/Peterande/D-FINE.git .
 RUN uv pip install --system -r requirements.txt
-RUN uv pip install --system onnx onnxruntime onnxsim
+RUN uv pip install --system onnx onnxruntime onnxsim onnxscript
 # Create output directory and download checkpoint
 RUN mkdir -p output
 ARG MODEL_SIZE
@@ -1447,9 +1522,9 @@ FROM python:3.11 AS build
 RUN apt-get update && apt-get install --no-install-recommends -y libgl1 && rm -rf /var/lib/apt/lists/*
 COPY --from=ghcr.io/astral-sh/uv:0.8.0 /uv /bin/
 WORKDIR /rfdetr
-RUN uv pip install --system rfdetr onnx onnxruntime onnxsim onnx-graphsurgeon
+RUN uv pip install --system rfdetr[onnxexport] torch==2.8.0 onnxscript
 ARG MODEL_SIZE
-RUN python3 -c "from rfdetr import RFDETR${MODEL_SIZE}; x = RFDETR${MODEL_SIZE}(resolution=320); x.export()"
+RUN python3 -c "from rfdetr import RFDETR${MODEL_SIZE}; x = RFDETR${MODEL_SIZE}(resolution=320); x.export(simplify=True)"
 FROM scratch
 ARG MODEL_SIZE
 COPY --from=build /rfdetr/output/inference_model.onnx /rfdetr-${MODEL_SIZE}.onnx
@@ -1497,7 +1572,7 @@ COPY --from=ghcr.io/astral-sh/uv:0.8.0 /uv /bin/
 WORKDIR /yolov9
 ADD https://github.com/WongKinYiu/yolov9.git .
 RUN uv pip install --system -r requirements.txt
-RUN uv pip install --system onnx==1.18.0 onnxruntime onnx-simplifier>=0.4.1
+RUN uv pip install --system onnx==1.18.0 onnxruntime onnx-simplifier>=0.4.1 onnxscript
 ARG MODEL_SIZE
 ARG IMG_SIZE
 ADD https://github.com/WongKinYiu/yolov9/releases/download/v0.1/yolov9-${MODEL_SIZE}-converted.pt yolov9-${MODEL_SIZE}.pt

docs/docs/configuration/reference.md

@@ -240,11 +240,13 @@ birdseye:
     scaling_factor: 2.0
     # Optional: Maximum number of cameras to show at one time, showing the most recent (default: show all cameras)
     max_cameras: 1
+  # Optional: Frames-per-second to re-send the last composed Birdseye frame when idle (no motion or active updates). (default: shown below)
+  idle_heartbeat_fps: 0.0
 
 # Optional: ffmpeg configuration
 # More information about presets at https://docs.frigate.video/configuration/ffmpeg_presets
 ffmpeg:
-  # Optional: ffmpeg binry path (default: shown below)
+  # Optional: ffmpeg binary path (default: shown below)
   # can also be set to `7.0` or `5.0` to specify one of the included versions
   # or can be set to any path that holds `bin/ffmpeg` & `bin/ffprobe`
   path: "default"
@@ -427,6 +429,15 @@ review:
     alerts: True
     # Optional: Enable GenAI review summaries for detections (default: shown below)
     detections: False
+    # Optional: Activity Context Prompt to give context to the GenAI what activity is and is not suspicious.
+    # It is important to be direct and detailed. See documentation for the default prompt structure.
+    activity_context_prompt: """Define what is and is not suspicious
+"""
+    # Optional: Image source for GenAI (default: preview)
+    # Options: "preview" (uses cached preview frames at ~180p) or "recordings" (extracts frames from recordings at 480p)
+    # Using "recordings" provides better image quality but uses more tokens per image.
+    # Frame count is automatically calculated based on context window size, aspect ratio, and image source (capped at 20 frames).
+    image_source: preview
     # Optional: Additional concerns that the GenAI should make note of (default: None)
     additional_concerns:
       - Animals in the garden
@@ -628,7 +639,7 @@ face_recognition:
   # Optional: Min face recognitions for the sub label to be applied to the person object (default: shown below)
   min_faces: 1
   # Optional: Number of images of recognized faces to save for training (default: shown below)
-  save_attempts: 100
+  save_attempts: 200
   # Optional: Apply a blur quality filter to adjust confidence based on the blur level of the image (default: shown below)
   blur_confidence_filter: True
   # Optional: Set the model size used face recognition. (default: shown below)
@@ -669,20 +680,18 @@ lpr:
   # Optional: List of regex replacement rules to normalize detected plates (default: shown below)
   replace_rules: {}
 
-# Optional: Configuration for AI generated tracked object descriptions
+# Optional: Configuration for AI / LLM provider
 # WARNING: Depending on the provider, this will send thumbnails over the internet
-# to Google or OpenAI's LLMs to generate descriptions. It can be overridden at
+# to Google or OpenAI's LLMs to generate descriptions. GenAI features can be configured at
-# the camera level (enabled: False) to enhance privacy for indoor cameras.
+# the camera level to enhance privacy for indoor cameras.
 genai:
-  # Optional: Enable AI description generation (default: shown below)
+  # Required: Provider must be one of ollama, gemini, or openai
-  enabled: False
-  # Required if enabled: Provider must be one of ollama, gemini, or openai
   provider: ollama
   # Required if provider is ollama. May also be used for an OpenAI API compatible backend with the openai provider.
   base_url: http://localhost::11434
   # Required if gemini or openai
   api_key: "{FRIGATE_GENAI_API_KEY}"
-  # Required if enabled: The model to use with the provider.
+  # Required: The model to use with the provider.
   model: gemini-1.5-flash
   # Optional additional args to pass to the GenAI Provider (default: None)
   provider_options:
@@ -801,6 +810,8 @@ cameras:
       # NOTE: This must be different than any camera names, but can match with another zone on another
       #       camera.
       front_steps:
+        # Optional: A friendly name or descriptive text for the zones
+        friendly_name: ""
         # Required: List of x,y coordinates to define the polygon of the zone.
         # NOTE: Presence in a zone is evaluated only based on the bottom center of the objects bounding box.
         coordinates: 0.033,0.306,0.324,0.138,0.439,0.185,0.042,0.428
@@ -920,10 +931,13 @@ cameras:
         type: thumbnail
         # Reference data for matching, either an event ID for `thumbnail` or a text string for `description`. (default: none)
         data: 1751565549.853251-b69j73
-        # Similarity threshold for triggering. (default: none)
+        # Similarity threshold for triggering. (default: shown below)
-        threshold: 0.7
+        threshold: 0.8
         # List of actions to perform when the trigger fires. (default: none)
-        # Available options: `notification` (send a webpush notification)
+        # Available options:
+        # - `notification` (send a webpush notification)
+        # - `sub_label` (add trigger friendly name as a sub label to the triggering tracked object)
+        # - `attribute` (add trigger's name and similarity score as a data attribute to the triggering tracked object)
         actions:
           - notification
 

docs/docs/configuration/restream.md

@@ -24,6 +24,11 @@ birdseye:
   restream: True
 ```
 
+:::tip 
+
+To improve connection speed when using Birdseye via restream you can enable a small idle heartbeat by setting `birdseye.idle_heartbeat_fps` to a low value (e.g. `1–2`).  This makes Frigate periodically push the last frame even when no motion is detected, reducing initial connection latency.
+
+:::
 ### Securing Restream With Authentication
 
 The go2rtc restream can be secured with RTSP based username / password authentication. Ex:

docs/docs/configuration/semantic_search.md

@@ -78,7 +78,7 @@ Switching between V1 and V2 requires reindexing your embeddings. The embeddings
 
 ### GPU Acceleration
 
-The CLIP models are downloaded in ONNX format, and the `large` model can be accelerated using GPU / NPU hardware, when available. This depends on the Docker build that is used. You can also target a specific device in a multi-GPU installation.
+The CLIP models are downloaded in ONNX format, and the `large` model can be accelerated using GPU hardware, when available. This depends on the Docker build that is used. You can also target a specific device in a multi-GPU installation.
 
 ```yaml
 semantic_search:
@@ -90,7 +90,7 @@ semantic_search:
 
 :::info
 
-If the correct build is used for your GPU / NPU and the `large` model is configured, then the GPU / NPU will be detected and used automatically.
+If the correct build is used for your GPU / NPU and the `large` model is configured, then the GPU will be detected and used automatically.
 Specify the `device` option to target a specific GPU in a multi-GPU system (see [onnxruntime's provider options](https://onnxruntime.ai/docs/execution-providers/)).
 If you do not specify a device, the first available GPU will be used.
 
@@ -119,7 +119,7 @@ Semantic Search must be enabled to use Triggers.
 
 ### Configuration
 
-Triggers are defined within the `semantic_search` configuration for each camera in your Frigate configuration file or through the UI. Each trigger consists of a `friendly_name`, a `type` (either `thumbnail` or `description`), a `data` field (the reference image event ID or text), a `threshold` for similarity matching, and a list of `actions` to perform when the trigger fires.
+Triggers are defined within the `semantic_search` configuration for each camera in your Frigate configuration file or through the UI. Each trigger consists of a `friendly_name`, a `type` (either `thumbnail` or `description`), a `data` field (the reference image event ID or text), a `threshold` for similarity matching, and a list of `actions` to perform when the trigger fires - `notification`, `sub_label`, and `attribute`.
 
 Triggers are best configured through the Frigate UI.
 
@@ -128,17 +128,20 @@ Triggers are best configured through the Frigate UI.
 1. Navigate to the **Settings** page and select the **Triggers** tab.
 2. Choose a camera from the dropdown menu to view or manage its triggers.
 3. Click **Add Trigger** to create a new trigger or use the pencil icon to edit an existing one.
-4. In the **Create Trigger** dialog:
+4. In the **Create Trigger** wizard:
-   - Enter a **Name** for the trigger (e.g., "red_car_alert").
+   - Enter a **Name** for the trigger (e.g., "Red Car Alert").
    - Enter a descriptive **Friendly Name** for the trigger (e.g., "Red car on the driveway camera").
    - Select the **Type** (`Thumbnail` or `Description`).
    - For `Thumbnail`, select an image to trigger this action when a similar thumbnail image is detected, based on the threshold.
    - For `Description`, enter text to trigger this action when a similar tracked object description is detected.
    - Set the **Threshold** for similarity matching.
    - Select **Actions** to perform when the trigger fires.
+     If native webpush notifications are enabled, check the `Send Notification` box to send a notification.
+     Check the `Add Sub Label` box to add the trigger's friendly name as a sub label to any triggering tracked objects.
+     Check the `Add Attribute` box to add the trigger's internal ID (e.g., "red_car_alert") to a data attribute on the tracked object that can be processed via the API or MQTT.
 5. Save the trigger to update the configuration and store the embedding in the database.
 
-When a trigger fires, the UI highlights the trigger with a blue outline for 3 seconds for easy identification.
+When a trigger fires, the UI highlights the trigger with a blue dot for 3 seconds for easy identification. Additionally, the UI will show the last date/time and tracked object ID that activated your trigger. The last triggered timestamp is not saved to the database or persisted through restarts of Frigate.
 
 ### Usage and Best Practices
 

docs/docs/configuration/zones.md

@@ -27,6 +27,7 @@ cameras:
           - entire_yard
     zones:
       entire_yard:
+        friendly_name: Entire yard # You can use characters from any language text
         coordinates: ...
 ```
 
@@ -44,8 +45,10 @@ cameras:
           - edge_yard
     zones:
       edge_yard:
+        friendly_name: Edge yard # You can use characters from any language text
         coordinates: ...
       inner_yard:
+        friendly_name: Inner yard # You can use characters from any language text
         coordinates: ...
 ```
 
@@ -59,6 +62,7 @@ cameras:
         - entire_yard
     zones:
       entire_yard:
+        friendly_name: Entire yard
         coordinates: ...
 ```
 
@@ -82,6 +86,7 @@ cameras:
 
 Only car objects can trigger the `front_yard_street` zone and only person can trigger the `entire_yard`. Objects will be tracked for any `person` that enter anywhere in the yard, and for cars only if they enter the street.
 
+
 ### Zone Loitering
 
 Sometimes objects are expected to be passing through a zone, but an object loitering in an area is unexpected. Zones can be configured to have a minimum loitering time after which the object will be considered in the zone.

docs/docs/frigate/hardware.md

@@ -3,6 +3,8 @@ id: hardware
 title: Recommended hardware
 ---
 
+import CommunityBadge from '@site/src/components/CommunityBadge';
+
 ## Cameras
 
 Cameras that output H.264 video and AAC audio will offer the most compatibility with all features of Frigate and Home Assistant. It is also helpful if your camera supports multiple substreams to allow different resolutions to be used for detection, streaming, and recordings without re-encoding.
@@ -59,7 +61,7 @@ Frigate supports multiple different detectors that work on different types of ha
 
   - [Supports primarily ssdlite and mobilenet model architectures](../../configuration/object_detectors#edge-tpu-detector)
 
-- [MemryX](#memryx-mx3): The MX3 M.2 accelerator module is available in m.2 format allowing for a wide range of compatibility with devices.
+- <CommunityBadge /> [MemryX](#memryx-mx3): The MX3 M.2 accelerator module is available in m.2 format allowing for a wide range of compatibility with devices.
   - [Supports many model architectures](../../configuration/object_detectors#memryx-mx3)
   - Runs best with tiny, small, or medium-size models
 
@@ -78,46 +80,32 @@ Frigate supports multiple different detectors that work on different types of ha
 
 **Intel**
 
-- [OpenVino](#openvino---intel): OpenVino can run on Intel Arc GPUs, Intel integrated GPUs, and Intel CPUs to provide efficient object detection.
+- [OpenVino](#openvino---intel): OpenVino can run on Intel Arc GPUs, Intel integrated GPUs, and Intel NPUs to provide efficient object detection.
   - [Supports majority of model architectures](../../configuration/object_detectors#openvino-supported-models)
   - Runs best with tiny, small, or medium models
 
 **Nvidia**
 
-- [TensortRT](#tensorrt---nvidia-gpu): TensorRT can run on Nvidia GPUs and Jetson devices.
+- [TensortRT](#tensorrt---nvidia-gpu): TensorRT can run on Nvidia GPUs to provide efficient object detection.
+
   - [Supports majority of model architectures via ONNX](../../configuration/object_detectors#onnx-supported-models)
   - Runs well with any size models including large
 
-**Rockchip**
+- <CommunityBadge /> [Jetson](#nvidia-jetson): Jetson devices are supported via the TensorRT or ONNX detectors when running Jetpack 6.
+
+**Rockchip** <CommunityBadge />
 
 - [RKNN](#rockchip-platform): RKNN models can run on Rockchip devices with included NPUs to provide efficient object detection.
   - [Supports limited model architectures](../../configuration/object_detectors#choosing-a-model)
   - Runs best with tiny or small size models
   - Runs efficiently on low power hardware
 
-**Synaptics**
+**Synaptics** <CommunityBadge />
 
 - [Synaptics](#synaptics): synap models can run on Synaptics devices(e.g astra machina) with included NPUs to provide efficient object detection.
 
 :::
 
-### Synaptics
-
-- **Synaptics** Default model is **mobilenet**
-
-| Name             | Synaptics SL1680 Inference Time |
-| ---------------- | ------------------------------- |
-| ssd mobilenet    | ~ 25 ms                         |
-| yolov5m          | ~ 118 ms                        |
-
-### AXERA
-
-- **AXEngine** Default model is **yolov9**
-
-| Name             | AXERA AX650N/AX8850N Inference Time |
-| ---------------- | ----------------------------------- |
-| yolov9-tiny      | ~ 4 ms                              |
-
 ### Hailo-8
 
 Frigate supports both the Hailo-8 and Hailo-8L AI Acceleration Modules on compatible hardware platforms—including the Raspberry Pi 5 with the PCIe hat from the AI kit. The Hailo detector integration in Frigate automatically identifies your hardware type and selects the appropriate default model when a custom model isn’t provided.
@@ -150,6 +138,7 @@ The OpenVINO detector type is able to run on:
 
 - 6th Gen Intel Platforms and newer that have an iGPU
 - x86 hosts with an Intel Arc GPU
+- Intel NPUs
 - Most modern AMD CPUs (though this is officially not supported by Intel)
 - x86 & Arm64 hosts via CPU (generally not recommended)
 
@@ -174,7 +163,8 @@ Inference speeds vary greatly depending on the CPU or GPU used, some known examp
 | Intel UHD 770  | ~ 15 ms                    | t-320: ~ 16 ms s-320: ~ 20 ms s-640: ~ 40 ms      | 320: ~ 20 ms 640: ~ 46 ms |                        |                                    |
 | Intel N100     | ~ 15 ms                    | s-320: 30 ms                                      | 320: ~ 25 ms              |                        | Can only run one detector instance |
 | Intel N150     | ~ 15 ms                    | t-320: 16 ms s-320: 24 ms                         |                           |                        |                                    |
-| Intel Iris XE  | ~ 10 ms                    | s-320: 12 ms s-640: 30 ms                         | 320: ~ 18 ms 640: ~ 50 ms |                        |                                    |
+| Intel Iris XE  | ~ 10 ms                    | t-320: 6 ms t-640: 14 ms s-320: 8 ms s-640: 16 ms | 320: ~ 10 ms 640: ~ 20 ms | 320-n: 33 ms           |                                    |
+| Intel NPU      | ~ 6 ms                     | s-320: 11 ms                                      | 320: ~ 14 ms 640: ~ 34 ms | 320-n: 40 ms           |                                    |
 | Intel Arc A310 | ~ 5 ms                     | t-320: 7 ms t-640: 11 ms s-320: 8 ms s-640: 15 ms | 320: ~ 8 ms 640: ~ 14 ms  |                        |                                    |
 | Intel Arc A380 | ~ 6 ms                     |                                                   | 320: ~ 10 ms 640: ~ 22 ms | 336: 20 ms 448: 27 ms  |                                    |
 | Intel Arc A750 | ~ 4 ms                     |                                                   | 320: ~ 8 ms               |                        |                                    |
@@ -267,7 +257,7 @@ Inference speeds may vary depending on the host platform. The above data was mea
 
 ### Nvidia Jetson
 
-Frigate supports all Jetson boards, from the inexpensive Jetson Nano to the powerful Jetson Orin AGX. It will [make use of the Jetson's hardware media engine](/configuration/hardware_acceleration_video#nvidia-jetson-orin-agx-orin-nx-orin-nano-xavier-agx-xavier-nx-tx2-tx1-nano) when configured with the [appropriate presets](/configuration/ffmpeg_presets#hwaccel-presets), and will make use of the Jetson's GPU and DLA for object detection when configured with the [TensorRT detector](/configuration/object_detectors#nvidia-tensorrt-detector).
+Jetson devices are supported via the TensorRT or ONNX detectors when running Jetpack 6. It will [make use of the Jetson's hardware media engine](/configuration/hardware_acceleration_video#nvidia-jetson-orin-agx-orin-nx-orin-nano-xavier-agx-xavier-nx-tx2-tx1-nano) when configured with the [appropriate presets](/configuration/ffmpeg_presets#hwaccel-presets), and will make use of the Jetson's GPU and DLA for object detection when configured with the [TensorRT detector](/configuration/object_detectors#nvidia-tensorrt-detector).
 
 Inference speed will vary depending on the YOLO model, jetson platform and jetson nvpmodel (GPU/DLA/EMC clock speed). It is typically 20-40 ms for most models. The DLA is more efficient than the GPU, but not faster, so using the DLA will reduce power consumption but will slightly increase inference time.
 
@@ -288,6 +278,15 @@ Frigate supports hardware video processing on all Rockchip boards. However, hard
 
 The inference time of a rk3588 with all 3 cores enabled is typically 25-30 ms for yolo-nas s.
 
+### Synaptics
+
+- **Synaptics** Default model is **mobilenet**
+
+| Name          | Synaptics SL1680 Inference Time |
+| ------------- | ------------------------------- |
+| ssd mobilenet | ~ 25 ms                         |
+| yolov5m       | ~ 118 ms                        |
+
 ## What does Frigate use the CPU for and what does it use a detector for? (ELI5 Version)
 
 This is taken from a [user question on reddit](https://www.reddit.com/r/homeassistant/comments/q8mgau/comment/hgqbxh5/?utm_source=share&utm_medium=web2x&context=3). Modified slightly for clarity.
@@ -309,3 +308,11 @@ Basically - When you increase the resolution and/or the frame rate of the stream
 YES! The Coral does not help with decoding video streams.
 
 Decompressing video streams takes a significant amount of CPU power. Video compression uses key frames (also known as I-frames) to send a full frame in the video stream. The following frames only include the difference from the key frame, and the CPU has to compile each frame by merging the differences with the key frame. [More detailed explanation](https://support.video.ibm.com/hc/en-us/articles/18106203580316-Keyframes-InterFrame-Video-Compression). Higher resolutions and frame rates mean more processing power is needed to decode the video stream, so try and set them on the camera to avoid unnecessary decoding work.
+
+### AXERA
+
+- **AXEngine** Default model is **yolov9**
+
+| Name             | AXERA AX650N/AX8850N Inference Time |
+| ---------------- | ----------------------------------- |
+| yolov9-tiny      | ~ 4 ms                              |

docs/docs/frigate/installation.md

@@ -56,7 +56,7 @@ services:
     volumes:
       - /path/to/your/config:/config
       - /path/to/your/storage:/media/frigate
-      - type: tmpfs # Optional: 1GB of memory, reduces SSD/SD Card wear
+      - type: tmpfs # Recommended: 1GB of memory
         target: /tmp/cache
         tmpfs:
           size: 1000000000
@@ -340,12 +340,13 @@ services:
       - /dev/bus/usb:/dev/bus/usb # Passes the USB Coral, needs to be modified for other versions
       - /dev/apex_0:/dev/apex_0 # Passes a PCIe Coral, follow driver instructions here https://coral.ai/docs/m2/get-started/#2a-on-linux
       - /dev/video11:/dev/video11 # For Raspberry Pi 4B
-      - /dev/dri/renderD128:/dev/dri/renderD128 # For intel hwaccel, needs to be updated for your hardware
+      - /dev/dri/renderD128:/dev/dri/renderD128 # AMD / Intel GPU, needs to be updated for your hardware
+      - /dev/accel:/dev/accel # Intel NPU
     volumes:
       - /etc/localtime:/etc/localtime:ro
       - /path/to/your/config:/config
       - /path/to/your/storage:/media/frigate
-      - type: tmpfs # Optional: 1GB of memory, reduces SSD/SD Card wear
+      - type: tmpfs # Recommended: 1GB of memory
         target: /tmp/cache
         tmpfs:
           size: 1000000000

docs/docs/frigate/updating.md

@@ -5,7 +5,7 @@ title: Updating
 
 # Updating Frigate
 
-The current stable version of Frigate is **0.16.1**. The release notes and any breaking changes for this version can be found on the [Frigate GitHub releases page](https://github.com/blakeblackshear/frigate/releases/tag/v0.16.1).
+The current stable version of Frigate is **0.16.2**. The release notes and any breaking changes for this version can be found on the [Frigate GitHub releases page](https://github.com/blakeblackshear/frigate/releases/tag/v0.16.2).
 
 Keeping Frigate up to date ensures you benefit from the latest features, performance improvements, and bug fixes. The update process varies slightly depending on your installation method (Docker, Home Assistant Addon, etc.). Below are instructions for the most common setups.
 
@@ -33,21 +33,21 @@ If you’re running Frigate via Docker (recommended method), follow these steps:
 2. **Update and Pull the Latest Image**:
 
    - If using Docker Compose:
-     - Edit your `docker-compose.yml` file to specify the desired version tag (e.g., `0.16.1` instead of `0.15.2`). For example:
+     - Edit your `docker-compose.yml` file to specify the desired version tag (e.g., `0.16.2` instead of `0.15.2`). For example:
        ```yaml
        services:
          frigate:
-           image: ghcr.io/blakeblackshear/frigate:0.16.1
+           image: ghcr.io/blakeblackshear/frigate:0.16.2
        ```
      - Then pull the image:
        ```bash
-       docker pull ghcr.io/blakeblackshear/frigate:0.16.1
+       docker pull ghcr.io/blakeblackshear/frigate:0.16.2
        ```
      - **Note for `stable` Tag Users**: If your `docker-compose.yml` uses the `stable` tag (e.g., `ghcr.io/blakeblackshear/frigate:stable`), you don’t need to update the tag manually. The `stable` tag always points to the latest stable release after pulling.
    - If using `docker run`:
-     - Pull the image with the appropriate tag (e.g., `0.16.1`, `0.16.1-tensorrt`, or `stable`):
+     - Pull the image with the appropriate tag (e.g., `0.16.2`, `0.16.2-tensorrt`, or `stable`):
        ```bash
-       docker pull ghcr.io/blakeblackshear/frigate:0.16.1
+       docker pull ghcr.io/blakeblackshear/frigate:0.16.2
        ```
 
 3. **Start the Container**:

docs/docs/integrations/mqtt.md

@@ -159,9 +159,49 @@ Message published for updates to tracked object metadata, for example:
 }
 ```
 
+#### Object Classification Update
+
+Message published when [object classification](/configuration/custom_classification/object_classification) reaches consensus on a classification result.
+
+**Sub label type:**
+
+```json
+{
+  "type": "classification",
+  "id": "1607123955.475377-mxklsc",
+  "camera": "front_door_cam",
+  "timestamp": 1607123958.748393,
+  "model": "person_classifier",
+  "sub_label": "delivery_person",
+  "score": 0.87
+}
+```
+
+**Attribute type:**
+
+```json
+{
+  "type": "classification",
+  "id": "1607123955.475377-mxklsc",
+  "camera": "front_door_cam",
+  "timestamp": 1607123958.748393,
+  "model": "helmet_detector",
+  "attribute": "yes",
+  "score": 0.92
+}
+```
+
 ### `frigate/reviews`
 
-Message published for each changed review item. The first message is published when the `detection` or `alert` is initiated. When additional objects are detected or when a zone change occurs, it will publish a, `update` message with the same id. When the review activity has ended a final `end` message is published.
+Message published for each changed review item. The first message is published when the `detection` or `alert` is initiated.
+
+An `update` with the same ID will be published when:
+
+- The severity changes from `detection` to `alert`
+- Additional objects are detected
+- An object is recognized via face, lpr, etc.
+
+When the review activity has ended a final `end` message is published.
 
 ```json
 {
@@ -301,6 +341,11 @@ Publishes transcribed text for audio detected on this camera.
 
 **NOTE:** Requires audio detection and transcription to be enabled
 
+### `frigate/<camera_name>/classification/<model_name>`
+
+Publishes the current state detected by a state classification model for the camera. The topic name includes the model name as configured in your classification settings.
+The published value is the detected state class name (e.g., `open`, `closed`, `on`, `off`). The state is only published when it changes, helping to reduce unnecessary MQTT traffic.
+
 ### `frigate/<camera_name>/enabled/set`
 
 Topic to turn Frigate's processing of a camera on and off. Expected values are `ON` and `OFF`.

docs/docs/plus/annotating.md

@@ -42,6 +42,7 @@ Misidentified objects should have a correct label added. For example, if a perso
 | `w`               | Add box                       |
 | `d`               | Toggle difficult              |
 | `s`               | Switch to the next label      |
+| `Shift + s`       | Switch to the previous label  |
 | `tab`             | Select next largest box       |
 | `del`             | Delete current box            |
 | `esc`             | Deselect/Cancel               |

docs/docusaurus.config.ts

@@ -10,7 +10,7 @@ const config: Config = {
   baseUrl: "/",
   onBrokenLinks: "throw",
   onBrokenMarkdownLinks: "warn",
-  favicon: "img/favicon.ico",
+  favicon: "img/branding/favicon.ico",
   organizationName: "blakeblackshear",
   projectName: "frigate",
   themes: [
@@ -116,8 +116,8 @@ const config: Config = {
       title: "Frigate",
       logo: {
         alt: "Frigate",
-        src: "img/logo.svg",
+        src: "img/branding/logo.svg",
-        srcDark: "img/logo-dark.svg",
+        srcDark: "img/branding/logo-dark.svg",
       },
       items: [
         {
@@ -170,7 +170,7 @@ const config: Config = {
           ],
         },
       ],
-      copyright: `Copyright © ${new Date().getFullYear()} Blake Blackshear`,
+      copyright: `Copyright © ${new Date().getFullYear()} Frigate LLC`,
     },
   },
   plugins: [

docs/src/components/CommunityBadge/index.jsx

@@ -0,0 +1,23 @@
+import React from "react";
+
+export default function CommunityBadge() {
+  return (
+    <span
+      title="This detector is maintained by community members who provide code, maintenance, and support. See the contributing boards documentation for more information."
+      style={{
+        display: "inline-block",
+        backgroundColor: "#f1f3f5",
+        color: "#24292f",
+        fontSize: "11px",
+        fontWeight: 600,
+        padding: "2px 6px",
+        borderRadius: "3px",
+        border: "1px solid #d1d9e0",
+        marginLeft: "4px",
+        cursor: "help",
+      }}
+    >
+      Community Supported
+    </span>
+  );
+}

docs/static/img/branding/LICENSE.md

@@ -0,0 +1,30 @@
+# COPYRIGHT AND TRADEMARK NOTICE
+
+The images, logos, and icons contained in this directory (the "Brand Assets") are
+proprietary to Frigate LLC and are NOT covered by the MIT License governing the
+rest of this repository.
+
+1. TRADEMARK STATUS
+   The "Frigate" name and the accompanying logo are common law trademarks™ of
+   Frigate LLC. Frigate LLC reserves all rights to these marks.
+
+2. LIMITED PERMISSION FOR USE
+   Permission is hereby granted to display these Brand Assets strictly for the
+   following purposes:
+   a. To execute the software interface on a local machine.
+   b. To identify the software in documentation or reviews (nominative use).
+
+3. RESTRICTIONS
+   You may NOT:
+   a. Use these Brand Assets to represent a derivative work (fork) as an official
+   product of Frigate LLC.
+   b. Use these Brand Assets in a way that implies endorsement, sponsorship, or
+   commercial affiliation with Frigate LLC.
+   c. Modify or alter the Brand Assets.
+
+If you fork this repository with the intent to distribute a modified or competing
+version of the software, you must replace these Brand Assets with your own
+original content.
+
+ALL RIGHTS RESERVED.
+Copyright (c) 2025 Frigate LLC.

frigate/api/app.py

@@ -37,7 +37,6 @@ from frigate.stats.prometheus import get_metrics, update_metrics
 from frigate.util.builtin import (
     clean_camera_user_pass,
     flatten_config_data,
-    get_tz_modifiers,
     process_config_query_string,
     update_yaml_file_bulk,
 )
@@ -48,6 +47,7 @@ from frigate.util.services import (
     restart_frigate,
     vainfo_hwaccel,
 )
+from frigate.util.time import get_tz_modifiers
 from frigate.version import VERSION
 
 logger = logging.getLogger(__name__)
@@ -179,6 +179,36 @@ def config(request: Request):
     return JSONResponse(content=config)
 
 
+@router.get("/config/raw_paths", dependencies=[Depends(require_role(["admin"]))])
+def config_raw_paths(request: Request):
+    """Admin-only endpoint that returns camera paths and go2rtc streams without credential masking."""
+    config_obj: FrigateConfig = request.app.frigate_config
+
+    raw_paths = {"cameras": {}, "go2rtc": {"streams": {}}}
+
+    # Extract raw camera ffmpeg input paths
+    for camera_name, camera in config_obj.cameras.items():
+        raw_paths["cameras"][camera_name] = {
+            "ffmpeg": {
+                "inputs": [
+                    {"path": input.path, "roles": input.roles}
+                    for input in camera.ffmpeg.inputs
+                ]
+            }
+        }
+
+    # Extract raw go2rtc stream URLs
+    go2rtc_config = config_obj.go2rtc.model_dump(
+        mode="json", warnings="none", exclude_none=True
+    )
+    for stream_name, stream in go2rtc_config.get("streams", {}).items():
+        if stream is None:
+            continue
+        raw_paths["go2rtc"]["streams"][stream_name] = stream
+
+    return JSONResponse(content=raw_paths)
+
+
 @router.get("/config/raw")
 def config_raw():
     config_file = find_config_file()
@@ -387,20 +417,29 @@ def config_set(request: Request, body: AppConfigSetBody):
         old_config: FrigateConfig = request.app.frigate_config
         request.app.frigate_config = config
 
-        if body.update_topic and body.update_topic.startswith("config/cameras/"):
+        if body.update_topic:
-            _, _, camera, field = body.update_topic.split("/")
+            if body.update_topic.startswith("config/cameras/"):
+                _, _, camera, field = body.update_topic.split("/")
 
-            if field == "add":
+                if field == "add":
-                settings = config.cameras[camera]
+                    settings = config.cameras[camera]
-            elif field == "remove":
+                elif field == "remove":
-                settings = old_config.cameras[camera]
+                    settings = old_config.cameras[camera]
+                else:
+                    settings = config.get_nested_object(body.update_topic)
+
+                request.app.config_publisher.publish_update(
+                    CameraConfigUpdateTopic(CameraConfigUpdateEnum[field], camera),
+                    settings,
+                )
             else:
+                # Generic handling for global config updates
                 settings = config.get_nested_object(body.update_topic)
 
-            request.app.config_publisher.publish_update(
+                # Publish None for removal, actual config for add/update
-                CameraConfigUpdateTopic(CameraConfigUpdateEnum[field], camera),
+                request.app.config_publisher.publisher.publish(
-                settings,
+                    body.update_topic, settings
-            )
+                )
 
     return JSONResponse(
         content=(
@@ -688,7 +727,11 @@ def timeline(camera: str = "all", limit: int = 100, source_id: Optional[str] = N
         clauses.append((Timeline.camera == camera))
 
     if source_id:
-        clauses.append((Timeline.source_id == source_id))
+        source_ids = [sid.strip() for sid in source_id.split(",")]
+        if len(source_ids) == 1:
+            clauses.append((Timeline.source_id == source_ids[0]))
+        else:
+            clauses.append((Timeline.source_id.in_(source_ids)))
 
     if len(clauses) == 0:
         clauses.append((True))

frigate/api/auth.py

@@ -35,6 +35,23 @@ logger = logging.getLogger(__name__)
 router = APIRouter(tags=[Tags.auth])
 
 
+@router.get("/auth/first_time_login")
+def first_time_login(request: Request):
+    """Return whether the admin first-time login help flag is set in config.
+
+    This endpoint is intentionally unauthenticated so the login page can
+    query it before a user is authenticated.
+    """
+    auth_config = request.app.frigate_config.auth
+
+    return JSONResponse(
+        content={
+            "admin_first_time_login": auth_config.admin_first_time_login
+            or auth_config.reset_admin_password
+        }
+    )
+
+
 class RateLimiter:
     _limit = ""
 
@@ -515,6 +532,11 @@ def login(request: Request, body: AppPostLoginBody):
         set_jwt_cookie(
             response, JWT_COOKIE_NAME, encoded_jwt, expiration, JWT_COOKIE_SECURE
         )
+        # Clear admin_first_time_login flag after successful admin login so the
+        # UI stops showing the first-time login documentation link.
+        if role == "admin":
+            request.app.frigate_config.auth.admin_first_time_login = False
+
         return response
     return JSONResponse(content={"message": "Login failed"}, status_code=401)
 

frigate/api/camera.py

@@ -3,11 +3,17 @@
 import json
 import logging
 import re
+from importlib.util import find_spec
+from pathlib import Path
 from urllib.parse import quote_plus
 
+import httpx
 import requests
-from fastapi import APIRouter, Depends, Request, Response
+from fastapi import APIRouter, Depends, Query, Request, Response
 from fastapi.responses import JSONResponse
+from onvif import ONVIFCamera, ONVIFError
+from zeep.exceptions import Fault, TransportError
+from zeep.transports import AsyncTransport
 
 from frigate.api.auth import require_role
 from frigate.api.defs.tags import Tags
@@ -199,19 +205,30 @@ def ffprobe(request: Request, paths: str = "", detailed: bool = False):
             request.app.frigate_config.ffmpeg, path.strip(), detailed=detailed
         )
 
-        result = {
+        if ffprobe.returncode != 0:
-            "return_code": ffprobe.returncode,
+            try:
-            "stderr": (
+                stderr_decoded = ffprobe.stderr.decode("utf-8")
-                ffprobe.stderr.decode("unicode_escape").strip()
+            except UnicodeDecodeError:
-                if ffprobe.returncode != 0
+                try:
-                else ""
+                    stderr_decoded = ffprobe.stderr.decode("unicode_escape")
-            ),
+                except Exception:
-            "stdout": (
+                    stderr_decoded = str(ffprobe.stderr)
-                json.loads(ffprobe.stdout.decode("unicode_escape").strip())
+
-                if ffprobe.returncode == 0
+            stderr_lines = [
-                else ""
+                line.strip() for line in stderr_decoded.split("\n") if line.strip()
-            ),
+            ]
-        }
+
+            result = {
+                "return_code": ffprobe.returncode,
+                "stderr": stderr_lines,
+                "stdout": "",
+            }
+        else:
+            result = {
+                "return_code": ffprobe.returncode,
+                "stderr": [],
+                "stdout": json.loads(ffprobe.stdout.decode("unicode_escape").strip()),
+            }
 
         # Add detailed metadata if requested and probe was successful
         if detailed and ffprobe.returncode == 0 and result["stdout"]:
@@ -441,3 +458,537 @@ def _extract_fps(r_frame_rate: str) -> float | None:
         return round(float(num) / float(den), 2)
     except (ValueError, ZeroDivisionError):
         return None
+
+
+@router.get(
+    "/onvif/probe",
+    dependencies=[Depends(require_role(["admin"]))],
+    summary="Probe ONVIF device",
+    description=(
+        "Probe an ONVIF device to determine capabilities and optionally test available stream URIs. "
+        "Query params: host (required), port (default 80), username, password, test (boolean), "
+        "auth_type (basic or digest, default basic)."
+    ),
+)
+async def onvif_probe(
+    request: Request,
+    host: str = Query(None),
+    port: int = Query(80),
+    username: str = Query(""),
+    password: str = Query(""),
+    test: bool = Query(False),
+    auth_type: str = Query("basic"),  # Add auth_type parameter
+):
+    """
+    Probe a single ONVIF device to determine capabilities.
+
+    Connects to an ONVIF device and queries for:
+    - Device information (manufacturer, model)
+    - Media profiles count
+    - PTZ support
+    - Available presets
+    - Autotracking support
+
+    Query Parameters:
+        host: Device host/IP address (required)
+        port: Device port (default 80)
+        username: ONVIF username (optional)
+        password: ONVIF password (optional)
+        test: run ffprobe on the stream (optional)
+        auth_type: Authentication type - "basic" or "digest" (default "basic")
+
+    Returns:
+        JSON with device capabilities information
+    """
+    if not host:
+        return JSONResponse(
+            content={"success": False, "message": "host parameter is required"},
+            status_code=400,
+        )
+
+    # Validate host format
+    if not _is_valid_host(host):
+        return JSONResponse(
+            content={"success": False, "message": "Invalid host format"},
+            status_code=400,
+        )
+
+    # Validate auth_type
+    if auth_type not in ["basic", "digest"]:
+        return JSONResponse(
+            content={
+                "success": False,
+                "message": "auth_type must be 'basic' or 'digest'",
+            },
+            status_code=400,
+        )
+
+    onvif_camera = None
+
+    try:
+        logger.debug(f"Probing ONVIF device at {host}:{port} with {auth_type} auth")
+
+        try:
+            wsdl_base = None
+            spec = find_spec("onvif")
+            if spec and getattr(spec, "origin", None):
+                wsdl_base = str(Path(spec.origin).parent / "wsdl")
+        except Exception:
+            wsdl_base = None
+
+        onvif_camera = ONVIFCamera(
+            host, port, username or "", password or "", wsdl_dir=wsdl_base
+        )
+
+        # Configure digest authentication if requested
+        if auth_type == "digest" and username and password:
+            # Create httpx client with digest auth
+            auth = httpx.DigestAuth(username, password)
+            client = httpx.AsyncClient(auth=auth, timeout=10.0)
+
+            # Replace the transport in the zeep client
+            transport = AsyncTransport(client=client)
+
+            # Update the xaddr before setting transport
+            await onvif_camera.update_xaddrs()
+
+            # Replace transport in all services
+            if hasattr(onvif_camera, "devicemgmt"):
+                onvif_camera.devicemgmt.zeep_client.transport = transport
+            if hasattr(onvif_camera, "media"):
+                onvif_camera.media.zeep_client.transport = transport
+            if hasattr(onvif_camera, "ptz"):
+                onvif_camera.ptz.zeep_client.transport = transport
+
+            logger.debug("Configured digest authentication")
+        else:
+            await onvif_camera.update_xaddrs()
+
+        # Get device information
+        device_info = {
+            "manufacturer": "Unknown",
+            "model": "Unknown",
+            "firmware_version": "Unknown",
+        }
+        try:
+            device_service = await onvif_camera.create_devicemgmt_service()
+
+            # Update transport for device service if digest auth
+            if auth_type == "digest" and username and password:
+                auth = httpx.DigestAuth(username, password)
+                client = httpx.AsyncClient(auth=auth, timeout=10.0)
+                transport = AsyncTransport(client=client)
+                device_service.zeep_client.transport = transport
+
+            device_info_resp = await device_service.GetDeviceInformation()
+            manufacturer = getattr(device_info_resp, "Manufacturer", None) or (
+                device_info_resp.get("Manufacturer")
+                if isinstance(device_info_resp, dict)
+                else None
+            )
+            model = getattr(device_info_resp, "Model", None) or (
+                device_info_resp.get("Model")
+                if isinstance(device_info_resp, dict)
+                else None
+            )
+            firmware = getattr(device_info_resp, "FirmwareVersion", None) or (
+                device_info_resp.get("FirmwareVersion")
+                if isinstance(device_info_resp, dict)
+                else None
+            )
+            device_info.update(
+                {
+                    "manufacturer": manufacturer or "Unknown",
+                    "model": model or "Unknown",
+                    "firmware_version": firmware or "Unknown",
+                }
+            )
+        except Exception as e:
+            logger.debug(f"Failed to get device info: {e}")
+
+        # Get media profiles
+        profiles = []
+        profiles_count = 0
+        first_profile_token = None
+        ptz_config_token = None
+        try:
+            media_service = await onvif_camera.create_media_service()
+
+            # Update transport for media service if digest auth
+            if auth_type == "digest" and username and password:
+                auth = httpx.DigestAuth(username, password)
+                client = httpx.AsyncClient(auth=auth, timeout=10.0)
+                transport = AsyncTransport(client=client)
+                media_service.zeep_client.transport = transport
+
+            profiles = await media_service.GetProfiles()
+            profiles_count = len(profiles) if profiles else 0
+            if profiles and len(profiles) > 0:
+                p = profiles[0]
+                first_profile_token = getattr(p, "token", None) or (
+                    p.get("token") if isinstance(p, dict) else None
+                )
+                # Get PTZ configuration token from the profile
+                ptz_configuration = getattr(p, "PTZConfiguration", None) or (
+                    p.get("PTZConfiguration") if isinstance(p, dict) else None
+                )
+                if ptz_configuration:
+                    ptz_config_token = getattr(ptz_configuration, "token", None) or (
+                        ptz_configuration.get("token")
+                        if isinstance(ptz_configuration, dict)
+                        else None
+                    )
+        except Exception as e:
+            logger.debug(f"Failed to get media profiles: {e}")
+
+        # Check PTZ support and capabilities
+        ptz_supported = False
+        presets_count = 0
+        autotrack_supported = False
+
+        try:
+            ptz_service = await onvif_camera.create_ptz_service()
+
+            # Update transport for PTZ service if digest auth
+            if auth_type == "digest" and username and password:
+                auth = httpx.DigestAuth(username, password)
+                client = httpx.AsyncClient(auth=auth, timeout=10.0)
+                transport = AsyncTransport(client=client)
+                ptz_service.zeep_client.transport = transport
+
+            # Check if PTZ service is available
+            try:
+                await ptz_service.GetServiceCapabilities()
+                ptz_supported = True
+                logger.debug("PTZ service is available")
+            except Exception as e:
+                logger.debug(f"PTZ service not available: {e}")
+                ptz_supported = False
+
+            # Try to get presets if PTZ is supported and we have a profile
+            if ptz_supported and first_profile_token:
+                try:
+                    presets_resp = await ptz_service.GetPresets(
+                        {"ProfileToken": first_profile_token}
+                    )
+                    presets_count = len(presets_resp) if presets_resp else 0
+                    logger.debug(f"Found {presets_count} presets")
+                except Exception as e:
+                    logger.debug(f"Failed to get presets: {e}")
+                    presets_count = 0
+
+            # Check for autotracking support - requires both FOV relative movement and MoveStatus
+            if ptz_supported and first_profile_token and ptz_config_token:
+                # First check for FOV relative movement support
+                pt_r_fov_supported = False
+                try:
+                    config_request = ptz_service.create_type("GetConfigurationOptions")
+                    config_request.ConfigurationToken = ptz_config_token
+                    ptz_config = await ptz_service.GetConfigurationOptions(
+                        config_request
+                    )
+
+                    if ptz_config:
+                        # Check for pt-r-fov support
+                        spaces = getattr(ptz_config, "Spaces", None) or (
+                            ptz_config.get("Spaces")
+                            if isinstance(ptz_config, dict)
+                            else None
+                        )
+
+                        if spaces:
+                            rel_pan_tilt_space = getattr(
+                                spaces, "RelativePanTiltTranslationSpace", None
+                            ) or (
+                                spaces.get("RelativePanTiltTranslationSpace")
+                                if isinstance(spaces, dict)
+                                else None
+                            )
+
+                            if rel_pan_tilt_space:
+                                # Look for FOV space
+                                for i, space in enumerate(rel_pan_tilt_space):
+                                    uri = None
+                                    if isinstance(space, dict):
+                                        uri = space.get("URI")
+                                    else:
+                                        uri = getattr(space, "URI", None)
+
+                                    if uri and "TranslationSpaceFov" in uri:
+                                        pt_r_fov_supported = True
+                                        logger.debug(
+                                            "FOV relative movement (pt-r-fov) supported"
+                                        )
+                                        break
+
+                        logger.debug(f"PTZ config spaces: {ptz_config}")
+                except Exception as e:
+                    logger.debug(f"Failed to check FOV relative movement: {e}")
+                    pt_r_fov_supported = False
+
+                # Now check for MoveStatus support via GetServiceCapabilities
+                if pt_r_fov_supported:
+                    try:
+                        service_capabilities_request = ptz_service.create_type(
+                            "GetServiceCapabilities"
+                        )
+                        service_capabilities = await ptz_service.GetServiceCapabilities(
+                            service_capabilities_request
+                        )
+
+                        # Look for MoveStatus in the capabilities
+                        move_status_capable = False
+                        if service_capabilities:
+                            # Try to find MoveStatus key recursively
+                            def find_move_status(obj, key="MoveStatus"):
+                                if isinstance(obj, dict):
+                                    if key in obj:
+                                        return obj[key]
+                                    for v in obj.values():
+                                        result = find_move_status(v, key)
+                                        if result is not None:
+                                            return result
+                                elif hasattr(obj, key):
+                                    return getattr(obj, key)
+                                elif hasattr(obj, "__dict__"):
+                                    for v in vars(obj).values():
+                                        result = find_move_status(v, key)
+                                        if result is not None:
+                                            return result
+                                return None
+
+                            move_status_value = find_move_status(service_capabilities)
+
+                            # MoveStatus should return "true" if supported
+                            if isinstance(move_status_value, bool):
+                                move_status_capable = move_status_value
+                            elif isinstance(move_status_value, str):
+                                move_status_capable = (
+                                    move_status_value.lower() == "true"
+                                )
+
+                            logger.debug(f"MoveStatus capability: {move_status_value}")
+
+                        # Autotracking is supported if both conditions are met
+                        autotrack_supported = pt_r_fov_supported and move_status_capable
+
+                        if autotrack_supported:
+                            logger.debug(
+                                "Autotracking fully supported (pt-r-fov + MoveStatus)"
+                            )
+                        else:
+                            logger.debug(
+                                f"Autotracking not fully supported - pt-r-fov: {pt_r_fov_supported}, MoveStatus: {move_status_capable}"
+                            )
+                    except Exception as e:
+                        logger.debug(f"Failed to check MoveStatus support: {e}")
+                        autotrack_supported = False
+
+        except Exception as e:
+            logger.debug(f"Failed to probe PTZ service: {e}")
+
+        result = {
+            "success": True,
+            "host": host,
+            "port": port,
+            "manufacturer": device_info["manufacturer"],
+            "model": device_info["model"],
+            "firmware_version": device_info["firmware_version"],
+            "profiles_count": profiles_count,
+            "ptz_supported": ptz_supported,
+            "presets_count": presets_count,
+            "autotrack_supported": autotrack_supported,
+        }
+
+        # Gather RTSP candidates
+        rtsp_candidates: list[dict] = []
+        try:
+            media_service = await onvif_camera.create_media_service()
+
+            # Update transport for media service if digest auth
+            if auth_type == "digest" and username and password:
+                auth = httpx.DigestAuth(username, password)
+                client = httpx.AsyncClient(auth=auth, timeout=10.0)
+                transport = AsyncTransport(client=client)
+                media_service.zeep_client.transport = transport
+
+            if profiles_count and media_service:
+                for p in profiles or []:
+                    token = getattr(p, "token", None) or (
+                        p.get("token") if isinstance(p, dict) else None
+                    )
+                    if not token:
+                        continue
+                    try:
+                        stream_setup = {
+                            "Stream": "RTP-Unicast",
+                            "Transport": {"Protocol": "RTSP"},
+                        }
+                        stream_req = {
+                            "ProfileToken": token,
+                            "StreamSetup": stream_setup,
+                        }
+                        stream_uri_resp = await media_service.GetStreamUri(stream_req)
+                        uri = (
+                            stream_uri_resp.get("Uri")
+                            if isinstance(stream_uri_resp, dict)
+                            else getattr(stream_uri_resp, "Uri", None)
+                        )
+                        if uri:
+                            logger.debug(
+                                f"GetStreamUri returned for token {token}: {uri}"
+                            )
+                            # If credentials were provided, do NOT add the unauthenticated URI.
+                            try:
+                                if isinstance(uri, str) and uri.startswith("rtsp://"):
+                                    if username and password and "@" not in uri:
+                                        # Inject URL-encoded credentials and add only the
+                                        # authenticated version.
+                                        cred = f"{quote_plus(username)}:{quote_plus(password)}@"
+                                        injected = uri.replace(
+                                            "rtsp://", f"rtsp://{cred}", 1
+                                        )
+                                        rtsp_candidates.append(
+                                            {
+                                                "source": "GetStreamUri",
+                                                "profile_token": token,
+                                                "uri": injected,
+                                            }
+                                        )
+                                    else:
+                                        # No credentials provided or URI already contains
+                                        # credentials — add the URI as returned.
+                                        rtsp_candidates.append(
+                                            {
+                                                "source": "GetStreamUri",
+                                                "profile_token": token,
+                                                "uri": uri,
+                                            }
+                                        )
+                                else:
+                                    # Non-RTSP URIs (e.g., http-flv) — add as returned.
+                                    rtsp_candidates.append(
+                                        {
+                                            "source": "GetStreamUri",
+                                            "profile_token": token,
+                                            "uri": uri,
+                                        }
+                                    )
+                            except Exception as e:
+                                logger.debug(
+                                    f"Skipping stream URI for token {token} due to processing error: {e}"
+                                )
+                                continue
+                    except Exception:
+                        logger.debug(
+                            f"GetStreamUri failed for token {token}", exc_info=True
+                        )
+                        continue
+
+            # Add common RTSP patterns as fallback
+            if not rtsp_candidates:
+                common_paths = [
+                    "/h264",
+                    "/live.sdp",
+                    "/media.amp",
+                    "/Streaming/Channels/101",
+                    "/Streaming/Channels/1",
+                    "/stream1",
+                    "/cam/realmonitor?channel=1&subtype=0",
+                    "/11",
+                ]
+                # Use URL-encoded credentials for pattern fallback URIs when provided
+                auth_str = (
+                    f"{quote_plus(username)}:{quote_plus(password)}@"
+                    if username and password
+                    else ""
+                )
+                rtsp_port = 554
+                for path in common_paths:
+                    uri = f"rtsp://{auth_str}{host}:{rtsp_port}{path}"
+                    rtsp_candidates.append({"source": "pattern", "uri": uri})
+        except Exception:
+            logger.debug("Failed to collect RTSP candidates")
+
+        # Optionally test RTSP candidates using ffprobe_stream
+        tested_candidates = []
+        if test and rtsp_candidates:
+            for c in rtsp_candidates:
+                uri = c["uri"]
+                to_test = [uri]
+                try:
+                    if (
+                        username
+                        and password
+                        and isinstance(uri, str)
+                        and uri.startswith("rtsp://")
+                        and "@" not in uri
+                    ):
+                        cred = f"{quote_plus(username)}:{quote_plus(password)}@"
+                        cred_uri = uri.replace("rtsp://", f"rtsp://{cred}", 1)
+                        if cred_uri not in to_test:
+                            to_test.append(cred_uri)
+                except Exception:
+                    pass
+
+                for test_uri in to_test:
+                    try:
+                        probe = ffprobe_stream(
+                            request.app.frigate_config.ffmpeg, test_uri, detailed=False
+                        )
+                        print(probe)
+                        ok = probe is not None and getattr(probe, "returncode", 1) == 0
+                        tested_candidates.append(
+                            {
+                                "uri": test_uri,
+                                "source": c.get("source"),
+                                "ok": ok,
+                                "profile_token": c.get("profile_token"),
+                            }
+                        )
+                    except Exception as e:
+                        logger.debug(f"Unable to probe stream: {e}")
+                        tested_candidates.append(
+                            {
+                                "uri": test_uri,
+                                "source": c.get("source"),
+                                "ok": False,
+                                "profile_token": c.get("profile_token"),
+                            }
+                        )
+
+        result["rtsp_candidates"] = rtsp_candidates
+        if test:
+            result["rtsp_tested"] = tested_candidates
+
+        logger.debug(f"ONVIF probe successful: {result}")
+        return JSONResponse(content=result)
+
+    except ONVIFError as e:
+        logger.warning(f"ONVIF error probing {host}:{port}: {e}")
+        return JSONResponse(
+            content={"success": False, "message": "ONVIF error"},
+            status_code=400,
+        )
+    except (Fault, TransportError) as e:
+        logger.warning(f"Connection error probing {host}:{port}: {e}")
+        return JSONResponse(
+            content={"success": False, "message": "Connection error"},
+            status_code=503,
+        )
+    except Exception as e:
+        logger.warning(f"Error probing ONVIF device at {host}:{port}, {e}")
+        return JSONResponse(
+            content={"success": False, "message": "Probe failed"},
+            status_code=500,
+        )
+
+    finally:
+        # Best-effort cleanup of ONVIF camera client session
+        if onvif_camera is not None:
+            try:
+                # Check if the camera has a close method and call it
+                if hasattr(onvif_camera, "close"):
+                    await onvif_camera.close()
+            except Exception as e:
+                logger.debug(f"Error closing ONVIF camera session: {e}")

frigate/api/classification.py

@@ -3,7 +3,9 @@
 import datetime
 import logging
 import os
+import random
 import shutil
+import string
 from typing import Any
 
 import cv2
@@ -17,6 +19,8 @@ from frigate.api.auth import require_role
 from frigate.api.defs.request.classification_body import (
     AudioTranscriptionBody,
     DeleteFaceImagesBody,
+    GenerateObjectExamplesBody,
+    GenerateStateExamplesBody,
     RenameFaceBody,
 )
 from frigate.api.defs.response.classification_response import (
@@ -27,10 +31,16 @@ from frigate.api.defs.response.generic_response import GenericResponse
 from frigate.api.defs.tags import Tags
 from frigate.config import FrigateConfig
 from frigate.config.camera import DetectConfig
-from frigate.const import CLIPS_DIR, FACE_DIR
+from frigate.const import CLIPS_DIR, FACE_DIR, MODEL_CACHE_DIR
 from frigate.embeddings import EmbeddingsContext
 from frigate.models import Event
-from frigate.util.path import get_event_snapshot
+from frigate.util.classification import (
+    collect_object_classification_examples,
+    collect_state_classification_examples,
+    get_dataset_image_count,
+    read_training_metadata,
+)
+from frigate.util.file import get_event_snapshot
 
 logger = logging.getLogger(__name__)
 
@@ -104,9 +114,18 @@ def reclassify_face(request: Request, body: dict = None):
     context: EmbeddingsContext = request.app.embeddings
     response = context.reprocess_face(training_file)
 
+    if not isinstance(response, dict):
+        return JSONResponse(
+            status_code=500,
+            content={
+                "success": False,
+                "message": "Could not process request.",
+            },
+        )
+
     return JSONResponse(
+        status_code=200 if response.get("success", True) else 400,
         content=response,
-        status_code=200,
     )
 
 
@@ -159,8 +178,7 @@ def train_face(request: Request, name: str, body: dict = None):
     new_name = f"{sanitized_name}-{datetime.datetime.now().timestamp()}.webp"
     new_file_folder = os.path.join(FACE_DIR, f"{sanitized_name}")
 
-    if not os.path.exists(new_file_folder):
+    os.makedirs(new_file_folder, exist_ok=True)
-        os.mkdir(new_file_folder)
 
     if training_file_name:
         shutil.move(training_file, os.path.join(new_file_folder, new_name))
@@ -548,23 +566,59 @@ def get_classification_dataset(name: str):
     dataset_dir = os.path.join(CLIPS_DIR, sanitize_filename(name), "dataset")
 
     if not os.path.exists(dataset_dir):
-        return JSONResponse(status_code=200, content={})
+        return JSONResponse(
+            status_code=200, content={"categories": {}, "training_metadata": None}
+        )
 
-    for name in os.listdir(dataset_dir):
+    for category_name in os.listdir(dataset_dir):
-        category_dir = os.path.join(dataset_dir, name)
+        category_dir = os.path.join(dataset_dir, category_name)
 
         if not os.path.isdir(category_dir):
             continue
 
-        dataset_dict[name] = []
+        dataset_dict[category_name] = []
 
         for file in filter(
             lambda f: (f.lower().endswith((".webp", ".png", ".jpg", ".jpeg"))),
             os.listdir(category_dir),
         ):
-            dataset_dict[name].append(file)
+            dataset_dict[category_name].append(file)
 
-    return JSONResponse(status_code=200, content=dataset_dict)
+    # Get training metadata
+    metadata = read_training_metadata(sanitize_filename(name))
+    current_image_count = get_dataset_image_count(sanitize_filename(name))
+
+    if metadata is None:
+        training_metadata = {
+            "has_trained": False,
+            "last_training_date": None,
+            "last_training_image_count": 0,
+            "current_image_count": current_image_count,
+            "new_images_count": current_image_count,
+            "dataset_changed": current_image_count > 0,
+        }
+    else:
+        last_training_count = metadata.get("last_training_image_count", 0)
+        # Dataset has changed if count is different (either added or deleted images)
+        dataset_changed = current_image_count != last_training_count
+        # Only show positive count for new images (ignore deletions in the count display)
+        new_images_count = max(0, current_image_count - last_training_count)
+        training_metadata = {
+            "has_trained": True,
+            "last_training_date": metadata.get("last_training_date"),
+            "last_training_image_count": last_training_count,
+            "current_image_count": current_image_count,
+            "new_images_count": new_images_count,
+            "dataset_changed": dataset_changed,
+        }
+
+    return JSONResponse(
+        status_code=200,
+        content={
+            "categories": dataset_dict,
+            "training_metadata": training_metadata,
+        },
+    )
 
 
 @router.get(
@@ -655,12 +709,106 @@ def delete_classification_dataset_images(
         if os.path.isfile(file_path):
             os.unlink(file_path)
 
+    if os.path.exists(folder) and not os.listdir(folder):
+        os.rmdir(folder)
+
     return JSONResponse(
-        content=({"success": True, "message": "Successfully deleted faces."}),
+        content=({"success": True, "message": "Successfully deleted images."}),
         status_code=200,
     )
 
 
+@router.put(
+    "/classification/{name}/dataset/{old_category}/rename",
+    response_model=GenericResponse,
+    dependencies=[Depends(require_role(["admin"]))],
+    summary="Rename a classification category",
+    description="""Renames a classification category for a given classification model.
+    The old category must exist and the new name must be valid. Returns a success message or an error if the name is invalid.""",
+)
+def rename_classification_category(
+    request: Request, name: str, old_category: str, body: dict = None
+):
+    config: FrigateConfig = request.app.frigate_config
+
+    if name not in config.classification.custom:
+        return JSONResponse(
+            content=(
+                {
+                    "success": False,
+                    "message": f"{name} is not a known classification model.",
+                }
+            ),
+            status_code=404,
+        )
+
+    json: dict[str, Any] = body or {}
+    new_category = sanitize_filename(json.get("new_category", ""))
+
+    if not new_category:
+        return JSONResponse(
+            content=(
+                {
+                    "success": False,
+                    "message": "New category name is required.",
+                }
+            ),
+            status_code=400,
+        )
+
+    old_folder = os.path.join(
+        CLIPS_DIR, sanitize_filename(name), "dataset", sanitize_filename(old_category)
+    )
+    new_folder = os.path.join(
+        CLIPS_DIR, sanitize_filename(name), "dataset", new_category
+    )
+
+    if not os.path.exists(old_folder):
+        return JSONResponse(
+            content=(
+                {
+                    "success": False,
+                    "message": f"Category {old_category} does not exist.",
+                }
+            ),
+            status_code=404,
+        )
+
+    if os.path.exists(new_folder):
+        return JSONResponse(
+            content=(
+                {
+                    "success": False,
+                    "message": f"Category {new_category} already exists.",
+                }
+            ),
+            status_code=400,
+        )
+
+    try:
+        os.rename(old_folder, new_folder)
+        return JSONResponse(
+            content=(
+                {
+                    "success": True,
+                    "message": f"Successfully renamed category to {new_category}.",
+                }
+            ),
+            status_code=200,
+        )
+    except Exception as e:
+        logger.error(f"Error renaming category: {e}")
+        return JSONResponse(
+            content=(
+                {
+                    "success": False,
+                    "message": "Failed to rename category",
+                }
+            ),
+            status_code=500,
+        )
+
+
 @router.post(
     "/classification/{name}/dataset/categorize",
     response_model=GenericResponse,
@@ -701,13 +849,14 @@ def categorize_classification_image(request: Request, name: str, body: dict = No
             status_code=404,
         )
 
-    new_name = f"{category}-{datetime.datetime.now().timestamp()}.png"
+    random_id = "".join(random.choices(string.ascii_lowercase + string.digits, k=6))
+    timestamp = datetime.datetime.now().timestamp()
+    new_name = f"{category}-{timestamp}-{random_id}.png"
     new_file_folder = os.path.join(
         CLIPS_DIR, sanitize_filename(name), "dataset", category
     )
 
-    if not os.path.exists(new_file_folder):
+    os.makedirs(new_file_folder, exist_ok=True)
-        os.mkdir(new_file_folder)
 
     # use opencv because webp images can not be used to train
     img = cv2.imread(training_file)
@@ -715,7 +864,7 @@ def categorize_classification_image(request: Request, name: str, body: dict = No
     os.unlink(training_file)
 
     return JSONResponse(
-        content=({"success": True, "message": "Successfully deleted faces."}),
+        content=({"success": True, "message": "Successfully categorized image."}),
         status_code=200,
     )
 
@@ -753,6 +902,87 @@ def delete_classification_train_images(request: Request, name: str, body: dict =
             os.unlink(file_path)
 
     return JSONResponse(
-        content=({"success": True, "message": "Successfully deleted faces."}),
+        content=({"success": True, "message": "Successfully deleted images."}),
+        status_code=200,
+    )
+
+
+@router.post(
+    "/classification/generate_examples/state",
+    response_model=GenericResponse,
+    dependencies=[Depends(require_role(["admin"]))],
+    summary="Generate state classification examples",
+)
+async def generate_state_examples(request: Request, body: GenerateStateExamplesBody):
+    """Generate examples for state classification."""
+    model_name = sanitize_filename(body.model_name)
+    cameras_normalized = {
+        camera_name: tuple(crop)
+        for camera_name, crop in body.cameras.items()
+        if camera_name in request.app.frigate_config.cameras
+    }
+
+    collect_state_classification_examples(model_name, cameras_normalized)
+
+    return JSONResponse(
+        content={"success": True, "message": "Example generation completed"},
+        status_code=200,
+    )
+
+
+@router.post(
+    "/classification/generate_examples/object",
+    response_model=GenericResponse,
+    dependencies=[Depends(require_role(["admin"]))],
+    summary="Generate object classification examples",
+)
+async def generate_object_examples(request: Request, body: GenerateObjectExamplesBody):
+    """Generate examples for object classification."""
+    model_name = sanitize_filename(body.model_name)
+    collect_object_classification_examples(model_name, body.label)
+
+    return JSONResponse(
+        content={"success": True, "message": "Example generation completed"},
+        status_code=200,
+    )
+
+
+@router.delete(
+    "/classification/{name}",
+    response_model=GenericResponse,
+    dependencies=[Depends(require_role(["admin"]))],
+    summary="Delete a classification model",
+    description="""Deletes a specific classification model and all its associated data.
+    Works even if the model is not in the config (e.g., partially created during wizard).
+    Returns a success message.""",
+)
+def delete_classification_model(request: Request, name: str):
+    sanitized_name = sanitize_filename(name)
+
+    # Delete the classification model's data directory in clips
+    data_dir = os.path.join(CLIPS_DIR, sanitized_name)
+    if os.path.exists(data_dir):
+        try:
+            shutil.rmtree(data_dir)
+            logger.info(f"Deleted classification data directory for {name}")
+        except Exception as e:
+            logger.debug(f"Failed to delete data directory for {name}: {e}")
+
+    # Delete the classification model's files in model_cache
+    model_dir = os.path.join(MODEL_CACHE_DIR, sanitized_name)
+    if os.path.exists(model_dir):
+        try:
+            shutil.rmtree(model_dir)
+            logger.info(f"Deleted classification model directory for {name}")
+        except Exception as e:
+            logger.debug(f"Failed to delete model directory for {name}: {e}")
+
+    return JSONResponse(
+        content=(
+            {
+                "success": True,
+                "message": f"Successfully deleted classification model {name}.",
+            }
+        ),
         status_code=200,
     )

frigate/api/defs/request/classification_body.py

@@ -1,17 +1,31 @@
-from typing import List
+from typing import Dict, List, Tuple
 
 from pydantic import BaseModel, Field
 
 
 class RenameFaceBody(BaseModel):
-    new_name: str
+    new_name: str = Field(description="New name for the face")
 
 
 class AudioTranscriptionBody(BaseModel):
-    event_id: str
+    event_id: str = Field(description="ID of the event to transcribe audio for")
 
 
 class DeleteFaceImagesBody(BaseModel):
     ids: List[str] = Field(
         description="List of image filenames to delete from the face folder"
     )
+
+
+class GenerateStateExamplesBody(BaseModel):
+    model_name: str = Field(description="Name of the classification model")
+    cameras: Dict[str, Tuple[float, float, float, float]] = Field(
+        description="Dictionary mapping camera names to normalized crop coordinates in [x1, y1, x2, y2] format (values 0-1)"
+    )
+
+
+class GenerateObjectExamplesBody(BaseModel):
+    model_name: str = Field(description="Name of the classification model")
+    label: str = Field(
+        description="Object label to collect examples for (e.g., 'person', 'car')"
+    )

frigate/api/event.py

@@ -2,6 +2,7 @@
 
 import base64
 import datetime
+import json
 import logging
 import os
 import random
@@ -57,8 +58,8 @@ from frigate.const import CLIPS_DIR, TRIGGER_DIR
 from frigate.embeddings import EmbeddingsContext
 from frigate.models import Event, ReviewSegment, Timeline, Trigger
 from frigate.track.object_processing import TrackedObject
-from frigate.util.builtin import get_tz_modifiers
+from frigate.util.file import get_event_thumbnail_bytes
-from frigate.util.path import get_event_thumbnail_bytes
+from frigate.util.time import get_dst_transitions, get_tz_modifiers
 
 logger = logging.getLogger(__name__)
 
@@ -813,7 +814,6 @@ def events_summary(
     allowed_cameras: List[str] = Depends(get_allowed_cameras_for_filter),
 ):
     tz_name = params.timezone
-    hour_modifier, minute_modifier, seconds_offset = get_tz_modifiers(tz_name)
     has_clip = params.has_clip
     has_snapshot = params.has_snapshot
 
@@ -828,33 +828,91 @@ def events_summary(
     if len(clauses) == 0:
         clauses.append((True))
 
-    groups = (
+    time_range_query = (
         Event.select(
-            Event.camera,
+            fn.MIN(Event.start_time).alias("min_time"),
-            Event.label,
+            fn.MAX(Event.start_time).alias("max_time"),
-            Event.sub_label,
-            Event.data,
-            fn.strftime(
-                "%Y-%m-%d",
-                fn.datetime(
-                    Event.start_time, "unixepoch", hour_modifier, minute_modifier
-                ),
-            ).alias("day"),
-            Event.zones,
-            fn.COUNT(Event.id).alias("count"),
         )
         .where(reduce(operator.and_, clauses) & (Event.camera << allowed_cameras))
-        .group_by(
+        .dicts()
-            Event.camera,
+        .get()
-            Event.label,
-            Event.sub_label,
-            Event.data,
-            (Event.start_time + seconds_offset).cast("int") / (3600 * 24),
-            Event.zones,
-        )
     )
 
-    return JSONResponse(content=[e for e in groups.dicts()])
+    min_time = time_range_query.get("min_time")
+    max_time = time_range_query.get("max_time")
+
+    if min_time is None or max_time is None:
+        return JSONResponse(content=[])
+
+    dst_periods = get_dst_transitions(tz_name, min_time, max_time)
+
+    grouped: dict[tuple, dict] = {}
+
+    for period_start, period_end, period_offset in dst_periods:
+        hours_offset = int(period_offset / 60 / 60)
+        minutes_offset = int(period_offset / 60 - hours_offset * 60)
+        period_hour_modifier = f"{hours_offset} hour"
+        period_minute_modifier = f"{minutes_offset} minute"
+
+        period_groups = (
+            Event.select(
+                Event.camera,
+                Event.label,
+                Event.sub_label,
+                Event.data,
+                fn.strftime(
+                    "%Y-%m-%d",
+                    fn.datetime(
+                        Event.start_time,
+                        "unixepoch",
+                        period_hour_modifier,
+                        period_minute_modifier,
+                    ),
+                ).alias("day"),
+                Event.zones,
+                fn.COUNT(Event.id).alias("count"),
+            )
+            .where(
+                reduce(operator.and_, clauses)
+                & (Event.camera << allowed_cameras)
+                & (Event.start_time >= period_start)
+                & (Event.start_time <= period_end)
+            )
+            .group_by(
+                Event.camera,
+                Event.label,
+                Event.sub_label,
+                Event.data,
+                (Event.start_time + period_offset).cast("int") / (3600 * 24),
+                Event.zones,
+            )
+            .namedtuples()
+        )
+
+        for g in period_groups:
+            key = (
+                g.camera,
+                g.label,
+                g.sub_label,
+                json.dumps(g.data, sort_keys=True) if g.data is not None else None,
+                g.day,
+                json.dumps(g.zones, sort_keys=True) if g.zones is not None else None,
+            )
+
+            if key in grouped:
+                grouped[key]["count"] += int(g.count or 0)
+            else:
+                grouped[key] = {
+                    "camera": g.camera,
+                    "label": g.label,
+                    "sub_label": g.sub_label,
+                    "data": g.data,
+                    "day": g.day,
+                    "zones": g.zones,
+                    "count": int(g.count or 0),
+                }
+
+    return JSONResponse(content=sorted(grouped.values(), key=lambda x: x["day"]))
 
 
 @router.get(
@@ -1723,9 +1781,8 @@ def create_trigger_embedding(
                 logger.debug(
                     f"Writing thumbnail for trigger with data {body.data} in {camera_name}."
                 )
-            except Exception as e:
+            except Exception:
-                logger.error(e.with_traceback())
+                logger.exception(
-                logger.error(
                     f"Failed to write thumbnail for trigger with data {body.data} in {camera_name}"
                 )
 
@@ -1749,8 +1806,8 @@ def create_trigger_embedding(
             status_code=200,
         )
 
-    except Exception as e:
+    except Exception:
-        logger.error(e.with_traceback())
+        logger.exception("Error creating trigger embedding")
         return JSONResponse(
             content={
                 "success": False,
@@ -1859,9 +1916,8 @@ def update_trigger_embedding(
                     logger.debug(
                         f"Deleted thumbnail for trigger with data {trigger.data} in {camera_name}."
                     )
-                except Exception as e:
+                except Exception:
-                    logger.error(e.with_traceback())
+                    logger.exception(
-                    logger.error(
                         f"Failed to delete thumbnail for trigger with data {trigger.data} in {camera_name}"
                     )
 
@@ -1900,9 +1956,8 @@ def update_trigger_embedding(
                 logger.debug(
                     f"Writing thumbnail for trigger with data {body.data} in {camera_name}."
                 )
-            except Exception as e:
+            except Exception:
-                logger.error(e.with_traceback())
+                logger.exception(
-                logger.error(
                     f"Failed to write thumbnail for trigger with data {body.data} in {camera_name}"
                 )
 
@@ -1914,8 +1969,8 @@ def update_trigger_embedding(
             status_code=200,
         )
 
-    except Exception as e:
+    except Exception:
-        logger.error(e.with_traceback())
+        logger.exception("Error updating trigger embedding")
         return JSONResponse(
             content={
                 "success": False,
@@ -1975,9 +2030,8 @@ def delete_trigger_embedding(
             logger.debug(
                 f"Deleted thumbnail for trigger with data {trigger.data} in {camera_name}."
             )
-        except Exception as e:
+        except Exception:
-            logger.error(e.with_traceback())
+            logger.exception(
-            logger.error(
                 f"Failed to delete thumbnail for trigger with data {trigger.data} in {camera_name}"
             )
 
@@ -1989,8 +2043,8 @@ def delete_trigger_embedding(
             status_code=200,
         )
 
-    except Exception as e:
+    except Exception:
-        logger.error(e.with_traceback())
+        logger.exception("Error deleting trigger embedding")
         return JSONResponse(
             content={
                 "success": False,

frigate/api/export.py

@@ -9,6 +9,7 @@ from typing import List
 import psutil
 from fastapi import APIRouter, Depends, Request
 from fastapi.responses import JSONResponse
+from pathvalidate import sanitize_filepath
 from peewee import DoesNotExist
 from playhouse.shortcuts import model_to_dict
 
@@ -26,14 +27,14 @@ from frigate.api.defs.response.export_response import (
 )
 from frigate.api.defs.response.generic_response import GenericResponse
 from frigate.api.defs.tags import Tags
-from frigate.const import EXPORT_DIR
+from frigate.const import CLIPS_DIR, EXPORT_DIR
 from frigate.models import Export, Previews, Recordings
 from frigate.record.export import (
     PlaybackFactorEnum,
     PlaybackSourceEnum,
     RecordingExporter,
 )
-from frigate.util.builtin import is_current_hour
+from frigate.util.time import is_current_hour
 
 logger = logging.getLogger(__name__)
 
@@ -88,7 +89,14 @@ def export_recording(
     playback_factor = body.playback
     playback_source = body.source
     friendly_name = body.name
-    existing_image = body.image_path
+    existing_image = sanitize_filepath(body.image_path) if body.image_path else None
+
+    # Ensure that existing_image is a valid path
+    if existing_image and not existing_image.startswith(CLIPS_DIR):
+        return JSONResponse(
+            content=({"success": False, "message": "Invalid image path"}),
+            status_code=400,
+        )
 
     if playback_source == "recordings":
         recordings_count = (

frigate/api/media.py

@@ -44,9 +44,9 @@ from frigate.const import (
 )
 from frigate.models import Event, Previews, Recordings, Regions, ReviewSegment
 from frigate.track.object_processing import TrackedObjectProcessor
-from frigate.util.builtin import get_tz_modifiers
+from frigate.util.file import get_event_thumbnail_bytes
 from frigate.util.image import get_image_from_recording
-from frigate.util.path import get_event_thumbnail_bytes
+from frigate.util.time import get_dst_transitions
 
 logger = logging.getLogger(__name__)
 
@@ -424,7 +424,6 @@ def all_recordings_summary(
     allowed_cameras: List[str] = Depends(get_allowed_cameras_for_filter),
 ):
     """Returns true/false by day indicating if recordings exist"""
-    hour_modifier, minute_modifier, seconds_offset = get_tz_modifiers(params.timezone)
 
     cameras = params.cameras
     if cameras != "all":
@@ -432,43 +431,72 @@ def all_recordings_summary(
         filtered = requested.intersection(allowed_cameras)
         if not filtered:
             return JSONResponse(content={})
-        cameras = ",".join(filtered)
+        camera_list = list(filtered)
     else:
-        cameras = allowed_cameras
+        camera_list = allowed_cameras
 
-    query = (
+    time_range_query = (
         Recordings.select(
-            fn.strftime(
+            fn.MIN(Recordings.start_time).alias("min_time"),
-                "%Y-%m-%d",
+            fn.MAX(Recordings.start_time).alias("max_time"),
-                fn.datetime(
-                    Recordings.start_time + seconds_offset,
-                    "unixepoch",
-                    hour_modifier,
-                    minute_modifier,
-                ),
-            ).alias("day")
         )
-        .group_by(
+        .where(Recordings.camera << camera_list)
-            fn.strftime(
+        .dicts()
-                "%Y-%m-%d",
+        .get()
-                fn.datetime(
-                    Recordings.start_time + seconds_offset,
-                    "unixepoch",
-                    hour_modifier,
-                    minute_modifier,
-                ),
-            )
-        )
-        .order_by(Recordings.start_time.desc())
     )
 
-    if params.cameras != "all":
+    min_time = time_range_query.get("min_time")
-        query = query.where(Recordings.camera << cameras.split(","))
+    max_time = time_range_query.get("max_time")
 
-    recording_days = query.namedtuples()
+    if min_time is None or max_time is None:
-    days = {day.day: True for day in recording_days}
+        return JSONResponse(content={})
 
-    return JSONResponse(content=days)
+    dst_periods = get_dst_transitions(params.timezone, min_time, max_time)
+
+    days: dict[str, bool] = {}
+
+    for period_start, period_end, period_offset in dst_periods:
+        hours_offset = int(period_offset / 60 / 60)
+        minutes_offset = int(period_offset / 60 - hours_offset * 60)
+        period_hour_modifier = f"{hours_offset} hour"
+        period_minute_modifier = f"{minutes_offset} minute"
+
+        period_query = (
+            Recordings.select(
+                fn.strftime(
+                    "%Y-%m-%d",
+                    fn.datetime(
+                        Recordings.start_time,
+                        "unixepoch",
+                        period_hour_modifier,
+                        period_minute_modifier,
+                    ),
+                ).alias("day")
+            )
+            .where(
+                (Recordings.camera << camera_list)
+                & (Recordings.end_time >= period_start)
+                & (Recordings.start_time <= period_end)
+            )
+            .group_by(
+                fn.strftime(
+                    "%Y-%m-%d",
+                    fn.datetime(
+                        Recordings.start_time,
+                        "unixepoch",
+                        period_hour_modifier,
+                        period_minute_modifier,
+                    ),
+                )
+            )
+            .order_by(Recordings.start_time.desc())
+            .namedtuples()
+        )
+
+        for g in period_query:
+            days[g.day] = True
+
+    return JSONResponse(content=dict(sorted(days.items())))
 
 
 @router.get(
@@ -476,61 +504,103 @@ def all_recordings_summary(
 )
 async def recordings_summary(camera_name: str, timezone: str = "utc"):
     """Returns hourly summary for recordings of given camera"""
-    hour_modifier, minute_modifier, seconds_offset = get_tz_modifiers(timezone)
+
-    recording_groups = (
+    time_range_query = (
         Recordings.select(
-            fn.strftime(
+            fn.MIN(Recordings.start_time).alias("min_time"),
-                "%Y-%m-%d %H",
+            fn.MAX(Recordings.start_time).alias("max_time"),
-                fn.datetime(
-                    Recordings.start_time, "unixepoch", hour_modifier, minute_modifier
-                ),
-            ).alias("hour"),
-            fn.SUM(Recordings.duration).alias("duration"),
-            fn.SUM(Recordings.motion).alias("motion"),
-            fn.SUM(Recordings.objects).alias("objects"),
         )
         .where(Recordings.camera == camera_name)
-        .group_by((Recordings.start_time + seconds_offset).cast("int") / 3600)
+        .dicts()
-        .order_by(Recordings.start_time.desc())
+        .get()
-        .namedtuples()
     )
 
-    event_groups = (
+    min_time = time_range_query.get("min_time")
-        Event.select(
+    max_time = time_range_query.get("max_time")
-            fn.strftime(
+
-                "%Y-%m-%d %H",
+    days: dict[str, dict] = {}
-                fn.datetime(
+
-                    Event.start_time, "unixepoch", hour_modifier, minute_modifier
+    if min_time is None or max_time is None:
-                ),
+        return JSONResponse(content=list(days.values()))
-            ).alias("hour"),
+
-            fn.COUNT(Event.id).alias("count"),
+    dst_periods = get_dst_transitions(timezone, min_time, max_time)
+
+    for period_start, period_end, period_offset in dst_periods:
+        hours_offset = int(period_offset / 60 / 60)
+        minutes_offset = int(period_offset / 60 - hours_offset * 60)
+        period_hour_modifier = f"{hours_offset} hour"
+        period_minute_modifier = f"{minutes_offset} minute"
+
+        recording_groups = (
+            Recordings.select(
+                fn.strftime(
+                    "%Y-%m-%d %H",
+                    fn.datetime(
+                        Recordings.start_time,
+                        "unixepoch",
+                        period_hour_modifier,
+                        period_minute_modifier,
+                    ),
+                ).alias("hour"),
+                fn.SUM(Recordings.duration).alias("duration"),
+                fn.SUM(Recordings.motion).alias("motion"),
+                fn.SUM(Recordings.objects).alias("objects"),
+            )
+            .where(
+                (Recordings.camera == camera_name)
+                & (Recordings.end_time >= period_start)
+                & (Recordings.start_time <= period_end)
+            )
+            .group_by((Recordings.start_time + period_offset).cast("int") / 3600)
+            .order_by(Recordings.start_time.desc())
+            .namedtuples()
         )
-        .where(Event.camera == camera_name, Event.has_clip)
-        .group_by((Event.start_time + seconds_offset).cast("int") / 3600)
-        .namedtuples()
-    )
 
-    event_map = {g.hour: g.count for g in event_groups}
+        event_groups = (
+            Event.select(
+                fn.strftime(
+                    "%Y-%m-%d %H",
+                    fn.datetime(
+                        Event.start_time,
+                        "unixepoch",
+                        period_hour_modifier,
+                        period_minute_modifier,
+                    ),
+                ).alias("hour"),
+                fn.COUNT(Event.id).alias("count"),
+            )
+            .where(Event.camera == camera_name, Event.has_clip)
+            .where(
+                (Event.start_time >= period_start) & (Event.start_time <= period_end)
+            )
+            .group_by((Event.start_time + period_offset).cast("int") / 3600)
+            .namedtuples()
+        )
 
-    days = {}
+        event_map = {g.hour: g.count for g in event_groups}
 
-    for recording_group in recording_groups:
+        for recording_group in recording_groups:
-        parts = recording_group.hour.split()
+            parts = recording_group.hour.split()
-        hour = parts[1]
+            hour = parts[1]
-        day = parts[0]
+            day = parts[0]
-        events_count = event_map.get(recording_group.hour, 0)
+            events_count = event_map.get(recording_group.hour, 0)
-        hour_data = {
+            hour_data = {
-            "hour": hour,
+                "hour": hour,
-            "events": events_count,
+                "events": events_count,
-            "motion": recording_group.motion,
+                "motion": recording_group.motion,
-            "objects": recording_group.objects,
+                "objects": recording_group.objects,
-            "duration": round(recording_group.duration),
+                "duration": round(recording_group.duration),
-        }
+            }
-        if day not in days:
+            if day in days:
-            days[day] = {"events": events_count, "hours": [hour_data], "day": day}
+                # merge counts if already present (edge-case at DST boundary)
-        else:
+                days[day]["events"] += events_count or 0
-            days[day]["events"] += events_count
+                days[day]["hours"].append(hour_data)
-            days[day]["hours"].append(hour_data)
+            else:
+                days[day] = {
+                    "events": events_count or 0,
+                    "hours": [hour_data],
+                    "day": day,
+                }
 
     return JSONResponse(content=list(days.values()))
 
@@ -589,7 +659,7 @@ async def no_recordings(
     )
     scale = params.scale
 
-    clauses = [(Recordings.start_time > after) & (Recordings.end_time < before)]
+    clauses = [(Recordings.end_time >= after) & (Recordings.start_time <= before)]
     if cameras != "all":
         camera_list = cameras.split(",")
         clauses.append((Recordings.camera << camera_list))
@@ -608,33 +678,39 @@ async def no_recordings(
     # Convert recordings to list of (start, end) tuples
     recordings = [(r["start_time"], r["end_time"]) for r in data]
 
-    # Generate all time segments
+    # Iterate through time segments and check if each has any recording
-    current = after
     no_recording_segments = []
-    current_start = None
+    current = after
+    current_gap_start = None
 
     while current < before:
-        segment_end = current + scale
+        segment_end = min(current + scale, before)
-        # Check if segment overlaps with any recording
+
+        # Check if this segment overlaps with any recording
         has_recording = any(
-            start <= segment_end and end >= current for start, end in recordings
+            rec_start < segment_end and rec_end > current
+            for rec_start, rec_end in recordings
         )
+
         if not has_recording:
-            if current_start is None:
+            # This segment has no recordings
-                current_start = current  # Start a new gap
+            if current_gap_start is None:
+                current_gap_start = current  # Start a new gap
         else:
-            if current_start is not None:
+            # This segment has recordings
+            if current_gap_start is not None:
                 # End the current gap and append it
                 no_recording_segments.append(
-                    {"start_time": int(current_start), "end_time": int(current)}
+                    {"start_time": int(current_gap_start), "end_time": int(current)}
                 )
-                current_start = None
+                current_gap_start = None
+
         current = segment_end
 
     # Append the last gap if it exists
-    if current_start is not None:
+    if current_gap_start is not None:
         no_recording_segments.append(
-            {"start_time": int(current_start), "end_time": int(before)}
+            {"start_time": int(current_gap_start), "end_time": int(before)}
         )
 
     return JSONResponse(content=no_recording_segments)
@@ -686,6 +762,15 @@ async def recording_clip(
         .order_by(Recordings.start_time.asc())
     )
 
+    if recordings.count() == 0:
+        return JSONResponse(
+            content={
+                "success": False,
+                "message": "No recordings found for the specified time range",
+            },
+            status_code=400,
+        )
+
     file_name = sanitize_filename(f"playlist_{camera_name}_{start_ts}-{end_ts}.txt")
     file_path = os.path.join(CACHE_DIR, file_name)
     with open(file_path, "w") as file:
@@ -764,6 +849,7 @@ async def vod_ts(camera_name: str, start_ts: float, end_ts: float):
 
     clips = []
     durations = []
+    min_duration_ms = 100  # Minimum 100ms to ensure at least one video frame
     max_duration_ms = MAX_SEGMENT_DURATION * 1000
 
     recording: Recordings
@@ -781,11 +867,11 @@ async def vod_ts(camera_name: str, start_ts: float, end_ts: float):
         if recording.end_time > end_ts:
             duration -= int((recording.end_time - end_ts) * 1000)
 
-        if duration <= 0:
+        if duration < min_duration_ms:
-            # skip if the clip has no valid duration
+            # skip if the clip has no valid duration (too short to contain frames)
             continue
 
-        if 0 < duration < max_duration_ms:
+        if min_duration_ms <= duration < max_duration_ms:
             clip["keyFrameDurations"] = [duration]
             clips.append(clip)
             durations.append(duration)

frigate/api/review.py

@@ -36,7 +36,7 @@ from frigate.config import FrigateConfig
 from frigate.embeddings import EmbeddingsContext
 from frigate.models import Recordings, ReviewSegment, UserReviewStatus
 from frigate.review.types import SeverityEnum
-from frigate.util.builtin import get_tz_modifiers
+from frigate.util.time import get_dst_transitions
 
 logger = logging.getLogger(__name__)
 
@@ -197,7 +197,6 @@ async def review_summary(
 
     user_id = current_user["username"]
 
-    hour_modifier, minute_modifier, seconds_offset = get_tz_modifiers(params.timezone)
     day_ago = (datetime.datetime.now() - datetime.timedelta(hours=24)).timestamp()
 
     cameras = params.cameras
@@ -329,89 +328,135 @@ async def review_summary(
             )
         clauses.append(reduce(operator.or_, label_clauses))
 
-    day_in_seconds = 60 * 60 * 24
+    # Find the time range of available data
-    last_month_query = (
+    time_range_query = (
         ReviewSegment.select(
-            fn.strftime(
+            fn.MIN(ReviewSegment.start_time).alias("min_time"),
-                "%Y-%m-%d",
+            fn.MAX(ReviewSegment.start_time).alias("max_time"),
-                fn.datetime(
-                    ReviewSegment.start_time,
-                    "unixepoch",
-                    hour_modifier,
-                    minute_modifier,
-                ),
-            ).alias("day"),
-            fn.SUM(
-                Case(
-                    None,
-                    [
-                        (
-                            (ReviewSegment.severity == SeverityEnum.alert)
-                            & (UserReviewStatus.has_been_reviewed == True),
-                            1,
-                        )
-                    ],
-                    0,
-                )
-            ).alias("reviewed_alert"),
-            fn.SUM(
-                Case(
-                    None,
-                    [
-                        (
-                            (ReviewSegment.severity == SeverityEnum.detection)
-                            & (UserReviewStatus.has_been_reviewed == True),
-                            1,
-                        )
-                    ],
-                    0,
-                )
-            ).alias("reviewed_detection"),
-            fn.SUM(
-                Case(
-                    None,
-                    [
-                        (
-                            (ReviewSegment.severity == SeverityEnum.alert),
-                            1,
-                        )
-                    ],
-                    0,
-                )
-            ).alias("total_alert"),
-            fn.SUM(
-                Case(
-                    None,
-                    [
-                        (
-                            (ReviewSegment.severity == SeverityEnum.detection),
-                            1,
-                        )
-                    ],
-                    0,
-                )
-            ).alias("total_detection"),
-        )
-        .left_outer_join(
-            UserReviewStatus,
-            on=(
-                (ReviewSegment.id == UserReviewStatus.review_segment)
-                & (UserReviewStatus.user_id == user_id)
-            ),
         )
         .where(reduce(operator.and_, clauses) if clauses else True)
-        .group_by(
+        .dicts()
-            (ReviewSegment.start_time + seconds_offset).cast("int") / day_in_seconds
+        .get()
-        )
-        .order_by(ReviewSegment.start_time.desc())
     )
 
+    min_time = time_range_query.get("min_time")
+    max_time = time_range_query.get("max_time")
+
     data = {
         "last24Hours": last_24_query,
     }
 
-    for e in last_month_query.dicts().iterator():
+    # If no data, return early
-        data[e["day"]] = e
+    if min_time is None or max_time is None:
+        return JSONResponse(content=data)
+
+    # Get DST transition periods
+    dst_periods = get_dst_transitions(params.timezone, min_time, max_time)
+
+    day_in_seconds = 60 * 60 * 24
+
+    # Query each DST period separately with the correct offset
+    for period_start, period_end, period_offset in dst_periods:
+        # Calculate hour/minute modifiers for this period
+        hours_offset = int(period_offset / 60 / 60)
+        minutes_offset = int(period_offset / 60 - hours_offset * 60)
+        period_hour_modifier = f"{hours_offset} hour"
+        period_minute_modifier = f"{minutes_offset} minute"
+
+        # Build clauses including time range for this period
+        period_clauses = clauses.copy()
+        period_clauses.append(
+            (ReviewSegment.start_time >= period_start)
+            & (ReviewSegment.start_time <= period_end)
+        )
+
+        period_query = (
+            ReviewSegment.select(
+                fn.strftime(
+                    "%Y-%m-%d",
+                    fn.datetime(
+                        ReviewSegment.start_time,
+                        "unixepoch",
+                        period_hour_modifier,
+                        period_minute_modifier,
+                    ),
+                ).alias("day"),
+                fn.SUM(
+                    Case(
+                        None,
+                        [
+                            (
+                                (ReviewSegment.severity == SeverityEnum.alert)
+                                & (UserReviewStatus.has_been_reviewed == True),
+                                1,
+                            )
+                        ],
+                        0,
+                    )
+                ).alias("reviewed_alert"),
+                fn.SUM(
+                    Case(
+                        None,
+                        [
+                            (
+                                (ReviewSegment.severity == SeverityEnum.detection)
+                                & (UserReviewStatus.has_been_reviewed == True),
+                                1,
+                            )
+                        ],
+                        0,
+                    )
+                ).alias("reviewed_detection"),
+                fn.SUM(
+                    Case(
+                        None,
+                        [
+                            (
+                                (ReviewSegment.severity == SeverityEnum.alert),
+                                1,
+                            )
+                        ],
+                        0,
+                    )
+                ).alias("total_alert"),
+                fn.SUM(
+                    Case(
+                        None,
+                        [
+                            (
+                                (ReviewSegment.severity == SeverityEnum.detection),
+                                1,
+                            )
+                        ],
+                        0,
+                    )
+                ).alias("total_detection"),
+            )
+            .left_outer_join(
+                UserReviewStatus,
+                on=(
+                    (ReviewSegment.id == UserReviewStatus.review_segment)
+                    & (UserReviewStatus.user_id == user_id)
+                ),
+            )
+            .where(reduce(operator.and_, period_clauses))
+            .group_by(
+                (ReviewSegment.start_time + period_offset).cast("int") / day_in_seconds
+            )
+            .order_by(ReviewSegment.start_time.desc())
+        )
+
+        # Merge results from this period
+        for e in period_query.dicts().iterator():
+            day_key = e["day"]
+            if day_key in data:
+                # Merge counts if day already exists (edge case at DST boundary)
+                data[day_key]["reviewed_alert"] += e["reviewed_alert"] or 0
+                data[day_key]["reviewed_detection"] += e["reviewed_detection"] or 0
+                data[day_key]["total_alert"] += e["total_alert"] or 0
+                data[day_key]["total_detection"] += e["total_detection"] or 0
+            else:
+                data[day_key] = e
 
     return JSONResponse(content=data)
 

frigate/app.py

@@ -488,6 +488,8 @@ class FrigateApp:
                     }
                 ).execute()
 
+                self.config.auth.admin_first_time_login = True
+
                 logger.info("********************************************************")
                 logger.info("********************************************************")
                 logger.info("***    Auth is enabled, but no users exist.          ***")

frigate/camera/maintainer.py

@@ -136,6 +136,7 @@ class CameraMaintainer(threading.Thread):
             self.ptz_metrics[name],
             self.region_grids[name],
             self.stop_event,
+            self.config.logger,
         )
         self.camera_processes[config.name] = camera_process
         camera_process.start()
@@ -156,7 +157,11 @@ class CameraMaintainer(threading.Thread):
             self.frame_manager.create(f"{config.name}_frame{i}", frame_size)
 
         capture_process = CameraCapture(
-            config, count, self.camera_metrics[name], self.stop_event
+            config,
+            count,
+            self.camera_metrics[name],
+            self.stop_event,
+            self.config.logger,
         )
         capture_process.daemon = True
         self.capture_processes[name] = capture_process

frigate/config/auth.py

@@ -38,6 +38,13 @@ class AuthConfig(FrigateBaseModel):
         default_factory=dict,
         title="Role to camera mappings. Empty list grants access to all cameras.",
     )
+    admin_first_time_login: Optional[bool] = Field(
+        default=False,
+        title="Internal field to expose first-time admin login flag to the UI",
+        description=(
+            "When true the UI may show a help link on the login page informing users how to sign in after an admin password reset. "
+        ),
+    )
 
     @field_validator("roles")
     @classmethod

frigate/config/camera/birdseye.py

@@ -55,6 +55,12 @@ class BirdseyeConfig(FrigateBaseModel):
     layout: BirdseyeLayoutConfig = Field(
         default_factory=BirdseyeLayoutConfig, title="Birdseye Layout Config"
     )
+    idle_heartbeat_fps: float = Field(
+        default=0.0,
+        ge=0.0,
+        le=10.0,
+        title="Idle heartbeat FPS (0 disables, max 10)",
+    )
 
 
 # uses BaseModel because some global attributes are not available at the camera level

frigate/config/camera/camera.py

@@ -177,6 +177,12 @@ class CameraConfig(FrigateBaseModel):
     def ffmpeg_cmds(self) -> list[dict[str, list[str]]]:
         return self._ffmpeg_cmds
 
+    def get_formatted_name(self) -> str:
+        """Return the friendly name if set, otherwise return a formatted version of the camera name."""
+        if self.friendly_name:
+            return self.friendly_name
+        return self.name.replace("_", " ").title() if self.name else ""
+
     def create_ffmpeg_cmds(self):
         if "_ffmpeg_cmds" in self:
             return

frigate/config/camera/review.py

@@ -1,10 +1,18 @@
+from enum import Enum
 from typing import Optional, Union
 
 from pydantic import Field, field_validator
 
 from ..base import FrigateBaseModel
 
-__all__ = ["ReviewConfig", "DetectionsConfig", "AlertsConfig"]
+__all__ = ["ReviewConfig", "DetectionsConfig", "AlertsConfig", "ImageSourceEnum"]
+
+
+class ImageSourceEnum(str, Enum):
+    """Image source options for GenAI Review."""
+
+    preview = "preview"
+    recordings = "recordings"
 
 
 DEFAULT_ALERT_OBJECTS = ["person", "car"]
@@ -77,6 +85,10 @@ class GenAIReviewConfig(FrigateBaseModel):
     )
     alerts: bool = Field(default=True, title="Enable GenAI for alerts.")
     detections: bool = Field(default=False, title="Enable GenAI for detections.")
+    image_source: ImageSourceEnum = Field(
+        default=ImageSourceEnum.preview,
+        title="Image source for review descriptions.",
+    )
     additional_concerns: list[str] = Field(
         default=[],
         title="Additional concerns that GenAI should make note of on this camera.",
@@ -93,13 +105,40 @@ class GenAIReviewConfig(FrigateBaseModel):
         default=None,
     )
     activity_context_prompt: str = Field(
-        default="""- **Zone context is critical**: Private enclosed spaces (back yards, back decks, fenced areas, inside garages) are resident territory where brief transient activity, routine tasks, and pet care are expected and normal. Front yards, driveways, and porches are semi-public but still resident spaces where deliveries, parking, and coming/going are routine. Consider whether the zone and activity align with normal residential use.
+        default="""### Normal Activity Indicators (Level 0)
-- **Person + Pet = Normal Activity**: When both "Person" and "Dog" (or "Cat") are detected together in residential zones, this is routine pet care activity (walking, letting out, playing, supervising). Assign Level 0 unless there are OTHER strong suspicious behaviors present (like testing doors, taking items, etc.). A person with their pet in a residential zone is baseline normal activity.
+- Known/verified people in any zone at any time
-- Brief appearances in private zones (back yards, garages) are normal residential patterns.
+- People with pets in residential areas
-- Normal residential activity includes: residents, family members, guests, deliveries, services, maintenance workers, routine property use (parking, unloading, mail pickup, trash removal).
+- Deliveries or services during daytime/evening (6 AM - 10 PM): carrying packages to doors/porches, placing items, leaving
-- Brief movement with legitimate items (bags, packages, tools, equipment) in appropriate zones is routine.
+- Services/maintenance workers with visible tools, uniforms, or service vehicles during daytime
-""",
+- Activity confined to public areas only (sidewalks, streets) without entering property at any time
-        title="Custom activity context prompt defining normal activity patterns for this property.",
+
+### Suspicious Activity Indicators (Level 1)
+- **Testing or attempting to open doors/windows/handles on vehicles or buildings** — ALWAYS Level 1 regardless of time or duration
+- **Unidentified person in private areas (driveways, near vehicles/buildings) during late night/early morning (11 PM - 5 AM)** — ALWAYS Level 1 regardless of activity or duration
+- Taking items that don't belong to them (packages, objects from porches/driveways)
+- Climbing or jumping fences/barriers to access property
+- Attempting to conceal actions or items from view
+- Prolonged loitering: remaining in same area without visible purpose throughout most of the sequence
+
+### Critical Threat Indicators (Level 2)
+- Holding break-in tools (crowbars, pry bars, bolt cutters)
+- Weapons visible (guns, knives, bats used aggressively)
+- Forced entry in progress
+- Physical aggression or violence
+- Active property damage or theft in progress
+
+### Assessment Guidance
+Evaluate in this order:
+
+1. **If person is verified/known** → Level 0 regardless of time or activity
+2. **If person is unidentified:**
+   - Check time: If late night/early morning (11 PM - 5 AM) AND in private areas (driveways, near vehicles/buildings) → Level 1
+   - Check actions: If testing doors/handles, taking items, climbing → Level 1
+   - Otherwise, if daytime/evening (6 AM - 10 PM) with clear legitimate purpose (delivery, service worker) → Level 0
+3. **Escalate to Level 2 if:** Weapons, break-in tools, forced entry in progress, violence, or active property damage visible (escalates from Level 0 or 1)
+
+The mere presence of an unidentified person in private areas during late night hours is inherently suspicious and warrants human review, regardless of what activity they appear to be doing or how brief the sequence is.""",
+        title="Custom activity context prompt defining normal and suspicious activity patterns for this property.",
     )
 
 

frigate/config/camera/zone.py

@@ -13,6 +13,9 @@ logger = logging.getLogger(__name__)
 
 
 class ZoneConfig(BaseModel):
+    friendly_name: Optional[str] = Field(
+        None, title="Zone friendly name used in the Frigate UI."
+    )
     filters: dict[str, FilterConfig] = Field(
         default_factory=dict, title="Zone filters."
     )
@@ -53,6 +56,12 @@ class ZoneConfig(BaseModel):
     def contour(self) -> np.ndarray:
         return self._contour
 
+    def get_formatted_name(self, zone_name: str) -> str:
+        """Return the friendly name if set, otherwise return a formatted version of the zone name."""
+        if self.friendly_name:
+            return self.friendly_name
+        return zone_name.replace("_", " ").title()
+
     @field_validator("objects", mode="before")
     @classmethod
     def validate_objects(cls, v):

frigate/config/classification.py

@@ -33,6 +33,8 @@ class TriggerType(str, Enum):
 
 class TriggerAction(str, Enum):
     NOTIFICATION = "notification"
+    SUB_LABEL = "sub_label"
+    ATTRIBUTE = "attribute"
 
 
 class ObjectClassificationType(str, Enum):
@@ -69,7 +71,7 @@ class BirdClassificationConfig(FrigateBaseModel):
 
 
 class CustomClassificationStateCameraConfig(FrigateBaseModel):
-    crop: list[int, int, int, int] = Field(
+    crop: list[float, float, float, float] = Field(
         title="Crop of image frame on this camera to run classification on."
     )
 
@@ -197,7 +199,9 @@ class FaceRecognitionConfig(FrigateBaseModel):
         title="Min face recognitions for the sub label to be applied to the person object.",
     )
     save_attempts: int = Field(
-        default=100, ge=0, title="Number of face attempts to save in the train tab."
+        default=200,
+        ge=0,
+        title="Number of face attempts to save in the recent recognitions tab.",
     )
     blur_confidence_filter: bool = Field(
         default=True, title="Apply blur quality filter to face confidence."

frigate/config/config.py

@@ -792,6 +792,10 @@ class FrigateConfig(FrigateBaseModel):
             # copy over auth and proxy config in case auth needs to be enforced
             safe_config["auth"] = config.get("auth", {})
             safe_config["proxy"] = config.get("proxy", {})
+
+            # copy over database config for auth and so a new db is not created
+            safe_config["database"] = config.get("database", {})
+
             return cls.parse_object(safe_config, **context)
 
         # Validate and return the config dict.

frigate/data_processing/common/audio_transcription/model.py

@@ -4,7 +4,6 @@ import logging
 import os
 
 import sherpa_onnx
-from faster_whisper.utils import download_model
 
 from frigate.comms.inter_process import InterProcessRequestor
 from frigate.const import MODEL_CACHE_DIR
@@ -25,6 +24,9 @@ class AudioTranscriptionModelRunner:
 
         if model_size == "large":
             # use the Whisper download function instead of our own
+            # Import dynamically to avoid crashes on systems without AVX support
+            from faster_whisper.utils import download_model
+
             logger.debug("Downloading Whisper audio transcription model")
             download_model(
                 size_or_id="small" if device == "cuda" else "tiny",

frigate/data_processing/common/license_plate/mixin.py

@@ -14,8 +14,8 @@ from typing import Any, List, Optional, Tuple
 
 import cv2
 import numpy as np
-from Levenshtein import distance, jaro_winkler
 from pyclipper import ET_CLOSEDPOLYGON, JT_ROUND, PyclipperOffset
+from rapidfuzz.distance import JaroWinkler, Levenshtein
 from shapely.geometry import Polygon
 
 from frigate.comms.event_metadata_updater import (
@@ -1123,7 +1123,9 @@ class LicensePlateProcessingMixin:
         for i, plate in enumerate(plates):
             merged = False
             for j, cluster in enumerate(clusters):
-                sims = [jaro_winkler(plate["plate"], v["plate"]) for v in cluster]
+                sims = [
+                    JaroWinkler.similarity(plate["plate"], v["plate"]) for v in cluster
+                ]
                 if len(sims) > 0:
                     avg_sim = sum(sims) / len(sims)
                     if avg_sim >= self.cluster_threshold:
@@ -1500,7 +1502,7 @@ class LicensePlateProcessingMixin:
                     and current_time - data["last_seen"]
                     <= self.config.cameras[camera].lpr.expire_time
                 ):
-                    similarity = jaro_winkler(data["plate"], top_plate)
+                    similarity = JaroWinkler.similarity(data["plate"], top_plate)
                     if similarity >= self.similarity_threshold:
                         plate_id = existing_id
                         logger.debug(
@@ -1580,7 +1582,8 @@ class LicensePlateProcessingMixin:
                     for label, plates_list in self.lpr_config.known_plates.items()
                     if any(
                         re.match(f"^{plate}$", rep_plate)
-                        or distance(plate, rep_plate) <= self.lpr_config.match_distance
+                        or Levenshtein.distance(plate, rep_plate)
+                        <= self.lpr_config.match_distance
                         for plate in plates_list
                     )
                 ),

frigate/data_processing/post/audio_transcription.py

@@ -6,10 +6,8 @@ import threading
 import time
 from typing import Optional
 
-from faster_whisper import WhisperModel
 from peewee import DoesNotExist
 
-from frigate.comms.embeddings_updater import EmbeddingsRequestEnum
 from frigate.comms.inter_process import InterProcessRequestor
 from frigate.config import FrigateConfig
 from frigate.const import (
@@ -32,11 +30,13 @@ class AudioTranscriptionPostProcessor(PostProcessorApi):
         self,
         config: FrigateConfig,
         requestor: InterProcessRequestor,
+        embeddings,
         metrics: DataProcessorMetrics,
     ):
         super().__init__(config, metrics, None)
         self.config = config
         self.requestor = requestor
+        self.embeddings = embeddings
         self.recognizer = None
         self.transcription_lock = threading.Lock()
         self.transcription_thread = None
@@ -50,6 +50,9 @@ class AudioTranscriptionPostProcessor(PostProcessorApi):
 
     def __build_recognizer(self) -> None:
         try:
+            # Import dynamically to avoid crashes on systems without AVX support
+            from faster_whisper import WhisperModel
+
             self.recognizer = WhisperModel(
                 model_size_or_path="small",
                 device="cuda"
@@ -128,10 +131,7 @@ class AudioTranscriptionPostProcessor(PostProcessorApi):
             )
 
             # Embed the description
-            self.requestor.send_data(
+            self.embeddings.embed_description(event_id, transcription)
-                EmbeddingsRequestEnum.embed_description.value,
-                {"id": event_id, "description": transcription},
-            )
 
         except DoesNotExist:
             logger.debug("No recording found for audio transcription post-processing")

frigate/data_processing/post/object_descriptions.py

@@ -20,8 +20,8 @@ from frigate.genai import GenAIClient
 from frigate.models import Event
 from frigate.types import TrackedObjectUpdateTypesEnum
 from frigate.util.builtin import EventsPerSecond, InferenceSpeed
+from frigate.util.file import get_event_thumbnail_bytes
 from frigate.util.image import create_thumbnail, ensure_jpeg_bytes
-from frigate.util.path import get_event_thumbnail_bytes
 
 if TYPE_CHECKING:
     from frigate.embeddings import Embeddings

frigate/data_processing/post/review_descriptions.py

@@ -3,6 +3,7 @@
 import copy
 import datetime
 import logging
+import math
 import os
 import shutil
 import threading
@@ -10,22 +11,28 @@ from pathlib import Path
 from typing import Any
 
 import cv2
+from peewee import DoesNotExist
 
 from frigate.comms.embeddings_updater import EmbeddingsRequestEnum
 from frigate.comms.inter_process import InterProcessRequestor
 from frigate.config import FrigateConfig
-from frigate.config.camera.review import GenAIReviewConfig
+from frigate.config.camera import CameraConfig
+from frigate.config.camera.review import GenAIReviewConfig, ImageSourceEnum
 from frigate.const import CACHE_DIR, CLIPS_DIR, UPDATE_REVIEW_DESCRIPTION
 from frigate.data_processing.types import PostProcessDataEnum
 from frigate.genai import GenAIClient
-from frigate.models import ReviewSegment
+from frigate.models import Recordings, ReviewSegment
 from frigate.util.builtin import EventsPerSecond, InferenceSpeed
+from frigate.util.image import get_image_from_recording
 
 from ..post.api import PostProcessorApi
 from ..types import DataProcessorMetrics
 
 logger = logging.getLogger(__name__)
 
+RECORDING_BUFFER_EXTENSION_PERCENT = 0.10
+MIN_RECORDING_DURATION = 10
+
 
 class ReviewDescriptionProcessor(PostProcessorApi):
     def __init__(
@@ -43,20 +50,53 @@ class ReviewDescriptionProcessor(PostProcessorApi):
         self.review_descs_dps = EventsPerSecond()
         self.review_descs_dps.start()
 
-    def calculate_frame_count(self) -> int:
+    def calculate_frame_count(
-        """Calculate optimal number of frames based on context size."""
+        self,
-        # With our preview images (height of 180px) each image should be ~100 tokens per image
+        camera: str,
-        # We want to be conservative to not have too long of query times with too many images
+        image_source: ImageSourceEnum = ImageSourceEnum.preview,
-        context_size = self.genai_client.get_context_size()
+        height: int = 480,
+    ) -> int:
+        """Calculate optimal number of frames based on context size, image source, and resolution.
 
-        if context_size > 10000:
+        Token usage varies by resolution: larger images (ultrawide aspect ratios) use more tokens.
-            return 20
+        Estimates ~1 token per 1250 pixels. Targets 98% context utilization with safety margin.
-        elif context_size > 6000:
+        Capped at 20 frames.
-            return 16
+        """
-        elif context_size > 4000:
+        context_size = self.genai_client.get_context_size()
-            return 12
+        camera_config = self.config.cameras[camera]
+
+        detect_width = camera_config.detect.width
+        detect_height = camera_config.detect.height
+        aspect_ratio = detect_width / detect_height
+
+        if image_source == ImageSourceEnum.recordings:
+            if aspect_ratio >= 1:
+                # Landscape or square: constrain height
+                width = int(height * aspect_ratio)
+            else:
+                # Portrait: constrain width
+                width = height
+                height = int(width / aspect_ratio)
         else:
-            return 8
+            if aspect_ratio >= 1:
+                # Landscape or square: constrain height
+                target_height = 180
+                width = int(target_height * aspect_ratio)
+                height = target_height
+            else:
+                # Portrait: constrain width
+                target_width = 180
+                width = target_width
+                height = int(target_width / aspect_ratio)
+
+        pixels_per_image = width * height
+        tokens_per_image = pixels_per_image / 1250
+        prompt_tokens = 3500
+        response_tokens = 300
+        available_tokens = context_size - prompt_tokens - response_tokens
+        max_frames = int(available_tokens / tokens_per_image)
+
+        return min(max(max_frames, 3), 20)
 
     def process_data(self, data, data_type):
         self.metrics.review_desc_dps.value = self.review_descs_dps.eps()
@@ -88,36 +128,61 @@ class ReviewDescriptionProcessor(PostProcessorApi):
             ):
                 return
 
-            frames = self.get_cache_frames(
+            image_source = camera_config.review.genai.image_source
-                camera, final_data["start_time"], final_data["end_time"]
-            )
 
-            if not frames:
+            if image_source == ImageSourceEnum.recordings:
-                frames = [final_data["thumb_path"]]
+                duration = final_data["end_time"] - final_data["start_time"]
+                buffer_extension = min(5, duration * RECORDING_BUFFER_EXTENSION_PERCENT)
 
-            thumbs = []
+                # Ensure minimum total duration for short review items
+                # This provides better context for brief events
+                total_duration = duration + (2 * buffer_extension)
+                if total_duration < MIN_RECORDING_DURATION:
+                    # Expand buffer to reach minimum duration, still respecting max of 5s per side
+                    additional_buffer_per_side = (MIN_RECORDING_DURATION - duration) / 2
+                    buffer_extension = min(5, additional_buffer_per_side)
 
-            for idx, thumb_path in enumerate(frames):
+                thumbs = self.get_recording_frames(
-                thumb_data = cv2.imread(thumb_path)
+                    camera,
-                ret, jpg = cv2.imencode(
+                    final_data["start_time"] - buffer_extension,
-                    ".jpg", thumb_data, [int(cv2.IMWRITE_JPEG_QUALITY), 100]
+                    final_data["end_time"] + buffer_extension,
+                    height=480,  # Use 480p for good balance between quality and token usage
                 )
 
-                if ret:
+                if not thumbs:
-                    thumbs.append(jpg.tobytes())
+                    # Fallback to preview frames if no recordings available
-
+                    logger.warning(
-                if camera_config.review.genai.debug_save_thumbnails:
+                        f"No recording frames found for {camera}, falling back to preview frames"
-                    id = data["after"]["id"]
+                    )
-                    Path(os.path.join(CLIPS_DIR, "genai-requests", f"{id}")).mkdir(
+                    thumbs = self.get_preview_frames_as_bytes(
+                        camera,
+                        final_data["start_time"],
+                        final_data["end_time"],
+                        final_data["thumb_path"],
+                        id,
+                        camera_config.review.genai.debug_save_thumbnails,
+                    )
+                elif camera_config.review.genai.debug_save_thumbnails:
+                    # Save debug thumbnails for recordings
+                    Path(os.path.join(CLIPS_DIR, "genai-requests", id)).mkdir(
                         parents=True, exist_ok=True
                     )
-                    shutil.copy(
+                    for idx, frame_bytes in enumerate(thumbs):
-                        thumb_path,
+                        with open(
-                        os.path.join(
+                            os.path.join(CLIPS_DIR, f"genai-requests/{id}/{idx}.jpg"),
-                            CLIPS_DIR,
+                            "wb",
-                            f"genai-requests/{id}/{idx}.webp",
+                        ) as f:
-                        ),
+                            f.write(frame_bytes)
-                    )
+            else:
+                # Use preview frames
+                thumbs = self.get_preview_frames_as_bytes(
+                    camera,
+                    final_data["start_time"],
+                    final_data["end_time"],
+                    final_data["thumb_path"],
+                    id,
+                    camera_config.review.genai.debug_save_thumbnails,
+                )
 
             # kickoff analysis
             self.review_descs_dps.update()
@@ -127,11 +192,12 @@ class ReviewDescriptionProcessor(PostProcessorApi):
                     self.requestor,
                     self.genai_client,
                     self.review_desc_speed,
-                    camera,
+                    camera_config,
                     final_data,
                     thumbs,
                     camera_config.review.genai,
                     list(self.config.model.merged_labelmap.values()),
+                    self.config.model.all_attributes,
                 ),
             ).start()
 
@@ -217,7 +283,7 @@ class ReviewDescriptionProcessor(PostProcessorApi):
             all_frames.append(os.path.join(preview_dir, file))
 
         frame_count = len(all_frames)
-        desired_frame_count = self.calculate_frame_count()
+        desired_frame_count = self.calculate_frame_count(camera)
 
         if frame_count <= desired_frame_count:
             return all_frames
@@ -231,48 +297,179 @@ class ReviewDescriptionProcessor(PostProcessorApi):
 
         return selected_frames
 
+    def get_recording_frames(
+        self,
+        camera: str,
+        start_time: float,
+        end_time: float,
+        height: int = 480,
+    ) -> list[bytes]:
+        """Get frames from recordings at specified timestamps."""
+        duration = end_time - start_time
+        desired_frame_count = self.calculate_frame_count(
+            camera, ImageSourceEnum.recordings, height
+        )
+
+        # Calculate evenly spaced timestamps throughout the duration
+        if desired_frame_count == 1:
+            timestamps = [start_time + duration / 2]
+        else:
+            step = duration / (desired_frame_count - 1)
+            timestamps = [start_time + (i * step) for i in range(desired_frame_count)]
+
+        def extract_frame_from_recording(ts: float) -> bytes | None:
+            """Extract a single frame from recording at given timestamp."""
+            try:
+                recording = (
+                    Recordings.select(
+                        Recordings.path,
+                        Recordings.start_time,
+                    )
+                    .where((ts >= Recordings.start_time) & (ts <= Recordings.end_time))
+                    .where(Recordings.camera == camera)
+                    .order_by(Recordings.start_time.desc())
+                    .limit(1)
+                    .get()
+                )
+
+                time_in_segment = ts - recording.start_time
+                return get_image_from_recording(
+                    self.config.ffmpeg,
+                    recording.path,
+                    time_in_segment,
+                    "mjpeg",
+                    height=height,
+                )
+            except DoesNotExist:
+                return None
+
+        frames = []
+
+        for timestamp in timestamps:
+            try:
+                # Try to extract frame at exact timestamp
+                image_data = extract_frame_from_recording(timestamp)
+
+                if not image_data:
+                    # Try with rounded timestamp as fallback
+                    rounded_timestamp = math.ceil(timestamp)
+                    image_data = extract_frame_from_recording(rounded_timestamp)
+
+                if image_data:
+                    frames.append(image_data)
+                else:
+                    logger.warning(
+                        f"No recording found for {camera} at timestamp {timestamp}"
+                    )
+            except Exception as e:
+                logger.error(
+                    f"Error extracting frame from recording for {camera} at {timestamp}: {e}"
+                )
+                continue
+
+        return frames
+
+    def get_preview_frames_as_bytes(
+        self,
+        camera: str,
+        start_time: float,
+        end_time: float,
+        thumb_path_fallback: str,
+        review_id: str,
+        save_debug: bool,
+    ) -> list[bytes]:
+        """Get preview frames and convert them to JPEG bytes.
+
+        Args:
+            camera: Camera name
+            start_time: Start timestamp
+            end_time: End timestamp
+            thumb_path_fallback: Fallback thumbnail path if no preview frames found
+            review_id: Review item ID for debug saving
+            save_debug: Whether to save debug thumbnails
+
+        Returns:
+            List of JPEG image bytes
+        """
+        frame_paths = self.get_cache_frames(camera, start_time, end_time)
+        if not frame_paths:
+            frame_paths = [thumb_path_fallback]
+
+        thumbs = []
+        for idx, thumb_path in enumerate(frame_paths):
+            thumb_data = cv2.imread(thumb_path)
+            ret, jpg = cv2.imencode(
+                ".jpg", thumb_data, [int(cv2.IMWRITE_JPEG_QUALITY), 100]
+            )
+            if ret:
+                thumbs.append(jpg.tobytes())
+
+            if save_debug:
+                Path(os.path.join(CLIPS_DIR, "genai-requests", review_id)).mkdir(
+                    parents=True, exist_ok=True
+                )
+                shutil.copy(
+                    thumb_path,
+                    os.path.join(CLIPS_DIR, f"genai-requests/{review_id}/{idx}.webp"),
+                )
+
+        return thumbs
+
 
 @staticmethod
 def run_analysis(
     requestor: InterProcessRequestor,
     genai_client: GenAIClient,
     review_inference_speed: InferenceSpeed,
-    camera: str,
+    camera_config: CameraConfig,
     final_data: dict[str, str],
     thumbs: list[bytes],
     genai_config: GenAIReviewConfig,
     labelmap_objects: list[str],
+    attribute_labels: list[str],
 ) -> None:
     start = datetime.datetime.now().timestamp()
+
+    # Format zone names using zone config friendly names if available
+    formatted_zones = []
+    for zone_name in final_data["data"]["zones"]:
+        if zone_name in camera_config.zones:
+            formatted_zones.append(
+                camera_config.zones[zone_name].get_formatted_name(zone_name)
+            )
+
     analytics_data = {
         "id": final_data["id"],
-        "camera": camera,
+        "camera": camera_config.get_formatted_name(),
-        "zones": final_data["data"]["zones"],
+        "zones": formatted_zones,
         "start": datetime.datetime.fromtimestamp(final_data["start_time"]).strftime(
             "%A, %I:%M %p"
         ),
         "duration": round(final_data["end_time"] - final_data["start_time"]),
     }
 
-    objects = []
+    unified_objects = []
-    named_objects = []
 
     objects_list = final_data["data"]["objects"]
     sub_labels_list = final_data["data"]["sub_labels"]
 
+    for i, verified_label in enumerate(final_data["data"]["verified_objects"]):
+        object_type = verified_label.replace("-verified", "").replace("_", " ")
+        name = sub_labels_list[i].replace("_", " ").title()
+        unified_objects.append(f"{name} ({object_type})")
+
     for label in objects_list:
         if "-verified" in label:
             continue
         elif label in labelmap_objects:
-            objects.append(label.replace("_", " ").title())
+            object_type = label.replace("_", " ").title()
 
-    for i, verified_label in enumerate(final_data["data"]["verified_objects"]):
+            if label in attribute_labels:
-        named_objects.append(
+                unified_objects.append(f"{object_type} (delivery/service)")
-            f"{sub_labels_list[i].replace('_', ' ').title()} ({verified_label.replace('-verified', '')})"
+            else:
-        )
+                unified_objects.append(object_type)
 
-    analytics_data["objects"] = objects
+    analytics_data["unified_objects"] = unified_objects
-    analytics_data["recognized_objects"] = named_objects
 
     metadata = genai_client.generate_review_description(
         analytics_data,

frigate/data_processing/post/semantic_trigger.py

@@ -10,6 +10,10 @@ import cv2
 import numpy as np
 from peewee import DoesNotExist
 
+from frigate.comms.event_metadata_updater import (
+    EventMetadataPublisher,
+    EventMetadataTypeEnum,
+)
 from frigate.comms.inter_process import InterProcessRequestor
 from frigate.config import FrigateConfig
 from frigate.const import CONFIG_DIR
@@ -18,7 +22,7 @@ from frigate.db.sqlitevecq import SqliteVecQueueDatabase
 from frigate.embeddings.util import ZScoreNormalization
 from frigate.models import Event, Trigger
 from frigate.util.builtin import cosine_distance
-from frigate.util.path import get_event_thumbnail_bytes
+from frigate.util.file import get_event_thumbnail_bytes
 
 from ..post.api import PostProcessorApi
 from ..types import DataProcessorMetrics
@@ -34,6 +38,7 @@ class SemanticTriggerProcessor(PostProcessorApi):
         db: SqliteVecQueueDatabase,
         config: FrigateConfig,
         requestor: InterProcessRequestor,
+        sub_label_publisher: EventMetadataPublisher,
         metrics: DataProcessorMetrics,
         embeddings,
     ):
@@ -41,6 +46,7 @@ class SemanticTriggerProcessor(PostProcessorApi):
         self.db = db
         self.embeddings = embeddings
         self.requestor = requestor
+        self.sub_label_publisher = sub_label_publisher
         self.trigger_embeddings: list[np.ndarray] = []
 
         self.thumb_stats = ZScoreNormalization()
@@ -184,14 +190,44 @@ class SemanticTriggerProcessor(PostProcessorApi):
                     ),
                 )
 
+                friendly_name = (
+                    self.config.cameras[camera]
+                    .semantic_search.triggers[trigger["name"]]
+                    .friendly_name
+                )
+
                 if (
                     self.config.cameras[camera]
                     .semantic_search.triggers[trigger["name"]]
                     .actions
                 ):
-                    # TODO: handle actions for the trigger
+                    # handle actions for the trigger
                     # notifications already handled by webpush
-                    pass
+                    if (
+                        "sub_label"
+                        in self.config.cameras[camera]
+                        .semantic_search.triggers[trigger["name"]]
+                        .actions
+                    ):
+                        self.sub_label_publisher.publish(
+                            (event_id, friendly_name, similarity),
+                            EventMetadataTypeEnum.sub_label,
+                        )
+                    if (
+                        "attribute"
+                        in self.config.cameras[camera]
+                        .semantic_search.triggers[trigger["name"]]
+                        .actions
+                    ):
+                        self.sub_label_publisher.publish(
+                            (
+                                event_id,
+                                trigger["name"],
+                                trigger["type"],
+                                similarity,
+                            ),
+                            EventMetadataTypeEnum.attribute.value,
+                        )
 
             if WRITE_DEBUG_IMAGES:
                 try:

frigate/data_processing/real_time/custom_classification.py

@@ -1,6 +1,7 @@
 """Real time processor that works with classification tflite models."""
 
 import datetime
+import json
 import logging
 import os
 from typing import Any
@@ -21,6 +22,7 @@ from frigate.config.classification import (
 )
 from frigate.const import CLIPS_DIR, MODEL_CACHE_DIR
 from frigate.log import redirect_output_to_logger
+from frigate.types import TrackedObjectUpdateTypesEnum
 from frigate.util.builtin import EventsPerSecond, InferenceSpeed, load_labels
 from frigate.util.object import box_overlaps, calculate_region
 
@@ -34,6 +36,8 @@ except ModuleNotFoundError:
 
 logger = logging.getLogger(__name__)
 
+MAX_OBJECT_CLASSIFICATIONS = 16
+
 
 class CustomStateClassificationProcessor(RealTimeProcessorApi):
     def __init__(
@@ -53,9 +57,18 @@ class CustomStateClassificationProcessor(RealTimeProcessorApi):
         self.tensor_output_details: dict[str, Any] | None = None
         self.labelmap: dict[int, str] = {}
         self.classifications_per_second = EventsPerSecond()
-        self.inference_speed = InferenceSpeed(
+        self.state_history: dict[str, dict[str, Any]] = {}
-            self.metrics.classification_speeds[self.model_config.name]
+
-        )
+        if (
+            self.metrics
+            and self.model_config.name in self.metrics.classification_speeds
+        ):
+            self.inference_speed = InferenceSpeed(
+                self.metrics.classification_speeds[self.model_config.name]
+            )
+        else:
+            self.inference_speed = None
+
         self.last_run = datetime.datetime.now().timestamp()
         self.__build_detector()
 
@@ -83,12 +96,50 @@ class CustomStateClassificationProcessor(RealTimeProcessorApi):
 
     def __update_metrics(self, duration: float) -> None:
         self.classifications_per_second.update()
-        self.inference_speed.update(duration)
+        if self.inference_speed:
+            self.inference_speed.update(duration)
+
+    def verify_state_change(self, camera: str, detected_state: str) -> str | None:
+        """
+        Verify state change requires 3 consecutive identical states before publishing.
+        Returns state to publish or None if verification not complete.
+        """
+        if camera not in self.state_history:
+            self.state_history[camera] = {
+                "current_state": None,
+                "pending_state": None,
+                "consecutive_count": 0,
+            }
+
+        verification = self.state_history[camera]
+
+        if detected_state == verification["current_state"]:
+            verification["pending_state"] = None
+            verification["consecutive_count"] = 0
+            return None
+
+        if detected_state == verification["pending_state"]:
+            verification["consecutive_count"] += 1
+
+            if verification["consecutive_count"] >= 3:
+                verification["current_state"] = detected_state
+                verification["pending_state"] = None
+                verification["consecutive_count"] = 0
+                return detected_state
+        else:
+            verification["pending_state"] = detected_state
+            verification["consecutive_count"] = 1
+            logger.debug(
+                f"New state '{detected_state}' detected for {camera}, need {3 - verification['consecutive_count']} more consecutive detections"
+            )
+
+        return None
 
     def process_frame(self, frame_data: dict[str, Any], frame: np.ndarray):
-        self.metrics.classification_cps[
+        if self.metrics and self.model_config.name in self.metrics.classification_cps:
-            self.model_config.name
+            self.metrics.classification_cps[
-        ].value = self.classifications_per_second.eps()
+                self.model_config.name
+            ].value = self.classifications_per_second.eps()
         camera = frame_data.get("camera")
 
         if camera not in self.model_config.state_config.cameras:
@@ -96,10 +147,10 @@ class CustomStateClassificationProcessor(RealTimeProcessorApi):
 
         camera_config = self.model_config.state_config.cameras[camera]
         crop = [
-            camera_config.crop[0],
+            camera_config.crop[0] * self.config.cameras[camera].detect.width,
-            camera_config.crop[1],
+            camera_config.crop[1] * self.config.cameras[camera].detect.height,
-            camera_config.crop[2],
+            camera_config.crop[2] * self.config.cameras[camera].detect.width,
-            camera_config.crop[3],
+            camera_config.crop[3] * self.config.cameras[camera].detect.height,
         ]
         should_run = False
 
@@ -121,6 +172,19 @@ class CustomStateClassificationProcessor(RealTimeProcessorApi):
                 self.last_run = now
                 should_run = True
 
+        # Shortcut: always run if we have a pending state verification to complete
+        if (
+            not should_run
+            and camera in self.state_history
+            and self.state_history[camera]["pending_state"] is not None
+            and now > self.last_run + 0.5
+        ):
+            self.last_run = now
+            should_run = True
+            logger.debug(
+                f"Running verification check for pending state: {self.state_history[camera]['pending_state']} ({self.state_history[camera]['consecutive_count']}/3)"
+            )
+
         if not should_run:
             return
 
@@ -165,6 +229,9 @@ class CustomStateClassificationProcessor(RealTimeProcessorApi):
             self.tensor_output_details[0]["index"]
         )[0]
         probs = res / res.sum(axis=0)
+        logger.debug(
+            f"{self.model_config.name} Ran state classification with probabilities: {probs}"
+        )
         best_id = np.argmax(probs)
         score = round(probs[best_id], 2)
         self.__update_metrics(datetime.datetime.now().timestamp() - now)
@@ -178,10 +245,19 @@ class CustomStateClassificationProcessor(RealTimeProcessorApi):
             score,
         )
 
-        if score >= self.model_config.threshold:
+        if score < self.model_config.threshold:
+            logger.debug(
+                f"Score {score} below threshold {self.model_config.threshold}, skipping verification"
+            )
+            return
+
+        detected_state = self.labelmap[best_id]
+        verified_state = self.verify_state_change(camera, detected_state)
+
+        if verified_state is not None:
             self.requestor.send_data(
                 f"{camera}/classification/{self.model_config.name}",
-                self.labelmap[best_id],
+                verified_state,
             )
 
     def handle_request(self, topic, request_data):
@@ -210,6 +286,7 @@ class CustomObjectClassificationProcessor(RealTimeProcessorApi):
         config: FrigateConfig,
         model_config: CustomClassificationConfig,
         sub_label_publisher: EventMetadataPublisher,
+        requestor: InterProcessRequestor,
         metrics: DataProcessorMetrics,
     ):
         super().__init__(config, metrics)
@@ -218,14 +295,23 @@ class CustomObjectClassificationProcessor(RealTimeProcessorApi):
         self.train_dir = os.path.join(CLIPS_DIR, self.model_config.name, "train")
         self.interpreter: Interpreter | None = None
         self.sub_label_publisher = sub_label_publisher
+        self.requestor = requestor
         self.tensor_input_details: dict[str, Any] | None = None
         self.tensor_output_details: dict[str, Any] | None = None
-        self.detected_objects: dict[str, float] = {}
+        self.classification_history: dict[str, list[tuple[str, float, float]]] = {}
         self.labelmap: dict[int, str] = {}
         self.classifications_per_second = EventsPerSecond()
-        self.inference_speed = InferenceSpeed(
+
-            self.metrics.classification_speeds[self.model_config.name]
+        if (
-        )
+            self.metrics
+            and self.model_config.name in self.metrics.classification_speeds
+        ):
+            self.inference_speed = InferenceSpeed(
+                self.metrics.classification_speeds[self.model_config.name]
+            )
+        else:
+            self.inference_speed = None
+
         self.__build_detector()
 
     @redirect_output_to_logger(logger, logging.DEBUG)
@@ -251,12 +337,64 @@ class CustomObjectClassificationProcessor(RealTimeProcessorApi):
 
     def __update_metrics(self, duration: float) -> None:
         self.classifications_per_second.update()
-        self.inference_speed.update(duration)
+        if self.inference_speed:
+            self.inference_speed.update(duration)
+
+    def get_weighted_score(
+        self,
+        object_id: str,
+        current_label: str,
+        current_score: float,
+        current_time: float,
+    ) -> tuple[str | None, float]:
+        """
+        Determine weighted score based on history to prevent false positives/negatives.
+        Requires 60% of attempts to agree on a label before publishing.
+        Returns (weighted_label, weighted_score) or (None, 0.0) if no weighted score.
+        """
+        if object_id not in self.classification_history:
+            self.classification_history[object_id] = []
+
+        self.classification_history[object_id].append(
+            (current_label, current_score, current_time)
+        )
+
+        history = self.classification_history[object_id]
+
+        if len(history) < 3:
+            return None, 0.0
+
+        label_counts = {}
+        label_scores = {}
+        total_attempts = len(history)
+
+        for label, score, timestamp in history:
+            if label not in label_counts:
+                label_counts[label] = 0
+                label_scores[label] = []
+
+            label_counts[label] += 1
+            label_scores[label].append(score)
+
+        best_label = max(label_counts, key=label_counts.get)
+        best_count = label_counts[best_label]
+
+        consensus_threshold = total_attempts * 0.6
+        if best_count < consensus_threshold:
+            return None, 0.0
+
+        avg_score = sum(label_scores[best_label]) / len(label_scores[best_label])
+
+        if best_label == "none":
+            return None, 0.0
+
+        return best_label, avg_score
 
     def process_frame(self, obj_data, frame):
-        self.metrics.classification_cps[
+        if self.metrics and self.model_config.name in self.metrics.classification_cps:
-            self.model_config.name
+            self.metrics.classification_cps[
-        ].value = self.classifications_per_second.eps()
+                self.model_config.name
+            ].value = self.classifications_per_second.eps()
 
         if obj_data["false_positive"]:
             return
@@ -264,6 +402,21 @@ class CustomObjectClassificationProcessor(RealTimeProcessorApi):
         if obj_data["label"] not in self.model_config.object_config.objects:
             return
 
+        if obj_data.get("end_time") is not None:
+            return
+
+        if obj_data.get("stationary"):
+            return
+
+        object_id = obj_data["id"]
+
+        if (
+            object_id in self.classification_history
+            and len(self.classification_history[object_id])
+            >= MAX_OBJECT_CLASSIFICATIONS
+        ):
+            return
+
         now = datetime.datetime.now().timestamp()
         x, y, x2, y2 = calculate_region(
             frame.shape,
@@ -272,8 +425,8 @@ class CustomObjectClassificationProcessor(RealTimeProcessorApi):
             obj_data["box"][2],
             obj_data["box"][3],
             max(
-                obj_data["box"][1] - obj_data["box"][0],
+                obj_data["box"][2] - obj_data["box"][0],
-                obj_data["box"][3] - obj_data["box"][2],
+                obj_data["box"][3] - obj_data["box"][1],
             ),
             1.0,
         )
@@ -295,7 +448,7 @@ class CustomObjectClassificationProcessor(RealTimeProcessorApi):
             write_classification_attempt(
                 self.train_dir,
                 cv2.cvtColor(crop, cv2.COLOR_RGB2BGR),
-                obj_data["id"],
+                object_id,
                 now,
                 "unknown",
                 0.0,
@@ -309,48 +462,85 @@ class CustomObjectClassificationProcessor(RealTimeProcessorApi):
             self.tensor_output_details[0]["index"]
         )[0]
         probs = res / res.sum(axis=0)
+        logger.debug(
+            f"{self.model_config.name} Ran object classification with probabilities: {probs}"
+        )
         best_id = np.argmax(probs)
         score = round(probs[best_id], 2)
-        previous_score = self.detected_objects.get(obj_data["id"], 0.0)
         self.__update_metrics(datetime.datetime.now().timestamp() - now)
 
         write_classification_attempt(
             self.train_dir,
             cv2.cvtColor(crop, cv2.COLOR_RGB2BGR),
-            obj_data["id"],
+            object_id,
             now,
             self.labelmap[best_id],
             score,
+            max_files=200,
         )
 
         if score < self.model_config.threshold:
             logger.debug(f"Score {score} is less than threshold.")
             return
 
-        if score <= previous_score:
-            logger.debug(f"Score {score} is worse than previous score {previous_score}")
-            return
-
         sub_label = self.labelmap[best_id]
-        self.detected_objects[obj_data["id"]] = score
 
-        if (
+        consensus_label, consensus_score = self.get_weighted_score(
-            self.model_config.object_config.classification_type
+            object_id, sub_label, score, now
-            == ObjectClassificationType.sub_label
+        )
-        ):
+
-            if sub_label != "none":
+        if consensus_label is not None:
+            camera = obj_data["camera"]
+
+            if (
+                self.model_config.object_config.classification_type
+                == ObjectClassificationType.sub_label
+            ):
                 self.sub_label_publisher.publish(
-                    (obj_data["id"], sub_label, score),
+                    (object_id, consensus_label, consensus_score),
                     EventMetadataTypeEnum.sub_label,
                 )
-        elif (
+                self.requestor.send_data(
-            self.model_config.object_config.classification_type
+                    "tracked_object_update",
-            == ObjectClassificationType.attribute
+                    json.dumps(
-        ):
+                        {
-            self.sub_label_publisher.publish(
+                            "type": TrackedObjectUpdateTypesEnum.classification,
-                (obj_data["id"], self.model_config.name, sub_label, score),
+                            "id": object_id,
-                EventMetadataTypeEnum.attribute.value,
+                            "camera": camera,
-            )
+                            "timestamp": now,
+                            "model": self.model_config.name,
+                            "sub_label": consensus_label,
+                            "score": consensus_score,
+                        }
+                    ),
+                )
+            elif (
+                self.model_config.object_config.classification_type
+                == ObjectClassificationType.attribute
+            ):
+                self.sub_label_publisher.publish(
+                    (
+                        object_id,
+                        self.model_config.name,
+                        consensus_label,
+                        consensus_score,
+                    ),
+                    EventMetadataTypeEnum.attribute.value,
+                )
+                self.requestor.send_data(
+                    "tracked_object_update",
+                    json.dumps(
+                        {
+                            "type": TrackedObjectUpdateTypesEnum.classification,
+                            "id": object_id,
+                            "camera": camera,
+                            "timestamp": now,
+                            "model": self.model_config.name,
+                            "attribute": consensus_label,
+                            "score": consensus_score,
+                        }
+                    ),
+                )
 
     def handle_request(self, topic, request_data):
         if topic == EmbeddingsRequestEnum.reload_classification_model.value:
@@ -368,8 +558,8 @@ class CustomObjectClassificationProcessor(RealTimeProcessorApi):
             return None
 
     def expire_object(self, object_id, camera):
-        if object_id in self.detected_objects:
+        if object_id in self.classification_history:
-            self.detected_objects.pop(object_id)
+            self.classification_history.pop(object_id)
 
 
 @staticmethod
@@ -380,6 +570,7 @@ def write_classification_attempt(
     timestamp: float,
     label: str,
     score: float,
+    max_files: int = 100,
 ) -> None:
     if "-" in label:
         label = label.replace("-", "_")
@@ -395,5 +586,8 @@ def write_classification_attempt(
     )
 
     # delete oldest face image if maximum is reached
-    if len(files) > 100:
+    try:
-        os.unlink(os.path.join(folder, files[-1]))
+        if len(files) > max_files:
+            os.unlink(os.path.join(folder, files[-1]))
+    except FileNotFoundError:
+        pass

frigate/data_processing/real_time/face.py

@@ -166,6 +166,7 @@ class FaceRealTimeProcessor(RealTimeProcessorApi):
         camera = obj_data["camera"]
 
         if not self.config.cameras[camera].face_recognition.enabled:
+            logger.debug(f"Face recognition disabled for camera {camera}, skipping")
             return
 
         start = datetime.datetime.now().timestamp()
@@ -208,6 +209,7 @@ class FaceRealTimeProcessor(RealTimeProcessorApi):
             person_box = obj_data.get("box")
 
             if not person_box:
+                logger.debug(f"No person box available for {id}")
                 return
 
             rgb = cv2.cvtColor(frame, cv2.COLOR_YUV2RGB_I420)
@@ -233,7 +235,8 @@ class FaceRealTimeProcessor(RealTimeProcessorApi):
 
             try:
                 face_frame = cv2.cvtColor(face_frame, cv2.COLOR_RGB2BGR)
-            except Exception:
+            except Exception as e:
+                logger.debug(f"Failed to convert face frame color for {id}: {e}")
                 return
         else:
             # don't run for object without attributes
@@ -251,6 +254,7 @@ class FaceRealTimeProcessor(RealTimeProcessorApi):
 
             # no faces detected in this frame
             if not face:
+                logger.debug(f"No face attributes found for {id}")
                 return
 
             face_box = face.get("box")
@@ -274,6 +278,7 @@ class FaceRealTimeProcessor(RealTimeProcessorApi):
         res = self.recognizer.classify(face_frame)
 
         if not res:
+            logger.debug(f"Face recognizer returned no result for {id}")
             self.__update_metrics(datetime.datetime.now().timestamp() - start)
             return
 
@@ -330,6 +335,7 @@ class FaceRealTimeProcessor(RealTimeProcessorApi):
     def handle_request(self, topic, request_data) -> dict[str, Any] | None:
         if topic == EmbeddingsRequestEnum.clear_face_classifier.value:
             self.recognizer.clear()
+            return {"success": True, "message": "Face classifier cleared."}
         elif topic == EmbeddingsRequestEnum.recognize_face.value:
             img = cv2.imdecode(
                 np.frombuffer(base64.b64decode(request_data["image"]), dtype=np.uint8),
@@ -417,7 +423,10 @@ class FaceRealTimeProcessor(RealTimeProcessorApi):
             res = self.recognizer.classify(img)
 
             if not res:
-                return
+                return {
+                    "message": "Model is still training, please try again in a few moments.",
+                    "success": False,
+                }
 
             sub_label, score = res
 
@@ -436,6 +445,13 @@ class FaceRealTimeProcessor(RealTimeProcessorApi):
                 )
                 shutil.move(current_file, new_file)
 
+            return {
+                "message": f"Successfully reprocessed face. Result: {sub_label} (score: {score:.2f})",
+                "success": True,
+                "face_name": sub_label,
+                "score": score,
+            }
+
     def expire_object(self, object_id: str, camera: str):
         if object_id in self.person_face_history:
             self.person_face_history.pop(object_id)

frigate/detectors/detection_runners.py

@@ -3,6 +3,7 @@
 import logging
 import os
 import platform
+import threading
 from abc import ABC, abstractmethod
 from typing import Any
 
@@ -21,21 +22,25 @@ def is_arm64_platform() -> bool:
     return machine in ("aarch64", "arm64", "armv8", "armv7l")
 
 
-def get_ort_session_options() -> ort.SessionOptions | None:
+def get_ort_session_options(
+    is_complex_model: bool = False,
+) -> ort.SessionOptions | None:
     """Get ONNX Runtime session options with appropriate settings.
 
-    On ARM/RKNN platforms, use basic optimizations to avoid graph fusion issues
+    Args:
-    that can break certain models. On amd64, use default optimizations for better performance.
+        is_complex_model: Whether the model needs basic optimization to avoid graph fusion issues.
-    """
-    sess_options = None
 
-    if is_arm64_platform():
+    Returns:
+        SessionOptions with appropriate optimization level, or None for default settings.
+    """
+    if is_complex_model:
         sess_options = ort.SessionOptions()
         sess_options.graph_optimization_level = (
             ort.GraphOptimizationLevel.ORT_ENABLE_BASIC
         )
+        return sess_options
 
-    return sess_options
+    return None
 
 
 # Import OpenVINO only when needed to avoid circular dependencies
@@ -103,6 +108,21 @@ class BaseModelRunner(ABC):
 class ONNXModelRunner(BaseModelRunner):
     """Run ONNX models using ONNX Runtime."""
 
+    @staticmethod
+    def is_cpu_complex_model(model_type: str) -> bool:
+        """Check if model needs basic optimization level to avoid graph fusion issues.
+
+        Some models (like Jina-CLIP) have issues with aggressive optimizations like
+        SimplifiedLayerNormFusion that create or expect nodes that don't exist.
+        """
+        # Import here to avoid circular imports
+        from frigate.embeddings.types import EnrichmentModelTypeEnum
+
+        return model_type in [
+            EnrichmentModelTypeEnum.jina_v1.value,
+            EnrichmentModelTypeEnum.jina_v2.value,
+        ]
+
     @staticmethod
     def is_migraphx_complex_model(model_type: str) -> bool:
         # Import here to avoid circular imports
@@ -142,12 +162,12 @@ class CudaGraphRunner(BaseModelRunner):
     """
 
     @staticmethod
-    def is_complex_model(model_type: str) -> bool:
+    def is_model_supported(model_type: str) -> bool:
         # Import here to avoid circular imports
         from frigate.detectors.detector_config import ModelTypeEnum
         from frigate.embeddings.types import EnrichmentModelTypeEnum
 
-        return model_type in [
+        return model_type not in [
             ModelTypeEnum.yolonas.value,
             EnrichmentModelTypeEnum.paddleocr.value,
             EnrichmentModelTypeEnum.jina_v1.value,
@@ -215,11 +235,36 @@ class OpenVINOModelRunner(BaseModelRunner):
         # Import here to avoid circular imports
         from frigate.embeddings.types import EnrichmentModelTypeEnum
 
-        return model_type in [EnrichmentModelTypeEnum.paddleocr.value]
+        return model_type in [
+            EnrichmentModelTypeEnum.paddleocr.value,
+            EnrichmentModelTypeEnum.jina_v2.value,
+        ]
+
+    @staticmethod
+    def is_model_npu_supported(model_type: str) -> bool:
+        # Import here to avoid circular imports
+        from frigate.embeddings.types import EnrichmentModelTypeEnum
+
+        return model_type not in [
+            EnrichmentModelTypeEnum.paddleocr.value,
+            EnrichmentModelTypeEnum.jina_v1.value,
+            EnrichmentModelTypeEnum.jina_v2.value,
+            EnrichmentModelTypeEnum.arcface.value,
+        ]
 
     def __init__(self, model_path: str, device: str, model_type: str, **kwargs):
         self.model_path = model_path
         self.device = device
+        self.model_type = model_type
+
+        if device == "NPU" and not OpenVINOModelRunner.is_model_npu_supported(
+            model_type
+        ):
+            logger.warning(
+                f"OpenVINO model {model_type} is not supported on NPU, using GPU instead"
+            )
+            device = "GPU"
+
         self.complex_model = OpenVINOModelRunner.is_complex_model(model_type)
 
         if not os.path.isfile(model_path):
@@ -247,6 +292,10 @@ class OpenVINOModelRunner(BaseModelRunner):
         self.infer_request = self.compiled_model.create_infer_request()
         self.input_tensor: ov.Tensor | None = None
 
+        # Thread lock to prevent concurrent inference (needed for JinaV2 which shares
+        # one runner between text and vision embeddings called from different threads)
+        self._inference_lock = threading.Lock()
+
         if not self.complex_model:
             try:
                 input_shape = self.compiled_model.inputs[0].get_shape()
@@ -290,55 +339,81 @@ class OpenVINOModelRunner(BaseModelRunner):
         Returns:
             List of output tensors
         """
-        # Handle single input case for backward compatibility
+        # Lock prevents concurrent access to infer_request
-        if (
+        # Needed for JinaV2: genai thread (text) + embeddings thread (vision)
-            len(inputs) == 1
+        with self._inference_lock:
-            and len(self.compiled_model.inputs) == 1
+            from frigate.embeddings.types import EnrichmentModelTypeEnum
-            and self.input_tensor is not None
+
-        ):
+            if self.model_type in [EnrichmentModelTypeEnum.arcface.value]:
-            # Single input case - use the pre-allocated tensor for efficiency
+                # For face recognition models, create a fresh infer_request
-            input_data = list(inputs.values())[0]
+                # for each inference to avoid state pollution that causes incorrect results.
-            np.copyto(self.input_tensor.data, input_data)
+                self.infer_request = self.compiled_model.create_infer_request()
-            self.infer_request.infer(self.input_tensor)
+
-        else:
+            # Handle single input case for backward compatibility
-            if self.complex_model:
+            if (
+                len(inputs) == 1
+                and len(self.compiled_model.inputs) == 1
+                and self.input_tensor is not None
+            ):
+                # Single input case - use the pre-allocated tensor for efficiency
+                input_data = list(inputs.values())[0]
+                np.copyto(self.input_tensor.data, input_data)
+                self.infer_request.infer(self.input_tensor)
+            else:
+                if self.complex_model:
+                    try:
+                        # This ensures the model starts with a clean state for each sequence
+                        # Important for RNN models like PaddleOCR recognition
+                        self.infer_request.reset_state()
+                    except Exception:
+                        # this will raise an exception for models with AUTO set as the device
+                        pass
+
+                # Multiple inputs case - set each input by name
+                for input_name, input_data in inputs.items():
+                    # Find the input by name and its index
+                    input_port = None
+                    input_index = None
+                    for idx, port in enumerate(self.compiled_model.inputs):
+                        if port.get_any_name() == input_name:
+                            input_port = port
+                            input_index = idx
+                            break
+
+                    if input_port is None:
+                        raise ValueError(f"Input '{input_name}' not found in model")
+
+                    # Create tensor with the correct element type
+                    input_element_type = input_port.get_element_type()
+
+                    # Ensure input data matches the expected dtype to prevent type mismatches
+                    # that can occur with models like Jina-CLIP v2 running on OpenVINO
+                    expected_dtype = input_element_type.to_dtype()
+                    if input_data.dtype != expected_dtype:
+                        logger.debug(
+                            f"Converting input '{input_name}' from {input_data.dtype} to {expected_dtype}"
+                        )
+                        input_data = input_data.astype(expected_dtype)
+
+                    input_tensor = ov.Tensor(input_element_type, input_data.shape)
+                    np.copyto(input_tensor.data, input_data)
+
+                    # Set the input tensor for the specific port index
+                    self.infer_request.set_input_tensor(input_index, input_tensor)
+
+                # Run inference
                 try:
-                    # This ensures the model starts with a clean state for each sequence
+                    self.infer_request.infer()
-                    # Important for RNN models like PaddleOCR recognition
+                except Exception as e:
-                    self.infer_request.reset_state()
+                    logger.error(f"Error during OpenVINO inference: {e}")
-                except Exception:
+                    return []
-                    # this will raise an exception for models with AUTO set as the device
-                    pass
 
-            # Multiple inputs case - set each input by name
+            # Get all output tensors
-            for input_name, input_data in inputs.items():
+            outputs = []
-                # Find the input by name
+            for i in range(len(self.compiled_model.outputs)):
-                input_port = None
+                outputs.append(self.infer_request.get_output_tensor(i).data)
-                for port in self.compiled_model.inputs:
-                    if port.get_any_name() == input_name:
-                        input_port = port
-                        break
 
-                if input_port is None:
+            return outputs
-                    raise ValueError(f"Input '{input_name}' not found in model")
-
-                # Create tensor with the correct element type
-                input_element_type = input_port.get_element_type()
-                input_tensor = ov.Tensor(input_element_type, input_data.shape)
-                np.copyto(input_tensor.data, input_data)
-
-                # Set the input tensor
-                self.infer_request.set_input_tensor(input_tensor)
-
-            # Run inference
-            self.infer_request.infer()
-
-        # Get all output tensors
-        outputs = []
-        for i in range(len(self.compiled_model.outputs)):
-            outputs.append(self.infer_request.get_output_tensor(i).data)
-
-        return outputs
 
 
 class RKNNModelRunner(BaseModelRunner):
@@ -466,7 +541,7 @@ def get_optimized_runner(
             return OpenVINOModelRunner(model_path, device, model_type, **kwargs)
 
     if (
-        not CudaGraphRunner.is_complex_model(model_type)
+        CudaGraphRunner.is_model_supported(model_type)
         and providers[0] == "CUDAExecutionProvider"
     ):
         options[0] = {
@@ -494,7 +569,9 @@ def get_optimized_runner(
     return ONNXModelRunner(
         ort.InferenceSession(
             model_path,
-            sess_options=get_ort_session_options(),
+            sess_options=get_ort_session_options(
+                ONNXModelRunner.is_cpu_complex_model(model_type)
+            ),
             providers=providers,
             provider_options=options,
         )

frigate/detectors/plugins/memryx.py

@@ -17,7 +17,7 @@ from frigate.detectors.detector_config import (
     BaseDetectorConfig,
     ModelTypeEnum,
 )
-from frigate.util.model import post_process_yolo
+from frigate.util.file import FileLock
 
 logger = logging.getLogger(__name__)
 
@@ -177,44 +177,14 @@ class MemryXDetector(DetectionApi):
             logger.error(f"Failed to initialize MemryX model: {e}")
             raise
 
-    def _acquire_file_lock(self, lock_path: str, timeout: int = 60, poll: float = 0.2):
-        """
-        Create an exclusive lock file. Blocks (with polling) until it can acquire,
-        or raises TimeoutError. Uses only stdlib (os.O_EXCL).
-        """
-        start = time.time()
-        while True:
-            try:
-                fd = os.open(lock_path, os.O_CREAT | os.O_EXCL | os.O_RDWR)
-                os.close(fd)
-                return
-            except FileExistsError:
-                if time.time() - start > timeout:
-                    raise TimeoutError(f"Timeout waiting for lock: {lock_path}")
-                time.sleep(poll)
-
-    def _release_file_lock(self, lock_path: str):
-        """Best-effort removal of the lock file."""
-        try:
-            os.remove(lock_path)
-        except FileNotFoundError:
-            pass
-
-    def load_yolo_constants(self):
-        base = f"{self.cache_dir}/{self.model_folder}"
-        # constants for yolov9 post-processing
-        self.const_A = np.load(f"{base}/_model_22_Constant_9_output_0.npy")
-        self.const_B = np.load(f"{base}/_model_22_Constant_10_output_0.npy")
-        self.const_C = np.load(f"{base}/_model_22_Constant_12_output_0.npy")
-
     def check_and_prepare_model(self):
         if not os.path.exists(self.cache_dir):
             os.makedirs(self.cache_dir, exist_ok=True)
 
         lock_path = os.path.join(self.cache_dir, f".{self.model_folder}.lock")
-        self._acquire_file_lock(lock_path)
+        lock = FileLock(lock_path, timeout=60)
 
-        try:
+        with lock:
             # ---------- CASE 1: user provided a custom model path ----------
             if self.memx_model_path:
                 if not self.memx_model_path.endswith(".zip"):
@@ -258,7 +228,6 @@ class MemryXDetector(DetectionApi):
 
                 # Handle post model requirements by model type
                 if self.memx_model_type in [
-                    ModelTypeEnum.yologeneric,
                     ModelTypeEnum.yolonas,
                     ModelTypeEnum.ssd,
                 ]:
@@ -267,7 +236,10 @@ class MemryXDetector(DetectionApi):
                             f"No *_post.onnx file found in custom model zip for {self.memx_model_type.name}."
                         )
                     self.memx_post_model = post_candidates[0]
-                elif self.memx_model_type == ModelTypeEnum.yolox:
+                elif self.memx_model_type in [
+                    ModelTypeEnum.yolox,
+                    ModelTypeEnum.yologeneric,
+                ]:
                     # Explicitly ignore any post model even if present
                     self.memx_post_model = None
                 else:
@@ -295,8 +267,6 @@ class MemryXDetector(DetectionApi):
                 logger.info("Using cached models.")
                 self.memx_model_path = dfp_path
                 self.memx_post_model = post_path
-                if self.memx_model_type == ModelTypeEnum.yologeneric:
-                    self.load_yolo_constants()
                 return
 
             # ---------- CASE 3: download MemryX model (no cache) ----------
@@ -325,9 +295,6 @@ class MemryXDetector(DetectionApi):
                     else None
                 )
 
-                if self.memx_model_type == ModelTypeEnum.yologeneric:
-                    self.load_yolo_constants()
-
             finally:
                 if os.path.exists(zip_path):
                     try:
@@ -338,9 +305,6 @@ class MemryXDetector(DetectionApi):
                             f"Failed to remove downloaded zip {zip_path}: {e}"
                         )
 
-        finally:
-            self._release_file_lock(lock_path)
-
     def send_input(self, connection_id, tensor_input: np.ndarray):
         """Pre-process (if needed) and send frame to MemryX input queue"""
         if tensor_input is None:
@@ -625,127 +589,232 @@ class MemryXDetector(DetectionApi):
 
         self.output_queue.put(final_detections)
 
-    def onnx_reshape_with_allowzero(
+    def _generate_anchors(self, sizes=[80, 40, 20]):
-        self, data: np.ndarray, shape: np.ndarray, allowzero: int = 0
+        """Generate anchor points for YOLOv9 style processing"""
+        yscales = []
+        xscales = []
+        for s in sizes:
+            r = np.arange(s) + 0.5
+            yscales.append(np.repeat(r, s))
+            xscales.append(np.repeat(r[None, ...], s, axis=0).flatten())
+
+        yscales = np.concatenate(yscales)
+        xscales = np.concatenate(xscales)
+        anchors = np.stack([xscales, yscales], axis=1)
+        return anchors
+
+    def _generate_scales(self, sizes=[80, 40, 20]):
+        """Generate scaling factors for each detection level"""
+        factors = [8, 16, 32]
+        s = np.concatenate([np.ones([int(s * s)]) * f for s, f in zip(sizes, factors)])
+        return s[:, None]
+
+    @staticmethod
+    def _softmax(x: np.ndarray, axis: int) -> np.ndarray:
+        """Efficient softmax implementation"""
+        x = x - np.max(x, axis=axis, keepdims=True)
+        np.exp(x, out=x)
+        x /= np.sum(x, axis=axis, keepdims=True)
+        return x
+
+    def dfl(self, x: np.ndarray) -> np.ndarray:
+        """Distribution Focal Loss decoding - YOLOv9 style"""
+        x = x.reshape(-1, 4, 16)
+        weights = np.arange(16, dtype=np.float32)
+        p = self._softmax(x, axis=2)
+        p = p * weights[None, None, :]
+        out = np.sum(p, axis=2, keepdims=False)
+        return out
+
+    def dist2bbox(
+        self, x: np.ndarray, anchors: np.ndarray, scales: np.ndarray
     ) -> np.ndarray:
-        shape = shape.astype(int)
+        """Convert distances to bounding boxes - YOLOv9 style"""
-        input_shape = data.shape
+        lt = x[:, :2]
-        output_shape = []
+        rb = x[:, 2:]
 
-        for i, dim in enumerate(shape):
+        x1y1 = anchors - lt
-            if dim == 0 and allowzero == 0:
+        x2y2 = anchors + rb
-                output_shape.append(input_shape[i])  # Copy dimension from input
-            else:
-                output_shape.append(dim)
 
-        # Now let NumPy infer any -1 if needed
+        wh = x2y2 - x1y1
-        reshaped = np.reshape(data, output_shape)
+        c_xy = (x1y1 + x2y2) / 2
 
-        return reshaped
+        out = np.concatenate([c_xy, wh], axis=1)
+        out = out * scales
+        return out
+
+    def post_process_yolo_optimized(self, outputs):
+        """
+        Custom YOLOv9 post-processing optimized for MemryX ONNX outputs.
+        Implements DFL decoding, confidence filtering, and NMS in pure NumPy.
+        """
+        # YOLOv9 outputs: 6 outputs (lbox, lcls, mbox, mcls, sbox, scls)
+        conv_out1, conv_out2, conv_out3, conv_out4, conv_out5, conv_out6 = outputs
+
+        # Determine grid sizes based on input resolution
+        # YOLOv9 uses 3 detection heads with strides [8, 16, 32]
+        # Grid sizes = input_size / stride
+        sizes = [
+            self.memx_model_height
+            // 8,  # Large objects (e.g., 80 for 640x640, 40 for 320x320)
+            self.memx_model_height
+            // 16,  # Medium objects (e.g., 40 for 640x640, 20 for 320x320)
+            self.memx_model_height
+            // 32,  # Small objects (e.g., 20 for 640x640, 10 for 320x320)
+        ]
+
+        # Generate anchors and scales if not already done
+        if not hasattr(self, "anchors"):
+            self.anchors = self._generate_anchors(sizes)
+            self.scales = self._generate_scales(sizes)
+
+        # Process outputs in YOLOv9 format: reshape and moveaxis for ONNX format
+        lbox = np.moveaxis(conv_out1, 1, -1)  # Large boxes
+        lcls = np.moveaxis(conv_out2, 1, -1)  # Large classes
+        mbox = np.moveaxis(conv_out3, 1, -1)  # Medium boxes
+        mcls = np.moveaxis(conv_out4, 1, -1)  # Medium classes
+        sbox = np.moveaxis(conv_out5, 1, -1)  # Small boxes
+        scls = np.moveaxis(conv_out6, 1, -1)  # Small classes
+
+        # Determine number of classes dynamically from the class output shape
+        # lcls shape should be (batch, height, width, num_classes)
+        num_classes = lcls.shape[-1]
+
+        # Validate that all class outputs have the same number of classes
+        if not (mcls.shape[-1] == num_classes and scls.shape[-1] == num_classes):
+            raise ValueError(
+                f"Class output shapes mismatch: lcls={lcls.shape}, mcls={mcls.shape}, scls={scls.shape}"
+            )
+
+        # Concatenate boxes and classes
+        boxes = np.concatenate(
+            [
+                lbox.reshape(-1, 64),  # 64 is for 4 bbox coords * 16 DFL bins
+                mbox.reshape(-1, 64),
+                sbox.reshape(-1, 64),
+            ],
+            axis=0,
+        )
+
+        classes = np.concatenate(
+            [
+                lcls.reshape(-1, num_classes),
+                mcls.reshape(-1, num_classes),
+                scls.reshape(-1, num_classes),
+            ],
+            axis=0,
+        )
+
+        # Apply sigmoid to classes
+        classes = self.sigmoid(classes)
+
+        # Apply DFL to box predictions
+        boxes = self.dfl(boxes)
+
+        # YOLOv9 postprocessing with confidence filtering and NMS
+        confidence_thres = 0.4
+        iou_thres = 0.6
+
+        # Find the class with the highest score for each detection
+        max_scores = np.max(classes, axis=1)  # Maximum class score for each detection
+        class_ids = np.argmax(classes, axis=1)  # Index of the best class
+
+        # Filter out detections with scores below the confidence threshold
+        valid_indices = np.where(max_scores >= confidence_thres)[0]
+        if len(valid_indices) == 0:
+            # Return empty detections array
+            final_detections = np.zeros((20, 6), np.float32)
+            return final_detections
+
+        # Select only valid detections
+        valid_boxes = boxes[valid_indices]
+        valid_class_ids = class_ids[valid_indices]
+        valid_scores = max_scores[valid_indices]
+
+        # Convert distances to actual bounding boxes using anchors and scales
+        valid_boxes = self.dist2bbox(
+            valid_boxes, self.anchors[valid_indices], self.scales[valid_indices]
+        )
+
+        # Convert bounding box coordinates from (x_center, y_center, w, h) to (x_min, y_min, x_max, y_max)
+        x_center, y_center, width, height = (
+            valid_boxes[:, 0],
+            valid_boxes[:, 1],
+            valid_boxes[:, 2],
+            valid_boxes[:, 3],
+        )
+        x_min = x_center - width / 2
+        y_min = y_center - height / 2
+        x_max = x_center + width / 2
+        y_max = y_center + height / 2
+
+        # Convert to format expected by cv2.dnn.NMSBoxes: [x, y, width, height]
+        boxes_for_nms = []
+        scores_for_nms = []
+
+        for i in range(len(valid_indices)):
+            # Ensure coordinates are within bounds and positive
+            x_min_clipped = max(0, x_min[i])
+            y_min_clipped = max(0, y_min[i])
+            x_max_clipped = min(self.memx_model_width, x_max[i])
+            y_max_clipped = min(self.memx_model_height, y_max[i])
+
+            width_clipped = x_max_clipped - x_min_clipped
+            height_clipped = y_max_clipped - y_min_clipped
+
+            if width_clipped > 0 and height_clipped > 0:
+                boxes_for_nms.append(
+                    [x_min_clipped, y_min_clipped, width_clipped, height_clipped]
+                )
+                scores_for_nms.append(float(valid_scores[i]))
+
+        final_detections = np.zeros((20, 6), np.float32)
+
+        if len(boxes_for_nms) == 0:
+            return final_detections
+
+        # Apply NMS using OpenCV
+        indices = cv2.dnn.NMSBoxes(
+            boxes_for_nms, scores_for_nms, confidence_thres, iou_thres
+        )
+
+        if len(indices) > 0:
+            # Flatten indices if they are returned as a list of arrays
+            if isinstance(indices[0], list) or isinstance(indices[0], np.ndarray):
+                indices = [i[0] for i in indices]
+
+            # Limit to top 20 detections
+            indices = indices[:20]
+
+            # Convert to Frigate format: [class_id, confidence, y_min, x_min, y_max, x_max] (normalized)
+            for i, idx in enumerate(indices):
+                class_id = valid_class_ids[idx]
+                confidence = valid_scores[idx]
+
+                # Get the box coordinates
+                box = boxes_for_nms[idx]
+                x_min_norm = box[0] / self.memx_model_width
+                y_min_norm = box[1] / self.memx_model_height
+                x_max_norm = (box[0] + box[2]) / self.memx_model_width
+                y_max_norm = (box[1] + box[3]) / self.memx_model_height
+
+                final_detections[i] = [
+                    class_id,
+                    confidence,
+                    y_min_norm,  # Frigate expects y_min first
+                    x_min_norm,
+                    y_max_norm,
+                    x_max_norm,
+                ]
+
+        return final_detections
 
     def process_output(self, *outputs):
         """Output callback function -- receives frames from the MX3 and triggers post-processing"""
         if self.memx_model_type == ModelTypeEnum.yologeneric:
-            if not self.memx_post_model:
+            # Use complete YOLOv9-style postprocessing (includes NMS)
-                conv_out1 = outputs[0]
+            final_detections = self.post_process_yolo_optimized(outputs)
-                conv_out2 = outputs[1]
-                conv_out3 = outputs[2]
-                conv_out4 = outputs[3]
-                conv_out5 = outputs[4]
-                conv_out6 = outputs[5]
 
-                concat_1 = self.onnx_concat([conv_out1, conv_out2], axis=1)
-                concat_2 = self.onnx_concat([conv_out3, conv_out4], axis=1)
-                concat_3 = self.onnx_concat([conv_out5, conv_out6], axis=1)
-
-                shape = np.array([1, 144, -1], dtype=np.int64)
-
-                reshaped_1 = self.onnx_reshape_with_allowzero(
-                    concat_1, shape, allowzero=0
-                )
-                reshaped_2 = self.onnx_reshape_with_allowzero(
-                    concat_2, shape, allowzero=0
-                )
-                reshaped_3 = self.onnx_reshape_with_allowzero(
-                    concat_3, shape, allowzero=0
-                )
-
-                concat_4 = self.onnx_concat([reshaped_1, reshaped_2, reshaped_3], 2)
-
-                axis = 1
-                split_sizes = [64, 80]
-
-                # Calculate indices at which to split
-                indices = np.cumsum(split_sizes)[
-                    :-1
-                ]  # [64] — split before the second chunk
-
-                # Perform split along axis 1
-                split_0, split_1 = np.split(concat_4, indices, axis=axis)
-
-                num_boxes = 2100 if self.memx_model_height == 320 else 8400
-                shape1 = np.array([1, 4, 16, num_boxes])
-                reshape_4 = self.onnx_reshape_with_allowzero(
-                    split_0, shape1, allowzero=0
-                )
-
-                transpose_1 = reshape_4.transpose(0, 2, 1, 3)
-
-                axis = 1  # As per ONNX softmax node
-
-                # Subtract max for numerical stability
-                x_max = np.max(transpose_1, axis=axis, keepdims=True)
-                x_exp = np.exp(transpose_1 - x_max)
-                x_sum = np.sum(x_exp, axis=axis, keepdims=True)
-                softmax_output = x_exp / x_sum
-
-                # Weight W from the ONNX initializer (1, 16, 1, 1) with values 0 to 15
-                W = np.arange(16, dtype=np.float32).reshape(
-                    1, 16, 1, 1
-                )  # (1, 16, 1, 1)
-
-                # Apply 1x1 convolution: this is a weighted sum over channels
-                conv_output = np.sum(
-                    softmax_output * W, axis=1, keepdims=True
-                )  # shape: (1, 1, 4, 8400)
-
-                shape2 = np.array([1, 4, num_boxes])
-                reshape_5 = self.onnx_reshape_with_allowzero(
-                    conv_output, shape2, allowzero=0
-                )
-
-                # ONNX Slice — get first 2 channels: [0:2] along axis 1
-                slice_output1 = reshape_5[:, 0:2, :]  # Result: (1, 2, 8400)
-
-                # Slice channels 2 to 4 → axis = 1
-                slice_output2 = reshape_5[:, 2:4, :]
-
-                # Perform Subtraction
-                sub_output = self.const_A - slice_output1  # Equivalent to ONNX Sub
-
-                # Perform the ONNX-style Add
-                add_output = self.const_B + slice_output2
-
-                sub1 = add_output - sub_output
-
-                add1 = sub_output + add_output
-
-                div_output = add1 / 2.0
-
-                concat_5 = self.onnx_concat([div_output, sub1], axis=1)
-
-                # Expand B to (1, 1, 8400) so it can broadcast across axis=1 (4 channels)
-                const_C_expanded = self.const_C[:, np.newaxis, :]  # Shape: (1, 1, 8400)
-
-                # Perform ONNX-style element-wise multiplication
-                mul_output = concat_5 * const_C_expanded  # Result: (1, 4, 8400)
-
-                sigmoid_output = self.sigmoid(split_1)
-                outputs = self.onnx_concat([mul_output, sigmoid_output], axis=1)
-
-            final_detections = post_process_yolo(
-                outputs, self.memx_model_width, self.memx_model_height
-            )
             self.output_queue.put(final_detections)
 
         elif self.memx_model_type == ModelTypeEnum.yolonas:

frigate/embeddings/embeddings.py

@@ -29,7 +29,7 @@ from frigate.db.sqlitevecq import SqliteVecQueueDatabase
 from frigate.models import Event, Trigger
 from frigate.types import ModelStatusTypesEnum
 from frigate.util.builtin import EventsPerSecond, InferenceSpeed, serialize
-from frigate.util.path import get_event_thumbnail_bytes
+from frigate.util.file import get_event_thumbnail_bytes
 
 from .onnx.jina_v1_embedding import JinaV1ImageEmbedding, JinaV1TextEmbedding
 from .onnx.jina_v2_embedding import JinaV2Embedding
@@ -472,7 +472,7 @@ class Embeddings:
                                 )
                                 thumbnail_missing = True
                         except DoesNotExist:
-                            logger.warning(
+                            logger.debug(
                                 f"Event ID {trigger.data} for trigger {trigger_name} does not exist."
                             )
                             continue

frigate/embeddings/maintainer.py

@@ -9,6 +9,7 @@ from typing import Any
 
 from peewee import DoesNotExist
 
+from frigate.comms.config_updater import ConfigSubscriber
 from frigate.comms.detections_updater import DetectionSubscriber, DetectionTypeEnum
 from frigate.comms.embeddings_updater import (
     EmbeddingsRequestEnum,
@@ -61,8 +62,8 @@ from frigate.events.types import EventTypeEnum, RegenerateDescriptionEnum
 from frigate.genai import get_genai_client
 from frigate.models import Event, Recordings, ReviewSegment, Trigger
 from frigate.util.builtin import serialize
+from frigate.util.file import get_event_thumbnail_bytes
 from frigate.util.image import SharedMemoryFrameManager
-from frigate.util.path import get_event_thumbnail_bytes
 
 from .embeddings import Embeddings
 
@@ -95,6 +96,9 @@ class EmbeddingMaintainer(threading.Thread):
                 CameraConfigUpdateEnum.semantic_search,
             ],
         )
+        self.classification_config_subscriber = ConfigSubscriber(
+            "config/classification/custom/"
+        )
 
         # Configure Frigate DB
         db = SqliteVecQueueDatabase(
@@ -154,11 +158,13 @@ class EmbeddingMaintainer(threading.Thread):
         self.realtime_processors: list[RealTimeProcessorApi] = []
 
         if self.config.face_recognition.enabled:
+            logger.debug("Face recognition enabled, initializing FaceRealTimeProcessor")
             self.realtime_processors.append(
                 FaceRealTimeProcessor(
                     self.config, self.requestor, self.event_metadata_publisher, metrics
                 )
             )
+            logger.debug("FaceRealTimeProcessor initialized successfully")
 
         if self.config.classification.bird.enabled:
             self.realtime_processors.append(
@@ -189,6 +195,7 @@ class EmbeddingMaintainer(threading.Thread):
                     self.config,
                     model_config,
                     self.event_metadata_publisher,
+                    self.requestor,
                     self.metrics,
                 )
             )
@@ -220,7 +227,9 @@ class EmbeddingMaintainer(threading.Thread):
             for c in self.config.cameras.values()
         ):
             self.post_processors.append(
-                AudioTranscriptionPostProcessor(self.config, self.requestor, metrics)
+                AudioTranscriptionPostProcessor(
+                    self.config, self.requestor, self.embeddings, metrics
+                )
             )
 
         semantic_trigger_processor: SemanticTriggerProcessor | None = None
@@ -229,6 +238,7 @@ class EmbeddingMaintainer(threading.Thread):
                 db,
                 self.config,
                 self.requestor,
+                self.event_metadata_publisher,
                 metrics,
                 self.embeddings,
             )
@@ -255,6 +265,7 @@ class EmbeddingMaintainer(threading.Thread):
         """Maintain a SQLite-vec database for semantic search."""
         while not self.stop_event.is_set():
             self.config_updater.check_for_updates()
+            self._check_classification_config_updates()
             self._process_requests()
             self._process_updates()
             self._process_recordings_updates()
@@ -265,6 +276,7 @@ class EmbeddingMaintainer(threading.Thread):
             self._process_event_metadata()
 
         self.config_updater.stop()
+        self.classification_config_subscriber.stop()
         self.event_subscriber.stop()
         self.event_end_subscriber.stop()
         self.recordings_subscriber.stop()
@@ -275,6 +287,68 @@ class EmbeddingMaintainer(threading.Thread):
         self.requestor.stop()
         logger.info("Exiting embeddings maintenance...")
 
+    def _check_classification_config_updates(self) -> None:
+        """Check for classification config updates and add/remove processors."""
+        topic, model_config = self.classification_config_subscriber.check_for_update()
+
+        if topic:
+            model_name = topic.split("/")[-1]
+
+            if model_config is None:
+                self.realtime_processors = [
+                    processor
+                    for processor in self.realtime_processors
+                    if not (
+                        isinstance(
+                            processor,
+                            (
+                                CustomStateClassificationProcessor,
+                                CustomObjectClassificationProcessor,
+                            ),
+                        )
+                        and processor.model_config.name == model_name
+                    )
+                ]
+
+                logger.info(
+                    f"Successfully removed classification processor for model: {model_name}"
+                )
+            else:
+                self.config.classification.custom[model_name] = model_config
+
+                # Check if processor already exists
+                for processor in self.realtime_processors:
+                    if isinstance(
+                        processor,
+                        (
+                            CustomStateClassificationProcessor,
+                            CustomObjectClassificationProcessor,
+                        ),
+                    ):
+                        if processor.model_config.name == model_name:
+                            logger.debug(
+                                f"Classification processor for model {model_name} already exists, skipping"
+                            )
+                            return
+
+                if model_config.state_config is not None:
+                    processor = CustomStateClassificationProcessor(
+                        self.config, model_config, self.requestor, self.metrics
+                    )
+                else:
+                    processor = CustomObjectClassificationProcessor(
+                        self.config,
+                        model_config,
+                        self.event_metadata_publisher,
+                        self.requestor,
+                        self.metrics,
+                    )
+
+                self.realtime_processors.append(processor)
+                logger.info(
+                    f"Added classification processor for model: {model_name} (type: {type(processor).__name__})"
+                )
+
     def _process_requests(self) -> None:
         """Process embeddings requests"""
 
@@ -327,7 +401,14 @@ class EmbeddingMaintainer(threading.Thread):
 
         source_type, _, camera, frame_name, data = update
 
+        logger.debug(
+            f"Received update - source_type: {source_type}, camera: {camera}, data label: {data.get('label') if data else 'None'}"
+        )
+
         if not camera or source_type != EventTypeEnum.tracked_object:
+            logger.debug(
+                f"Skipping update - camera: {camera}, source_type: {source_type}"
+            )
             return
 
         if self.config.semantic_search.enabled:
@@ -337,6 +418,9 @@ class EmbeddingMaintainer(threading.Thread):
 
         # no need to process updated objects if no processors are active
         if len(self.realtime_processors) == 0 and len(self.post_processors) == 0:
+            logger.debug(
+                f"No processors active - realtime: {len(self.realtime_processors)}, post: {len(self.post_processors)}"
+            )
             return
 
         # Create our own thumbnail based on the bounding box and the frame time
@@ -345,6 +429,7 @@ class EmbeddingMaintainer(threading.Thread):
                 frame_name, camera_config.frame_shape_yuv
             )
         except FileNotFoundError:
+            logger.debug(f"Frame {frame_name} not found for camera {camera}")
             pass
 
         if yuv_frame is None:
@@ -353,7 +438,11 @@ class EmbeddingMaintainer(threading.Thread):
             )
             return
 
+        logger.debug(
+            f"Processing {len(self.realtime_processors)} realtime processors for object {data.get('id')} (label: {data.get('label')})"
+        )
         for processor in self.realtime_processors:
+            logger.debug(f"Calling process_frame on {processor.__class__.__name__}")
             processor.process_frame(data, yuv_frame)
 
         for processor in self.post_processors:

frigate/events/cleanup.py

@@ -12,7 +12,7 @@ from frigate.config import FrigateConfig
 from frigate.const import CLIPS_DIR
 from frigate.db.sqlitevecq import SqliteVecQueueDatabase
 from frigate.models import Event, Timeline
-from frigate.util.path import delete_event_snapshot, delete_event_thumbnail
+from frigate.util.file import delete_event_snapshot, delete_event_thumbnail
 
 logger = logging.getLogger(__name__)
 

frigate/ffmpeg_presets.py

@@ -150,10 +150,10 @@ PRESETS_HW_ACCEL_SCALE["preset-rk-h265"] = PRESETS_HW_ACCEL_SCALE[FFMPEG_HWACCEL
 PRESETS_HW_ACCEL_ENCODE_BIRDSEYE = {
     "preset-rpi-64-h264": "{0} -hide_banner {1} -c:v h264_v4l2m2m {2}",
     "preset-rpi-64-h265": "{0} -hide_banner {1} -c:v hevc_v4l2m2m {2}",
-    FFMPEG_HWACCEL_VAAPI: "{0} -hide_banner -hwaccel vaapi -hwaccel_output_format vaapi {3} {1} -c:v h264_vaapi -g 50 -bf 0 -profile:v high -level:v 4.1 -sei:v 0 -an -vf format=vaapi|nv12,hwupload {2}",
+    FFMPEG_HWACCEL_VAAPI: "{0} -hide_banner -hwaccel vaapi -hwaccel_output_format vaapi -hwaccel_device {3} {1} -c:v h264_vaapi -g 50 -bf 0 -profile:v high -level:v 4.1 -sei:v 0 -an -vf format=vaapi|nv12,hwupload {2}",
     "preset-intel-qsv-h264": "{0} -hide_banner {1} -c:v h264_qsv -g 50 -bf 0 -profile:v high -level:v 4.1 -async_depth:v 1 {2}",
     "preset-intel-qsv-h265": "{0} -hide_banner {1} -c:v h264_qsv -g 50 -bf 0 -profile:v main -level:v 4.1 -async_depth:v 1 {2}",
-    FFMPEG_HWACCEL_NVIDIA: "{0} -hide_banner {1} {3} -c:v h264_nvenc -g 50 -profile:v high -level:v auto -preset:v p2 -tune:v ll {2}",
+    FFMPEG_HWACCEL_NVIDIA: "{0} -hide_banner {1} -hwaccel device {3} -c:v h264_nvenc -g 50 -profile:v high -level:v auto -preset:v p2 -tune:v ll {2}",
     "preset-jetson-h264": "{0} -hide_banner {1} -c:v h264_nvmpi -profile high {2}",
     "preset-jetson-h265": "{0} -hide_banner {1} -c:v h264_nvmpi -profile main {2}",
     FFMPEG_HWACCEL_RKMPP: "{0} -hide_banner {1} -c:v h264_rkmpp -profile:v high {2}",
@@ -246,7 +246,7 @@ def parse_preset_hardware_acceleration_scale(
         ",hwdownload,format=nv12,eq=gamma=1.4:gamma_weight=0.5" in scale
         and os.environ.get("FFMPEG_DISABLE_GAMMA_EQUALIZER") is not None
     ):
-        scale.replace(
+        scale = scale.replace(
             ",hwdownload,format=nv12,eq=gamma=1.4:gamma_weight=0.5",
             ":format=nv12,hwdownload,format=nv12,format=yuv420p",
         )

frigate/genai/__init__.py

@@ -51,8 +51,7 @@ class GenAIClient:
         def get_concern_prompt() -> str:
             if concerns:
                 concern_list = "\n    - ".join(concerns)
-                return f"""
+                return f"""- `other_concerns` (list of strings): Include a list of any of the following concerns that are occurring:
-- `other_concerns` (list of strings): Include a list of any of the following concerns that are occurring:
     - {concern_list}"""
             else:
                 return ""
@@ -63,58 +62,68 @@ class GenAIClient:
             else:
                 return ""
 
-        def get_verified_objects() -> str:
+        def get_objects_list() -> str:
-            if review_data["recognized_objects"]:
+            if review_data["unified_objects"]:
-                return "  - " + "\n  - ".join(review_data["recognized_objects"])
+                return "\n- " + "\n- ".join(review_data["unified_objects"])
             else:
-                return "  None"
+                return "\n- (No objects detected)"
 
         context_prompt = f"""
-Please analyze the sequence of images ({len(thumbnails)} total) taken in chronological order from the perspective of the {review_data["camera"].replace("_", " ")} security camera.
+Your task is to analyze the sequence of images ({len(thumbnails)} total) taken in chronological order from the perspective of the {review_data["camera"]} security camera.
+
+## Normal Activity Patterns for This Property
 
-**Normal activity patterns for this property:**
 {activity_context_prompt}
 
+## Task Instructions
+
 Your task is to provide a clear, accurate description of the scene that:
 1. States exactly what is happening based on observable actions and movements.
-2. Evaluates whether the observable evidence suggests normal activity for this property or genuine security concerns.
+2. Evaluates the activity against the Normal and Suspicious Activity Indicators above.
-3. Assigns a potential_threat_level based on the definitions below, applying them consistently.
+3. Assigns a potential_threat_level (0, 1, or 2) based on the threat level indicators defined above, applying them consistently.
 
-**IMPORTANT: Start by checking if the activity matches the normal patterns above. If it does, assign Level 0. Only consider higher threat levels if the activity clearly deviates from normal patterns or shows genuine security concerns.**
+**Use the activity patterns above as guidance to calibrate your assessment. Match the activity against both normal and suspicious indicators, then use your judgment based on the complete context.**
+
+## Analysis Guidelines
 
 When forming your description:
-- **CRITICAL: Only describe objects explicitly listed in "Detected objects" below.** Do not infer or mention additional people, vehicles, or objects not present in the detected objects list, even if visual patterns suggest them. If only a car is detected, do not describe a person interacting with it unless "person" is also in the detected objects list.
+- **CRITICAL: Only describe objects explicitly listed in "Objects in Scene" below.** Do not infer or mention additional people, vehicles, or objects not present in this list, even if visual patterns suggest them. If only a car is listed, do not describe a person interacting with it unless "person" is also in the objects list.
 - **Only describe actions actually visible in the frames.** Do not assume or infer actions that you don't observe happening. If someone walks toward furniture but you never see them sit, do not say they sat. Stick to what you can see across the sequence.
 - Describe what you observe: actions, movements, interactions with objects and the environment. Include any observable environmental changes (e.g., lighting changes triggered by activity).
 - Note visible details such as clothing, items being carried or placed, tools or equipment present, and how they interact with the property or objects.
 - Consider the full sequence chronologically: what happens from start to finish, how duration and actions relate to the location and objects involved.
 - **Use the actual timestamp provided in "Activity started at"** below for time of day context—do not infer time from image brightness or darkness. Unusual hours (late night/early morning) should increase suspicion when the observable behavior itself appears questionable. However, recognize that some legitimate activities can occur at any hour.
-- Identify patterns that suggest genuine security concerns: testing doors/windows on vehicles or buildings, accessing unauthorized areas, attempting to conceal actions, extended loitering without apparent purpose, taking items, behavior that clearly doesn't align with the zone context and detected objects.
+- **Consider duration as a primary factor**: Apply the duration thresholds defined in the activity patterns above. Brief sequences during normal hours with apparent purpose typically indicate normal activity unless explicit suspicious actions are visible.
-- **Weigh all evidence holistically**: Start by checking if the activity matches the normal patterns above. If it does, assign Level 0. Only consider Level 1 if the activity clearly deviates from normal patterns or shows genuine security concerns that warrant attention.
+- **Weigh all evidence holistically**: Match the activity against the normal and suspicious patterns defined above, then evaluate based on the complete context (zone, objects, time, actions, duration). Apply the threat level indicators consistently. Use your judgment for edge cases.
+
+## Response Format
 
 Your response MUST be a flat JSON object with:
-- `title` (string): A concise, one-sentence title that captures the main activity. Include any verified recognized objects (from the "Verified recognized objects" list below) and key detected objects. Examples: "Joe walking dog in backyard", "Unknown person testing car doors at night".
+- `title` (string): A concise, direct title that describes the primary action or event in the sequence, not just what you literally see. Use spatial context when available to make titles more meaningful. When multiple objects/actions are present, prioritize whichever is most prominent or occurs first. Use names from "Objects in Scene" based on what you visually observe. If you see both a name and an unidentified object of the same type but visually observe only one person/object, use ONLY the name. Examples: "Joe walking dog", "Person taking out trash", "Vehicle arriving in driveway", "Joe accessing vehicle", "Person leaving porch for driveway".
-- `scene` (string): A narrative description of what happens across the sequence from start to finish. **Only describe actions you can actually observe happening in the frames provided.** Do not infer or assume actions that aren't visible (e.g., if you see someone walking but never see them sit, don't say they sat down). Include setting, detected objects, and their observable actions. Avoid speculation or filling in assumed behaviors. Your description should align with and support the threat level you assign.
+- `scene` (string): A narrative description of what happens across the sequence from start to finish, in chronological order. Start by describing how the sequence begins, then describe the progression of events. **Describe all significant movements and actions in the order they occur.** For example, if a vehicle arrives and then a person exits, describe both actions sequentially. **Only describe actions you can actually observe happening in the frames provided.** Do not infer or assume actions that aren't visible (e.g., if you see someone walking but never see them sit, don't say they sat down). Include setting, detected objects, and their observable actions. Avoid speculation or filling in assumed behaviors. Your description should align with and support the threat level you assign.
 - `confidence` (float): 0-1 confidence in your analysis. Higher confidence when objects/actions are clearly visible and context is unambiguous. Lower confidence when the sequence is unclear, objects are partially obscured, or context is ambiguous.
-- `potential_threat_level` (integer): 0, 1, or 2 as defined below. Your threat level must be consistent with your scene description and the guidance above.
+- `potential_threat_level` (integer): 0, 1, or 2 as defined in "Normal Activity Patterns for This Property" above. Your threat level must be consistent with your scene description and the guidance above.
 {get_concern_prompt()}
 
-Threat-level definitions:
+## Sequence Details
-- 0 — **Normal activity (DEFAULT)**: What you observe matches the normal activity patterns above or is consistent with expected activity for this property type. The observable evidence—considering zone context, detected objects, and timing together—supports a benign explanation. **Use this level for routine activities even if minor ambiguous elements exist.**
-- 1 — **Potentially suspicious**: Observable behavior raises genuine security concerns that warrant human review. The evidence doesn't support a routine explanation and clearly deviates from the normal patterns above. Examples: testing doors/windows on vehicles or structures, accessing areas that don't align with the activity, taking items that likely don't belong to them, behavior clearly inconsistent with the zone and context, or activity that lacks any visible legitimate indicators. **Only use this level when the activity clearly doesn't match normal patterns.**
-- 2 — **Immediate threat**: Clear evidence of forced entry, break-in, vandalism, aggression, weapons, theft in progress, or active property damage.
 
-Sequence details:
 - Frame 1 = earliest, Frame {len(thumbnails)} = latest
 - Activity started at {review_data["start"]} and lasted {review_data["duration"]} seconds
-- Detected objects: {", ".join(review_data["objects"])}
+- Zones involved: {", ".join(review_data["zones"]) if review_data["zones"] else "None"}
-- Verified recognized objects (use these names when describing these objects):
-{get_verified_objects()}
-- Zones involved: {", ".join(z.replace("_", " ").title() for z in review_data["zones"]) or "None"}
 
-**IMPORTANT:**
+## Objects in Scene
+
+Each line represents a detection state, not necessarily unique individuals. Parentheses indicate object type or category, use only the name/label in your response, not the parentheses.
+
+**CRITICAL: When you see both recognized and unrecognized entries of the same type (e.g., "Joe (person)" and "Person"), visually count how many distinct people/objects you actually see based on appearance and clothing. If you observe only ONE person throughout the sequence, use ONLY the recognized name (e.g., "Joe"). The same person may be recognized in some frames but not others. Only describe both if you visually see MULTIPLE distinct people with clearly different appearances.**
+
+**Note: Unidentified objects (without names) are NOT indicators of suspicious activity—they simply mean the system hasn't identified that object.**
+{get_objects_list()}
+
+## Important Notes
 - Values must be plain strings, floats, or integers — no nested objects, no extra commentary.
-- Only describe objects from the "Detected objects" list above. Do not hallucinate additional objects.
+- Only describe objects from the "Objects in Scene" list above. Do not hallucinate additional objects.
+- When describing people or vehicles, use the exact names provided.
 {get_language_prompt()}
 """
         logger.debug(
@@ -149,7 +158,8 @@ Sequence details:
             try:
                 metadata = ReviewMetadata.model_validate_json(clean_json)
 
-                if review_data["recognized_objects"]:
+                # If any verified objects (contain parentheses with name), set to 0
+                if any("(" in obj for obj in review_data["unified_objects"]):
                     metadata.potential_threat_level = 0
 
                 metadata.time = review_data["start"]

frigate/genai/ollama.py

@@ -1,7 +1,7 @@
 """Ollama Provider for Frigate AI."""
 
 import logging
-from typing import Optional
+from typing import Any, Optional
 
 from httpx import TimeoutException
 from ollama import Client as ApiClient
@@ -17,10 +17,24 @@ logger = logging.getLogger(__name__)
 class OllamaClient(GenAIClient):
     """Generative AI client for Frigate using Ollama."""
 
+    LOCAL_OPTIMIZED_OPTIONS = {
+        "options": {
+            "temperature": 0.5,
+            "repeat_penalty": 1.05,
+            "presence_penalty": 0.3,
+        },
+    }
+
     provider: ApiClient
+    provider_options: dict[str, Any]
 
     def _init_provider(self):
         """Initialize the client."""
+        self.provider_options = {
+            **self.LOCAL_OPTIMIZED_OPTIONS,
+            **self.genai_config.provider_options,
+        }
+
         try:
             client = ApiClient(host=self.genai_config.base_url, timeout=self.timeout)
             # ensure the model is available locally
@@ -48,10 +62,13 @@ class OllamaClient(GenAIClient):
                 self.genai_config.model,
                 prompt,
                 images=images if images else None,
-                **self.genai_config.provider_options,
+                **self.provider_options,
+            )
+            logger.debug(
+                f"Ollama tokens used: eval_count={result.get('eval_count')}, prompt_eval_count={result.get('prompt_eval_count')}"
             )
             return result["response"].strip()
-        except (TimeoutException, ResponseError) as e:
+        except (TimeoutException, ResponseError, ConnectionError) as e:
             logger.warning("Ollama returned an error: %s", str(e))
             return None
 

frigate/genai/openai.py

@@ -18,6 +18,7 @@ class OpenAIClient(GenAIClient):
     """Generative AI client for Frigate using OpenAI."""
 
     provider: OpenAI
+    context_size: Optional[int] = None
 
     def _init_provider(self):
         """Initialize the client."""
@@ -69,5 +70,33 @@ class OpenAIClient(GenAIClient):
 
     def get_context_size(self) -> int:
         """Get the context window size for OpenAI."""
-        # OpenAI GPT-4 Vision models have 128K token context window
+        if self.context_size is not None:
-        return 128000
+            return self.context_size
+
+        try:
+            models = self.provider.models.list()
+            for model in models.data:
+                if model.id == self.genai_config.model:
+                    if hasattr(model, "max_model_len") and model.max_model_len:
+                        self.context_size = model.max_model_len
+                        logger.debug(
+                            f"Retrieved context size {self.context_size} for model {self.genai_config.model}"
+                        )
+                        return self.context_size
+
+        except Exception as e:
+            logger.debug(
+                f"Failed to fetch model context size from API: {e}, using default"
+            )
+
+        # Default to 128K for ChatGPT models, 8K for others
+        model_name = self.genai_config.model.lower()
+        if "gpt" in model_name:
+            self.context_size = 128000
+        else:
+            self.context_size = 8192
+
+        logger.debug(
+            f"Using default context size {self.context_size} for model {self.genai_config.model}"
+        )
+        return self.context_size

frigate/object_detection/base.py

@@ -9,6 +9,7 @@ from multiprocessing import Queue, Value
 from multiprocessing.synchronize import Event as MpEvent
 
 import numpy as np
+import zmq
 
 from frigate.comms.object_detector_signaler import (
     ObjectDetectorPublisher,
@@ -377,6 +378,15 @@ class RemoteObjectDetector:
         if self.stop_event.is_set():
             return detections
 
+        # Drain any stale detection results from the ZMQ buffer before making a new request
+        # This prevents reading detection results from a previous request
+        # NOTE: This should never happen, but can in some rare cases
+        while True:
+            try:
+                self.detector_subscriber.socket.recv_string(flags=zmq.NOBLOCK)
+            except zmq.Again:
+                break
+
         # copy input to shared memory
         self.np_shm[:] = tensor_input[:]
         self.detection_queue.put(self.name)

frigate/output/birdseye.py

@@ -9,6 +9,7 @@ import os
 import queue
 import subprocess as sp
 import threading
+import time
 import traceback
 from typing import Any, Optional
 
@@ -791,6 +792,10 @@ class Birdseye:
         self.frame_manager = SharedMemoryFrameManager()
         self.stop_event = stop_event
         self.requestor = InterProcessRequestor()
+        self.idle_fps: float = self.config.birdseye.idle_heartbeat_fps
+        self._idle_interval: Optional[float] = (
+            (1.0 / self.idle_fps) if self.idle_fps > 0 else None
+        )
 
         if config.birdseye.restream:
             self.birdseye_buffer = self.frame_manager.create(
@@ -848,6 +853,15 @@ class Birdseye:
             if frame_layout_changed:
                 coordinates = self.birdseye_manager.get_camera_coordinates()
                 self.requestor.send_data(UPDATE_BIRDSEYE_LAYOUT, coordinates)
+        if self._idle_interval:
+            now = time.monotonic()
+            is_idle = len(self.birdseye_manager.camera_layout) == 0
+            if (
+                is_idle
+                and (now - self.birdseye_manager.last_output_time)
+                >= self._idle_interval
+            ):
+                self.__send_new_frame()
 
     def stop(self) -> None:
         self.converter.join()

frigate/record/cleanup.py

@@ -14,7 +14,8 @@ from frigate.config import CameraConfig, FrigateConfig, RetainModeEnum
 from frigate.const import CACHE_DIR, CLIPS_DIR, MAX_WAL_SIZE, RECORD_DIR
 from frigate.models import Previews, Recordings, ReviewSegment, UserReviewStatus
 from frigate.record.util import remove_empty_directories, sync_recordings
-from frigate.util.builtin import clear_and_unlink, get_tomorrow_at_time
+from frigate.util.builtin import clear_and_unlink
+from frigate.util.time import get_tomorrow_at_time
 
 logger = logging.getLogger(__name__)
 

frigate/record/export.py

@@ -28,7 +28,7 @@ from frigate.ffmpeg_presets import (
     parse_preset_hardware_acceleration_encode,
 )
 from frigate.models import Export, Previews, Recordings
-from frigate.util.builtin import is_current_hour
+from frigate.util.time import is_current_hour
 
 logger = logging.getLogger(__name__)
 

frigate/review/maintainer.py

@@ -407,6 +407,19 @@ class ReviewSegmentMaintainer(threading.Thread):
                 segment.last_detection_time = frame_time
 
             for object in activity.get_all_objects():
+                # Alert-level objects should always be added (they extend/upgrade the segment)
+                # Detection-level objects should only be added if:
+                #   - The segment is a detection segment (matching severity), OR
+                #   - The segment is an alert AND the object started before the alert ended
+                #     (objects starting after will be in the new detection segment)
+                is_alert_object = object in activity.categorized_objects["alerts"]
+
+                if not is_alert_object and segment.severity == SeverityEnum.alert:
+                    # This is a detection-level object
+                    # Only add if it started during the alert's active period
+                    if object["start_time"] > segment.last_alert_time:
+                        continue
+
                 if not object["sub_label"]:
                     segment.detections[object["id"]] = object["label"]
                 elif object["sub_label"][0] in self.config.model.all_attributes:

frigate/stats/util.py

@@ -25,6 +25,7 @@ from frigate.util.services import (
     get_intel_gpu_stats,
     get_jetson_stats,
     get_nvidia_gpu_stats,
+    get_openvino_npu_stats,
     get_rockchip_gpu_stats,
     get_rockchip_npu_stats,
     is_vaapi_amd_driver,
@@ -247,6 +248,10 @@ async def set_npu_usages(config: FrigateConfig, all_stats: dict[str, Any]) -> No
             # Rockchip NPU usage
             rk_usage = get_rockchip_npu_stats()
             stats["rockchip"] = rk_usage
+        elif detector.type == "openvino" and detector.device == "NPU":
+            # OpenVINO NPU usage
+            ov_usage = get_openvino_npu_stats()
+            stats["openvino"] = ov_usage
 
     if stats:
         all_stats["npu_usages"] = stats
@@ -357,7 +362,7 @@ def stats_snapshot(
             stats["embeddings"]["review_description_speed"] = round(
                 embeddings_metrics.review_desc_speed.value * 1000, 2
             )
-            stats["embeddings"]["review_descriptions"] = round(
+            stats["embeddings"]["review_description_events_per_second"] = round(
                 embeddings_metrics.review_desc_dps.value, 2
             )
 
@@ -365,7 +370,7 @@ def stats_snapshot(
             stats["embeddings"]["object_description_speed"] = round(
                 embeddings_metrics.object_desc_speed.value * 1000, 2
             )
-            stats["embeddings"]["object_descriptions"] = round(
+            stats["embeddings"]["object_description_events_per_second"] = round(
                 embeddings_metrics.object_desc_dps.value, 2
             )
 
@@ -373,7 +378,7 @@ def stats_snapshot(
             stats["embeddings"][f"{key}_classification_speed"] = round(
                 embeddings_metrics.classification_speeds[key].value * 1000, 2
             )
-            stats["embeddings"][f"{key}_classification"] = round(
+            stats["embeddings"][f"{key}_classification_events_per_second"] = round(
                 embeddings_metrics.classification_cps[key].value, 2
             )
 

frigate/storage.py

@@ -113,6 +113,7 @@ class StorageMaintainer(threading.Thread):
         recordings: Recordings = (
             Recordings.select(
                 Recordings.id,
+                Recordings.camera,
                 Recordings.start_time,
                 Recordings.end_time,
                 Recordings.segment_size,
@@ -137,7 +138,7 @@ class StorageMaintainer(threading.Thread):
         )
 
         event_start = 0
-        deleted_recordings = set()
+        deleted_recordings = []
         for recording in recordings:
             # check if 1 hour of storage has been reclaimed
             if deleted_segments_size > hourly_bandwidth:
@@ -172,7 +173,7 @@ class StorageMaintainer(threading.Thread):
             if not keep:
                 try:
                     clear_and_unlink(Path(recording.path), missing_ok=False)
-                    deleted_recordings.add(recording.id)
+                    deleted_recordings.append(recording)
                     deleted_segments_size += recording.segment_size
                 except FileNotFoundError:
                     # this file was not found so we must assume no space was cleaned up
@@ -186,6 +187,9 @@ class StorageMaintainer(threading.Thread):
             recordings = (
                 Recordings.select(
                     Recordings.id,
+                    Recordings.camera,
+                    Recordings.start_time,
+                    Recordings.end_time,
                     Recordings.path,
                     Recordings.segment_size,
                 )
@@ -201,7 +205,7 @@ class StorageMaintainer(threading.Thread):
                 try:
                     clear_and_unlink(Path(recording.path), missing_ok=False)
                     deleted_segments_size += recording.segment_size
-                    deleted_recordings.add(recording.id)
+                    deleted_recordings.append(recording)
                 except FileNotFoundError:
                     # this file was not found so we must assume no space was cleaned up
                     pass
@@ -211,7 +215,50 @@ class StorageMaintainer(threading.Thread):
         logger.debug(f"Expiring {len(deleted_recordings)} recordings")
         # delete up to 100,000 at a time
         max_deletes = 100000
-        deleted_recordings_list = list(deleted_recordings)
+
+        # Update has_clip for events that overlap with deleted recordings
+        if deleted_recordings:
+            # Group deleted recordings by camera
+            camera_recordings = {}
+            for recording in deleted_recordings:
+                if recording.camera not in camera_recordings:
+                    camera_recordings[recording.camera] = {
+                        "min_start": recording.start_time,
+                        "max_end": recording.end_time,
+                    }
+                else:
+                    camera_recordings[recording.camera]["min_start"] = min(
+                        camera_recordings[recording.camera]["min_start"],
+                        recording.start_time,
+                    )
+                    camera_recordings[recording.camera]["max_end"] = max(
+                        camera_recordings[recording.camera]["max_end"],
+                        recording.end_time,
+                    )
+
+            # Find all events that overlap with deleted recordings time range per camera
+            events_to_update = []
+            for camera, time_range in camera_recordings.items():
+                overlapping_events = Event.select(Event.id).where(
+                    Event.camera == camera,
+                    Event.has_clip == True,
+                    Event.start_time < time_range["max_end"],
+                    Event.end_time > time_range["min_start"],
+                )
+
+                for event in overlapping_events:
+                    events_to_update.append(event.id)
+
+            # Update has_clip to False for overlapping events
+            if events_to_update:
+                for i in range(0, len(events_to_update), max_deletes):
+                    batch = events_to_update[i : i + max_deletes]
+                    Event.update(has_clip=False).where(Event.id << batch).execute()
+                logger.debug(
+                    f"Updated has_clip to False for {len(events_to_update)} events"
+                )
+
+        deleted_recordings_list = [r.id for r in deleted_recordings]
         for i in range(0, len(deleted_recordings_list), max_deletes):
             Recordings.delete().where(
                 Recordings.id << deleted_recordings_list[i : i + max_deletes]

frigate/test/http_api/test_http_media.py

@@ -0,0 +1,379 @@
+"""Unit tests for recordings/media API endpoints."""
+
+from datetime import datetime, timezone
+from typing import Any
+
+import pytz
+from fastapi.testclient import TestClient
+
+from frigate.api.auth import get_allowed_cameras_for_filter, get_current_user
+from frigate.models import Recordings
+from frigate.test.http_api.base_http_test import BaseTestHttp
+
+
+class TestHttpMedia(BaseTestHttp):
+    """Test media API endpoints, particularly recordings with DST handling."""
+
+    def setUp(self):
+        """Set up test fixtures."""
+        super().setUp([Recordings])
+        self.app = super().create_app()
+
+        # Mock auth to bypass camera access for tests
+        async def mock_get_current_user(request: Any):
+            return {"username": "test_user", "role": "admin"}
+
+        self.app.dependency_overrides[get_current_user] = mock_get_current_user
+        self.app.dependency_overrides[get_allowed_cameras_for_filter] = lambda: [
+            "front_door",
+            "back_door",
+        ]
+
+    def tearDown(self):
+        """Clean up after tests."""
+        self.app.dependency_overrides.clear()
+        super().tearDown()
+
+    def test_recordings_summary_across_dst_spring_forward(self):
+        """
+        Test recordings summary across spring DST transition (spring forward).
+
+        In 2024, DST in America/New_York transitions on March 10, 2024 at 2:00 AM
+        Clocks spring forward from 2:00 AM to 3:00 AM (EST to EDT)
+        """
+        tz = pytz.timezone("America/New_York")
+
+        # March 9, 2024 at 12:00 PM EST (before DST)
+        march_9_noon = tz.localize(datetime(2024, 3, 9, 12, 0, 0)).timestamp()
+
+        # March 10, 2024 at 12:00 PM EDT (after DST transition)
+        march_10_noon = tz.localize(datetime(2024, 3, 10, 12, 0, 0)).timestamp()
+
+        # March 11, 2024 at 12:00 PM EDT (after DST)
+        march_11_noon = tz.localize(datetime(2024, 3, 11, 12, 0, 0)).timestamp()
+
+        with TestClient(self.app) as client:
+            # Insert recordings for each day
+            Recordings.insert(
+                id="recording_march_9",
+                path="/media/recordings/march_9.mp4",
+                camera="front_door",
+                start_time=march_9_noon,
+                end_time=march_9_noon + 3600,  # 1 hour recording
+                duration=3600,
+                motion=100,
+                objects=5,
+            ).execute()
+
+            Recordings.insert(
+                id="recording_march_10",
+                path="/media/recordings/march_10.mp4",
+                camera="front_door",
+                start_time=march_10_noon,
+                end_time=march_10_noon + 3600,
+                duration=3600,
+                motion=150,
+                objects=8,
+            ).execute()
+
+            Recordings.insert(
+                id="recording_march_11",
+                path="/media/recordings/march_11.mp4",
+                camera="front_door",
+                start_time=march_11_noon,
+                end_time=march_11_noon + 3600,
+                duration=3600,
+                motion=200,
+                objects=10,
+            ).execute()
+
+            # Test recordings summary with America/New_York timezone
+            response = client.get(
+                "/recordings/summary",
+                params={"timezone": "America/New_York", "cameras": "all"},
+            )
+
+            assert response.status_code == 200
+            summary = response.json()
+
+            # Verify we get exactly 3 days
+            assert len(summary) == 3, f"Expected 3 days, got {len(summary)}"
+
+            # Verify the correct dates are returned (API returns dict with True values)
+            assert "2024-03-09" in summary, f"Expected 2024-03-09 in {summary}"
+            assert "2024-03-10" in summary, f"Expected 2024-03-10 in {summary}"
+            assert "2024-03-11" in summary, f"Expected 2024-03-11 in {summary}"
+            assert summary["2024-03-09"] is True
+            assert summary["2024-03-10"] is True
+            assert summary["2024-03-11"] is True
+
+    def test_recordings_summary_across_dst_fall_back(self):
+        """
+        Test recordings summary across fall DST transition (fall back).
+
+        In 2024, DST in America/New_York transitions on November 3, 2024 at 2:00 AM
+        Clocks fall back from 2:00 AM to 1:00 AM (EDT to EST)
+        """
+        tz = pytz.timezone("America/New_York")
+
+        # November 2, 2024 at 12:00 PM EDT (before DST transition)
+        nov_2_noon = tz.localize(datetime(2024, 11, 2, 12, 0, 0)).timestamp()
+
+        # November 3, 2024 at 12:00 PM EST (after DST transition)
+        # Need to specify is_dst=False to get the time after fall back
+        nov_3_noon = tz.localize(
+            datetime(2024, 11, 3, 12, 0, 0), is_dst=False
+        ).timestamp()
+
+        # November 4, 2024 at 12:00 PM EST (after DST)
+        nov_4_noon = tz.localize(datetime(2024, 11, 4, 12, 0, 0)).timestamp()
+
+        with TestClient(self.app) as client:
+            # Insert recordings for each day
+            Recordings.insert(
+                id="recording_nov_2",
+                path="/media/recordings/nov_2.mp4",
+                camera="front_door",
+                start_time=nov_2_noon,
+                end_time=nov_2_noon + 3600,
+                duration=3600,
+                motion=100,
+                objects=5,
+            ).execute()
+
+            Recordings.insert(
+                id="recording_nov_3",
+                path="/media/recordings/nov_3.mp4",
+                camera="front_door",
+                start_time=nov_3_noon,
+                end_time=nov_3_noon + 3600,
+                duration=3600,
+                motion=150,
+                objects=8,
+            ).execute()
+
+            Recordings.insert(
+                id="recording_nov_4",
+                path="/media/recordings/nov_4.mp4",
+                camera="front_door",
+                start_time=nov_4_noon,
+                end_time=nov_4_noon + 3600,
+                duration=3600,
+                motion=200,
+                objects=10,
+            ).execute()
+
+            # Test recordings summary with America/New_York timezone
+            response = client.get(
+                "/recordings/summary",
+                params={"timezone": "America/New_York", "cameras": "all"},
+            )
+
+            assert response.status_code == 200
+            summary = response.json()
+
+            # Verify we get exactly 3 days
+            assert len(summary) == 3, f"Expected 3 days, got {len(summary)}"
+
+            # Verify the correct dates are returned (API returns dict with True values)
+            assert "2024-11-02" in summary, f"Expected 2024-11-02 in {summary}"
+            assert "2024-11-03" in summary, f"Expected 2024-11-03 in {summary}"
+            assert "2024-11-04" in summary, f"Expected 2024-11-04 in {summary}"
+            assert summary["2024-11-02"] is True
+            assert summary["2024-11-03"] is True
+            assert summary["2024-11-04"] is True
+
+    def test_recordings_summary_multiple_cameras_across_dst(self):
+        """
+        Test recordings summary with multiple cameras across DST boundary.
+        """
+        tz = pytz.timezone("America/New_York")
+
+        # March 9, 2024 at 10:00 AM EST (before DST)
+        march_9_morning = tz.localize(datetime(2024, 3, 9, 10, 0, 0)).timestamp()
+
+        # March 10, 2024 at 3:00 PM EDT (after DST transition)
+        march_10_afternoon = tz.localize(datetime(2024, 3, 10, 15, 0, 0)).timestamp()
+
+        with TestClient(self.app) as client:
+            # Insert recordings for front_door on March 9
+            Recordings.insert(
+                id="front_march_9",
+                path="/media/recordings/front_march_9.mp4",
+                camera="front_door",
+                start_time=march_9_morning,
+                end_time=march_9_morning + 3600,
+                duration=3600,
+                motion=100,
+                objects=5,
+            ).execute()
+
+            # Insert recordings for back_door on March 10
+            Recordings.insert(
+                id="back_march_10",
+                path="/media/recordings/back_march_10.mp4",
+                camera="back_door",
+                start_time=march_10_afternoon,
+                end_time=march_10_afternoon + 3600,
+                duration=3600,
+                motion=150,
+                objects=8,
+            ).execute()
+
+            # Test with all cameras
+            response = client.get(
+                "/recordings/summary",
+                params={"timezone": "America/New_York", "cameras": "all"},
+            )
+
+            assert response.status_code == 200
+            summary = response.json()
+
+            # Verify we get both days
+            assert len(summary) == 2, f"Expected 2 days, got {len(summary)}"
+            assert "2024-03-09" in summary
+            assert "2024-03-10" in summary
+            assert summary["2024-03-09"] is True
+            assert summary["2024-03-10"] is True
+
+    def test_recordings_summary_at_dst_transition_time(self):
+        """
+        Test recordings that span the exact DST transition time.
+        """
+        tz = pytz.timezone("America/New_York")
+
+        # March 10, 2024 at 1:00 AM EST (1 hour before DST transition)
+        # At 2:00 AM, clocks jump to 3:00 AM
+        before_transition = tz.localize(datetime(2024, 3, 10, 1, 0, 0)).timestamp()
+
+        # Recording that spans the transition (1:00 AM to 3:30 AM EDT)
+        # This is 1.5 hours of actual time but spans the "missing" hour
+        after_transition = tz.localize(datetime(2024, 3, 10, 3, 30, 0)).timestamp()
+
+        with TestClient(self.app) as client:
+            Recordings.insert(
+                id="recording_during_transition",
+                path="/media/recordings/transition.mp4",
+                camera="front_door",
+                start_time=before_transition,
+                end_time=after_transition,
+                duration=after_transition - before_transition,
+                motion=100,
+                objects=5,
+            ).execute()
+
+            response = client.get(
+                "/recordings/summary",
+                params={"timezone": "America/New_York", "cameras": "all"},
+            )
+
+            assert response.status_code == 200
+            summary = response.json()
+
+            # The recording should appear on March 10
+            assert len(summary) == 1
+            assert "2024-03-10" in summary
+            assert summary["2024-03-10"] is True
+
+    def test_recordings_summary_utc_timezone(self):
+        """
+        Test recordings summary with UTC timezone (no DST).
+        """
+        # Use UTC timestamps directly
+        march_9_utc = datetime(2024, 3, 9, 17, 0, 0, tzinfo=timezone.utc).timestamp()
+        march_10_utc = datetime(2024, 3, 10, 17, 0, 0, tzinfo=timezone.utc).timestamp()
+
+        with TestClient(self.app) as client:
+            Recordings.insert(
+                id="recording_march_9_utc",
+                path="/media/recordings/march_9_utc.mp4",
+                camera="front_door",
+                start_time=march_9_utc,
+                end_time=march_9_utc + 3600,
+                duration=3600,
+                motion=100,
+                objects=5,
+            ).execute()
+
+            Recordings.insert(
+                id="recording_march_10_utc",
+                path="/media/recordings/march_10_utc.mp4",
+                camera="front_door",
+                start_time=march_10_utc,
+                end_time=march_10_utc + 3600,
+                duration=3600,
+                motion=150,
+                objects=8,
+            ).execute()
+
+            # Test with UTC timezone
+            response = client.get(
+                "/recordings/summary", params={"timezone": "utc", "cameras": "all"}
+            )
+
+            assert response.status_code == 200
+            summary = response.json()
+
+            # Verify we get both days
+            assert len(summary) == 2
+            assert "2024-03-09" in summary
+            assert "2024-03-10" in summary
+            assert summary["2024-03-09"] is True
+            assert summary["2024-03-10"] is True
+
+    def test_recordings_summary_no_recordings(self):
+        """
+        Test recordings summary when no recordings exist.
+        """
+        with TestClient(self.app) as client:
+            response = client.get(
+                "/recordings/summary",
+                params={"timezone": "America/New_York", "cameras": "all"},
+            )
+
+            assert response.status_code == 200
+            summary = response.json()
+            assert len(summary) == 0
+
+    def test_recordings_summary_single_camera_filter(self):
+        """
+        Test recordings summary filtered to a single camera.
+        """
+        tz = pytz.timezone("America/New_York")
+        march_10_noon = tz.localize(datetime(2024, 3, 10, 12, 0, 0)).timestamp()
+
+        with TestClient(self.app) as client:
+            # Insert recordings for both cameras
+            Recordings.insert(
+                id="front_recording",
+                path="/media/recordings/front.mp4",
+                camera="front_door",
+                start_time=march_10_noon,
+                end_time=march_10_noon + 3600,
+                duration=3600,
+                motion=100,
+                objects=5,
+            ).execute()
+
+            Recordings.insert(
+                id="back_recording",
+                path="/media/recordings/back.mp4",
+                camera="back_door",
+                start_time=march_10_noon,
+                end_time=march_10_noon + 3600,
+                duration=3600,
+                motion=150,
+                objects=8,
+            ).execute()
+
+            # Test with only front_door camera
+            response = client.get(
+                "/recordings/summary",
+                params={"timezone": "America/New_York", "cameras": "front_door"},
+            )
+
+            assert response.status_code == 200
+            summary = response.json()
+            assert len(summary) == 1
+            assert "2024-03-10" in summary
+            assert summary["2024-03-10"] is True

frigate/timeline.py

@@ -142,6 +142,14 @@ class TimelineProcessor(threading.Thread):
                 timeline_entry[Timeline.data]["attribute"] = list(
                     event_data["attributes"].keys()
                 )[0]
+
+                if len(event_data["current_attributes"]) > 0:
+                    timeline_entry[Timeline.data]["attribute_box"] = to_relative_box(
+                        camera_config.detect.width,
+                        camera_config.detect.height,
+                        event_data["current_attributes"][0]["box"],
+                    )
+
                 save = True
         elif event_type == EventStateEnum.end:
             timeline_entry[Timeline.class_type] = "gone"

frigate/types.py

@@ -23,9 +23,11 @@ class ModelStatusTypesEnum(str, Enum):
     error = "error"
     training = "training"
     complete = "complete"
+    failed = "failed"
 
 
 class TrackedObjectUpdateTypesEnum(str, Enum):
     description = "description"
     face = "face"
     lpr = "lpr"
+    classification = "classification"

frigate/util/builtin.py

@@ -15,12 +15,9 @@ from collections.abc import Mapping
 from multiprocessing.sharedctypes import Synchronized
 from pathlib import Path
 from typing import Any, Dict, Optional, Tuple, Union
-from zoneinfo import ZoneInfoNotFoundError
 
 import numpy as np
-import pytz
 from ruamel.yaml import YAML
-from tzlocal import get_localzone
 
 from frigate.const import REGEX_HTTP_CAMERA_USER_PASS, REGEX_RTSP_CAMERA_USER_PASS
 
@@ -157,17 +154,6 @@ def load_labels(path: Optional[str], encoding="utf-8", prefill=91):
         return labels
 
 
-def get_tz_modifiers(tz_name: str) -> Tuple[str, str, float]:
-    seconds_offset = (
-        datetime.datetime.now(pytz.timezone(tz_name)).utcoffset().total_seconds()
-    )
-    hours_offset = int(seconds_offset / 60 / 60)
-    minutes_offset = int(seconds_offset / 60 - hours_offset * 60)
-    hour_modifier = f"{hours_offset} hour"
-    minute_modifier = f"{minutes_offset} minute"
-    return hour_modifier, minute_modifier, seconds_offset
-
-
 def to_relative_box(
     width: int, height: int, box: Tuple[int, int, int, int]
 ) -> Tuple[int | float, int | float, int | float, int | float]:
@@ -298,34 +284,6 @@ def find_by_key(dictionary, target_key):
     return None
 
 
-def get_tomorrow_at_time(hour: int) -> datetime.datetime:
-    """Returns the datetime of the following day at 2am."""
-    try:
-        tomorrow = datetime.datetime.now(get_localzone()) + datetime.timedelta(days=1)
-    except ZoneInfoNotFoundError:
-        tomorrow = datetime.datetime.now(datetime.timezone.utc) + datetime.timedelta(
-            days=1
-        )
-        logger.warning(
-            "Using utc for maintenance due to missing or incorrect timezone set"
-        )
-
-    return tomorrow.replace(hour=hour, minute=0, second=0).astimezone(
-        datetime.timezone.utc
-    )
-
-
-def is_current_hour(timestamp: int) -> bool:
-    """Returns if timestamp is in the current UTC hour."""
-    start_of_next_hour = (
-        datetime.datetime.now(datetime.timezone.utc).replace(
-            minute=0, second=0, microsecond=0
-        )
-        + datetime.timedelta(hours=1)
-    ).timestamp()
-    return timestamp < start_of_next_hour
-
-
 def clear_and_unlink(file: Path, missing_ok: bool = True) -> None:
     """clear file then unlink to avoid space retained by file descriptors."""
     if not missing_ok and not file.exists():

frigate/util/classification.py

@@ -1,13 +1,18 @@
 """Util for classification models."""
 
+import datetime
+import json
 import logging
 import os
+import random
+from collections import defaultdict
 
 import cv2
 import numpy as np
 
 from frigate.comms.embeddings_updater import EmbeddingsRequestEnum, EmbeddingsRequestor
 from frigate.comms.inter_process import InterProcessRequestor
+from frigate.config import FfmpegConfig
 from frigate.const import (
     CLIPS_DIR,
     MODEL_CACHE_DIR,
@@ -15,16 +20,105 @@ from frigate.const import (
     UPDATE_MODEL_STATE,
 )
 from frigate.log import redirect_output_to_logger
+from frigate.models import Event, Recordings, ReviewSegment
 from frigate.types import ModelStatusTypesEnum
+from frigate.util.file import get_event_thumbnail_bytes
+from frigate.util.image import get_image_from_recording
 from frigate.util.process import FrigateProcess
 
 BATCH_SIZE = 16
 EPOCHS = 50
 LEARNING_RATE = 0.001
+TRAINING_METADATA_FILE = ".training_metadata.json"
 
 logger = logging.getLogger(__name__)
 
 
+def write_training_metadata(model_name: str, image_count: int) -> None:
+    """
+    Write training metadata to a hidden file in the model's clips directory.
+
+    Args:
+        model_name: Name of the classification model
+        image_count: Number of images used in training
+    """
+    clips_model_dir = os.path.join(CLIPS_DIR, model_name)
+    os.makedirs(clips_model_dir, exist_ok=True)
+
+    metadata_path = os.path.join(clips_model_dir, TRAINING_METADATA_FILE)
+    metadata = {
+        "last_training_date": datetime.datetime.now().isoformat(),
+        "last_training_image_count": image_count,
+    }
+
+    try:
+        with open(metadata_path, "w") as f:
+            json.dump(metadata, f, indent=2)
+        logger.info(f"Wrote training metadata for {model_name}: {image_count} images")
+    except Exception as e:
+        logger.error(f"Failed to write training metadata for {model_name}: {e}")
+
+
+def read_training_metadata(model_name: str) -> dict[str, any] | None:
+    """
+    Read training metadata from the hidden file in the model's clips directory.
+
+    Args:
+        model_name: Name of the classification model
+
+    Returns:
+        Dictionary with last_training_date and last_training_image_count, or None if not found
+    """
+    clips_model_dir = os.path.join(CLIPS_DIR, model_name)
+    metadata_path = os.path.join(clips_model_dir, TRAINING_METADATA_FILE)
+
+    if not os.path.exists(metadata_path):
+        return None
+
+    try:
+        with open(metadata_path, "r") as f:
+            metadata = json.load(f)
+        return metadata
+    except Exception as e:
+        logger.error(f"Failed to read training metadata for {model_name}: {e}")
+        return None
+
+
+def get_dataset_image_count(model_name: str) -> int:
+    """
+    Count the total number of images in the model's dataset directory.
+
+    Args:
+        model_name: Name of the classification model
+
+    Returns:
+        Total count of images across all categories
+    """
+    dataset_dir = os.path.join(CLIPS_DIR, model_name, "dataset")
+
+    if not os.path.exists(dataset_dir):
+        return 0
+
+    total_count = 0
+    try:
+        for category in os.listdir(dataset_dir):
+            category_dir = os.path.join(dataset_dir, category)
+            if not os.path.isdir(category_dir):
+                continue
+
+            image_files = [
+                f
+                for f in os.listdir(category_dir)
+                if f.lower().endswith((".webp", ".png", ".jpg", ".jpeg"))
+            ]
+            total_count += len(image_files)
+    except Exception as e:
+        logger.error(f"Failed to count dataset images for {model_name}: {e}")
+        return 0
+
+    return total_count
+
+
 class ClassificationTrainingProcess(FrigateProcess):
     def __init__(self, model_name: str) -> None:
         super().__init__(
@@ -36,7 +130,8 @@ class ClassificationTrainingProcess(FrigateProcess):
 
     def run(self) -> None:
         self.pre_run_setup()
-        self.__train_classification_model()
+        success = self.__train_classification_model()
+        exit(0 if success else 1)
 
     def __generate_representative_dataset_factory(self, dataset_dir: str):
         def generate_representative_dataset():
@@ -59,87 +154,119 @@ class ClassificationTrainingProcess(FrigateProcess):
     @redirect_output_to_logger(logger, logging.DEBUG)
     def __train_classification_model(self) -> bool:
         """Train a classification model."""
+        try:
+            # import in the function so that tensorflow is not initialized multiple times
+            import tensorflow as tf
+            from tensorflow.keras import layers, models, optimizers
+            from tensorflow.keras.applications import MobileNetV2
+            from tensorflow.keras.preprocessing.image import ImageDataGenerator
 
-        # import in the function so that tensorflow is not initialized multiple times
+            dataset_dir = os.path.join(CLIPS_DIR, self.model_name, "dataset")
-        import tensorflow as tf
+            model_dir = os.path.join(MODEL_CACHE_DIR, self.model_name)
-        from tensorflow.keras import layers, models, optimizers
+            os.makedirs(model_dir, exist_ok=True)
-        from tensorflow.keras.applications import MobileNetV2
-        from tensorflow.keras.preprocessing.image import ImageDataGenerator
 
-        logger.info(f"Kicking off classification training for {self.model_name}.")
+            num_classes = len(
-        dataset_dir = os.path.join(CLIPS_DIR, self.model_name, "dataset")
+                [
-        model_dir = os.path.join(MODEL_CACHE_DIR, self.model_name)
+                    d
-        num_classes = len(
+                    for d in os.listdir(dataset_dir)
-            [
+                    if os.path.isdir(os.path.join(dataset_dir, d))
-                d
+                ]
-                for d in os.listdir(dataset_dir)
+            )
-                if os.path.isdir(os.path.join(dataset_dir, d))
-            ]
-        )
 
-        # Start with imagenet base model with 35% of channels in each layer
+            if num_classes < 2:
-        base_model = MobileNetV2(
+                logger.error(
-            input_shape=(224, 224, 3),
+                    f"Training failed for {self.model_name}: Need at least 2 classes, found {num_classes}"
-            include_top=False,
+                )
-            weights="imagenet",
+                return False
-            alpha=0.35,
-        )
-        base_model.trainable = False  # Freeze pre-trained layers
 
-        model = models.Sequential(
+            # Start with imagenet base model with 35% of channels in each layer
-            [
+            base_model = MobileNetV2(
-                base_model,
+                input_shape=(224, 224, 3),
-                layers.GlobalAveragePooling2D(),
+                include_top=False,
-                layers.Dense(128, activation="relu"),
+                weights="imagenet",
-                layers.Dropout(0.3),
+                alpha=0.35,
-                layers.Dense(num_classes, activation="softmax"),
+            )
-            ]
+            base_model.trainable = False  # Freeze pre-trained layers
-        )
 
-        model.compile(
+            model = models.Sequential(
-            optimizer=optimizers.Adam(learning_rate=LEARNING_RATE),
+                [
-            loss="categorical_crossentropy",
+                    base_model,
-            metrics=["accuracy"],
+                    layers.GlobalAveragePooling2D(),
-        )
+                    layers.Dense(128, activation="relu"),
+                    layers.Dropout(0.3),
+                    layers.Dense(num_classes, activation="softmax"),
+                ]
+            )
 
-        # create training set
+            model.compile(
-        datagen = ImageDataGenerator(rescale=1.0 / 255, validation_split=0.2)
+                optimizer=optimizers.Adam(learning_rate=LEARNING_RATE),
-        train_gen = datagen.flow_from_directory(
+                loss="categorical_crossentropy",
-            dataset_dir,
+                metrics=["accuracy"],
-            target_size=(224, 224),
+            )
-            batch_size=BATCH_SIZE,
-            class_mode="categorical",
-            subset="training",
-        )
 
-        # write labelmap
+            # create training set
-        class_indices = train_gen.class_indices
+            datagen = ImageDataGenerator(rescale=1.0 / 255, validation_split=0.2)
-        index_to_class = {v: k for k, v in class_indices.items()}
+            train_gen = datagen.flow_from_directory(
-        sorted_classes = [index_to_class[i] for i in range(len(index_to_class))]
+                dataset_dir,
-        with open(os.path.join(model_dir, "labelmap.txt"), "w") as f:
+                target_size=(224, 224),
-            for class_name in sorted_classes:
+                batch_size=BATCH_SIZE,
-                f.write(f"{class_name}\n")
+                class_mode="categorical",
+                subset="training",
+            )
 
-        # train the model
+            total_images = train_gen.samples
-        model.fit(train_gen, epochs=EPOCHS, verbose=0)
+            logger.debug(
+                f"Training {self.model_name}: {total_images} images across {num_classes} classes"
+            )
 
-        # convert model to tflite
+            # write labelmap
-        converter = tf.lite.TFLiteConverter.from_keras_model(model)
+            class_indices = train_gen.class_indices
-        converter.optimizations = [tf.lite.Optimize.DEFAULT]
+            index_to_class = {v: k for k, v in class_indices.items()}
-        converter.representative_dataset = (
+            sorted_classes = [index_to_class[i] for i in range(len(index_to_class))]
-            self.__generate_representative_dataset_factory(dataset_dir)
+            with open(os.path.join(model_dir, "labelmap.txt"), "w") as f:
-        )
+                for class_name in sorted_classes:
-        converter.target_spec.supported_ops = [tf.lite.OpsSet.TFLITE_BUILTINS_INT8]
+                    f.write(f"{class_name}\n")
-        converter.inference_input_type = tf.uint8
-        converter.inference_output_type = tf.uint8
-        tflite_model = converter.convert()
 
-        # write model
+            # train the model
-        with open(os.path.join(model_dir, "model.tflite"), "wb") as f:
+            logger.debug(f"Training {self.model_name} for {EPOCHS} epochs...")
-            f.write(tflite_model)
+            model.fit(train_gen, epochs=EPOCHS, verbose=0)
+            logger.debug(f"Converting {self.model_name} to TFLite...")
+
+            # convert model to tflite
+            converter = tf.lite.TFLiteConverter.from_keras_model(model)
+            converter.optimizations = [tf.lite.Optimize.DEFAULT]
+            converter.representative_dataset = (
+                self.__generate_representative_dataset_factory(dataset_dir)
+            )
+            converter.target_spec.supported_ops = [tf.lite.OpsSet.TFLITE_BUILTINS_INT8]
+            converter.inference_input_type = tf.uint8
+            converter.inference_output_type = tf.uint8
+            tflite_model = converter.convert()
+
+            # write model
+            model_path = os.path.join(model_dir, "model.tflite")
+            with open(model_path, "wb") as f:
+                f.write(tflite_model)
+
+            # verify model file was written successfully
+            if not os.path.exists(model_path) or os.path.getsize(model_path) == 0:
+                logger.error(
+                    f"Training failed for {self.model_name}: Model file was not created or is empty"
+                )
+                return False
+
+            # write training metadata with image count
+            dataset_image_count = get_dataset_image_count(self.model_name)
+            write_training_metadata(self.model_name, dataset_image_count)
+
+            logger.info(f"Finished training {self.model_name}")
+            return True
+
+        except Exception as e:
+            logger.error(f"Training failed for {self.model_name}: {e}", exc_info=True)
+            return False
 
 
-@staticmethod
 def kickoff_model_training(
     embeddingRequestor: EmbeddingsRequestor, model_name: str
 ) -> None:
@@ -159,16 +286,551 @@ def kickoff_model_training(
     training_process.start()
     training_process.join()
 
-    # reload model and mark training as complete
+    # check if training succeeded by examining the exit code
-    embeddingRequestor.send_data(
+    training_success = training_process.exitcode == 0
-        EmbeddingsRequestEnum.reload_classification_model.value,
+
-        {"model_name": model_name},
+    if training_success:
-    )
+        # reload model and mark training as complete
-    requestor.send_data(
+        embeddingRequestor.send_data(
-        UPDATE_MODEL_STATE,
+            EmbeddingsRequestEnum.reload_classification_model.value,
-        {
+            {"model_name": model_name},
-            "model": model_name,
+        )
-            "state": ModelStatusTypesEnum.complete,
+        requestor.send_data(
-        },
+            UPDATE_MODEL_STATE,
-    )
+            {
+                "model": model_name,
+                "state": ModelStatusTypesEnum.complete,
+            },
+        )
+    else:
+        logger.error(
+            f"Training subprocess failed for {model_name} (exit code: {training_process.exitcode})"
+        )
+        # mark training as failed so UI shows error state
+        # don't reload the model since it failed
+        requestor.send_data(
+            UPDATE_MODEL_STATE,
+            {
+                "model": model_name,
+                "state": ModelStatusTypesEnum.failed,
+            },
+        )
+
     requestor.stop()
+
+
+@staticmethod
+def collect_state_classification_examples(
+    model_name: str, cameras: dict[str, tuple[float, float, float, float]]
+) -> None:
+    """
+    Collect representative state classification examples from review items.
+
+    This function:
+    1. Queries review items from specified cameras
+    2. Selects 100 balanced timestamps across the data
+    3. Extracts keyframes from recordings (cropped to specified regions)
+    4. Selects 20 most visually distinct images
+    5. Saves them to the dataset directory
+
+    Args:
+        model_name: Name of the classification model
+        cameras: Dict mapping camera names to normalized crop coordinates [x1, y1, x2, y2] (0-1)
+    """
+    dataset_dir = os.path.join(CLIPS_DIR, model_name, "dataset")
+    temp_dir = os.path.join(dataset_dir, "temp")
+    os.makedirs(temp_dir, exist_ok=True)
+
+    # Step 1: Get review items for the cameras
+    camera_names = list(cameras.keys())
+    review_items = list(
+        ReviewSegment.select()
+        .where(ReviewSegment.camera.in_(camera_names))
+        .where(ReviewSegment.end_time.is_null(False))
+        .order_by(ReviewSegment.start_time.asc())
+    )
+
+    if not review_items:
+        logger.warning(f"No review items found for cameras: {camera_names}")
+        return
+
+    # Step 2: Create balanced timestamp selection (100 samples)
+    timestamps = _select_balanced_timestamps(review_items, target_count=100)
+
+    # Step 3: Extract keyframes from recordings with crops applied
+    keyframes = _extract_keyframes(
+        "/usr/lib/ffmpeg/7.0/bin/ffmpeg", timestamps, temp_dir, cameras
+    )
+
+    # Step 4: Select 24 most visually distinct images (they're already cropped)
+    distinct_images = _select_distinct_images(keyframes, target_count=24)
+
+    # Step 5: Save to train directory for later classification
+    train_dir = os.path.join(CLIPS_DIR, model_name, "train")
+    os.makedirs(train_dir, exist_ok=True)
+
+    saved_count = 0
+    for idx, image_path in enumerate(distinct_images):
+        dest_path = os.path.join(train_dir, f"example_{idx:03d}.jpg")
+        try:
+            img = cv2.imread(image_path)
+
+            if img is not None:
+                cv2.imwrite(dest_path, img)
+                saved_count += 1
+        except Exception as e:
+            logger.error(f"Failed to save image {image_path}: {e}")
+
+    import shutil
+
+    try:
+        shutil.rmtree(temp_dir)
+    except Exception as e:
+        logger.warning(f"Failed to clean up temp directory: {e}")
+
+
+def _select_balanced_timestamps(
+    review_items: list[ReviewSegment], target_count: int = 100
+) -> list[dict]:
+    """
+    Select balanced timestamps from review items.
+
+    Strategy:
+    - Group review items by camera and time of day
+    - Sample evenly across groups to ensure diversity
+    - For each selected review item, pick a random timestamp within its duration
+
+    Returns:
+        List of dicts with keys: camera, timestamp, review_item
+    """
+    # Group by camera and hour of day for temporal diversity
+    grouped = defaultdict(list)
+
+    for item in review_items:
+        camera = item.camera
+        # Group by 6-hour blocks for temporal diversity
+        hour_block = int(item.start_time // (6 * 3600))
+        key = f"{camera}_{hour_block}"
+        grouped[key].append(item)
+
+    # Calculate how many samples per group
+    num_groups = len(grouped)
+    if num_groups == 0:
+        return []
+
+    samples_per_group = max(1, target_count // num_groups)
+    timestamps = []
+
+    # Sample from each group
+    for group_items in grouped.values():
+        # Take samples_per_group items from this group
+        sample_size = min(samples_per_group, len(group_items))
+        sampled_items = random.sample(group_items, sample_size)
+
+        for item in sampled_items:
+            # Pick a random timestamp within the review item's duration
+            duration = item.end_time - item.start_time
+            if duration <= 0:
+                continue
+
+            # Sample from middle 80% to avoid edge artifacts
+            offset = random.uniform(duration * 0.1, duration * 0.9)
+            timestamp = item.start_time + offset
+
+            timestamps.append(
+                {
+                    "camera": item.camera,
+                    "timestamp": timestamp,
+                    "review_item": item,
+                }
+            )
+
+    # If we don't have enough, sample more from larger groups
+    while len(timestamps) < target_count and len(timestamps) < len(review_items):
+        for group_items in grouped.values():
+            if len(timestamps) >= target_count:
+                break
+
+            # Pick a random item not already sampled
+            item = random.choice(group_items)
+            duration = item.end_time - item.start_time
+            if duration <= 0:
+                continue
+
+            offset = random.uniform(duration * 0.1, duration * 0.9)
+            timestamp = item.start_time + offset
+
+            # Check if we already have a timestamp near this one
+            if not any(abs(t["timestamp"] - timestamp) < 1.0 for t in timestamps):
+                timestamps.append(
+                    {
+                        "camera": item.camera,
+                        "timestamp": timestamp,
+                        "review_item": item,
+                    }
+                )
+
+    return timestamps[:target_count]
+
+
+def _extract_keyframes(
+    ffmpeg_path: str,
+    timestamps: list[dict],
+    output_dir: str,
+    camera_crops: dict[str, tuple[float, float, float, float]],
+) -> list[str]:
+    """
+    Extract keyframes from recordings at specified timestamps and crop to specified regions.
+
+    Args:
+        ffmpeg_path: Path to ffmpeg binary
+        timestamps: List of timestamp dicts from _select_balanced_timestamps
+        output_dir: Directory to save extracted frames
+        camera_crops: Dict mapping camera names to normalized crop coordinates [x1, y1, x2, y2] (0-1)
+
+    Returns:
+        List of paths to successfully extracted and cropped keyframe images
+    """
+    keyframe_paths = []
+
+    for idx, ts_info in enumerate(timestamps):
+        camera = ts_info["camera"]
+        timestamp = ts_info["timestamp"]
+
+        if camera not in camera_crops:
+            logger.warning(f"No crop coordinates for camera {camera}")
+            continue
+
+        norm_x1, norm_y1, norm_x2, norm_y2 = camera_crops[camera]
+
+        try:
+            recording = (
+                Recordings.select()
+                .where(
+                    (timestamp >= Recordings.start_time)
+                    & (timestamp <= Recordings.end_time)
+                    & (Recordings.camera == camera)
+                )
+                .order_by(Recordings.start_time.desc())
+                .limit(1)
+                .get()
+            )
+        except Exception:
+            continue
+
+        relative_time = timestamp - recording.start_time
+
+        try:
+            config = FfmpegConfig(path="/usr/lib/ffmpeg/7.0")
+            image_data = get_image_from_recording(
+                config,
+                recording.path,
+                relative_time,
+                codec="mjpeg",
+                height=None,
+            )
+
+            if image_data:
+                nparr = np.frombuffer(image_data, np.uint8)
+                img = cv2.imdecode(nparr, cv2.IMREAD_COLOR)
+
+                if img is not None:
+                    height, width = img.shape[:2]
+
+                    x1 = int(norm_x1 * width)
+                    y1 = int(norm_y1 * height)
+                    x2 = int(norm_x2 * width)
+                    y2 = int(norm_y2 * height)
+
+                    x1_clipped = max(0, min(x1, width))
+                    y1_clipped = max(0, min(y1, height))
+                    x2_clipped = max(0, min(x2, width))
+                    y2_clipped = max(0, min(y2, height))
+
+                    if x2_clipped > x1_clipped and y2_clipped > y1_clipped:
+                        cropped = img[y1_clipped:y2_clipped, x1_clipped:x2_clipped]
+                        resized = cv2.resize(cropped, (224, 224))
+
+                        output_path = os.path.join(output_dir, f"frame_{idx:04d}.jpg")
+                        cv2.imwrite(output_path, resized)
+                        keyframe_paths.append(output_path)
+
+        except Exception as e:
+            logger.debug(
+                f"Failed to extract frame from {recording.path} at {relative_time}s: {e}"
+            )
+            continue
+
+    return keyframe_paths
+
+
+def _select_distinct_images(
+    image_paths: list[str], target_count: int = 20
+) -> list[str]:
+    """
+    Select the most visually distinct images from a set of keyframes.
+
+    Uses a greedy algorithm based on image histograms:
+    1. Start with a random image
+    2. Iteratively add the image that is most different from already selected images
+    3. Difference is measured using histogram comparison
+
+    Args:
+        image_paths: List of paths to candidate images
+        target_count: Number of distinct images to select
+
+    Returns:
+        List of paths to selected images
+    """
+    if len(image_paths) <= target_count:
+        return image_paths
+
+    histograms = {}
+    valid_paths = []
+
+    for path in image_paths:
+        try:
+            img = cv2.imread(path)
+
+            if img is None:
+                continue
+
+            hsv = cv2.cvtColor(img, cv2.COLOR_BGR2HSV)
+            hist = cv2.calcHist(
+                [hsv], [0, 1, 2], None, [8, 8, 8], [0, 180, 0, 256, 0, 256]
+            )
+            hist = cv2.normalize(hist, hist).flatten()
+            histograms[path] = hist
+            valid_paths.append(path)
+        except Exception as e:
+            logger.debug(f"Failed to process image {path}: {e}")
+            continue
+
+    if len(valid_paths) <= target_count:
+        return valid_paths
+
+    selected = []
+    first_image = random.choice(valid_paths)
+    selected.append(first_image)
+    remaining = [p for p in valid_paths if p != first_image]
+
+    while len(selected) < target_count and remaining:
+        max_min_distance = -1
+        best_candidate = None
+
+        for candidate in remaining:
+            min_distance = float("inf")
+
+            for selected_img in selected:
+                distance = cv2.compareHist(
+                    histograms[candidate],
+                    histograms[selected_img],
+                    cv2.HISTCMP_BHATTACHARYYA,
+                )
+                min_distance = min(min_distance, distance)
+
+            if min_distance > max_min_distance:
+                max_min_distance = min_distance
+                best_candidate = candidate
+
+        if best_candidate:
+            selected.append(best_candidate)
+            remaining.remove(best_candidate)
+        else:
+            break
+
+    return selected
+
+
+@staticmethod
+def collect_object_classification_examples(
+    model_name: str,
+    label: str,
+) -> None:
+    """
+    Collect representative object classification examples from event thumbnails.
+
+    This function:
+    1. Queries events for the specified label
+    2. Selects 100 balanced events across different cameras and times
+    3. Retrieves thumbnails for selected events (with 33% center crop applied)
+    4. Selects 24 most visually distinct thumbnails
+    5. Saves to dataset directory
+
+    Args:
+        model_name: Name of the classification model
+        label: Object label to collect (e.g., "person", "car")
+        cameras: List of camera names to collect examples from
+    """
+    dataset_dir = os.path.join(CLIPS_DIR, model_name, "dataset")
+    temp_dir = os.path.join(dataset_dir, "temp")
+    os.makedirs(temp_dir, exist_ok=True)
+
+    # Step 1: Query events for the specified label and cameras
+    events = list(
+        Event.select().where((Event.label == label)).order_by(Event.start_time.asc())
+    )
+
+    if not events:
+        logger.warning(f"No events found for label '{label}'")
+        return
+
+    logger.debug(f"Found {len(events)} events")
+
+    # Step 2: Select balanced events (100 samples)
+    selected_events = _select_balanced_events(events, target_count=100)
+    logger.debug(f"Selected {len(selected_events)} events")
+
+    # Step 3: Extract thumbnails from events
+    thumbnails = _extract_event_thumbnails(selected_events, temp_dir)
+    logger.debug(f"Successfully extracted {len(thumbnails)} thumbnails")
+
+    # Step 4: Select 24 most visually distinct thumbnails
+    distinct_images = _select_distinct_images(thumbnails, target_count=24)
+    logger.debug(f"Selected {len(distinct_images)} distinct images")
+
+    # Step 5: Save to train directory for later classification
+    train_dir = os.path.join(CLIPS_DIR, model_name, "train")
+    os.makedirs(train_dir, exist_ok=True)
+
+    saved_count = 0
+    for idx, image_path in enumerate(distinct_images):
+        dest_path = os.path.join(train_dir, f"example_{idx:03d}.jpg")
+        try:
+            img = cv2.imread(image_path)
+
+            if img is not None:
+                cv2.imwrite(dest_path, img)
+                saved_count += 1
+        except Exception as e:
+            logger.error(f"Failed to save image {image_path}: {e}")
+
+    import shutil
+
+    try:
+        shutil.rmtree(temp_dir)
+    except Exception as e:
+        logger.warning(f"Failed to clean up temp directory: {e}")
+
+    logger.debug(
+        f"Successfully collected {saved_count} classification examples in {train_dir}"
+    )
+
+
+def _select_balanced_events(
+    events: list[Event], target_count: int = 100
+) -> list[Event]:
+    """
+    Select balanced events from the event list.
+
+    Strategy:
+    - Group events by camera and time of day
+    - Sample evenly across groups to ensure diversity
+    - Prioritize events with higher scores
+
+    Returns:
+        List of selected events
+    """
+    grouped = defaultdict(list)
+
+    for event in events:
+        camera = event.camera
+        hour_block = int(event.start_time // (6 * 3600))
+        key = f"{camera}_{hour_block}"
+        grouped[key].append(event)
+
+    num_groups = len(grouped)
+    if num_groups == 0:
+        return []
+
+    samples_per_group = max(1, target_count // num_groups)
+    selected = []
+
+    for group_events in grouped.values():
+        sorted_events = sorted(
+            group_events,
+            key=lambda e: e.data.get("score", 0) if e.data else 0,
+            reverse=True,
+        )
+
+        sample_size = min(samples_per_group, len(sorted_events))
+        selected.extend(sorted_events[:sample_size])
+
+    if len(selected) < target_count:
+        remaining = [e for e in events if e not in selected]
+        remaining_sorted = sorted(
+            remaining,
+            key=lambda e: e.data.get("score", 0) if e.data else 0,
+            reverse=True,
+        )
+        needed = target_count - len(selected)
+        selected.extend(remaining_sorted[:needed])
+
+    return selected[:target_count]
+
+
+def _extract_event_thumbnails(events: list[Event], output_dir: str) -> list[str]:
+    """
+    Extract thumbnails from events and save to disk.
+
+    Args:
+        events: List of Event objects
+        output_dir: Directory to save thumbnails
+
+    Returns:
+        List of paths to successfully extracted thumbnail images
+    """
+    thumbnail_paths = []
+
+    for idx, event in enumerate(events):
+        try:
+            thumbnail_bytes = get_event_thumbnail_bytes(event)
+
+            if thumbnail_bytes:
+                nparr = np.frombuffer(thumbnail_bytes, np.uint8)
+                img = cv2.imdecode(nparr, cv2.IMREAD_COLOR)
+
+                if img is not None:
+                    height, width = img.shape[:2]
+
+                    crop_size = 1.0
+                    if event.data and "box" in event.data and "region" in event.data:
+                        box = event.data["box"]
+                        region = event.data["region"]
+
+                        if len(box) == 4 and len(region) == 4:
+                            box_w, box_h = box[2], box[3]
+                            region_w, region_h = region[2], region[3]
+
+                            box_area = (box_w * box_h) / (region_w * region_h)
+
+                            if box_area < 0.05:
+                                crop_size = 0.4
+                            elif box_area < 0.10:
+                                crop_size = 0.5
+                            elif box_area < 0.20:
+                                crop_size = 0.65
+                            elif box_area < 0.35:
+                                crop_size = 0.80
+                            else:
+                                crop_size = 0.95
+
+                    crop_width = int(width * crop_size)
+                    crop_height = int(height * crop_size)
+
+                    x1 = (width - crop_width) // 2
+                    y1 = (height - crop_height) // 2
+                    x2 = x1 + crop_width
+                    y2 = y1 + crop_height
+
+                    cropped = img[y1:y2, x1:x2]
+                    resized = cv2.resize(cropped, (224, 224))
+                    output_path = os.path.join(output_dir, f"thumbnail_{idx:04d}.jpg")
+                    cv2.imwrite(output_path, resized)
+                    thumbnail_paths.append(output_path)
+
+        except Exception as e:
+            logger.debug(f"Failed to extract thumbnail for event {event.id}: {e}")
+            continue
+
+    return thumbnail_paths

frigate/util/config.py

@@ -384,10 +384,10 @@ def migrate_017_0(config: dict[str, dict[str, Any]]) -> dict[str, dict[str, Any]
         new_object_config["genai"] = {}
 
         for key in global_genai.keys():
-            if key not in ["enabled", "model", "provider", "base_url", "api_key"]:
+            if key in ["model", "provider", "base_url", "api_key"]:
-                new_object_config["genai"][key] = global_genai[key]
-            else:
                 new_genai_config[key] = global_genai[key]
+            else:
+                new_object_config["genai"][key] = global_genai[key]
 
         config["genai"] = new_genai_config
 

frigate/util/downloader.py

@@ -1,7 +1,6 @@
 import logging
 import os
 import threading
-import time
 from pathlib import Path
 from typing import Callable, List
 
@@ -10,40 +9,11 @@ import requests
 from frigate.comms.inter_process import InterProcessRequestor
 from frigate.const import UPDATE_MODEL_STATE
 from frigate.types import ModelStatusTypesEnum
+from frigate.util.file import FileLock
 
 logger = logging.getLogger(__name__)
 
 
-class FileLock:
-    def __init__(self, path):
-        self.path = path
-        self.lock_file = f"{path}.lock"
-
-        # we have not acquired the lock yet so it should not exist
-        if os.path.exists(self.lock_file):
-            try:
-                os.remove(self.lock_file)
-            except Exception:
-                pass
-
-    def acquire(self):
-        parent_dir = os.path.dirname(self.lock_file)
-        os.makedirs(parent_dir, exist_ok=True)
-
-        while True:
-            try:
-                with open(self.lock_file, "x"):
-                    return
-            except FileExistsError:
-                time.sleep(0.1)
-
-    def release(self):
-        try:
-            os.remove(self.lock_file)
-        except FileNotFoundError:
-            pass
-
-
 class ModelDownloader:
     def __init__(
         self,
@@ -81,15 +51,13 @@ class ModelDownloader:
     def _download_models(self):
         for file_name in self.file_names:
             path = os.path.join(self.download_path, file_name)
-            lock = FileLock(path)
+            lock_path = f"{path}.lock"
+            lock = FileLock(lock_path, cleanup_stale_on_init=True)
 
             if not os.path.exists(path):
-                lock.acquire()
+                with lock:
-                try:
                     if not os.path.exists(path):
                         self.download_func(path)
-                finally:
-                    lock.release()
 
             self.requestor.send_data(
                 UPDATE_MODEL_STATE,

frigate/util/file.py

@@ -0,0 +1,276 @@
+"""Path and file utilities."""
+
+import base64
+import fcntl
+import logging
+import os
+import time
+from pathlib import Path
+from typing import Optional
+
+import cv2
+from numpy import ndarray
+
+from frigate.const import CLIPS_DIR, THUMB_DIR
+from frigate.models import Event
+
+logger = logging.getLogger(__name__)
+
+
+def get_event_thumbnail_bytes(event: Event) -> bytes | None:
+    if event.thumbnail:
+        return base64.b64decode(event.thumbnail)
+    else:
+        try:
+            with open(
+                os.path.join(THUMB_DIR, event.camera, f"{event.id}.webp"), "rb"
+            ) as f:
+                return f.read()
+        except Exception:
+            return None
+
+
+def get_event_snapshot(event: Event) -> ndarray:
+    media_name = f"{event.camera}-{event.id}"
+    return cv2.imread(f"{os.path.join(CLIPS_DIR, media_name)}.jpg")
+
+
+### Deletion
+
+
+def delete_event_images(event: Event) -> bool:
+    return delete_event_snapshot(event) and delete_event_thumbnail(event)
+
+
+def delete_event_snapshot(event: Event) -> bool:
+    media_name = f"{event.camera}-{event.id}"
+    media_path = Path(f"{os.path.join(CLIPS_DIR, media_name)}.jpg")
+
+    try:
+        media_path.unlink(missing_ok=True)
+        media_path = Path(f"{os.path.join(CLIPS_DIR, media_name)}-clean.webp")
+        media_path.unlink(missing_ok=True)
+        # also delete clean.png (legacy) for backward compatibility
+        media_path = Path(f"{os.path.join(CLIPS_DIR, media_name)}-clean.png")
+        media_path.unlink(missing_ok=True)
+        return True
+    except OSError:
+        return False
+
+
+def delete_event_thumbnail(event: Event) -> bool:
+    if event.thumbnail:
+        return True
+    else:
+        Path(os.path.join(THUMB_DIR, event.camera, f"{event.id}.webp")).unlink(
+            missing_ok=True
+        )
+        return True
+
+
+### File Locking
+
+
+class FileLock:
+    """
+    A file-based lock for coordinating access to resources across processes.
+
+    Uses fcntl.flock() for proper POSIX file locking on Linux. Supports timeouts,
+    stale lock detection, and can be used as a context manager.
+
+    Example:
+        ```python
+        # Using as a context manager (recommended)
+        with FileLock("/path/to/resource.lock", timeout=60):
+            # Critical section
+            do_something()
+
+        # Manual acquisition and release
+        lock = FileLock("/path/to/resource.lock")
+        if lock.acquire(timeout=60):
+            try:
+                do_something()
+            finally:
+                lock.release()
+        ```
+
+    Attributes:
+        lock_path: Path to the lock file
+        timeout: Maximum time to wait for lock acquisition (seconds)
+        poll_interval: Time to wait between lock acquisition attempts (seconds)
+        stale_timeout: Time after which a lock is considered stale (seconds)
+    """
+
+    def __init__(
+        self,
+        lock_path: str | Path,
+        timeout: int = 300,
+        poll_interval: float = 1.0,
+        stale_timeout: int = 600,
+        cleanup_stale_on_init: bool = False,
+    ):
+        """
+        Initialize a FileLock.
+
+        Args:
+            lock_path: Path to the lock file
+            timeout: Maximum time to wait for lock acquisition in seconds (default: 300)
+            poll_interval: Time to wait between lock attempts in seconds (default: 1.0)
+            stale_timeout: Time after which a lock is considered stale in seconds (default: 600)
+            cleanup_stale_on_init: Whether to clean up stale locks on initialization (default: False)
+        """
+        self.lock_path = Path(lock_path)
+        self.timeout = timeout
+        self.poll_interval = poll_interval
+        self.stale_timeout = stale_timeout
+        self._fd: Optional[int] = None
+        self._acquired = False
+
+        if cleanup_stale_on_init:
+            self._cleanup_stale_lock()
+
+    def _cleanup_stale_lock(self) -> bool:
+        """
+        Clean up a stale lock file if it exists and is old.
+
+        Returns:
+            True if lock was cleaned up, False otherwise
+        """
+        try:
+            if self.lock_path.exists():
+                # Check if lock file is older than stale_timeout
+                lock_age = time.time() - self.lock_path.stat().st_mtime
+                if lock_age > self.stale_timeout:
+                    logger.warning(
+                        f"Removing stale lock file: {self.lock_path} (age: {lock_age:.1f}s)"
+                    )
+                    self.lock_path.unlink()
+                    return True
+        except Exception as e:
+            logger.error(f"Error cleaning up stale lock: {e}")
+
+        return False
+
+    def is_stale(self) -> bool:
+        """
+        Check if the lock file is stale (older than stale_timeout).
+
+        Returns:
+            True if lock is stale, False otherwise
+        """
+        try:
+            if self.lock_path.exists():
+                lock_age = time.time() - self.lock_path.stat().st_mtime
+                return lock_age > self.stale_timeout
+        except Exception:
+            pass
+
+        return False
+
+    def acquire(self, timeout: Optional[int] = None) -> bool:
+        """
+        Acquire the file lock using fcntl.flock().
+
+        Args:
+            timeout: Maximum time to wait for lock in seconds (uses instance timeout if None)
+
+        Returns:
+            True if lock acquired, False if timeout or error
+        """
+        if self._acquired:
+            logger.warning(f"Lock already acquired: {self.lock_path}")
+            return True
+
+        if timeout is None:
+            timeout = self.timeout
+
+        # Ensure parent directory exists
+        self.lock_path.parent.mkdir(parents=True, exist_ok=True)
+
+        # Clean up stale lock before attempting to acquire
+        self._cleanup_stale_lock()
+
+        try:
+            self._fd = os.open(self.lock_path, os.O_CREAT | os.O_RDWR)
+
+            start_time = time.time()
+            while time.time() - start_time < timeout:
+                try:
+                    fcntl.flock(self._fd, fcntl.LOCK_EX | fcntl.LOCK_NB)
+                    self._acquired = True
+                    logger.debug(f"Acquired lock: {self.lock_path}")
+                    return True
+                except (OSError, IOError):
+                    # Lock is held by another process
+                    if time.time() - start_time >= timeout:
+                        logger.warning(f"Timeout waiting for lock: {self.lock_path}")
+                        os.close(self._fd)
+                        self._fd = None
+                        return False
+
+                    time.sleep(self.poll_interval)
+
+            # Timeout reached
+            if self._fd is not None:
+                os.close(self._fd)
+                self._fd = None
+            return False
+
+        except Exception as e:
+            logger.error(f"Error acquiring lock: {e}")
+            if self._fd is not None:
+                try:
+                    os.close(self._fd)
+                except Exception:
+                    pass
+                self._fd = None
+            return False
+
+    def release(self) -> None:
+        """
+        Release the file lock.
+
+        This closes the file descriptor and removes the lock file.
+        """
+        if not self._acquired:
+            return
+
+        try:
+            # Close file descriptor and release fcntl lock
+            if self._fd is not None:
+                try:
+                    fcntl.flock(self._fd, fcntl.LOCK_UN)
+                    os.close(self._fd)
+                except Exception as e:
+                    logger.warning(f"Error closing lock file descriptor: {e}")
+                finally:
+                    self._fd = None
+
+            # Remove lock file
+            if self.lock_path.exists():
+                self.lock_path.unlink()
+                logger.debug(f"Released lock: {self.lock_path}")
+
+        except FileNotFoundError:
+            # Lock file already removed, that's fine
+            pass
+        except Exception as e:
+            logger.error(f"Error releasing lock: {e}")
+        finally:
+            self._acquired = False
+
+    def __enter__(self):
+        """Context manager entry - acquire the lock."""
+        if not self.acquire():
+            raise TimeoutError(f"Failed to acquire lock: {self.lock_path}")
+        return self
+
+    def __exit__(self, exc_type, exc_val, exc_tb):
+        """Context manager exit - release the lock."""
+        self.release()
+        return False
+
+    def __del__(self):
+        """Destructor - ensure lock is released."""
+        if self._acquired:
+            self.release()

frigate/util/model.py

@@ -369,6 +369,10 @@ def get_ort_providers(
                     "enable_cpu_mem_arena": False,
                 }
             )
+        elif provider == "AzureExecutionProvider":
+            # Skip Azure provider - not typically available on local hardware
+            # and prevents fallback to OpenVINO when it's the first provider
+            continue
         else:
             providers.append(provider)
             options.append({})

frigate/util/path.py

@@ -1,62 +0,0 @@
-"""Path utilities."""
-
-import base64
-import os
-from pathlib import Path
-
-import cv2
-from numpy import ndarray
-
-from frigate.const import CLIPS_DIR, THUMB_DIR
-from frigate.models import Event
-
-
-def get_event_thumbnail_bytes(event: Event) -> bytes | None:
-    if event.thumbnail:
-        return base64.b64decode(event.thumbnail)
-    else:
-        try:
-            with open(
-                os.path.join(THUMB_DIR, event.camera, f"{event.id}.webp"), "rb"
-            ) as f:
-                return f.read()
-        except Exception:
-            return None
-
-
-def get_event_snapshot(event: Event) -> ndarray:
-    media_name = f"{event.camera}-{event.id}"
-    return cv2.imread(f"{os.path.join(CLIPS_DIR, media_name)}.jpg")
-
-
-### Deletion
-
-
-def delete_event_images(event: Event) -> bool:
-    return delete_event_snapshot(event) and delete_event_thumbnail(event)
-
-
-def delete_event_snapshot(event: Event) -> bool:
-    media_name = f"{event.camera}-{event.id}"
-    media_path = Path(f"{os.path.join(CLIPS_DIR, media_name)}.jpg")
-
-    try:
-        media_path.unlink(missing_ok=True)
-        media_path = Path(f"{os.path.join(CLIPS_DIR, media_name)}-clean.webp")
-        media_path.unlink(missing_ok=True)
-        # also delete clean.png (legacy) for backward compatibility
-        media_path = Path(f"{os.path.join(CLIPS_DIR, media_name)}-clean.png")
-        media_path.unlink(missing_ok=True)
-        return True
-    except OSError:
-        return False
-
-
-def delete_event_thumbnail(event: Event) -> bool:
-    if event.thumbnail:
-        return True
-    else:
-        Path(os.path.join(THUMB_DIR, event.camera, f"{event.id}.webp")).unlink(
-            missing_ok=True
-        )
-        return True

frigate/util/rknn_converter.py

@@ -1,6 +1,5 @@
 """RKNN model conversion utility for Frigate."""
 
-import fcntl
 import logging
 import os
 import subprocess
@@ -9,6 +8,8 @@ import time
 from pathlib import Path
 from typing import Optional
 
+from frigate.util.file import FileLock
+
 logger = logging.getLogger(__name__)
 
 MODEL_TYPE_CONFIGS = {
@@ -129,8 +130,13 @@ def get_soc_type() -> Optional[str]:
     """Get the SoC type from device tree."""
     try:
         with open("/proc/device-tree/compatible") as file:
-            soc = file.read().split(",")[-1].strip("\x00")
+            content = file.read()
-            return soc
+
+            # Check for Jetson devices
+            if "nvidia" in content:
+                return None
+
+            return content.split(",")[-1].strip("\x00")
     except FileNotFoundError:
         logger.debug("Could not determine SoC type from device tree")
         return None
@@ -245,112 +251,6 @@ def convert_onnx_to_rknn(
                 logger.warning(f"Failed to remove temporary ONNX file: {e}")
 
 
-def cleanup_stale_lock(lock_file_path: Path) -> bool:
-    """
-    Clean up a stale lock file if it exists and is old.
-
-    Args:
-        lock_file_path: Path to the lock file
-
-    Returns:
-        True if lock was cleaned up, False otherwise
-    """
-    try:
-        if lock_file_path.exists():
-            # Check if lock file is older than 10 minutes (stale)
-            lock_age = time.time() - lock_file_path.stat().st_mtime
-            if lock_age > 600:  # 10 minutes
-                logger.warning(
-                    f"Removing stale lock file: {lock_file_path} (age: {lock_age:.1f}s)"
-                )
-                lock_file_path.unlink()
-                return True
-    except Exception as e:
-        logger.error(f"Error cleaning up stale lock: {e}")
-
-    return False
-
-
-def acquire_conversion_lock(lock_file_path: Path, timeout: int = 300) -> bool:
-    """
-    Acquire a file-based lock for model conversion.
-
-    Args:
-        lock_file_path: Path to the lock file
-        timeout: Maximum time to wait for lock in seconds
-
-    Returns:
-        True if lock acquired, False if timeout or error
-    """
-    try:
-        lock_file_path.parent.mkdir(parents=True, exist_ok=True)
-        cleanup_stale_lock(lock_file_path)
-        lock_fd = os.open(lock_file_path, os.O_CREAT | os.O_RDWR)
-
-        # Try to acquire exclusive lock
-        start_time = time.time()
-        while time.time() - start_time < timeout:
-            try:
-                fcntl.flock(lock_fd, fcntl.LOCK_EX | fcntl.LOCK_NB)
-                # Lock acquired successfully
-                logger.debug(f"Acquired conversion lock: {lock_file_path}")
-                return True
-            except (OSError, IOError):
-                # Lock is held by another process, wait and retry
-                if time.time() - start_time >= timeout:
-                    logger.warning(
-                        f"Timeout waiting for conversion lock: {lock_file_path}"
-                    )
-                    os.close(lock_fd)
-                    return False
-
-                logger.debug("Waiting for conversion lock to be released...")
-                time.sleep(1)
-
-        os.close(lock_fd)
-        return False
-
-    except Exception as e:
-        logger.error(f"Error acquiring conversion lock: {e}")
-        return False
-
-
-def release_conversion_lock(lock_file_path: Path) -> None:
-    """
-    Release the conversion lock.
-
-    Args:
-        lock_file_path: Path to the lock file
-    """
-    try:
-        if lock_file_path.exists():
-            lock_file_path.unlink()
-            logger.debug(f"Released conversion lock: {lock_file_path}")
-    except Exception as e:
-        logger.error(f"Error releasing conversion lock: {e}")
-
-
-def is_lock_stale(lock_file_path: Path, max_age: int = 600) -> bool:
-    """
-    Check if a lock file is stale (older than max_age seconds).
-
-    Args:
-        lock_file_path: Path to the lock file
-        max_age: Maximum age in seconds before considering lock stale
-
-    Returns:
-        True if lock is stale, False otherwise
-    """
-    try:
-        if lock_file_path.exists():
-            lock_age = time.time() - lock_file_path.stat().st_mtime
-            return lock_age > max_age
-    except Exception:
-        pass
-
-    return False
-
-
 def wait_for_conversion_completion(
     model_type: str, rknn_path: Path, lock_file_path: Path, timeout: int = 300
 ) -> bool:
@@ -358,6 +258,7 @@ def wait_for_conversion_completion(
     Wait for another process to complete the conversion.
 
     Args:
+        model_type: Type of model being converted
         rknn_path: Path to the expected RKNN model
         lock_file_path: Path to the lock file to monitor
         timeout: Maximum time to wait in seconds
@@ -366,6 +267,8 @@ def wait_for_conversion_completion(
         True if RKNN model appears, False if timeout
     """
     start_time = time.time()
+    lock = FileLock(lock_file_path, stale_timeout=600)
+
     while time.time() - start_time < timeout:
         # Check if RKNN model appeared
         if rknn_path.exists():
@@ -385,11 +288,14 @@ def wait_for_conversion_completion(
                 return False
 
         # Check if lock is stale
-        if is_lock_stale(lock_file_path):
+        if lock.is_stale():
             logger.warning("Lock file is stale, attempting to clean up and retry...")
-            cleanup_stale_lock(lock_file_path)
+            lock._cleanup_stale_lock()
             # Try to acquire lock again
-            if acquire_conversion_lock(lock_file_path, timeout=60):
+            retry_lock = FileLock(
+                lock_file_path, timeout=60, cleanup_stale_on_init=True
+            )
+            if retry_lock.acquire():
                 try:
                     # Check if RKNN file appeared while waiting
                     if rknn_path.exists():
@@ -415,7 +321,7 @@ def wait_for_conversion_completion(
                     return False
 
                 finally:
-                    release_conversion_lock(lock_file_path)
+                    retry_lock.release()
 
         logger.debug("Waiting for RKNN model to appear...")
         time.sleep(1)
@@ -452,8 +358,9 @@ def auto_convert_model(
             return str(rknn_path)
 
         lock_file_path = base_path.parent / f"{base_name}.conversion.lock"
+        lock = FileLock(lock_file_path, timeout=300, cleanup_stale_on_init=True)
 
-        if acquire_conversion_lock(lock_file_path):
+        if lock.acquire():
             try:
                 if rknn_path.exists():
                     logger.info(
@@ -476,7 +383,7 @@ def auto_convert_model(
                     return None
 
             finally:
-                release_conversion_lock(lock_file_path)
+                lock.release()
         else:
             logger.info(
                 f"Another process is converting {model_path}, waiting for completion..."

frigate/util/services.py

@@ -9,6 +9,7 @@ import resource
 import shutil
 import signal
 import subprocess as sp
+import time
 import traceback
 from datetime import datetime
 from typing import Any, List, Optional, Tuple
@@ -388,6 +389,39 @@ def get_intel_gpu_stats(intel_gpu_device: Optional[str]) -> Optional[dict[str, s
         return results
 
 
+def get_openvino_npu_stats() -> Optional[dict[str, str]]:
+    """Get NPU stats using openvino."""
+    NPU_RUNTIME_PATH = "/sys/devices/pci0000:00/0000:00:0b.0/power/runtime_active_time"
+
+    try:
+        with open(NPU_RUNTIME_PATH, "r") as f:
+            initial_runtime = float(f.read().strip())
+
+        initial_time = time.time()
+
+        # Sleep for 1 second to get an accurate reading
+        time.sleep(1.0)
+
+        # Read runtime value again
+        with open(NPU_RUNTIME_PATH, "r") as f:
+            current_runtime = float(f.read().strip())
+
+        current_time = time.time()
+
+        # Calculate usage percentage
+        runtime_diff = current_runtime - initial_runtime
+        time_diff = (current_time - initial_time) * 1000.0  # Convert to milliseconds
+
+        if time_diff > 0:
+            usage = min(100.0, max(0.0, (runtime_diff / time_diff * 100.0)))
+        else:
+            usage = 0.0
+
+        return {"npu": f"{round(usage, 2)}", "mem": "-"}
+    except (FileNotFoundError, PermissionError, ValueError):
+        return None
+
+
 def get_rockchip_gpu_stats() -> Optional[dict[str, str]]:
     """Get GPU stats using rk."""
     try:
@@ -543,7 +577,7 @@ def ffprobe_stream(ffmpeg, path: str, detailed: bool = False) -> sp.CompletedPro
     if detailed and format_entries:
         ffprobe_cmd.extend(["-show_entries", f"format={format_entries}"])
 
-    ffprobe_cmd.extend(["-loglevel", "quiet", clean_path])
+    ffprobe_cmd.extend(["-loglevel", "error", clean_path])
 
     return sp.run(ffprobe_cmd, capture_output=True)
 

frigate/util/time.py

@@ -0,0 +1,100 @@
+"""Time utilities."""
+
+import datetime
+import logging
+from typing import Tuple
+from zoneinfo import ZoneInfoNotFoundError
+
+import pytz
+from tzlocal import get_localzone
+
+logger = logging.getLogger(__name__)
+
+
+def get_tz_modifiers(tz_name: str) -> Tuple[str, str, float]:
+    seconds_offset = (
+        datetime.datetime.now(pytz.timezone(tz_name)).utcoffset().total_seconds()
+    )
+    hours_offset = int(seconds_offset / 60 / 60)
+    minutes_offset = int(seconds_offset / 60 - hours_offset * 60)
+    hour_modifier = f"{hours_offset} hour"
+    minute_modifier = f"{minutes_offset} minute"
+    return hour_modifier, minute_modifier, seconds_offset
+
+
+def get_tomorrow_at_time(hour: int) -> datetime.datetime:
+    """Returns the datetime of the following day at 2am."""
+    try:
+        tomorrow = datetime.datetime.now(get_localzone()) + datetime.timedelta(days=1)
+    except ZoneInfoNotFoundError:
+        tomorrow = datetime.datetime.now(datetime.timezone.utc) + datetime.timedelta(
+            days=1
+        )
+        logger.warning(
+            "Using utc for maintenance due to missing or incorrect timezone set"
+        )
+
+    return tomorrow.replace(hour=hour, minute=0, second=0).astimezone(
+        datetime.timezone.utc
+    )
+
+
+def is_current_hour(timestamp: int) -> bool:
+    """Returns if timestamp is in the current UTC hour."""
+    start_of_next_hour = (
+        datetime.datetime.now(datetime.timezone.utc).replace(
+            minute=0, second=0, microsecond=0
+        )
+        + datetime.timedelta(hours=1)
+    ).timestamp()
+    return timestamp < start_of_next_hour
+
+
+def get_dst_transitions(
+    tz_name: str, start_time: float, end_time: float
+) -> list[tuple[float, float]]:
+    """
+    Find DST transition points and return time periods with consistent offsets.
+
+    Args:
+        tz_name: Timezone name (e.g., 'America/New_York')
+        start_time: Start timestamp (UTC)
+        end_time: End timestamp (UTC)
+
+    Returns:
+        List of (period_start, period_end, seconds_offset) tuples representing
+        continuous periods with the same UTC offset
+    """
+    try:
+        tz = pytz.timezone(tz_name)
+    except pytz.UnknownTimeZoneError:
+        # If timezone is invalid, return single period with no offset
+        return [(start_time, end_time, 0)]
+
+    periods = []
+    current = start_time
+
+    # Get initial offset
+    dt = datetime.datetime.utcfromtimestamp(current).replace(tzinfo=pytz.UTC)
+    local_dt = dt.astimezone(tz)
+    prev_offset = local_dt.utcoffset().total_seconds()
+    period_start = start_time
+
+    # Check each day for offset changes
+    while current <= end_time:
+        dt = datetime.datetime.utcfromtimestamp(current).replace(tzinfo=pytz.UTC)
+        local_dt = dt.astimezone(tz)
+        current_offset = local_dt.utcoffset().total_seconds()
+
+        if current_offset != prev_offset:
+            # Found a transition - close previous period
+            periods.append((period_start, current, prev_offset))
+            period_start = current
+            prev_offset = current_offset
+
+        current += 86400  # Check daily
+
+    # Add final period
+    periods.append((period_start, end_time, prev_offset))
+
+    return periods