Bump node-forge from 1.3.3 to 1.4.0 in /docs

Bumps [node-forge](https://github.com/digitalbazaar/forge) from 1.3.3 to 1.4.0.
- [Changelog](https://github.com/digitalbazaar/forge/blob/main/CHANGELOG.md)
- [Commits](https://github.com/digitalbazaar/forge/compare/v1.3.3...v1.4.0)

---
updated-dependencies:
- dependency-name: node-forge
  dependency-version: 1.4.0
  dependency-type: indirect
...

Signed-off-by: dependabot[bot] <support@github.com>

.github/workflows/pr_template_check.yml

@@ -16,7 +16,7 @@ jobs:
         uses: actions/github-script@v7
         with:
           script: |
-            const maintainers = ['blakeblackshear', 'NickM-27', 'hawkeye217'];
+            const maintainers = ['blakeblackshear', 'NickM-27', 'hawkeye217', 'dependabot[bot]'];
             const author = context.payload.pull_request.user.login;
 
             if (maintainers.includes(author)) {

docs/docs/configuration/custom_classification/object_classification.md

@@ -102,8 +102,19 @@ If examples for some of your classes do not appear in the grid, you can continue
 
 ### Improving the Model
 
+:::tip Diversity matters far more than volume
+
+Selecting dozens of nearly identical images is one of the fastest ways to degrade model performance. MobileNetV2 can overfit quickly when trained on homogeneous data — the model learns what *that exact moment* looked like rather than what actually defines the class. **This is why Frigate does not implement bulk training in the UI.**
+
+For more detail, see [Frigate Tip: Best Practices for Training Face and Custom Classification Models](https://github.com/blakeblackshear/frigate/discussions/21374).
+
+:::
+
+- **Start small and iterate**: Begin with a small, representative set of images per class. Models often begin working well with surprisingly few examples and improve naturally over time.
+- **Favor hard examples**: When images appear in the Recent Classifications tab, prioritize images scoring below 90–100% or those captured under new lighting, weather, or distance conditions.
+- **Avoid bulk training similar images**: Training large batches of images that already score 100% (or close) adds little new information and increases the risk of overfitting.
+- **The wizard is just the starting point**: You don’t need to find and label every class upfront. Missing classes will naturally appear in Recent Classifications, and those images tend to be more valuable because they represent new conditions and edge cases.
 - **Problem framing**: Keep classes visually distinct and relevant to the chosen object types.
-- **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

@@ -70,10 +70,21 @@ Once some images are assigned, training will begin automatically.
 
 ### Improving the Model
 
+:::tip Diversity matters far more than volume
+
+Selecting dozens of nearly identical images is one of the fastest ways to degrade model performance. MobileNetV2 can overfit quickly when trained on homogeneous data — the model learns what *that exact moment* looked like rather than what actually defines the state. This often leads to models that work perfectly under the original conditions but become unstable when day turns to night, weather changes, or seasonal lighting shifts. **This is why Frigate does not implement bulk training in the UI.**
+
+For more detail, see [Frigate Tip: Best Practices for Training Face and Custom Classification Models](https://github.com/blakeblackshear/frigate/discussions/21374).
+
+:::
+
+- **Start small and iterate**: Begin with a small, representative set of images per class. Models often begin working well with surprisingly few examples and improve naturally over time.
 - **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 Recent Classifications tab to gather balanced examples across times of day and weather.
 - **When to train**: Focus on cases where the model is entirely incorrect or flips between states when it should not. There's no need to train additional images when the model is already working consistently.
-- **Selecting training images**: Images scoring below 100% due to new conditions (e.g., first snow of the year, seasonal changes) or variations (e.g., objects temporarily in view, insects at night) are good candidates for training, as they represent scenarios different from the default state. Training these lower-scoring images that differ from existing training data helps prevent overfitting. Avoid training large quantities of images that look very similar, especially if they already score 100% as this can lead to overfitting.
+- **Favor hard examples**: When images appear in the Recent Classifications tab, prioritize images scoring below 90–100% or those captured under new conditions (e.g., first snow of the year, seasonal changes, objects temporarily in view, insects at night). These represent scenarios different from the default state and help prevent overfitting.
+- **Avoid bulk training similar images**: Training large batches of images that already score 100% (or close) adds little new information and increases the risk of overfitting.
+- **The wizard is just the starting point**: You don't need to find and label every state upfront. Missing states will naturally appear in Recent Classifications, and those images tend to be more valuable because they represent new conditions and edge cases.
 
 ## Debugging Classification Models
 

docs/package-lock.json

@@ -15952,9 +15952,9 @@
       }
     },
     "node_modules/node-forge": {
-      "version": "1.3.3",
-      "resolved": "https://registry.npmjs.org/node-forge/-/node-forge-1.3.3.tgz",
-      "integrity": "sha512-rLvcdSyRCyouf6jcOIPe/BgwG/d7hKjzMKOas33/pHEr6gbq18IK9zV7DiPvzsz0oBJPme6qr6H6kGZuI9/DZg==",
+      "version": "1.4.0",
+      "resolved": "https://registry.npmjs.org/node-forge/-/node-forge-1.4.0.tgz",
+      "integrity": "sha512-LarFH0+6VfriEhqMMcLX2F7SwSXeWwnEAJEsYm5QKWchiVYVvJyV9v7UDvUv+w5HO23ZpQTXDv/GxdDdMyOuoQ==",
       "license": "(BSD-3-Clause OR GPL-2.0)",
       "engines": {
         "node": ">= 6.13.0"

frigate/api/app.py

@@ -613,6 +613,34 @@ def config_set(request: Request, body: AppConfigSetBody):
 
                 try:
                     config = FrigateConfig.parse(new_raw_config)
+                except ValidationError as e:
+                    with open(config_file, "w") as f:
+                        f.write(old_raw_config)
+                        f.close()
+                    logger.error(
+                        f"Config Validation Error:\n\n{str(traceback.format_exc())}"
+                    )
+                    error_messages = []
+                    for err in e.errors():
+                        msg = err.get("msg", "")
+                        # Strip pydantic "Value error, " prefix for cleaner display
+                        if msg.startswith("Value error, "):
+                            msg = msg[len("Value error, ") :]
+                        error_messages.append(msg)
+                    message = (
+                        "; ".join(error_messages)
+                        if error_messages
+                        else "Check logs for error message."
+                    )
+                    return JSONResponse(
+                        content=(
+                            {
+                                "success": False,
+                                "message": f"Error saving config: {message}",
+                            }
+                        ),
+                        status_code=400,
+                    )
                 except Exception:
                     with open(config_file, "w") as f:
                         f.write(old_raw_config)

frigate/camera/__init__.py

@@ -1,26 +1,27 @@
 import multiprocessing as mp
-from multiprocessing.managers import SyncManager
+import queue
+from multiprocessing.managers import SyncManager, ValueProxy
 from multiprocessing.sharedctypes import Synchronized
 from multiprocessing.synchronize import Event
 
 
 class CameraMetrics:
-    camera_fps: Synchronized
-    detection_fps: Synchronized
-    detection_frame: Synchronized
-    process_fps: Synchronized
-    skipped_fps: Synchronized
-    read_start: Synchronized
-    audio_rms: Synchronized
-    audio_dBFS: Synchronized
+    camera_fps: ValueProxy[float]
+    detection_fps: ValueProxy[float]
+    detection_frame: ValueProxy[float]
+    process_fps: ValueProxy[float]
+    skipped_fps: ValueProxy[float]
+    read_start: ValueProxy[float]
+    audio_rms: ValueProxy[float]
+    audio_dBFS: ValueProxy[float]
 
-    frame_queue: mp.Queue
+    frame_queue: queue.Queue
 
-    process_pid: Synchronized
-    capture_process_pid: Synchronized
-    ffmpeg_pid: Synchronized
-    reconnects_last_hour: Synchronized
-    stalls_last_hour: Synchronized
+    process_pid: ValueProxy[int]
+    capture_process_pid: ValueProxy[int]
+    ffmpeg_pid: ValueProxy[int]
+    reconnects_last_hour: ValueProxy[int]
+    stalls_last_hour: ValueProxy[int]
 
     def __init__(self, manager: SyncManager):
         self.camera_fps = manager.Value("d", 0)
@@ -56,14 +57,14 @@ class PTZMetrics:
     reset: Event
 
     def __init__(self, *, autotracker_enabled: bool):
-        self.autotracker_enabled = mp.Value("i", autotracker_enabled)
+        self.autotracker_enabled = mp.Value("i", autotracker_enabled)  # type: ignore[assignment]
 
-        self.start_time = mp.Value("d", 0)
-        self.stop_time = mp.Value("d", 0)
-        self.frame_time = mp.Value("d", 0)
-        self.zoom_level = mp.Value("d", 0)
-        self.max_zoom = mp.Value("d", 0)
-        self.min_zoom = mp.Value("d", 0)
+        self.start_time = mp.Value("d", 0)  # type: ignore[assignment]
+        self.stop_time = mp.Value("d", 0)  # type: ignore[assignment]
+        self.frame_time = mp.Value("d", 0)  # type: ignore[assignment]
+        self.zoom_level = mp.Value("d", 0)  # type: ignore[assignment]
+        self.max_zoom = mp.Value("d", 0)  # type: ignore[assignment]
+        self.min_zoom = mp.Value("d", 0)  # type: ignore[assignment]
 
         self.tracking_active = mp.Event()
         self.motor_stopped = mp.Event()

frigate/camera/activity_manager.py

@@ -37,6 +37,9 @@ class CameraActivityManager:
             self.__init_camera(camera_config)
 
     def __init_camera(self, camera_config: CameraConfig) -> None:
+        if camera_config.name is None:
+            return
+
         self.last_camera_activity[camera_config.name] = {}
         self.camera_all_object_counts[camera_config.name] = Counter()
         self.camera_active_object_counts[camera_config.name] = Counter()
@@ -114,7 +117,7 @@ class CameraActivityManager:
         self.last_camera_activity = new_activity
 
     def compare_camera_activity(
-        self, camera: str, new_activity: dict[str, Any]
+        self, camera: str, new_activity: list[dict[str, Any]]
     ) -> None:
         all_objects = Counter(
             obj["label"].replace("-verified", "") for obj in new_activity
@@ -175,6 +178,9 @@ class AudioActivityManager:
             self.__init_camera(camera_config)
 
     def __init_camera(self, camera_config: CameraConfig) -> None:
+        if camera_config.name is None:
+            return
+
         self.current_audio_detections[camera_config.name] = {}
 
     def update_activity(self, new_activity: dict[str, dict[str, Any]]) -> None:
@@ -202,7 +208,7 @@ class AudioActivityManager:
 
     def compare_audio_activity(
         self, camera: str, new_detections: list[tuple[str, float]], now: float
-    ) -> None:
+    ) -> bool:
         camera_config = self.config.cameras.get(camera)
         if camera_config is None:
             return False

frigate/camera/maintainer.py

@@ -102,7 +102,7 @@ class CameraMaintainer(threading.Thread):
                 f"recommend increasing it to at least {shm_stats['min_shm']}MB."
             )
 
-        return shm_stats["shm_frame_count"]
+        return int(shm_stats["shm_frame_count"])
 
     def __start_camera_processor(
         self, name: str, config: CameraConfig, runtime: bool = False
@@ -152,10 +152,10 @@ class CameraMaintainer(threading.Thread):
             camera_stop_event,
             self.config.logger,
         )
-        self.camera_processes[config.name] = camera_process
+        self.camera_processes[name] = camera_process
         camera_process.start()
-        self.camera_metrics[config.name].process_pid.value = camera_process.pid
-        logger.info(f"Camera processor started for {config.name}: {camera_process.pid}")
+        self.camera_metrics[name].process_pid.value = camera_process.pid
+        logger.info(f"Camera processor started for {name}: {camera_process.pid}")
 
     def __start_camera_capture(
         self, name: str, config: CameraConfig, runtime: bool = False
@@ -219,7 +219,7 @@ class CameraMaintainer(threading.Thread):
             logger.info(f"Closing frame queue for {camera}")
             empty_and_close_queue(self.camera_metrics[camera].frame_queue)
 
-    def run(self):
+    def run(self) -> None:
         self.__init_historical_regions()
 
         # start camera processes

frigate/camera/state.py

@@ -31,26 +31,26 @@ logger = logging.getLogger(__name__)
 class CameraState:
     def __init__(
         self,
-        name,
+        name: str,
         config: FrigateConfig,
         frame_manager: SharedMemoryFrameManager,
         ptz_autotracker_thread: PtzAutoTrackerThread,
-    ):
+    ) -> None:
         self.name = name
         self.config = config
         self.camera_config = config.cameras[name]
         self.frame_manager = frame_manager
         self.best_objects: dict[str, TrackedObject] = {}
         self.tracked_objects: dict[str, TrackedObject] = {}
-        self.frame_cache = {}
-        self.zone_objects = defaultdict(list)
+        self.frame_cache: dict[float, dict[str, Any]] = {}
+        self.zone_objects: defaultdict[str, list[Any]] = defaultdict(list)
         self._current_frame = np.zeros(self.camera_config.frame_shape_yuv, np.uint8)
         self.current_frame_lock = threading.Lock()
         self.current_frame_time = 0.0
-        self.motion_boxes = []
-        self.regions = []
-        self.previous_frame_id = None
-        self.callbacks = defaultdict(list)
+        self.motion_boxes: list[tuple[int, int, int, int]] = []
+        self.regions: list[tuple[int, int, int, int]] = []
+        self.previous_frame_id: str | None = None
+        self.callbacks: defaultdict[str, list[Callable]] = defaultdict(list)
         self.ptz_autotracker_thread = ptz_autotracker_thread
         self.prev_enabled = self.camera_config.enabled
 
@@ -62,10 +62,10 @@ class CameraState:
             motion_boxes = self.motion_boxes.copy()
             regions = self.regions.copy()
 
-        frame_copy = cv2.cvtColor(frame_copy, cv2.COLOR_YUV2BGR_I420)
+        frame_copy = cv2.cvtColor(frame_copy, cv2.COLOR_YUV2BGR_I420)  # type: ignore[assignment]
         # draw on the frame
         if draw_options.get("mask"):
-            mask_overlay = np.where(self.camera_config.motion.rasterized_mask == [0])
+            mask_overlay = np.where(self.camera_config.motion.rasterized_mask == [0])  # type: ignore[attr-defined]
             frame_copy[mask_overlay] = [0, 0, 0]
 
         if draw_options.get("bounding_boxes"):
@@ -97,7 +97,7 @@ class CameraState:
                     and obj["id"]
                     == self.ptz_autotracker_thread.ptz_autotracker.tracked_object[
                         self.name
-                    ].obj_data["id"]
+                    ].obj_data["id"]  # type: ignore[attr-defined]
                     and obj["frame_time"] == frame_time
                 ):
                     thickness = 5
@@ -109,10 +109,12 @@ class CameraState:
                     if (
                         self.camera_config.onvif.autotracking.zooming
                         != ZoomingModeEnum.disabled
+                        and self.camera_config.detect.width is not None
+                        and self.camera_config.detect.height is not None
                     ):
                         max_target_box = self.ptz_autotracker_thread.ptz_autotracker.tracked_object_metrics[
                             self.name
-                        ]["max_target_box"]
+                        ]["max_target_box"]  # type: ignore[index]
                         side_length = max_target_box * (
                             max(
                                 self.camera_config.detect.width,
@@ -221,14 +223,14 @@ class CameraState:
                 )
 
         if draw_options.get("timestamp"):
-            color = self.camera_config.timestamp_style.color
+            ts_color = self.camera_config.timestamp_style.color
             draw_timestamp(
                 frame_copy,
                 frame_time,
                 self.camera_config.timestamp_style.format,
                 font_effect=self.camera_config.timestamp_style.effect,
                 font_thickness=self.camera_config.timestamp_style.thickness,
-                font_color=(color.blue, color.green, color.red),
+                font_color=(ts_color.blue, ts_color.green, ts_color.red),
                 position=self.camera_config.timestamp_style.position,
             )
 
@@ -273,10 +275,10 @@ class CameraState:
 
         return frame_copy
 
-    def finished(self, obj_id):
+    def finished(self, obj_id: str) -> None:
         del self.tracked_objects[obj_id]
 
-    def on(self, event_type: str, callback: Callable):
+    def on(self, event_type: str, callback: Callable[..., Any]) -> None:
         self.callbacks[event_type].append(callback)
 
     def update(
@@ -286,7 +288,7 @@ class CameraState:
         current_detections: dict[str, dict[str, Any]],
         motion_boxes: list[tuple[int, int, int, int]],
         regions: list[tuple[int, int, int, int]],
-    ):
+    ) -> None:
         current_frame = self.frame_manager.get(
             frame_name, self.camera_config.frame_shape_yuv
         )
@@ -313,7 +315,7 @@ class CameraState:
                 f"{self.name}: New object, adding {frame_time} to frame cache for {id}"
             )
             self.frame_cache[frame_time] = {
-                "frame": np.copy(current_frame),
+                "frame": np.copy(current_frame),  # type: ignore[arg-type]
                 "object_id": id,
             }
 
@@ -356,7 +358,8 @@ class CameraState:
             if thumb_update and current_frame is not None:
                 # ensure this frame is stored in the cache
                 if (
-                    updated_obj.thumbnail_data["frame_time"] == frame_time
+                    updated_obj.thumbnail_data is not None
+                    and updated_obj.thumbnail_data["frame_time"] == frame_time
                     and frame_time not in self.frame_cache
                 ):
                     logger.debug(
@@ -397,7 +400,7 @@ class CameraState:
 
         # TODO: can i switch to looking this up and only changing when an event ends?
         # maintain best objects
-        camera_activity: dict[str, list[Any]] = {
+        camera_activity: dict[str, Any] = {
             "motion": len(motion_boxes) > 0,
             "objects": [],
         }
@@ -411,10 +414,7 @@ class CameraState:
                 sub_label = None
 
                 if obj.obj_data.get("sub_label"):
-                    if (
-                        obj.obj_data.get("sub_label")[0]
-                        in self.config.model.all_attributes
-                    ):
+                    if obj.obj_data["sub_label"][0] in self.config.model.all_attributes:
                         label = obj.obj_data["sub_label"][0]
                     else:
                         label = f"{object_type}-verified"
@@ -449,14 +449,19 @@ class CameraState:
                 # if the object is a higher score than the current best score
                 # or the current object is older than desired, use the new object
                 if (
-                    is_better_thumbnail(
+                    current_best.thumbnail_data is not None
+                    and obj.thumbnail_data is not None
+                    and is_better_thumbnail(
                         object_type,
                         current_best.thumbnail_data,
                         obj.thumbnail_data,
                         self.camera_config.frame_shape,
                     )
-                    or (now - current_best.thumbnail_data["frame_time"])
-                    > self.camera_config.best_image_timeout
+                    or (
+                        current_best.thumbnail_data is not None
+                        and (now - current_best.thumbnail_data["frame_time"])
+                        > self.camera_config.best_image_timeout
+                    )
                 ):
                     self.send_mqtt_snapshot(obj, object_type)
             else:
@@ -472,7 +477,9 @@ class CameraState:
             if obj.thumbnail_data is not None
         }
         current_best_frames = {
-            obj.thumbnail_data["frame_time"] for obj in self.best_objects.values()
+            obj.thumbnail_data["frame_time"]
+            for obj in self.best_objects.values()
+            if obj.thumbnail_data is not None
         }
         thumb_frames_to_delete = [
             t
@@ -540,7 +547,7 @@ class CameraState:
             with open(
                 os.path.join(
                     CLIPS_DIR,
-                    f"{self.camera_config.name}-{event_id}-clean.webp",
+                    f"{self.name}-{event_id}-clean.webp",
                 ),
                 "wb",
             ) as p:
@@ -549,7 +556,7 @@ class CameraState:
         # create thumbnail with max height of 175 and save
         width = int(175 * img_frame.shape[1] / img_frame.shape[0])
         thumb = cv2.resize(img_frame, dsize=(width, 175), interpolation=cv2.INTER_AREA)
-        thumb_path = os.path.join(THUMB_DIR, self.camera_config.name)
+        thumb_path = os.path.join(THUMB_DIR, self.name)
         os.makedirs(thumb_path, exist_ok=True)
         cv2.imwrite(os.path.join(thumb_path, f"{event_id}.webp"), thumb)
 

frigate/comms/webpush.py

@@ -542,9 +542,9 @@ class WebPushClient(Communicator):
 
         self.check_registrations()
 
-        reasoning: str = payload.get("reasoning", "")
+        text: str = payload.get("message") or payload.get("reasoning", "")
         title = f"{camera_name}: Monitoring Alert"
-        message = (reasoning[:197] + "...") if len(reasoning) > 200 else reasoning
+        message = (text[:197] + "...") if len(text) > 200 else text
 
         logger.debug(f"Sending camera monitoring push notification for {camera_name}")
 

frigate/config/camera/detect.py

@@ -71,6 +71,7 @@ class DetectConfig(FrigateBaseModel):
         default=None,
         title="Minimum initialization frames",
         description="Number of consecutive detection hits required before creating a tracked object. Increase to reduce false initializations. Default value is fps divided by 2.",
+        ge=2,
     )
     max_disappeared: Optional[int] = Field(
         default=None,

frigate/config/camera/review.py

@@ -188,7 +188,7 @@ class ReviewConfig(FrigateBaseModel):
     detections: DetectionsConfig = Field(
         default_factory=DetectionsConfig,
         title="Detections config",
-        description="Settings for creating detection events (non-alert) and how long to keep them.",
+        description="Settings for which tracked objects generate detections (non-alert) and how detections are retained.",
     )
     genai: GenAIReviewConfig = Field(
         default_factory=GenAIReviewConfig,

frigate/config/camera/updater.py

@@ -32,6 +32,7 @@ class CameraConfigUpdateEnum(str, Enum):
     face_recognition = "face_recognition"
     lpr = "lpr"
     snapshots = "snapshots"
+    timestamp_style = "timestamp_style"
     zones = "zones"
 
 
@@ -133,6 +134,8 @@ class CameraConfigUpdateSubscriber:
             config.snapshots = updated_config
         elif update_type == CameraConfigUpdateEnum.onvif:
             config.onvif = updated_config
+        elif update_type == CameraConfigUpdateEnum.timestamp_style:
+            config.timestamp_style = updated_config
         elif update_type == CameraConfigUpdateEnum.zones:
             config.zones = updated_config
 

frigate/config/config.py

@@ -25,6 +25,7 @@ from frigate.plus import PlusApi
 from frigate.util.builtin import (
     deep_merge,
     get_ffmpeg_arg_list,
+    load_labels,
 )
 from frigate.util.config import (
     CURRENT_CONFIG_VERSION,
@@ -40,7 +41,7 @@ from frigate.util.services import auto_detect_hwaccel
 from .auth import AuthConfig
 from .base import FrigateBaseModel
 from .camera import CameraConfig, CameraLiveConfig
-from .camera.audio import AudioConfig
+from .camera.audio import AudioConfig, AudioFilterConfig
 from .camera.birdseye import BirdseyeConfig
 from .camera.detect import DetectConfig
 from .camera.ffmpeg import FfmpegConfig
@@ -473,7 +474,7 @@ class FrigateConfig(FrigateBaseModel):
     live: CameraLiveConfig = Field(
         default_factory=CameraLiveConfig,
         title="Live playback",
-        description="Settings used by the Web UI to control live stream resolution and quality.",
+        description="Settings to control the jsmpeg live stream resolution and quality. This does not affect restreamed cameras that use go2rtc for live view.",
     )
     motion: Optional[MotionConfig] = Field(
         default=None,
@@ -613,6 +614,21 @@ class FrigateConfig(FrigateBaseModel):
         if self.ffmpeg.hwaccel_args == "auto":
             self.ffmpeg.hwaccel_args = auto_detect_hwaccel()
 
+        # Populate global audio filters for all audio labels
+        all_audio_labels = {
+            label
+            for label in load_labels("/audio-labelmap.txt", prefill=521).values()
+            if label
+        }
+
+        if self.audio.filters is None:
+            self.audio.filters = {}
+
+        for key in sorted(all_audio_labels - self.audio.filters.keys()):
+            self.audio.filters[key] = AudioFilterConfig()
+
+        self.audio.filters = dict(sorted(self.audio.filters.items()))
+
         # Global config to propagate down to camera level
         global_config = self.model_dump(
             include={
@@ -748,7 +764,7 @@ class FrigateConfig(FrigateBaseModel):
                 )
 
             # Default min_initialized configuration
-            min_initialized = int(camera_config.detect.fps / 2)
+            min_initialized = max(int(camera_config.detect.fps / 2), 2)
             if camera_config.detect.min_initialized is None:
                 camera_config.detect.min_initialized = min_initialized
 
@@ -791,6 +807,16 @@ class FrigateConfig(FrigateBaseModel):
                 camera_config.review.genai.enabled
             )
 
+            if camera_config.audio.filters is None:
+                camera_config.audio.filters = {}
+
+            for key in sorted(all_audio_labels - camera_config.audio.filters.keys()):
+                camera_config.audio.filters[key] = AudioFilterConfig()
+
+            camera_config.audio.filters = dict(
+                sorted(camera_config.audio.filters.items())
+            )
+
             # Add default filters
             object_keys = camera_config.objects.track
             if camera_config.objects.filters is None:

frigate/data_processing/common/audio_transcription/model.py

@@ -53,7 +53,7 @@ class AudioTranscriptionModelRunner:
                 self.downloader = ModelDownloader(
                     model_name="sherpa-onnx",
                     download_path=download_path,
-                    file_names=self.model_files.keys(),
+                    file_names=list(self.model_files.keys()),
                     download_func=self.__download_models,
                 )
                 self.downloader.ensure_model_files()

frigate/data_processing/common/face/model.py

@@ -21,7 +21,7 @@ class FaceRecognizer(ABC):
 
     def __init__(self, config: FrigateConfig) -> None:
         self.config = config
-        self.landmark_detector: cv2.face.FacemarkLBF = None
+        self.landmark_detector: cv2.face.Facemark | None = None
         self.init_landmark_detector()
 
     @abstractmethod
@@ -38,13 +38,14 @@ class FaceRecognizer(ABC):
     def classify(self, face_image: np.ndarray) -> tuple[str, float] | None:
         pass
 
-    @redirect_output_to_logger(logger, logging.DEBUG)
+    @redirect_output_to_logger(logger, logging.DEBUG)  # type: ignore[misc]
     def init_landmark_detector(self) -> None:
         landmark_model = os.path.join(MODEL_CACHE_DIR, "facedet/landmarkdet.yaml")
 
         if os.path.exists(landmark_model):
-            self.landmark_detector = cv2.face.createFacemarkLBF()
-            self.landmark_detector.loadModel(landmark_model)
+            landmark_detector = cv2.face.createFacemarkLBF()
+            landmark_detector.loadModel(landmark_model)
+            self.landmark_detector = landmark_detector
 
     def align_face(
         self,
@@ -52,8 +53,10 @@ class FaceRecognizer(ABC):
         output_width: int,
         output_height: int,
     ) -> np.ndarray:
-        # landmark is run on grayscale images
+        if not self.landmark_detector:
+            raise ValueError("Landmark detector not initialized")
 
+        # landmark is run on grayscale images
         if image.ndim == 3:
             land_image = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
         else:
@@ -131,8 +134,11 @@ class FaceRecognizer(ABC):
 
 
 def similarity_to_confidence(
-    cosine_similarity: float, median=0.3, range_width=0.6, slope_factor=12
-):
+    cosine_similarity: float,
+    median: float = 0.3,
+    range_width: float = 0.6,
+    slope_factor: float = 12,
+) -> float:
     """
     Default sigmoid function to map cosine similarity to confidence.
 
@@ -151,14 +157,14 @@ def similarity_to_confidence(
     bias = median
 
     # Calculate confidence
-    confidence = 1 / (1 + np.exp(-slope * (cosine_similarity - bias)))
+    confidence: float = 1 / (1 + np.exp(-slope * (cosine_similarity - bias)))
     return confidence
 
 
 class FaceNetRecognizer(FaceRecognizer):
     def __init__(self, config: FrigateConfig):
         super().__init__(config)
-        self.mean_embs: dict[int, np.ndarray] = {}
+        self.mean_embs: dict[str, np.ndarray] = {}
         self.face_embedder: FaceNetEmbedding = FaceNetEmbedding()
         self.model_builder_queue: queue.Queue | None = None
 
@@ -168,7 +174,7 @@ class FaceNetRecognizer(FaceRecognizer):
     def run_build_task(self) -> None:
         self.model_builder_queue = queue.Queue()
 
-        def build_model():
+        def build_model() -> None:
             face_embeddings_map: dict[str, list[np.ndarray]] = {}
             idx = 0
 
@@ -187,7 +193,7 @@ class FaceNetRecognizer(FaceRecognizer):
                     img = cv2.imread(os.path.join(face_folder, image))
 
                     if img is None:
-                        continue
+                        continue  # type: ignore[unreachable]
 
                     img = self.align_face(img, img.shape[1], img.shape[0])
                     emb = self.face_embedder([img])[0].squeeze()
@@ -195,12 +201,13 @@ class FaceNetRecognizer(FaceRecognizer):
 
                 idx += 1
 
+            assert self.model_builder_queue is not None
             self.model_builder_queue.put(face_embeddings_map)
 
         thread = threading.Thread(target=build_model, daemon=True)
         thread.start()
 
-    def build(self):
+    def build(self) -> None:
         if not self.landmark_detector:
             self.init_landmark_detector()
             return None
@@ -226,7 +233,7 @@ class FaceNetRecognizer(FaceRecognizer):
 
         logger.debug("Finished building ArcFace model")
 
-    def classify(self, face_image):
+    def classify(self, face_image: np.ndarray) -> tuple[str, float] | None:
         if not self.landmark_detector:
             return None
 
@@ -245,7 +252,7 @@ class FaceNetRecognizer(FaceRecognizer):
         img = self.align_face(face_image, face_image.shape[1], face_image.shape[0])
         embedding = self.face_embedder([img])[0].squeeze()
 
-        score = 0
+        score: float = 0
         label = ""
 
         for name, mean_emb in self.mean_embs.items():
@@ -268,7 +275,7 @@ class FaceNetRecognizer(FaceRecognizer):
 class ArcFaceRecognizer(FaceRecognizer):
     def __init__(self, config: FrigateConfig):
         super().__init__(config)
-        self.mean_embs: dict[int, np.ndarray] = {}
+        self.mean_embs: dict[str, np.ndarray] = {}
         self.face_embedder: ArcfaceEmbedding = ArcfaceEmbedding(config.face_recognition)
         self.model_builder_queue: queue.Queue | None = None
 
@@ -278,7 +285,7 @@ class ArcFaceRecognizer(FaceRecognizer):
     def run_build_task(self) -> None:
         self.model_builder_queue = queue.Queue()
 
-        def build_model():
+        def build_model() -> None:
             face_embeddings_map: dict[str, list[np.ndarray]] = {}
             idx = 0
 
@@ -297,20 +304,21 @@ class ArcFaceRecognizer(FaceRecognizer):
                     img = cv2.imread(os.path.join(face_folder, image))
 
                     if img is None:
-                        continue
+                        continue  # type: ignore[unreachable]
 
                     img = self.align_face(img, img.shape[1], img.shape[0])
-                    emb = self.face_embedder([img])[0].squeeze()
+                    emb = self.face_embedder([img])[0].squeeze()  # type: ignore[arg-type]
                     face_embeddings_map[name].append(emb)
 
                 idx += 1
 
+            assert self.model_builder_queue is not None
             self.model_builder_queue.put(face_embeddings_map)
 
         thread = threading.Thread(target=build_model, daemon=True)
         thread.start()
 
-    def build(self):
+    def build(self) -> None:
         if not self.landmark_detector:
             self.init_landmark_detector()
             return None
@@ -336,7 +344,7 @@ class ArcFaceRecognizer(FaceRecognizer):
 
         logger.debug("Finished building ArcFace model")
 
-    def classify(self, face_image):
+    def classify(self, face_image: np.ndarray) -> tuple[str, float] | None:
         if not self.landmark_detector:
             return None
 
@@ -353,9 +361,9 @@ class ArcFaceRecognizer(FaceRecognizer):
 
         # align face and run recognition
         img = self.align_face(face_image, face_image.shape[1], face_image.shape[0])
-        embedding = self.face_embedder([img])[0].squeeze()
+        embedding = self.face_embedder([img])[0].squeeze()  # type: ignore[arg-type]
 
-        score = 0
+        score: float = 0
         label = ""
 
         for name, mean_emb in self.mean_embs.items():

frigate/data_processing/common/license_plate/mixin.py

@@ -10,7 +10,7 @@ import random
 import re
 import string
 from pathlib import Path
-from typing import Any, List, Optional, Tuple
+from typing import Any, List, Tuple
 
 import cv2
 import numpy as np
@@ -22,19 +22,35 @@ from frigate.comms.event_metadata_updater import (
     EventMetadataPublisher,
     EventMetadataTypeEnum,
 )
+from frigate.comms.inter_process import InterProcessRequestor
+from frigate.config import FrigateConfig
+from frigate.config.classification import LicensePlateRecognitionConfig
 from frigate.const import CLIPS_DIR, MODEL_CACHE_DIR
+from frigate.data_processing.common.license_plate.model import LicensePlateModelRunner
 from frigate.embeddings.onnx.lpr_embedding import LPR_EMBEDDING_SIZE
 from frigate.types import TrackedObjectUpdateTypesEnum
 from frigate.util.builtin import EventsPerSecond, InferenceSpeed
 from frigate.util.image import area
 
+from ...types import DataProcessorMetrics
+
 logger = logging.getLogger(__name__)
 
 WRITE_DEBUG_IMAGES = False
 
 
 class LicensePlateProcessingMixin:
-    def __init__(self, *args, **kwargs):
+    # Attributes expected from consuming classes (set before super().__init__)
+    config: FrigateConfig
+    metrics: DataProcessorMetrics
+    model_runner: LicensePlateModelRunner
+    lpr_config: LicensePlateRecognitionConfig
+    requestor: InterProcessRequestor
+    detected_license_plates: dict[str, dict[str, Any]]
+    camera_current_cars: dict[str, list[str]]
+    sub_label_publisher: EventMetadataPublisher
+
+    def __init__(self, *args: Any, **kwargs: Any) -> None:
         super().__init__(*args, **kwargs)
         self.plate_rec_speed = InferenceSpeed(self.metrics.alpr_speed)
         self.plates_rec_second = EventsPerSecond()
@@ -97,7 +113,7 @@ class LicensePlateProcessingMixin:
             )
 
         try:
-            outputs = self.model_runner.detection_model([normalized_image])[0]
+            outputs = self.model_runner.detection_model([normalized_image])[0]  # type: ignore[arg-type]
         except Exception as e:
             logger.warning(f"Error running LPR box detection model: {e}")
             return []
@@ -105,18 +121,18 @@ class LicensePlateProcessingMixin:
         outputs = outputs[0, :, :]
 
         if False:
-            current_time = int(datetime.datetime.now().timestamp())
+            current_time = int(datetime.datetime.now().timestamp())  # type: ignore[unreachable]
             cv2.imwrite(
                 f"debug/frames/probability_map_{current_time}.jpg",
                 (outputs * 255).astype(np.uint8),
             )
 
         boxes, _ = self._boxes_from_bitmap(outputs, outputs > self.mask_thresh, w, h)
-        return self._filter_polygon(boxes, (h, w))
+        return self._filter_polygon(boxes, (h, w))  # type: ignore[return-value,arg-type]
 
     def _classify(
         self, images: List[np.ndarray]
-    ) -> Tuple[List[np.ndarray], List[Tuple[str, float]]]:
+    ) -> Tuple[List[np.ndarray], List[Tuple[str, float]]] | None:
         """
         Classify the orientation or category of each detected license plate.
 
@@ -138,15 +154,15 @@ class LicensePlateProcessingMixin:
                 norm_images.append(norm_img)
 
         try:
-            outputs = self.model_runner.classification_model(norm_images)
+            outputs = self.model_runner.classification_model(norm_images)  # type: ignore[arg-type]
         except Exception as e:
             logger.warning(f"Error running LPR classification model: {e}")
-            return
+            return None
 
         return self._process_classification_output(images, outputs)
 
     def _recognize(
-        self, camera: string, images: List[np.ndarray]
+        self, camera: str, images: List[np.ndarray]
     ) -> Tuple[List[str], List[List[float]]]:
         """
         Recognize the characters on the detected license plates using the recognition model.
@@ -179,7 +195,7 @@ class LicensePlateProcessingMixin:
                 norm_images.append(norm_image)
 
         try:
-            outputs = self.model_runner.recognition_model(norm_images)
+            outputs = self.model_runner.recognition_model(norm_images)  # type: ignore[arg-type]
         except Exception as e:
             logger.warning(f"Error running LPR recognition model: {e}")
             return [], []
@@ -410,7 +426,8 @@ class LicensePlateProcessingMixin:
             )
 
             if sorted_data:
-                return map(list, zip(*sorted_data))
+                plates, confs, areas_list = zip(*sorted_data)
+                return list(plates), list(confs), list(areas_list)
 
         return [], [], []
 
@@ -532,7 +549,7 @@ class LicensePlateProcessingMixin:
         # Add the last box
         merged_boxes.append(current_box)
 
-        return np.array(merged_boxes, dtype=np.int32)
+        return np.array(merged_boxes, dtype=np.int32)  # type: ignore[return-value]
 
     def _boxes_from_bitmap(
         self, output: np.ndarray, mask: np.ndarray, dest_width: int, dest_height: int
@@ -560,38 +577,42 @@ class LicensePlateProcessingMixin:
         boxes = []
         scores = []
 
-        for index in range(len(contours)):
-            contour = contours[index]
+        for index in range(len(contours)):  # type: ignore[arg-type]
+            contour = contours[index]  # type: ignore[index]
 
             # get minimum bounding box (rotated rectangle) around the contour and the smallest side length.
             points, sside = self._get_min_boxes(contour)
             if sside < self.min_size:
                 continue
 
-            points = np.array(points, dtype=np.float32)
+            points = np.array(points, dtype=np.float32)  # type: ignore[assignment]
 
             score = self._box_score(output, contour)
             if self.box_thresh > score:
                 continue
 
-            points = self._expand_box(points)
+            points = self._expand_box(points)  # type: ignore[assignment]
 
             # Get the minimum area rectangle again after expansion
-            points, sside = self._get_min_boxes(points.reshape(-1, 1, 2))
+            points, sside = self._get_min_boxes(points.reshape(-1, 1, 2))  # type: ignore[attr-defined]
             if sside < self.min_size + 2:
                 continue
 
-            points = np.array(points, dtype=np.float32)
+            points = np.array(points, dtype=np.float32)  # type: ignore[assignment]
 
             # normalize and clip box coordinates to fit within the destination image size.
-            points[:, 0] = np.clip(
-                np.round(points[:, 0] / width * dest_width), 0, dest_width
+            points[:, 0] = np.clip(  # type: ignore[call-overload]
+                np.round(points[:, 0] / width * dest_width),  # type: ignore[call-overload]
+                0,
+                dest_width,
             )
-            points[:, 1] = np.clip(
-                np.round(points[:, 1] / height * dest_height), 0, dest_height
+            points[:, 1] = np.clip(  # type: ignore[call-overload]
+                np.round(points[:, 1] / height * dest_height),  # type: ignore[call-overload]
+                0,
+                dest_height,
             )
 
-            boxes.append(points.astype("int32"))
+            boxes.append(points.astype("int32"))  # type: ignore[attr-defined]
             scores.append(score)
 
         return np.array(boxes, dtype="int32"), scores
@@ -632,7 +653,7 @@ class LicensePlateProcessingMixin:
         x1, y1 = np.clip(contour.min(axis=0), 0, [w - 1, h - 1])
         x2, y2 = np.clip(contour.max(axis=0), 0, [w - 1, h - 1])
         mask = np.zeros((y2 - y1 + 1, x2 - x1 + 1), dtype=np.uint8)
-        cv2.fillPoly(mask, [contour - [x1, y1]], 1)
+        cv2.fillPoly(mask, [contour - [x1, y1]], 1)  # type: ignore[call-overload]
         return cv2.mean(bitmap[y1 : y2 + 1, x1 : x2 + 1], mask)[0]
 
     @staticmethod
@@ -690,7 +711,7 @@ class LicensePlateProcessingMixin:
         Returns:
             bool: Whether the polygon is valid or not.
         """
-        return (
+        return bool(
             point[:, 0].min() >= 0
             and point[:, 0].max() < width
             and point[:, 1].min() >= 0
@@ -735,7 +756,7 @@ class LicensePlateProcessingMixin:
         return np.array([tl, tr, br, bl])
 
     @staticmethod
-    def _sort_boxes(boxes):
+    def _sort_boxes(boxes: list[np.ndarray]) -> list[np.ndarray]:
         """
         Sort polygons based on their position in the image. If boxes are close in vertical
         position (within 5 pixels), sort them by horizontal position.
@@ -837,16 +858,16 @@ class LicensePlateProcessingMixin:
         results = [["", 0.0]] * len(images)
         indices = np.argsort(np.array([x.shape[1] / x.shape[0] for x in images]))
 
-        outputs = np.stack(outputs)
+        stacked_outputs = np.stack(outputs)
 
-        outputs = [
-            (labels[idx], outputs[i, idx])
-            for i, idx in enumerate(outputs.argmax(axis=1))
+        stacked_outputs = [
+            (labels[idx], stacked_outputs[i, idx])
+            for i, idx in enumerate(stacked_outputs.argmax(axis=1))
         ]
 
         for i in range(0, len(images), self.batch_size):
-            for j in range(len(outputs)):
-                label, score = outputs[j]
+            for j in range(len(stacked_outputs)):
+                label, score = stacked_outputs[j]
                 results[indices[i + j]] = [label, score]
                 # make sure we have high confidence if we need to flip a box
                 if "180" in label and score >= 0.7:
@@ -854,10 +875,10 @@ class LicensePlateProcessingMixin:
                         images[indices[i + j]], cv2.ROTATE_180
                     )
 
-        return images, results
+        return images, results  # type: ignore[return-value]
 
     def _preprocess_recognition_image(
-        self, camera: string, image: np.ndarray, max_wh_ratio: float
+        self, camera: str, image: np.ndarray, max_wh_ratio: float
     ) -> np.ndarray:
         """
         Preprocess an image for recognition by dynamically adjusting its width.
@@ -925,7 +946,7 @@ class LicensePlateProcessingMixin:
         input_w = int(input_h * max_wh_ratio)
 
         # check for model-specific input width
-        model_input_w = self.model_runner.recognition_model.runner.get_input_width()
+        model_input_w = self.model_runner.recognition_model.runner.get_input_width()  # type: ignore[union-attr]
         if isinstance(model_input_w, int) and model_input_w > 0:
             input_w = model_input_w
 
@@ -945,7 +966,7 @@ class LicensePlateProcessingMixin:
         padded_image[:, :, :resized_w] = resized_image
 
         if False:
-            current_time = int(datetime.datetime.now().timestamp() * 1000)
+            current_time = int(datetime.datetime.now().timestamp() * 1000)  # type: ignore[unreachable]
             cv2.imwrite(
                 f"debug/frames/preprocessed_recognition_{current_time}.jpg",
                 image,
@@ -983,8 +1004,9 @@ class LicensePlateProcessingMixin:
                 np.linalg.norm(points[1] - points[2]),
             )
         )
-        pts_std = np.float32(
-            [[0, 0], [crop_width, 0], [crop_width, crop_height], [0, crop_height]]
+        pts_std = np.array(
+            [[0, 0], [crop_width, 0], [crop_width, crop_height], [0, crop_height]],
+            dtype=np.float32,
         )
         matrix = cv2.getPerspectiveTransform(points, pts_std)
         image = cv2.warpPerspective(
@@ -1000,15 +1022,15 @@ class LicensePlateProcessingMixin:
         return image
 
     def _detect_license_plate(
-        self, camera: string, input: np.ndarray
-    ) -> tuple[int, int, int, int]:
+        self, camera: str, input: np.ndarray
+    ) -> tuple[int, int, int, int] | None:
         """
         Use a lightweight YOLOv9 model to detect license plates for users without Frigate+
 
         Return the dimensions of the detected plate as [x1, y1, x2, y2].
         """
         try:
-            predictions = self.model_runner.yolov9_detection_model(input)
+            predictions = self.model_runner.yolov9_detection_model(input)  # type: ignore[arg-type]
         except Exception as e:
             logger.warning(f"Error running YOLOv9 license plate detection model: {e}")
             return None
@@ -1073,7 +1095,7 @@ class LicensePlateProcessingMixin:
             logger.debug(
                 f"{camera}: Found license plate. Bounding box: {expanded_box.astype(int)}"
             )
-            return tuple(expanded_box.astype(int))
+            return tuple(expanded_box.astype(int))  # type: ignore[return-value]
         else:
             return None  # No detection above the threshold
 
@@ -1097,7 +1119,7 @@ class LicensePlateProcessingMixin:
                 f"  Variant {i + 1}: '{p['plate']}' (conf: {p['conf']:.3f}, area: {p['area']})"
             )
 
-        clusters = []
+        clusters: list[list[dict[str, Any]]] = []
         for i, plate in enumerate(plates):
             merged = False
             for j, cluster in enumerate(clusters):
@@ -1132,7 +1154,7 @@ class LicensePlateProcessingMixin:
             )
 
         # Best cluster: largest size, tiebroken by max conf
-        def cluster_score(c):
+        def cluster_score(c: list[dict[str, Any]]) -> tuple[int, float]:
             return (len(c), max(v["conf"] for v in c))
 
         best_cluster_idx = max(
@@ -1178,7 +1200,7 @@ class LicensePlateProcessingMixin:
 
     def lpr_process(
         self, obj_data: dict[str, Any], frame: np.ndarray, dedicated_lpr: bool = False
-    ):
+    ) -> None:
         """Look for license plates in image."""
         self.metrics.alpr_pps.value = self.plates_rec_second.eps()
         self.metrics.yolov9_lpr_pps.value = self.plates_det_second.eps()
@@ -1195,7 +1217,7 @@ class LicensePlateProcessingMixin:
             rgb = cv2.cvtColor(frame, cv2.COLOR_YUV2BGR_I420)
 
             # apply motion mask
-            rgb[self.config.cameras[obj_data].motion.rasterized_mask == 0] = [0, 0, 0]
+            rgb[self.config.cameras[camera].motion.rasterized_mask == 0] = [0, 0, 0]  # type: ignore[attr-defined]
 
             if WRITE_DEBUG_IMAGES:
                 cv2.imwrite(
@@ -1261,7 +1283,7 @@ class LicensePlateProcessingMixin:
                 "stationary", False
             ):
                 logger.debug(
-                    f"{camera}: Skipping LPR for non-stationary {obj_data['label']} object {id} with no position changes.  (Detected in {self.config.cameras[camera].detect.min_initialized + 1} concurrent frames, threshold to run is {self.config.cameras[camera].detect.min_initialized + 2} frames)"
+                    f"{camera}: Skipping LPR for non-stationary {obj_data['label']} object {id} with no position changes.  (Detected in {self.config.cameras[camera].detect.min_initialized + 1} concurrent frames, threshold to run is {self.config.cameras[camera].detect.min_initialized + 2} frames)"  # type: ignore[operator]
                 )
                 return
 
@@ -1288,7 +1310,7 @@ class LicensePlateProcessingMixin:
                     if time_since_stationary > self.stationary_scan_duration:
                         return
 
-            license_plate: Optional[dict[str, Any]] = None
+            license_plate = None
 
             if "license_plate" not in self.config.cameras[camera].objects.track:
                 logger.debug(f"{camera}: Running manual license_plate detection.")
@@ -1301,7 +1323,7 @@ class LicensePlateProcessingMixin:
                 rgb = cv2.cvtColor(frame, cv2.COLOR_YUV2BGR_I420)
 
                 # apply motion mask
-                rgb[self.config.cameras[camera].motion.rasterized_mask == 0] = [0, 0, 0]
+                rgb[self.config.cameras[camera].motion.rasterized_mask == 0] = [0, 0, 0]  # type: ignore[attr-defined]
 
                 left, top, right, bottom = car_box
                 car = rgb[top:bottom, left:right]
@@ -1378,10 +1400,10 @@ class LicensePlateProcessingMixin:
                         if attr.get("label") != "license_plate":
                             continue
 
-                        if license_plate is None or attr.get(
+                        if license_plate is None or attr.get(  # type: ignore[unreachable]
                             "score", 0.0
                         ) > license_plate.get("score", 0.0):
-                            license_plate = attr
+                            license_plate = attr  # type: ignore[assignment]
 
                     # no license plates detected in this frame
                     if not license_plate:
@@ -1389,9 +1411,9 @@ class LicensePlateProcessingMixin:
 
                 # we are using dedicated lpr with frigate+
                 if obj_data.get("label") == "license_plate":
-                    license_plate = obj_data
+                    license_plate = obj_data  # type: ignore[assignment]
 
-                license_plate_box = license_plate.get("box")
+                license_plate_box = license_plate.get("box")  # type: ignore[attr-defined]
 
                 # check that license plate is valid
                 if (
@@ -1420,7 +1442,7 @@ class LicensePlateProcessingMixin:
                     0, [license_plate_frame.shape[1], license_plate_frame.shape[0]] * 2
                 )
 
-                plate_box = tuple(int(x) for x in expanded_box)
+                plate_box = tuple(int(x) for x in expanded_box)  # type: ignore[assignment]
 
                 # Crop using the expanded box
                 license_plate_frame = license_plate_frame[
@@ -1596,7 +1618,7 @@ class LicensePlateProcessingMixin:
             sub_label = next(
                 (
                     label
-                    for label, plates_list in self.lpr_config.known_plates.items()
+                    for label, plates_list in self.lpr_config.known_plates.items()  # type: ignore[union-attr]
                     if any(
                         re.match(f"^{plate}$", rep_plate)
                         or Levenshtein.distance(plate, rep_plate)
@@ -1649,14 +1671,16 @@ class LicensePlateProcessingMixin:
             frame_bgr = cv2.cvtColor(frame, cv2.COLOR_YUV2BGR_I420)
             _, encoded_img = cv2.imencode(".jpg", frame_bgr)
             self.sub_label_publisher.publish(
-                (base64.b64encode(encoded_img).decode("ASCII"), id, camera),
+                (base64.b64encode(encoded_img.tobytes()).decode("ASCII"), id, camera),
                 EventMetadataTypeEnum.save_lpr_snapshot.value,
             )
 
-    def handle_request(self, topic, request_data) -> dict[str, Any] | None:
-        return
+    def handle_request(
+        self, topic: str, request_data: dict[str, Any]
+    ) -> dict[str, Any] | None:
+        return None
 
-    def lpr_expire(self, object_id: str, camera: str):
+    def lpr_expire(self, object_id: str, camera: str) -> None:
         if object_id in self.detected_license_plates:
             self.detected_license_plates.pop(object_id)
 
@@ -1673,7 +1697,7 @@ class CTCDecoder:
     for each decoded character sequence.
     """
 
-    def __init__(self, character_dict_path=None):
+    def __init__(self, character_dict_path: str | None = None) -> None:
         """
         Initializes the CTCDecoder.
         :param character_dict_path: Path to the character dictionary file.

frigate/data_processing/common/license_plate/model.py

@@ -1,3 +1,4 @@
+from frigate.comms.inter_process import InterProcessRequestor
 from frigate.embeddings.onnx.lpr_embedding import (
     LicensePlateDetector,
     PaddleOCRClassification,
@@ -9,7 +10,12 @@ from ...types import DataProcessorModelRunner
 
 
 class LicensePlateModelRunner(DataProcessorModelRunner):
-    def __init__(self, requestor, device: str = "CPU", model_size: str = "small"):
+    def __init__(
+        self,
+        requestor: InterProcessRequestor,
+        device: str = "CPU",
+        model_size: str = "small",
+    ):
         super().__init__(requestor, device, model_size)
         self.detection_model = PaddleOCRDetection(
             model_size=model_size, requestor=requestor, device=device

frigate/data_processing/post/api.py

@@ -17,7 +17,7 @@ class PostProcessorApi(ABC):
         self,
         config: FrigateConfig,
         metrics: DataProcessorMetrics,
-        model_runner: DataProcessorModelRunner,
+        model_runner: DataProcessorModelRunner | None,
     ) -> None:
         self.config = config
         self.metrics = metrics
@@ -41,7 +41,7 @@ class PostProcessorApi(ABC):
     @abstractmethod
     def handle_request(
         self, topic: str, request_data: dict[str, Any]
-    ) -> dict[str, Any] | None:
+    ) -> dict[str, Any] | str | None:
         """Handle metadata requests.
         Args:
             request_data (dict): containing data about requested change to process.

frigate/data_processing/post/audio_transcription.py

@@ -4,7 +4,7 @@ import logging
 import os
 import threading
 import time
-from typing import Optional
+from typing import Any, Optional
 
 from peewee import DoesNotExist
 
@@ -17,6 +17,7 @@ from frigate.const import (
     UPDATE_EVENT_DESCRIPTION,
 )
 from frigate.data_processing.types import PostProcessDataEnum
+from frigate.embeddings.embeddings import Embeddings
 from frigate.types import TrackedObjectUpdateTypesEnum
 from frigate.util.audio import get_audio_from_recording
 
@@ -31,7 +32,7 @@ class AudioTranscriptionPostProcessor(PostProcessorApi):
         self,
         config: FrigateConfig,
         requestor: InterProcessRequestor,
-        embeddings,
+        embeddings: Embeddings,
         metrics: DataProcessorMetrics,
     ):
         super().__init__(config, metrics, None)
@@ -40,7 +41,7 @@ class AudioTranscriptionPostProcessor(PostProcessorApi):
         self.embeddings = embeddings
         self.recognizer = None
         self.transcription_lock = threading.Lock()
-        self.transcription_thread = None
+        self.transcription_thread: threading.Thread | None = None
         self.transcription_running = False
 
         # faster-whisper handles model downloading automatically
@@ -69,7 +70,7 @@ class AudioTranscriptionPostProcessor(PostProcessorApi):
             self.recognizer = None
 
     def process_data(
-        self, data: dict[str, any], data_type: PostProcessDataEnum
+        self, data: dict[str, Any], data_type: PostProcessDataEnum
     ) -> None:
         """Transcribe audio from a recording.
 
@@ -141,13 +142,13 @@ class AudioTranscriptionPostProcessor(PostProcessorApi):
         except Exception as e:
             logger.error(f"Error in audio transcription post-processing: {e}")
 
-    def __transcribe_audio(self, audio_data: bytes) -> Optional[tuple[str, float]]:
+    def __transcribe_audio(self, audio_data: bytes) -> Optional[str]:
         """Transcribe WAV audio data using faster-whisper."""
         if not self.recognizer:
             logger.debug("Recognizer not initialized")
             return None
 
-        try:
+        try:  # type: ignore[unreachable]
             # Save audio data to a temporary wav (faster-whisper expects a file)
             temp_wav = os.path.join(CACHE_DIR, f"temp_audio_{int(time.time())}.wav")
             with open(temp_wav, "wb") as f:
@@ -176,7 +177,7 @@ class AudioTranscriptionPostProcessor(PostProcessorApi):
             logger.error(f"Error transcribing audio: {e}")
             return None
 
-    def _transcription_wrapper(self, event: dict[str, any]) -> None:
+    def _transcription_wrapper(self, event: dict[str, Any]) -> None:
         """Wrapper to run transcription and reset running flag when done."""
         try:
             self.process_data(
@@ -194,7 +195,7 @@ class AudioTranscriptionPostProcessor(PostProcessorApi):
 
             self.requestor.send_data(UPDATE_AUDIO_TRANSCRIPTION_STATE, "idle")
 
-    def handle_request(self, topic: str, request_data: dict[str, any]) -> str | None:
+    def handle_request(self, topic: str, request_data: dict[str, Any]) -> str | None:
         if topic == "transcribe_audio":
             event = request_data["event"]
 

frigate/data_processing/post/license_plate.py

@@ -29,7 +29,7 @@ from .api import PostProcessorApi
 logger = logging.getLogger(__name__)
 
 
-class LicensePlatePostProcessor(LicensePlateProcessingMixin, PostProcessorApi):
+class LicensePlatePostProcessor(LicensePlateProcessingMixin, PostProcessorApi):  # type: ignore[misc]
     def __init__(
         self,
         config: FrigateConfig,
@@ -71,7 +71,7 @@ class LicensePlatePostProcessor(LicensePlateProcessingMixin, PostProcessorApi):
         # don't run LPR post processing for now
         return
 
-        event_id = data["event_id"]
+        event_id = data["event_id"]  # type: ignore[unreachable]
         camera_name = data["camera"]
 
         if data_type == PostProcessDataEnum.recording:
@@ -225,7 +225,7 @@ class LicensePlatePostProcessor(LicensePlateProcessingMixin, PostProcessorApi):
         logger.debug(f"Post processing plate: {event_id}, {frame_time}")
         self.lpr_process(keyframe_obj_data, frame)
 
-    def handle_request(self, topic, request_data) -> dict[str, Any] | None:
+    def handle_request(self, topic: str, request_data: dict) -> dict[str, Any] | None:
         if topic == EmbeddingsRequestEnum.reprocess_plate.value:
             event = request_data["event"]
 
@@ -242,3 +242,5 @@ class LicensePlatePostProcessor(LicensePlateProcessingMixin, PostProcessorApi):
                 "message": "Successfully requested reprocessing of license plate.",
                 "success": True,
             }
+
+        return None

frigate/data_processing/post/object_descriptions.py

@@ -24,7 +24,7 @@ from frigate.util.file import get_event_thumbnail_bytes, load_event_snapshot_ima
 from frigate.util.image import create_thumbnail, ensure_jpeg_bytes
 
 if TYPE_CHECKING:
-    from frigate.embeddings import Embeddings
+    from frigate.embeddings.embeddings import Embeddings
 
 from ..post.api import PostProcessorApi
 from ..types import DataProcessorMetrics
@@ -139,7 +139,7 @@ class ObjectDescriptionProcessor(PostProcessorApi):
         ):
             self._process_genai_description(event, camera_config, thumbnail)
         else:
-            self.cleanup_event(event.id)
+            self.cleanup_event(str(event.id))
 
     def __regenerate_description(self, event_id: str, source: str, force: bool) -> None:
         """Regenerate the description for an event."""
@@ -149,17 +149,17 @@ class ObjectDescriptionProcessor(PostProcessorApi):
             logger.error(f"Event {event_id} not found for description regeneration")
             return
 
-        if self.genai_client is None:
-            logger.error("GenAI not enabled")
-            return
-
-        camera_config = self.config.cameras[event.camera]
+        camera_config = self.config.cameras[str(event.camera)]
         if not camera_config.objects.genai.enabled and not force:
             logger.error(f"GenAI not enabled for camera {event.camera}")
             return
 
         thumbnail = get_event_thumbnail_bytes(event)
 
+        if thumbnail is None:
+            logger.error("No thumbnail available for %s", event.id)
+            return
+
         # ensure we have a jpeg to pass to the model
         thumbnail = ensure_jpeg_bytes(thumbnail)
 
@@ -187,7 +187,9 @@ class ObjectDescriptionProcessor(PostProcessorApi):
             )
         )
 
-        self._genai_embed_description(event, embed_image)
+        self._genai_embed_description(
+            event, [img for img in embed_image if img is not None]
+        )
 
     def process_data(self, frame_data: dict, data_type: PostProcessDataEnum) -> None:
         """Process a frame update."""
@@ -241,7 +243,7 @@ class ObjectDescriptionProcessor(PostProcessorApi):
             # Crop snapshot based on region
             # provide full image if region doesn't exist (manual events)
             height, width = img.shape[:2]
-            x1_rel, y1_rel, width_rel, height_rel = event.data.get(
+            x1_rel, y1_rel, width_rel, height_rel = event.data.get(  # type: ignore[attr-defined]
                 "region", [0, 0, 1, 1]
             )
             x1, y1 = int(x1_rel * width), int(y1_rel * height)
@@ -258,14 +260,16 @@ class ObjectDescriptionProcessor(PostProcessorApi):
             return None
 
     def _process_genai_description(
-        self, event: Event, camera_config: CameraConfig, thumbnail
+        self, event: Event, camera_config: CameraConfig, thumbnail: bytes
     ) -> None:
+        event_id = str(event.id)
+
         if event.has_snapshot and camera_config.objects.genai.use_snapshot:
             snapshot_image = self._read_and_crop_snapshot(event)
             if not snapshot_image:
                 return
 
-        num_thumbnails = len(self.tracked_events.get(event.id, []))
+        num_thumbnails = len(self.tracked_events.get(event_id, []))
 
         # ensure we have a jpeg to pass to the model
         thumbnail = ensure_jpeg_bytes(thumbnail)
@@ -277,7 +281,7 @@ class ObjectDescriptionProcessor(PostProcessorApi):
             else (
                 [
                     data["thumbnail"][:] if data.get("thumbnail") else None
-                    for data in self.tracked_events[event.id]
+                    for data in self.tracked_events[event_id]
                     if data.get("thumbnail")
                 ]
                 if num_thumbnails > 0
@@ -286,22 +290,22 @@ class ObjectDescriptionProcessor(PostProcessorApi):
         )
 
         if camera_config.objects.genai.debug_save_thumbnails and num_thumbnails > 0:
-            logger.debug(f"Saving {num_thumbnails} thumbnails for event {event.id}")
+            logger.debug(f"Saving {num_thumbnails} thumbnails for event {event_id}")
 
-            Path(os.path.join(CLIPS_DIR, f"genai-requests/{event.id}")).mkdir(
+            Path(os.path.join(CLIPS_DIR, f"genai-requests/{event_id}")).mkdir(
                 parents=True, exist_ok=True
             )
 
-            for idx, data in enumerate(self.tracked_events[event.id], 1):
+            for idx, data in enumerate(self.tracked_events[event_id], 1):
                 jpg_bytes: bytes | None = data["thumbnail"]
 
                 if jpg_bytes is None:
-                    logger.warning(f"Unable to save thumbnail {idx} for {event.id}.")
+                    logger.warning(f"Unable to save thumbnail {idx} for {event_id}.")
                 else:
                     with open(
                         os.path.join(
                             CLIPS_DIR,
-                            f"genai-requests/{event.id}/{idx}.jpg",
+                            f"genai-requests/{event_id}/{idx}.jpg",
                         ),
                         "wb",
                     ) as j:
@@ -310,7 +314,7 @@ class ObjectDescriptionProcessor(PostProcessorApi):
         # Generate the description. Call happens in a thread since it is network bound.
         threading.Thread(
             target=self._genai_embed_description,
-            name=f"_genai_embed_description_{event.id}",
+            name=f"_genai_embed_description_{event_id}",
             daemon=True,
             args=(
                 event,
@@ -319,12 +323,12 @@ class ObjectDescriptionProcessor(PostProcessorApi):
         ).start()
 
         # Clean up tracked events and early request state
-        self.cleanup_event(event.id)
+        self.cleanup_event(event_id)
 
     def _genai_embed_description(self, event: Event, thumbnails: list[bytes]) -> None:
         """Embed the description for an event."""
         start = datetime.datetime.now().timestamp()
-        camera_config = self.config.cameras[event.camera]
+        camera_config = self.config.cameras[str(event.camera)]
         description = self.genai_client.generate_object_description(
             camera_config, thumbnails, event
         )
@@ -346,7 +350,7 @@ class ObjectDescriptionProcessor(PostProcessorApi):
 
         # Embed the description
         if self.config.semantic_search.enabled:
-            self.embeddings.embed_description(event.id, description)
+            self.embeddings.embed_description(str(event.id), description)
 
             # Check semantic trigger for this description
             if self.semantic_trigger_processor is not None:

frigate/data_processing/post/review_descriptions.py

@@ -48,8 +48,8 @@ class ReviewDescriptionProcessor(PostProcessorApi):
         self.metrics = metrics
         self.genai_client = client
         self.review_desc_speed = InferenceSpeed(self.metrics.review_desc_speed)
-        self.review_descs_dps = EventsPerSecond()
-        self.review_descs_dps.start()
+        self.review_desc_dps = EventsPerSecond()
+        self.review_desc_dps.start()
 
     def calculate_frame_count(
         self,
@@ -59,7 +59,7 @@ class ReviewDescriptionProcessor(PostProcessorApi):
     ) -> int:
         """Calculate optimal number of frames based on context size, image source, and resolution.
 
-        Token usage varies by resolution: larger images (ultrawide aspect ratios) use more tokens.
+        Token usage varies by resolution: larger images (ultra-wide aspect ratios) use more tokens.
         Estimates ~1 token per 1250 pixels. Targets 98% context utilization with safety margin.
         Capped at 20 frames.
         """
@@ -68,7 +68,11 @@ class ReviewDescriptionProcessor(PostProcessorApi):
 
         detect_width = camera_config.detect.width
         detect_height = camera_config.detect.height
-        aspect_ratio = detect_width / detect_height
+
+        if not detect_width or not detect_height:
+            aspect_ratio = 16 / 9
+        else:
+            aspect_ratio = detect_width / detect_height
 
         if image_source == ImageSourceEnum.recordings:
             if aspect_ratio >= 1:
@@ -99,8 +103,10 @@ class ReviewDescriptionProcessor(PostProcessorApi):
 
         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()
+    def process_data(
+        self, data: dict[str, Any], data_type: PostProcessDataEnum
+    ) -> None:
+        self.metrics.review_desc_dps.value = self.review_desc_dps.eps()
 
         if data_type != PostProcessDataEnum.review:
             return
@@ -186,7 +192,7 @@ class ReviewDescriptionProcessor(PostProcessorApi):
                 )
 
             # kickoff analysis
-            self.review_descs_dps.update()
+            self.review_desc_dps.update()
             threading.Thread(
                 target=run_analysis,
                 args=(
@@ -202,7 +208,7 @@ class ReviewDescriptionProcessor(PostProcessorApi):
                 ),
             ).start()
 
-    def handle_request(self, topic, request_data):
+    def handle_request(self, topic: str, request_data: dict[str, Any]) -> str | None:
         if topic == EmbeddingsRequestEnum.summarize_review.value:
             start_ts = request_data["start_ts"]
             end_ts = request_data["end_ts"]
@@ -327,7 +333,7 @@ class ReviewDescriptionProcessor(PostProcessorApi):
         file_start = f"preview_{camera}-"
         start_file = f"{file_start}{start_time}.webp"
         end_file = f"{file_start}{end_time}.webp"
-        all_frames = []
+        all_frames: list[str] = []
 
         for file in sorted(os.listdir(preview_dir)):
             if not file.startswith(file_start):
@@ -465,7 +471,7 @@ class ReviewDescriptionProcessor(PostProcessorApi):
             thumb_data = cv2.imread(thumb_path)
 
             if thumb_data is None:
-                logger.warning(
+                logger.warning(  # type: ignore[unreachable]
                     "Could not read preview frame at %s, skipping", thumb_path
                 )
                 continue
@@ -488,13 +494,12 @@ class ReviewDescriptionProcessor(PostProcessorApi):
         return thumbs
 
 
-@staticmethod
 def run_analysis(
     requestor: InterProcessRequestor,
     genai_client: GenAIClient,
     review_inference_speed: InferenceSpeed,
     camera_config: CameraConfig,
-    final_data: dict[str, str],
+    final_data: dict[str, Any],
     thumbs: list[bytes],
     genai_config: GenAIReviewConfig,
     labelmap_objects: list[str],

frigate/data_processing/post/semantic_trigger.py

@@ -19,6 +19,7 @@ from frigate.config import FrigateConfig
 from frigate.const import CONFIG_DIR
 from frigate.data_processing.types import PostProcessDataEnum
 from frigate.db.sqlitevecq import SqliteVecQueueDatabase
+from frigate.embeddings.embeddings import Embeddings
 from frigate.embeddings.util import ZScoreNormalization
 from frigate.models import Event, Trigger
 from frigate.util.builtin import cosine_distance
@@ -40,8 +41,8 @@ class SemanticTriggerProcessor(PostProcessorApi):
         requestor: InterProcessRequestor,
         sub_label_publisher: EventMetadataPublisher,
         metrics: DataProcessorMetrics,
-        embeddings,
-    ):
+        embeddings: Embeddings,
+    ) -> None:
         super().__init__(config, metrics, None)
         self.db = db
         self.embeddings = embeddings
@@ -236,11 +237,14 @@ class SemanticTriggerProcessor(PostProcessorApi):
                     return
 
                 # Skip the event if not an object
-                if event.data.get("type") != "object":
+                if event.data.get("type") != "object":  # type: ignore[attr-defined]
                     return
 
                 thumbnail_bytes = get_event_thumbnail_bytes(event)
 
+                if thumbnail_bytes is None:
+                    return
+
                 nparr = np.frombuffer(thumbnail_bytes, np.uint8)
                 thumbnail = cv2.imdecode(nparr, cv2.IMREAD_COLOR)
 
@@ -262,8 +266,10 @@ class SemanticTriggerProcessor(PostProcessorApi):
                     thumbnail,
                 )
 
-    def handle_request(self, topic, request_data):
+    def handle_request(
+        self, topic: str, request_data: dict[str, Any]
+    ) -> dict[str, Any] | str | None:
         return None
 
-    def expire_object(self, object_id, camera):
+    def expire_object(self, object_id: str, camera: str) -> None:
         pass

frigate/data_processing/real_time/audio_transcription.py

@@ -4,7 +4,7 @@ import logging
 import os
 import queue
 import threading
-from typing import Optional
+from typing import Any, Optional
 
 import numpy as np
 
@@ -39,11 +39,11 @@ class AudioTranscriptionRealTimeProcessor(RealTimeProcessorApi):
         self.config = config
         self.camera_config = camera_config
         self.requestor = requestor
-        self.stream = None
-        self.whisper_model = None
+        self.stream: Any = None
+        self.whisper_model: FasterWhisperASR | None = None
         self.model_runner = model_runner
-        self.transcription_segments = []
-        self.audio_queue = queue.Queue()
+        self.transcription_segments: list[str] = []
+        self.audio_queue: queue.Queue[tuple[dict[str, Any], np.ndarray]] = queue.Queue()
         self.stop_event = stop_event
 
     def __build_recognizer(self) -> None:
@@ -142,10 +142,10 @@ class AudioTranscriptionRealTimeProcessor(RealTimeProcessorApi):
             logger.error(f"Error processing audio stream: {e}")
             return None
 
-    def process_frame(self, obj_data: dict[str, any], frame: np.ndarray) -> None:
+    def process_frame(self, obj_data: dict[str, Any], frame: np.ndarray) -> None:
         pass
 
-    def process_audio(self, obj_data: dict[str, any], audio: np.ndarray) -> bool | None:
+    def process_audio(self, obj_data: dict[str, Any], audio: np.ndarray) -> bool | None:
         if audio is None or audio.size == 0:
             logger.debug("No audio data provided for transcription")
             return None
@@ -269,13 +269,13 @@ class AudioTranscriptionRealTimeProcessor(RealTimeProcessorApi):
         )
 
     def handle_request(
-        self, topic: str, request_data: dict[str, any]
-    ) -> dict[str, any] | None:
+        self, topic: str, request_data: dict[str, Any]
+    ) -> dict[str, Any] | None:
         if topic == "clear_audio_recognizer":
             self.stream = None
             self.__build_recognizer()
             return {"message": "Audio recognizer cleared and rebuilt", "success": True}
         return None
 
-    def expire_object(self, object_id: str) -> None:
+    def expire_object(self, object_id: str, camera: str) -> None:
         pass

frigate/data_processing/real_time/bird.py

@@ -14,7 +14,7 @@ from frigate.comms.event_metadata_updater import (
 from frigate.config import FrigateConfig
 from frigate.const import MODEL_CACHE_DIR
 from frigate.log import suppress_stderr_during
-from frigate.util.object import calculate_region
+from frigate.util.image import calculate_region
 
 from ..types import DataProcessorMetrics
 from .api import RealTimeProcessorApi
@@ -35,10 +35,10 @@ class BirdRealTimeProcessor(RealTimeProcessorApi):
         metrics: DataProcessorMetrics,
     ):
         super().__init__(config, metrics)
-        self.interpreter: Interpreter = None
+        self.interpreter: Interpreter | None = None
         self.sub_label_publisher = sub_label_publisher
-        self.tensor_input_details: dict[str, Any] = None
-        self.tensor_output_details: dict[str, Any] = None
+        self.tensor_input_details: list[dict[str, Any]] | None = None
+        self.tensor_output_details: list[dict[str, Any]] | None = None
         self.detected_birds: dict[str, float] = {}
         self.labelmap: dict[int, str] = {}
 
@@ -61,7 +61,7 @@ class BirdRealTimeProcessor(RealTimeProcessorApi):
             self.downloader = ModelDownloader(
                 model_name="bird",
                 download_path=download_path,
-                file_names=self.model_files.keys(),
+                file_names=list(self.model_files.keys()),
                 download_func=self.__download_models,
                 complete_func=self.__build_detector,
             )
@@ -102,8 +102,12 @@ class BirdRealTimeProcessor(RealTimeProcessorApi):
                 i += 1
                 line = f.readline()
 
-    def process_frame(self, obj_data, frame):
-        if not self.interpreter:
+    def process_frame(self, obj_data: dict[str, Any], frame: np.ndarray) -> None:
+        if (
+            not self.interpreter
+            or not self.tensor_input_details
+            or not self.tensor_output_details
+        ):
             return
 
         if obj_data["label"] != "bird":
@@ -145,7 +149,7 @@ class BirdRealTimeProcessor(RealTimeProcessorApi):
             self.tensor_output_details[0]["index"]
         )[0]
         probs = res / res.sum(axis=0)
-        best_id = np.argmax(probs)
+        best_id = int(np.argmax(probs))
 
         if best_id == 964:
             logger.debug("No bird classification was detected.")
@@ -179,9 +183,11 @@ class BirdRealTimeProcessor(RealTimeProcessorApi):
         self.config.classification = payload
         logger.debug("Bird classification config updated dynamically")
 
-    def handle_request(self, topic, request_data):
+    def handle_request(
+        self, topic: str, request_data: dict[str, Any]
+    ) -> dict[str, Any] | None:
         return None
 
-    def expire_object(self, object_id, camera):
+    def expire_object(self, object_id: str, camera: str) -> None:
         if object_id in self.detected_birds:
             self.detected_birds.pop(object_id)

frigate/data_processing/real_time/custom_classification.py

@@ -24,7 +24,8 @@ from frigate.const import CLIPS_DIR, MODEL_CACHE_DIR
 from frigate.log import suppress_stderr_during
 from frigate.types import TrackedObjectUpdateTypesEnum
 from frigate.util.builtin import EventsPerSecond, InferenceSpeed, load_labels
-from frigate.util.object import box_overlaps, calculate_region
+from frigate.util.image import calculate_region
+from frigate.util.object import box_overlaps
 
 from ..types import DataProcessorMetrics
 from .api import RealTimeProcessorApi
@@ -49,12 +50,16 @@ class CustomStateClassificationProcessor(RealTimeProcessorApi):
     ):
         super().__init__(config, metrics)
         self.model_config = model_config
+
+        if not self.model_config.name:
+            raise ValueError("Custom classification model name must be set.")
+
         self.requestor = requestor
         self.model_dir = os.path.join(MODEL_CACHE_DIR, self.model_config.name)
         self.train_dir = os.path.join(CLIPS_DIR, self.model_config.name, "train")
-        self.interpreter: Interpreter = None
-        self.tensor_input_details: dict[str, Any] | None = None
-        self.tensor_output_details: dict[str, Any] | None = None
+        self.interpreter: Interpreter | None = None
+        self.tensor_input_details: list[dict[str, Any]] | None = None
+        self.tensor_output_details: list[dict[str, Any]] | None = None
         self.labelmap: dict[int, str] = {}
         self.classifications_per_second = EventsPerSecond()
         self.state_history: dict[str, dict[str, Any]] = {}
@@ -63,7 +68,7 @@ class CustomStateClassificationProcessor(RealTimeProcessorApi):
             self.metrics
             and self.model_config.name in self.metrics.classification_speeds
         ):
-            self.inference_speed = InferenceSpeed(
+            self.inference_speed: InferenceSpeed | None = InferenceSpeed(
                 self.metrics.classification_speeds[self.model_config.name]
             )
         else:
@@ -172,12 +177,20 @@ class CustomStateClassificationProcessor(RealTimeProcessorApi):
 
         return None
 
-    def process_frame(self, frame_data: dict[str, Any], frame: np.ndarray):
+    def process_frame(self, frame_data: dict[str, Any], frame: np.ndarray) -> None:
+        if (
+            not self.model_config.name
+            or not self.model_config.state_config
+            or not self.tensor_input_details
+            or not self.tensor_output_details
+        ):
+            return
+
         if self.metrics and self.model_config.name in self.metrics.classification_cps:
             self.metrics.classification_cps[
                 self.model_config.name
             ].value = self.classifications_per_second.eps()
-        camera = frame_data.get("camera")
+        camera = str(frame_data.get("camera"))
 
         if camera not in self.model_config.state_config.cameras:
             return
@@ -283,7 +296,7 @@ class CustomStateClassificationProcessor(RealTimeProcessorApi):
         logger.debug(
             f"{self.model_config.name} Ran state classification with probabilities: {probs}"
         )
-        best_id = np.argmax(probs)
+        best_id = int(np.argmax(probs))
         score = round(probs[best_id], 2)
         self.__update_metrics(datetime.datetime.now().timestamp() - now)
 
@@ -319,7 +332,9 @@ class CustomStateClassificationProcessor(RealTimeProcessorApi):
                 verified_state,
             )
 
-    def handle_request(self, topic, request_data):
+    def handle_request(
+        self, topic: str, request_data: dict[str, Any]
+    ) -> dict[str, Any] | None:
         if topic == EmbeddingsRequestEnum.reload_classification_model.value:
             if request_data.get("model_name") == self.model_config.name:
                 self.__build_detector()
@@ -335,7 +350,7 @@ class CustomStateClassificationProcessor(RealTimeProcessorApi):
         else:
             return None
 
-    def expire_object(self, object_id, camera):
+    def expire_object(self, object_id: str, camera: str) -> None:
         pass
 
 
@@ -350,13 +365,17 @@ class CustomObjectClassificationProcessor(RealTimeProcessorApi):
     ):
         super().__init__(config, metrics)
         self.model_config = model_config
+
+        if not self.model_config.name:
+            raise ValueError("Custom classification model name must be set.")
+
         self.model_dir = os.path.join(MODEL_CACHE_DIR, self.model_config.name)
         self.train_dir = os.path.join(CLIPS_DIR, self.model_config.name, "train")
-        self.interpreter: Interpreter = None
+        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.tensor_input_details: list[dict[str, Any]] | None = None
+        self.tensor_output_details: list[dict[str, Any]] | None = None
         self.classification_history: dict[str, list[tuple[str, float, float]]] = {}
         self.labelmap: dict[int, str] = {}
         self.classifications_per_second = EventsPerSecond()
@@ -365,7 +384,7 @@ class CustomObjectClassificationProcessor(RealTimeProcessorApi):
             self.metrics
             and self.model_config.name in self.metrics.classification_speeds
         ):
-            self.inference_speed = InferenceSpeed(
+            self.inference_speed: InferenceSpeed | None = InferenceSpeed(
                 self.metrics.classification_speeds[self.model_config.name]
             )
         else:
@@ -431,8 +450,8 @@ class CustomObjectClassificationProcessor(RealTimeProcessorApi):
             )
             return None, 0.0
 
-        label_counts = {}
-        label_scores = {}
+        label_counts: dict[str, int] = {}
+        label_scores: dict[str, list[float]] = {}
         total_attempts = len(history)
 
         for label, score, timestamp in history:
@@ -443,7 +462,7 @@ class CustomObjectClassificationProcessor(RealTimeProcessorApi):
             label_counts[label] += 1
             label_scores[label].append(score)
 
-        best_label = max(label_counts, key=label_counts.get)
+        best_label = max(label_counts, key=lambda k: label_counts[k])
         best_count = label_counts[best_label]
 
         consensus_threshold = total_attempts * 0.6
@@ -470,7 +489,15 @@ class CustomObjectClassificationProcessor(RealTimeProcessorApi):
         )
         return best_label, avg_score
 
-    def process_frame(self, obj_data, frame):
+    def process_frame(self, obj_data: dict[str, Any], frame: np.ndarray) -> None:
+        if (
+            not self.model_config.name
+            or not self.model_config.object_config
+            or not self.tensor_input_details
+            or not self.tensor_output_details
+        ):
+            return
+
         if self.metrics and self.model_config.name in self.metrics.classification_cps:
             self.metrics.classification_cps[
                 self.model_config.name
@@ -555,7 +582,7 @@ class CustomObjectClassificationProcessor(RealTimeProcessorApi):
         logger.debug(
             f"{self.model_config.name} Ran object classification with probabilities: {probs}"
         )
-        best_id = np.argmax(probs)
+        best_id = int(np.argmax(probs))
         score = round(probs[best_id], 2)
         self.__update_metrics(datetime.datetime.now().timestamp() - now)
 
@@ -650,7 +677,7 @@ class CustomObjectClassificationProcessor(RealTimeProcessorApi):
                     ),
                 )
 
-    def handle_request(self, topic, request_data):
+    def handle_request(self, topic: str, request_data: dict) -> dict | None:
         if topic == EmbeddingsRequestEnum.reload_classification_model.value:
             if request_data.get("model_name") == self.model_config.name:
                 self.__build_detector()
@@ -666,12 +693,11 @@ class CustomObjectClassificationProcessor(RealTimeProcessorApi):
         else:
             return None
 
-    def expire_object(self, object_id, camera):
+    def expire_object(self, object_id: str, camera: str) -> None:
         if object_id in self.classification_history:
             self.classification_history.pop(object_id)
 
 
-@staticmethod
 def write_classification_attempt(
     folder: str,
     frame: np.ndarray,

frigate/data_processing/real_time/face.py

@@ -52,11 +52,11 @@ class FaceRealTimeProcessor(RealTimeProcessorApi):
         self.face_config = config.face_recognition
         self.requestor = requestor
         self.sub_label_publisher = sub_label_publisher
-        self.face_detector: cv2.FaceDetectorYN = None
+        self.face_detector: cv2.FaceDetectorYN | None = None
         self.requires_face_detection = "face" not in self.config.objects.all_objects
         self.person_face_history: dict[str, list[tuple[str, float, int]]] = {}
         self.camera_current_people: dict[str, list[str]] = {}
-        self.recognizer: FaceRecognizer | None = None
+        self.recognizer: FaceRecognizer
         self.faces_per_second = EventsPerSecond()
         self.inference_speed = InferenceSpeed(self.metrics.face_rec_speed)
 
@@ -78,7 +78,7 @@ class FaceRealTimeProcessor(RealTimeProcessorApi):
             self.downloader = ModelDownloader(
                 model_name="facedet",
                 download_path=download_path,
-                file_names=self.model_files.keys(),
+                file_names=list(self.model_files.keys()),
                 download_func=self.__download_models,
                 complete_func=self.__build_detector,
             )
@@ -134,7 +134,7 @@ class FaceRealTimeProcessor(RealTimeProcessorApi):
 
     def __detect_face(
         self, input: np.ndarray, threshold: float
-    ) -> tuple[int, int, int, int]:
+    ) -> tuple[int, int, int, int] | None:
         """Detect faces in input image."""
         if not self.face_detector:
             return None
@@ -153,7 +153,7 @@ class FaceRealTimeProcessor(RealTimeProcessorApi):
         faces = self.face_detector.detect(input)
 
         if faces is None or faces[1] is None:
-            return None
+            return None  # type: ignore[unreachable]
 
         face = None
 
@@ -168,7 +168,7 @@ class FaceRealTimeProcessor(RealTimeProcessorApi):
             h: int = int(raw_bbox[3] / scale_factor)
             bbox = (x, y, x + w, y + h)
 
-            if face is None or area(bbox) > area(face):
+            if face is None or area(bbox) > area(face):  # type: ignore[unreachable]
                 face = bbox
 
         return face
@@ -177,7 +177,7 @@ class FaceRealTimeProcessor(RealTimeProcessorApi):
         self.faces_per_second.update()
         self.inference_speed.update(duration)
 
-    def process_frame(self, obj_data: dict[str, Any], frame: np.ndarray):
+    def process_frame(self, obj_data: dict[str, Any], frame: np.ndarray) -> None:
         """Look for faces in image."""
         self.metrics.face_rec_fps.value = self.faces_per_second.eps()
         camera = obj_data["camera"]
@@ -349,7 +349,9 @@ class FaceRealTimeProcessor(RealTimeProcessorApi):
 
         self.__update_metrics(datetime.datetime.now().timestamp() - start)
 
-    def handle_request(self, topic, request_data) -> dict[str, Any] | None:
+    def handle_request(
+        self, topic: str, request_data: dict[str, Any]
+    ) -> dict[str, Any] | None:
         if topic == EmbeddingsRequestEnum.clear_face_classifier.value:
             self.recognizer.clear()
             return {"success": True, "message": "Face classifier cleared."}
@@ -432,7 +434,7 @@ class FaceRealTimeProcessor(RealTimeProcessorApi):
                 img = cv2.imread(current_file)
 
             if img is None:
-                return {
+                return {  # type: ignore[unreachable]
                     "message": "Invalid image file.",
                     "success": False,
                 }
@@ -469,7 +471,9 @@ class FaceRealTimeProcessor(RealTimeProcessorApi):
                 "score": score,
             }
 
-    def expire_object(self, object_id: str, camera: str):
+        return None
+
+    def expire_object(self, object_id: str, camera: str) -> None:
         if object_id in self.person_face_history:
             self.person_face_history.pop(object_id)
 
@@ -478,7 +482,7 @@ class FaceRealTimeProcessor(RealTimeProcessorApi):
 
     def weighted_average(
         self, results_list: list[tuple[str, float, int]], max_weight: int = 4000
-    ):
+    ) -> tuple[str | None, float]:
         """
         Calculates a robust weighted average, capping the area weight and giving more weight to higher scores.
 
@@ -493,8 +497,8 @@ class FaceRealTimeProcessor(RealTimeProcessorApi):
             return None, 0.0
 
         counts: dict[str, int] = {}
-        weighted_scores: dict[str, int] = {}
-        total_weights: dict[str, int] = {}
+        weighted_scores: dict[str, float] = {}
+        total_weights: dict[str, float] = {}
 
         for name, score, face_area in results_list:
             if name == "unknown":
@@ -509,7 +513,7 @@ class FaceRealTimeProcessor(RealTimeProcessorApi):
             counts[name] += 1
 
             # Capped weight based on face area
-            weight = min(face_area, max_weight)
+            weight: float = min(face_area, max_weight)
 
             # Score-based weighting (higher scores get more weight)
             weight *= (score - self.face_config.unknown_score) * 10
@@ -519,7 +523,7 @@ class FaceRealTimeProcessor(RealTimeProcessorApi):
         if not weighted_scores:
             return None, 0.0
 
-        best_name = max(weighted_scores, key=weighted_scores.get)
+        best_name = max(weighted_scores, key=lambda k: weighted_scores[k])
 
         # If the number of faces for this person < min_faces, we are not confident it is a correct result
         if counts[best_name] < self.face_config.min_faces:

frigate/data_processing/real_time/license_plate.py

@@ -61,14 +61,16 @@ class LicensePlateRealTimeProcessor(LicensePlateProcessingMixin, RealTimeProcess
         self,
         obj_data: dict[str, Any],
         frame: np.ndarray,
-        dedicated_lpr: bool | None = False,
-    ):
+        dedicated_lpr: bool = False,
+    ) -> None:
         """Look for license plates in image."""
         self.lpr_process(obj_data, frame, dedicated_lpr)
 
-    def handle_request(self, topic, request_data) -> dict[str, Any] | None:
-        return
+    def handle_request(
+        self, topic: str, request_data: dict[str, Any]
+    ) -> dict[str, Any] | None:
+        return None
 
-    def expire_object(self, object_id: str, camera: str):
+    def expire_object(self, object_id: str, camera: str) -> None:
         """Expire lpr objects."""
         self.lpr_expire(object_id, camera)

frigate/data_processing/types.py

@@ -1,8 +1,10 @@
 """Embeddings types."""
 
+from __future__ import annotations
+
 from enum import Enum
-from multiprocessing.managers import SyncManager
-from multiprocessing.sharedctypes import Synchronized
+from multiprocessing.managers import DictProxy, SyncManager, ValueProxy
+from typing import Any
 
 import sherpa_onnx
 
@@ -10,22 +12,22 @@ from frigate.data_processing.real_time.whisper_online import FasterWhisperASR
 
 
 class DataProcessorMetrics:
-    image_embeddings_speed: Synchronized
-    image_embeddings_eps: Synchronized
-    text_embeddings_speed: Synchronized
-    text_embeddings_eps: Synchronized
-    face_rec_speed: Synchronized
-    face_rec_fps: Synchronized
-    alpr_speed: Synchronized
-    alpr_pps: Synchronized
-    yolov9_lpr_speed: Synchronized
-    yolov9_lpr_pps: Synchronized
-    review_desc_speed: Synchronized
-    review_desc_dps: Synchronized
-    object_desc_speed: Synchronized
-    object_desc_dps: Synchronized
-    classification_speeds: dict[str, Synchronized]
-    classification_cps: dict[str, Synchronized]
+    image_embeddings_speed: ValueProxy[float]
+    image_embeddings_eps: ValueProxy[float]
+    text_embeddings_speed: ValueProxy[float]
+    text_embeddings_eps: ValueProxy[float]
+    face_rec_speed: ValueProxy[float]
+    face_rec_fps: ValueProxy[float]
+    alpr_speed: ValueProxy[float]
+    alpr_pps: ValueProxy[float]
+    yolov9_lpr_speed: ValueProxy[float]
+    yolov9_lpr_pps: ValueProxy[float]
+    review_desc_speed: ValueProxy[float]
+    review_desc_dps: ValueProxy[float]
+    object_desc_speed: ValueProxy[float]
+    object_desc_dps: ValueProxy[float]
+    classification_speeds: DictProxy[str, ValueProxy[float]]
+    classification_cps: DictProxy[str, ValueProxy[float]]
 
     def __init__(self, manager: SyncManager, custom_classification_models: list[str]):
         self.image_embeddings_speed = manager.Value("d", 0.0)
@@ -52,7 +54,7 @@ class DataProcessorMetrics:
 
 
 class DataProcessorModelRunner:
-    def __init__(self, requestor, device: str = "CPU", model_size: str = "large"):
+    def __init__(self, requestor: Any, device: str = "CPU", model_size: str = "large"):
         self.requestor = requestor
         self.device = device
         self.model_size = model_size

frigate/db/sqlitevecq.py

@@ -1,18 +1,21 @@
 import re
 import sqlite3
+from typing import Any
 
 from playhouse.sqliteq import SqliteQueueDatabase
 
 
 class SqliteVecQueueDatabase(SqliteQueueDatabase):
-    def __init__(self, *args, load_vec_extension: bool = False, **kwargs) -> None:
+    def __init__(
+        self, *args: Any, load_vec_extension: bool = False, **kwargs: Any
+    ) -> None:
         self.load_vec_extension: bool = load_vec_extension
         # no extension necessary, sqlite will load correctly for each platform
         self.sqlite_vec_path = "/usr/local/lib/vec0"
         super().__init__(*args, **kwargs)
 
-    def _connect(self, *args, **kwargs) -> sqlite3.Connection:
-        conn: sqlite3.Connection = super()._connect(*args, **kwargs)
+    def _connect(self, *args: Any, **kwargs: Any) -> sqlite3.Connection:
+        conn: sqlite3.Connection = super()._connect(*args, **kwargs)  # type: ignore[misc]
         if self.load_vec_extension:
             self._load_vec_extension(conn)
 
@@ -27,7 +30,7 @@ class SqliteVecQueueDatabase(SqliteQueueDatabase):
         conn.enable_load_extension(False)
 
     def _register_regexp(self, conn: sqlite3.Connection) -> None:
-        def regexp(expr: str, item: str) -> bool:
+        def regexp(expr: str, item: str | None) -> bool:
             if item is None:
                 return False
             try:

frigate/detectors/detector_config.py

@@ -47,7 +47,7 @@ class ModelTypeEnum(str, Enum):
 class ModelConfig(BaseModel):
     path: Optional[str] = Field(
         None,
-        title="Custom Object detection model path",
+        title="Custom object detector model path",
         description="Path to a custom detection model file (or plus://<model_id> for Frigate+ models).",
     )
     labelmap_path: Optional[str] = Field(

frigate/detectors/plugins/memryx.py

@@ -317,7 +317,7 @@ class MemryXDetector(DetectionApi):
                             f"Failed to remove downloaded zip {zip_path}: {e}"
                         )
 
-    def send_input(self, connection_id, tensor_input: np.ndarray):
+    def send_input(self, connection_id, tensor_input: np.ndarray) -> None:
         """Pre-process (if needed) and send frame to MemryX input queue"""
         if tensor_input is None:
             raise ValueError("[send_input] No image data provided for inference")

frigate/events/audio.py

@@ -2,17 +2,19 @@
 
 import datetime
 import logging
+import subprocess
 import threading
 import time
 from multiprocessing.managers import DictProxy
 from multiprocessing.synchronize import Event as MpEvent
-from typing import Tuple
+from typing import Any, Tuple
 
 import numpy as np
 
 from frigate.comms.detections_updater import DetectionPublisher, DetectionTypeEnum
 from frigate.comms.inter_process import InterProcessRequestor
-from frigate.config import CameraConfig, CameraInput, FfmpegConfig, FrigateConfig
+from frigate.config import CameraConfig, CameraInput, FrigateConfig
+from frigate.config.camera.ffmpeg import CameraFfmpegConfig
 from frigate.config.camera.updater import (
     CameraConfigUpdateEnum,
     CameraConfigUpdateSubscriber,
@@ -35,8 +37,7 @@ from frigate.data_processing.real_time.audio_transcription import (
 )
 from frigate.ffmpeg_presets import parse_preset_input
 from frigate.log import LogPipe, suppress_stderr_during
-from frigate.object_detection.base import load_labels
-from frigate.util.builtin import get_ffmpeg_arg_list
+from frigate.util.builtin import get_ffmpeg_arg_list, load_labels
 from frigate.util.ffmpeg import start_or_restart_ffmpeg, stop_ffmpeg
 from frigate.util.process import FrigateProcess
 
@@ -49,7 +50,7 @@ except ModuleNotFoundError:
 logger = logging.getLogger(__name__)
 
 
-def get_ffmpeg_command(ffmpeg: FfmpegConfig) -> list[str]:
+def get_ffmpeg_command(ffmpeg: CameraFfmpegConfig) -> list[str]:
     ffmpeg_input: CameraInput = [i for i in ffmpeg.inputs if "audio" in i.roles][0]
     input_args = get_ffmpeg_arg_list(ffmpeg.global_args) + (
         parse_preset_input(ffmpeg_input.input_args, 1)
@@ -102,9 +103,11 @@ class AudioProcessor(FrigateProcess):
         threading.current_thread().name = "process:audio_manager"
 
         if self.config.audio_transcription.enabled:
-            self.transcription_model_runner = AudioTranscriptionModelRunner(
-                self.config.audio_transcription.device,
-                self.config.audio_transcription.model_size,
+            self.transcription_model_runner: AudioTranscriptionModelRunner | None = (
+                AudioTranscriptionModelRunner(
+                    self.config.audio_transcription.device or "AUTO",
+                    self.config.audio_transcription.model_size,
+                )
             )
         else:
             self.transcription_model_runner = None
@@ -118,7 +121,7 @@ class AudioProcessor(FrigateProcess):
                 self.config,
                 self.camera_metrics,
                 self.transcription_model_runner,
-                self.stop_event,
+                self.stop_event,  # type: ignore[arg-type]
             )
             audio_threads.append(audio_thread)
             audio_thread.start()
@@ -162,7 +165,7 @@ class AudioEventMaintainer(threading.Thread):
         self.logger = logging.getLogger(f"audio.{self.camera_config.name}")
         self.ffmpeg_cmd = get_ffmpeg_command(self.camera_config.ffmpeg)
         self.logpipe = LogPipe(f"ffmpeg.{self.camera_config.name}.audio")
-        self.audio_listener = None
+        self.audio_listener: subprocess.Popen[Any] | None = None
         self.audio_transcription_model_runner = audio_transcription_model_runner
         self.transcription_processor = None
         self.transcription_thread = None
@@ -171,7 +174,7 @@ class AudioEventMaintainer(threading.Thread):
         self.requestor = InterProcessRequestor()
         self.config_subscriber = CameraConfigUpdateSubscriber(
             None,
-            {self.camera_config.name: self.camera_config},
+            {str(self.camera_config.name): self.camera_config},
             [
                 CameraConfigUpdateEnum.audio,
                 CameraConfigUpdateEnum.enabled,
@@ -180,7 +183,10 @@ class AudioEventMaintainer(threading.Thread):
         )
         self.detection_publisher = DetectionPublisher(DetectionTypeEnum.audio.value)
 
-        if self.config.audio_transcription.enabled:
+        if (
+            self.config.audio_transcription.enabled
+            and self.audio_transcription_model_runner is not None
+        ):
             # init the transcription processor for this camera
             self.transcription_processor = AudioTranscriptionRealTimeProcessor(
                 config=self.config,
@@ -200,11 +206,11 @@ class AudioEventMaintainer(threading.Thread):
 
         self.was_enabled = camera.enabled
 
-    def detect_audio(self, audio) -> None:
+    def detect_audio(self, audio: np.ndarray) -> None:
         if not self.camera_config.audio.enabled or self.stop_event.is_set():
             return
 
-        audio_as_float = audio.astype(np.float32)
+        audio_as_float: np.ndarray = audio.astype(np.float32)
         rms, dBFS = self.calculate_audio_levels(audio_as_float)
 
         self.camera_metrics[self.camera_config.name].audio_rms.value = rms
@@ -261,7 +267,7 @@ class AudioEventMaintainer(threading.Thread):
             else:
                 self.transcription_processor.check_unload_model()
 
-    def calculate_audio_levels(self, audio_as_float: np.float32) -> Tuple[float, float]:
+    def calculate_audio_levels(self, audio_as_float: np.ndarray) -> Tuple[float, float]:
         # Calculate RMS (Root-Mean-Square) which represents the average signal amplitude
         # Note: np.float32 isn't serializable, we must use np.float64 to publish the message
         rms = np.sqrt(np.mean(np.absolute(np.square(audio_as_float))))
@@ -296,6 +302,10 @@ class AudioEventMaintainer(threading.Thread):
             self.logpipe.dump()
             self.start_or_restart_ffmpeg()
 
+        if self.audio_listener is None or self.audio_listener.stdout is None:
+            log_and_restart()
+            return
+
         try:
             chunk = self.audio_listener.stdout.read(self.chunk_size)
 
@@ -341,7 +351,10 @@ class AudioEventMaintainer(threading.Thread):
                     self.requestor.send_data(
                         EXPIRE_AUDIO_ACTIVITY, self.camera_config.name
                     )
-                    stop_ffmpeg(self.audio_listener, self.logger)
+
+                    if self.audio_listener:
+                        stop_ffmpeg(self.audio_listener, self.logger)
+
                     self.audio_listener = None
                 self.was_enabled = enabled
                 continue
@@ -367,7 +380,7 @@ class AudioEventMaintainer(threading.Thread):
 
 
 class AudioTfl:
-    def __init__(self, stop_event: threading.Event, num_threads=2):
+    def __init__(self, stop_event: threading.Event, num_threads: int = 2) -> None:
         self.stop_event = stop_event
         self.num_threads = num_threads
         self.labels = load_labels("/audio-labelmap.txt", prefill=521)
@@ -382,7 +395,7 @@ class AudioTfl:
         self.tensor_input_details = self.interpreter.get_input_details()
         self.tensor_output_details = self.interpreter.get_output_details()
 
-    def _detect_raw(self, tensor_input):
+    def _detect_raw(self, tensor_input: np.ndarray) -> np.ndarray:
         self.interpreter.set_tensor(self.tensor_input_details[0]["index"], tensor_input)
         self.interpreter.invoke()
         detections = np.zeros((20, 6), np.float32)
@@ -410,8 +423,10 @@ class AudioTfl:
 
         return detections
 
-    def detect(self, tensor_input, threshold=AUDIO_MIN_CONFIDENCE):
-        detections = []
+    def detect(
+        self, tensor_input: np.ndarray, threshold: float = AUDIO_MIN_CONFIDENCE
+    ) -> list[tuple[str, float, tuple[float, float, float, float]]]:
+        detections: list[tuple[str, float, tuple[float, float, float, float]]] = []
 
         if self.stop_event.is_set():
             return detections

frigate/events/cleanup.py

@@ -29,7 +29,7 @@ class EventCleanup(threading.Thread):
         self.stop_event = stop_event
         self.db = db
         self.camera_keys = list(self.config.cameras.keys())
-        self.removed_camera_labels: list[str] = None
+        self.removed_camera_labels: list[Event] | None = None
         self.camera_labels: dict[str, dict[str, Any]] = {}
 
     def get_removed_camera_labels(self) -> list[Event]:
@@ -37,7 +37,7 @@ class EventCleanup(threading.Thread):
         if self.removed_camera_labels is None:
             self.removed_camera_labels = list(
                 Event.select(Event.label)
-                .where(Event.camera.not_in(self.camera_keys))
+                .where(Event.camera.not_in(self.camera_keys))  # type: ignore[arg-type,call-arg,misc]
                 .distinct()
                 .execute()
             )
@@ -61,7 +61,7 @@ class EventCleanup(threading.Thread):
                 ),
             }
 
-        return self.camera_labels[camera]["labels"]
+        return self.camera_labels[camera]["labels"]  # type: ignore[no-any-return]
 
     def expire_snapshots(self) -> list[str]:
         ## Expire events from unlisted cameras based on the global config
@@ -74,7 +74,9 @@ class EventCleanup(threading.Thread):
         # loop over object types in db
         for event in distinct_labels:
             # get expiration time for this label
-            expire_days = retain_config.objects.get(event.label, retain_config.default)
+            expire_days = retain_config.objects.get(
+                str(event.label), retain_config.default
+            )
 
             expire_after = (
                 datetime.datetime.now() - datetime.timedelta(days=expire_days)
@@ -87,7 +89,7 @@ class EventCleanup(threading.Thread):
                     Event.thumbnail,
                 )
                 .where(
-                    Event.camera.not_in(self.camera_keys),
+                    Event.camera.not_in(self.camera_keys),  # type: ignore[arg-type,call-arg,misc]
                     Event.start_time < expire_after,
                     Event.label == event.label,
                     Event.retain_indefinitely == False,
@@ -109,16 +111,16 @@ class EventCleanup(threading.Thread):
 
             # update the clips attribute for the db entry
             query = Event.select(Event.id).where(
-                Event.camera.not_in(self.camera_keys),
+                Event.camera.not_in(self.camera_keys),  # type: ignore[arg-type,call-arg,misc]
                 Event.start_time < expire_after,
                 Event.label == event.label,
                 Event.retain_indefinitely == False,
             )
 
-            events_to_update = []
+            events_to_update: list[str] = []
 
             for event in query.iterator():
-                events_to_update.append(event.id)
+                events_to_update.append(str(event.id))
                 if len(events_to_update) >= CHUNK_SIZE:
                     logger.debug(
                         f"Updating {update_params} for {len(events_to_update)} events"
@@ -150,7 +152,7 @@ class EventCleanup(threading.Thread):
             for event in distinct_labels:
                 # get expiration time for this label
                 expire_days = retain_config.objects.get(
-                    event.label, retain_config.default
+                    str(event.label), retain_config.default
                 )
 
                 expire_after = (
@@ -177,7 +179,7 @@ class EventCleanup(threading.Thread):
                 # only snapshots are stored in /clips
                 # so no need to delete mp4 files
                 for event in expired_events:
-                    events_to_update.append(event.id)
+                    events_to_update.append(str(event.id))
                     deleted = delete_event_snapshot(event)
 
                     if not deleted:
@@ -214,7 +216,7 @@ class EventCleanup(threading.Thread):
                 Event.camera,
             )
             .where(
-                Event.camera.not_in(self.camera_keys),
+                Event.camera.not_in(self.camera_keys),  # type: ignore[arg-type,call-arg,misc]
                 Event.start_time < expire_after,
                 Event.retain_indefinitely == False,
             )
@@ -245,7 +247,7 @@ class EventCleanup(threading.Thread):
 
         # update the clips attribute for the db entry
         query = Event.select(Event.id).where(
-            Event.camera.not_in(self.camera_keys),
+            Event.camera.not_in(self.camera_keys),  # type: ignore[arg-type,call-arg,misc]
             Event.start_time < expire_after,
             Event.retain_indefinitely == False,
         )
@@ -358,7 +360,7 @@ class EventCleanup(threading.Thread):
 
             logger.debug(f"Found {len(events_to_delete)} events that can be expired")
             if len(events_to_delete) > 0:
-                ids_to_delete = [e.id for e in events_to_delete]
+                ids_to_delete = [str(e.id) for e in events_to_delete]
                 for i in range(0, len(ids_to_delete), CHUNK_SIZE):
                     chunk = ids_to_delete[i : i + CHUNK_SIZE]
                     logger.debug(f"Deleting {len(chunk)} events from the database")

frigate/events/maintainer.py

@@ -2,7 +2,7 @@ import logging
 import threading
 from multiprocessing import Queue
 from multiprocessing.synchronize import Event as MpEvent
-from typing import Dict
+from typing import Any, Dict
 
 from frigate.comms.events_updater import EventEndPublisher, EventUpdateSubscriber
 from frigate.config import FrigateConfig
@@ -15,7 +15,7 @@ from frigate.util.builtin import to_relative_box
 logger = logging.getLogger(__name__)
 
 
-def should_update_db(prev_event: Event, current_event: Event) -> bool:
+def should_update_db(prev_event: dict[str, Any], current_event: dict[str, Any]) -> bool:
     """If current_event has updated fields and (clip or snapshot)."""
     # If event is ending and was previously saved, always update to set end_time
     # This ensures events are properly ended even when alerts/detections are disabled
@@ -47,7 +47,9 @@ def should_update_db(prev_event: Event, current_event: Event) -> bool:
     return False
 
 
-def should_update_state(prev_event: Event, current_event: Event) -> bool:
+def should_update_state(
+    prev_event: dict[str, Any], current_event: dict[str, Any]
+) -> bool:
     """If current event should update state, but not necessarily update the db."""
     if prev_event["stationary"] != current_event["stationary"]:
         return True
@@ -74,7 +76,7 @@ class EventProcessor(threading.Thread):
         super().__init__(name="event_processor")
         self.config = config
         self.timeline_queue = timeline_queue
-        self.events_in_process: Dict[str, Event] = {}
+        self.events_in_process: Dict[str, dict[str, Any]] = {}
         self.stop_event = stop_event
 
         self.event_receiver = EventUpdateSubscriber()
@@ -92,7 +94,7 @@ class EventProcessor(threading.Thread):
             if update == None:
                 continue
 
-            source_type, event_type, camera, _, event_data = update
+            source_type, event_type, camera, _, event_data = update  # type: ignore[misc]
 
             logger.debug(
                 f"Event received: {source_type} {event_type} {camera} {event_data['id']}"
@@ -140,7 +142,7 @@ class EventProcessor(threading.Thread):
         self,
         event_type: str,
         camera: str,
-        event_data: Event,
+        event_data: dict[str, Any],
     ) -> None:
         """handle tracked object event updates."""
         updated_db = False
@@ -150,8 +152,13 @@ class EventProcessor(threading.Thread):
             camera_config = self.config.cameras.get(camera)
             if camera_config is None:
                 return
+
             width = camera_config.detect.width
             height = camera_config.detect.height
+
+            if width is None or height is None:
+                return
+
             first_detector = list(self.config.detectors.values())[0]
 
             start_time = event_data["start_time"]
@@ -222,8 +229,12 @@ class EventProcessor(threading.Thread):
                 Event.thumbnail: event_data.get("thumbnail"),
                 Event.has_clip: event_data["has_clip"],
                 Event.has_snapshot: event_data["has_snapshot"],
-                Event.model_hash: first_detector.model.model_hash,
-                Event.model_type: first_detector.model.model_type,
+                Event.model_hash: first_detector.model.model_hash
+                if first_detector.model
+                else None,
+                Event.model_type: first_detector.model.model_type
+                if first_detector.model
+                else None,
                 Event.detector_type: first_detector.type,
                 Event.data: {
                     "box": box,
@@ -287,10 +298,10 @@ class EventProcessor(threading.Thread):
 
         if event_type == EventStateEnum.end:
             del self.events_in_process[event_data["id"]]
-            self.event_end_publisher.publish((event_data["id"], camera, updated_db))
+            self.event_end_publisher.publish((event_data["id"], camera, updated_db))  # type: ignore[arg-type]
 
     def handle_external_detection(
-        self, event_type: EventStateEnum, event_data: Event
+        self, event_type: EventStateEnum, event_data: dict[str, Any]
     ) -> None:
         # Skip replay cameras
         if event_data.get("camera", "").startswith(REPLAY_CAMERA_PREFIX):

frigate/ffmpeg_presets.py

@@ -3,7 +3,7 @@
 import logging
 import os
 from enum import Enum
-from typing import Any
+from typing import Any, Optional
 
 from frigate.const import (
     FFMPEG_HVC1_ARGS,
@@ -215,7 +215,7 @@ def parse_preset_hardware_acceleration_decode(
     width: int,
     height: int,
     gpu: int,
-) -> list[str]:
+) -> Optional[list[str]]:
     """Return the correct preset if in preset format otherwise return None."""
     if not isinstance(arg, str):
         return None
@@ -242,9 +242,9 @@ def parse_preset_hardware_acceleration_scale(
     else:
         scale = PRESETS_HW_ACCEL_SCALE.get(arg, PRESETS_HW_ACCEL_SCALE["default"])
 
-    scale = scale.format(fps, width, height).split(" ")
-    scale.extend(detect_args)
-    return scale
+    scale_args = scale.format(fps, width, height).split(" ")
+    scale_args.extend(detect_args)
+    return scale_args
 
 
 class EncodeTypeEnum(str, Enum):
@@ -420,7 +420,7 @@ PRESETS_INPUT = {
 }
 
 
-def parse_preset_input(arg: Any, detect_fps: int) -> list[str]:
+def parse_preset_input(arg: Any, detect_fps: int) -> Optional[list[str]]:
     """Return the correct preset if in preset format otherwise return None."""
     if not isinstance(arg, str):
         return None
@@ -465,6 +465,16 @@ PRESETS_RECORD_OUTPUT = {
         "-c:a",
         "aac",
     ],
+    # NOTE: This preset originally used "-c:a copy" to pass through audio
+    # without re-encoding. FFmpeg 7.x introduced a threaded pipeline where
+    # demuxing, encoding, and muxing run in parallel via a Scheduler. This
+    # broke audio streamcopy from RTSP sources: packets are demuxed correctly
+    # but silently dropped before reaching the muxer (0 bytes written). The
+    # issue is specific to RTSP + streamcopy; file inputs and transcoding both
+    # work. Transcoding AAC audio is very lightweight (~30KiB per 10s segment)
+    # and adds negligible CPU overhead, so this is an acceptable workaround.
+    # The benefits of FFmpeg 7.x — particularly the removal of gamma correction
+    # hacks required by earlier versions — outweigh this trade-off.
     "preset-record-generic-audio-copy": [
         "-f",
         "segment",
@@ -476,8 +486,10 @@ PRESETS_RECORD_OUTPUT = {
         "1",
         "-strftime",
         "1",
-        "-c",
+        "-c:v",
         "copy",
+        "-c:a",
+        "aac",
     ],
     "preset-record-mjpeg": [
         "-f",
@@ -530,7 +542,9 @@ PRESETS_RECORD_OUTPUT = {
 }
 
 
-def parse_preset_output_record(arg: Any, force_record_hvc1: bool) -> list[str]:
+def parse_preset_output_record(
+    arg: Any, force_record_hvc1: bool
+) -> Optional[list[str]]:
     """Return the correct preset if in preset format otherwise return None."""
     if not isinstance(arg, str):
         return None

frigate/genai/__init__.py

@@ -5,7 +5,7 @@ import importlib
 import logging
 import os
 import re
-from typing import Any, Optional
+from typing import Any, Callable, Optional
 
 import numpy as np
 from playhouse.shortcuts import model_to_dict
@@ -31,10 +31,10 @@ __all__ = [
 PROVIDERS = {}
 
 
-def register_genai_provider(key: GenAIProviderEnum):
+def register_genai_provider(key: GenAIProviderEnum) -> Callable:
     """Register a GenAI provider."""
 
-    def decorator(cls):
+    def decorator(cls: type) -> type:
         PROVIDERS[key] = cls
         return cls
 
@@ -297,7 +297,7 @@ Guidelines:
         """Generate a description for the frame."""
         try:
             prompt = camera_config.objects.genai.object_prompts.get(
-                event.label,
+                str(event.label),
                 camera_config.objects.genai.prompt,
             ).format(**model_to_dict(event))
         except KeyError as e:
@@ -307,7 +307,7 @@ Guidelines:
         logger.debug(f"Sending images to genai provider with prompt: {prompt}")
         return self._send(prompt, thumbnails)
 
-    def _init_provider(self):
+    def _init_provider(self) -> Any:
         """Initialize the client."""
         return None
 
@@ -402,7 +402,7 @@ Guidelines:
         }
 
 
-def load_providers():
+def load_providers() -> None:
     package_dir = os.path.dirname(__file__)
     for filename in os.listdir(package_dir):
         if filename.endswith(".py") and filename != "__init__.py":

frigate/genai/azure-openai.py

@@ -3,7 +3,7 @@
 import base64
 import json
 import logging
-from typing import Any, Optional
+from typing import Any, AsyncGenerator, Optional
 from urllib.parse import parse_qs, urlparse
 
 from openai import AzureOpenAI
@@ -20,10 +20,10 @@ class OpenAIClient(GenAIClient):
 
     provider: AzureOpenAI
 
-    def _init_provider(self):
+    def _init_provider(self) -> AzureOpenAI | None:
         """Initialize the client."""
         try:
-            parsed_url = urlparse(self.genai_config.base_url)
+            parsed_url = urlparse(self.genai_config.base_url or "")
             query_params = parse_qs(parsed_url.query)
             api_version = query_params.get("api-version", [None])[0]
             azure_endpoint = f"{parsed_url.scheme}://{parsed_url.netloc}/"
@@ -79,7 +79,7 @@ class OpenAIClient(GenAIClient):
             logger.warning("Azure OpenAI returned an error: %s", str(e))
             return None
         if len(result.choices) > 0:
-            return result.choices[0].message.content.strip()
+            return str(result.choices[0].message.content.strip())
         return None
 
     def get_context_size(self) -> int:
@@ -113,7 +113,7 @@ class OpenAIClient(GenAIClient):
                 if openai_tool_choice is not None:
                     request_params["tool_choice"] = openai_tool_choice
 
-            result = self.provider.chat.completions.create(**request_params)
+            result = self.provider.chat.completions.create(**request_params)  # type: ignore[call-overload]
 
             if (
                 result is None
@@ -181,7 +181,7 @@ class OpenAIClient(GenAIClient):
         messages: list[dict[str, Any]],
         tools: Optional[list[dict[str, Any]]] = None,
         tool_choice: Optional[str] = "auto",
-    ):
+    ) -> AsyncGenerator[tuple[str, Any], None]:
         """
         Stream chat with tools; yields content deltas then final message.
 
@@ -214,7 +214,7 @@ class OpenAIClient(GenAIClient):
             tool_calls_by_index: dict[int, dict[str, Any]] = {}
             finish_reason = "stop"
 
-            stream = self.provider.chat.completions.create(**request_params)
+            stream = self.provider.chat.completions.create(**request_params)  # type: ignore[call-overload]
 
             for chunk in stream:
                 if not chunk or not chunk.choices:

frigate/genai/gemini.py

@@ -2,10 +2,11 @@
 
 import json
 import logging
-from typing import Any, Optional
+from typing import Any, AsyncGenerator, Optional
 
 from google import genai
 from google.genai import errors, types
+from google.genai.types import FunctionCallingConfigMode
 
 from frigate.config import GenAIProviderEnum
 from frigate.genai import GenAIClient, register_genai_provider
@@ -19,10 +20,10 @@ class GeminiClient(GenAIClient):
 
     provider: genai.Client
 
-    def _init_provider(self):
+    def _init_provider(self) -> genai.Client:
         """Initialize the client."""
         # Merge provider_options into HttpOptions
-        http_options_dict = {
+        http_options_dict: dict[str, Any] = {
             "timeout": int(self.timeout * 1000),  # requires milliseconds
             "retry_options": types.HttpRetryOptions(
                 attempts=3,
@@ -49,12 +50,12 @@ class GeminiClient(GenAIClient):
         response_format: Optional[dict] = None,
     ) -> Optional[str]:
         """Submit a request to Gemini."""
-        contents = [
+        contents = [prompt] + [
             types.Part.from_bytes(data=img, mime_type="image/jpeg") for img in images
-        ] + [prompt]
+        ]
         try:
             # Merge runtime_options into generation_config if provided
-            generation_config_dict = {"candidate_count": 1}
+            generation_config_dict: dict[str, Any] = {"candidate_count": 1}
             generation_config_dict.update(self.genai_config.runtime_options)
 
             if response_format and response_format.get("type") == "json_schema":
@@ -65,7 +66,7 @@ class GeminiClient(GenAIClient):
 
             response = self.provider.models.generate_content(
                 model=self.genai_config.model,
-                contents=contents,
+                contents=contents,  # type: ignore[arg-type]
                 config=types.GenerateContentConfig(
                     **generation_config_dict,
                 ),
@@ -78,6 +79,8 @@ class GeminiClient(GenAIClient):
             return None
 
         try:
+            if response.text is None:
+                return None
             description = response.text.strip()
         except (ValueError, AttributeError):
             # No description was generated
@@ -102,7 +105,7 @@ class GeminiClient(GenAIClient):
         """
         try:
             # Convert messages to Gemini format
-            gemini_messages = []
+            gemini_messages: list[types.Content] = []
             for msg in messages:
                 role = msg.get("role", "user")
                 content = msg.get("content", "")
@@ -110,7 +113,11 @@ class GeminiClient(GenAIClient):
                 # Map roles to Gemini format
                 if role == "system":
                     # Gemini doesn't have system role, prepend to first user message
-                    if gemini_messages and gemini_messages[0].role == "user":
+                    if (
+                        gemini_messages
+                        and gemini_messages[0].role == "user"
+                        and gemini_messages[0].parts
+                    ):
                         gemini_messages[0].parts[
                             0
                         ].text = f"{content}\n\n{gemini_messages[0].parts[0].text}"
@@ -136,7 +143,7 @@ class GeminiClient(GenAIClient):
                         types.Content(
                             role="function",
                             parts=[
-                                types.Part.from_function_response(function_response)
+                                types.Part.from_function_response(function_response)  # type: ignore[misc,call-arg,arg-type]
                             ],
                         )
                     )
@@ -171,19 +178,25 @@ class GeminiClient(GenAIClient):
             if tool_choice:
                 if tool_choice == "none":
                     tool_config = types.ToolConfig(
-                        function_calling_config=types.FunctionCallingConfig(mode="NONE")
+                        function_calling_config=types.FunctionCallingConfig(
+                            mode=FunctionCallingConfigMode.NONE
+                        )
                     )
                 elif tool_choice == "auto":
                     tool_config = types.ToolConfig(
-                        function_calling_config=types.FunctionCallingConfig(mode="AUTO")
+                        function_calling_config=types.FunctionCallingConfig(
+                            mode=FunctionCallingConfigMode.AUTO
+                        )
                     )
                 elif tool_choice == "required":
                     tool_config = types.ToolConfig(
-                        function_calling_config=types.FunctionCallingConfig(mode="ANY")
+                        function_calling_config=types.FunctionCallingConfig(
+                            mode=FunctionCallingConfigMode.ANY
+                        )
                     )
 
             # Build request config
-            config_params = {"candidate_count": 1}
+            config_params: dict[str, Any] = {"candidate_count": 1}
 
             if gemini_tools:
                 config_params["tools"] = gemini_tools
@@ -197,7 +210,7 @@ class GeminiClient(GenAIClient):
 
             response = self.provider.models.generate_content(
                 model=self.genai_config.model,
-                contents=gemini_messages,
+                contents=gemini_messages,  # type: ignore[arg-type]
                 config=types.GenerateContentConfig(**config_params),
             )
 
@@ -291,7 +304,7 @@ class GeminiClient(GenAIClient):
         messages: list[dict[str, Any]],
         tools: Optional[list[dict[str, Any]]] = None,
         tool_choice: Optional[str] = "auto",
-    ):
+    ) -> AsyncGenerator[tuple[str, Any], None]:
         """
         Stream chat with tools; yields content deltas then final message.
 
@@ -299,7 +312,7 @@ class GeminiClient(GenAIClient):
         """
         try:
             # Convert messages to Gemini format
-            gemini_messages = []
+            gemini_messages: list[types.Content] = []
             for msg in messages:
                 role = msg.get("role", "user")
                 content = msg.get("content", "")
@@ -307,7 +320,11 @@ class GeminiClient(GenAIClient):
                 # Map roles to Gemini format
                 if role == "system":
                     # Gemini doesn't have system role, prepend to first user message
-                    if gemini_messages and gemini_messages[0].role == "user":
+                    if (
+                        gemini_messages
+                        and gemini_messages[0].role == "user"
+                        and gemini_messages[0].parts
+                    ):
                         gemini_messages[0].parts[
                             0
                         ].text = f"{content}\n\n{gemini_messages[0].parts[0].text}"
@@ -333,7 +350,7 @@ class GeminiClient(GenAIClient):
                         types.Content(
                             role="function",
                             parts=[
-                                types.Part.from_function_response(function_response)
+                                types.Part.from_function_response(function_response)  # type: ignore[misc,call-arg,arg-type]
                             ],
                         )
                     )
@@ -368,19 +385,25 @@ class GeminiClient(GenAIClient):
             if tool_choice:
                 if tool_choice == "none":
                     tool_config = types.ToolConfig(
-                        function_calling_config=types.FunctionCallingConfig(mode="NONE")
+                        function_calling_config=types.FunctionCallingConfig(
+                            mode=FunctionCallingConfigMode.NONE
+                        )
                     )
                 elif tool_choice == "auto":
                     tool_config = types.ToolConfig(
-                        function_calling_config=types.FunctionCallingConfig(mode="AUTO")
+                        function_calling_config=types.FunctionCallingConfig(
+                            mode=FunctionCallingConfigMode.AUTO
+                        )
                     )
                 elif tool_choice == "required":
                     tool_config = types.ToolConfig(
-                        function_calling_config=types.FunctionCallingConfig(mode="ANY")
+                        function_calling_config=types.FunctionCallingConfig(
+                            mode=FunctionCallingConfigMode.ANY
+                        )
                     )
 
             # Build request config
-            config_params = {"candidate_count": 1}
+            config_params: dict[str, Any] = {"candidate_count": 1}
 
             if gemini_tools:
                 config_params["tools"] = gemini_tools
@@ -399,7 +422,7 @@ class GeminiClient(GenAIClient):
 
             stream = await self.provider.aio.models.generate_content_stream(
                 model=self.genai_config.model,
-                contents=gemini_messages,
+                contents=gemini_messages,  # type: ignore[arg-type]
                 config=types.GenerateContentConfig(**config_params),
             )
 

frigate/genai/llama_cpp.py

@@ -4,7 +4,7 @@ import base64
 import io
 import json
 import logging
-from typing import Any, Optional
+from typing import Any, AsyncGenerator, Optional
 
 import httpx
 import numpy as np
@@ -23,7 +23,7 @@ def _to_jpeg(img_bytes: bytes) -> bytes | None:
     try:
         img = Image.open(io.BytesIO(img_bytes))
         if img.mode != "RGB":
-            img = img.convert("RGB")
+            img = img.convert("RGB")  # type: ignore[assignment]
         buf = io.BytesIO()
         img.save(buf, format="JPEG", quality=85)
         return buf.getvalue()
@@ -36,10 +36,10 @@ def _to_jpeg(img_bytes: bytes) -> bytes | None:
 class LlamaCppClient(GenAIClient):
     """Generative AI client for Frigate using llama.cpp server."""
 
-    provider: str  # base_url
+    provider: str | None  # base_url
     provider_options: dict[str, Any]
 
-    def _init_provider(self):
+    def _init_provider(self) -> str | None:
         """Initialize the client."""
         self.provider_options = {
             **self.genai_config.provider_options,
@@ -75,7 +75,7 @@ class LlamaCppClient(GenAIClient):
                 content.append(
                     {
                         "type": "image_url",
-                        "image_url": {
+                        "image_url": {  # type: ignore[dict-item]
                             "url": f"data:image/jpeg;base64,{encoded_image}",
                         },
                     }
@@ -111,7 +111,7 @@ class LlamaCppClient(GenAIClient):
             ):
                 choice = result["choices"][0]
                 if "message" in choice and "content" in choice["message"]:
-                    return choice["message"]["content"].strip()
+                    return str(choice["message"]["content"].strip())
             return None
         except Exception as e:
             logger.warning("llama.cpp returned an error: %s", str(e))
@@ -229,7 +229,7 @@ class LlamaCppClient(GenAIClient):
             content.append(
                 {
                     "prompt_string": "<__media__>\n",
-                    "multimodal_data": [encoded],
+                    "multimodal_data": [encoded],  # type: ignore[dict-item]
                 }
             )
 
@@ -367,7 +367,7 @@ class LlamaCppClient(GenAIClient):
         messages: list[dict[str, Any]],
         tools: Optional[list[dict[str, Any]]] = None,
         tool_choice: Optional[str] = "auto",
-    ):
+    ) -> AsyncGenerator[tuple[str, Any], None]:
         """Stream chat with tools via OpenAI-compatible streaming API."""
         if self.provider is None:
             logger.warning(

frigate/genai/ollama.py

@@ -2,7 +2,7 @@
 
 import json
 import logging
-from typing import Any, Optional
+from typing import Any, AsyncGenerator, Optional
 
 from httpx import RemoteProtocolError, TimeoutException
 from ollama import AsyncClient as OllamaAsyncClient
@@ -28,10 +28,10 @@ class OllamaClient(GenAIClient):
         },
     }
 
-    provider: ApiClient
+    provider: ApiClient | None
     provider_options: dict[str, Any]
 
-    def _init_provider(self):
+    def _init_provider(self) -> ApiClient | None:
         """Initialize the client."""
         self.provider_options = {
             **self.LOCAL_OPTIMIZED_OPTIONS,
@@ -73,7 +73,7 @@ class OllamaClient(GenAIClient):
             "exclusiveMinimum",
             "exclusiveMaximum",
         }
-        result = {}
+        result: dict[str, Any] = {}
         for key, value in schema.items():
             if not _is_properties and key in STRIP_KEYS:
                 continue
@@ -122,7 +122,7 @@ class OllamaClient(GenAIClient):
             logger.debug(
                 f"Ollama tokens used: eval_count={result.get('eval_count')}, prompt_eval_count={result.get('prompt_eval_count')}"
             )
-            return result["response"].strip()
+            return str(result["response"]).strip()
         except (
             TimeoutException,
             ResponseError,
@@ -263,7 +263,7 @@ class OllamaClient(GenAIClient):
         messages: list[dict[str, Any]],
         tools: Optional[list[dict[str, Any]]] = None,
         tool_choice: Optional[str] = "auto",
-    ):
+    ) -> AsyncGenerator[tuple[str, Any], None]:
         """Stream chat with tools; yields content deltas then final message.
 
         When tools are provided, Ollama streaming does not include tool_calls

frigate/genai/openai.py

@@ -3,7 +3,7 @@
 import base64
 import json
 import logging
-from typing import Any, Optional
+from typing import Any, AsyncGenerator, Optional
 
 from httpx import TimeoutException
 from openai import OpenAI
@@ -21,7 +21,7 @@ class OpenAIClient(GenAIClient):
     provider: OpenAI
     context_size: Optional[int] = None
 
-    def _init_provider(self):
+    def _init_provider(self) -> OpenAI:
         """Initialize the client."""
         # Extract context_size from provider_options as it's not a valid OpenAI client parameter
         # It will be used in get_context_size() instead
@@ -44,7 +44,12 @@ class OpenAIClient(GenAIClient):
     ) -> Optional[str]:
         """Submit a request to OpenAI."""
         encoded_images = [base64.b64encode(image).decode("utf-8") for image in images]
-        messages_content = []
+        messages_content: list[dict] = [
+            {
+                "type": "text",
+                "text": prompt,
+            }
+        ]
         for image in encoded_images:
             messages_content.append(
                 {
@@ -55,12 +60,6 @@ class OpenAIClient(GenAIClient):
                     },
                 }
             )
-        messages_content.append(
-            {
-                "type": "text",
-                "text": prompt,
-            }
-        )
         try:
             request_params = {
                 "model": self.genai_config.model,
@@ -81,7 +80,7 @@ class OpenAIClient(GenAIClient):
                 and hasattr(result, "choices")
                 and len(result.choices) > 0
             ):
-                return result.choices[0].message.content.strip()
+                return str(result.choices[0].message.content.strip())
             return None
         except (TimeoutException, Exception) as e:
             logger.warning("OpenAI returned an error: %s", str(e))
@@ -171,7 +170,7 @@ class OpenAIClient(GenAIClient):
                 }
                 request_params.update(provider_opts)
 
-            result = self.provider.chat.completions.create(**request_params)
+            result = self.provider.chat.completions.create(**request_params)  # type: ignore[call-overload]
 
             if (
                 result is None
@@ -245,7 +244,7 @@ class OpenAIClient(GenAIClient):
         messages: list[dict[str, Any]],
         tools: Optional[list[dict[str, Any]]] = None,
         tool_choice: Optional[str] = "auto",
-    ):
+    ) -> AsyncGenerator[tuple[str, Any], None]:
         """
         Stream chat with tools; yields content deltas then final message.
 
@@ -287,7 +286,7 @@ class OpenAIClient(GenAIClient):
             tool_calls_by_index: dict[int, dict[str, Any]] = {}
             finish_reason = "stop"
 
-            stream = self.provider.chat.completions.create(**request_params)
+            stream = self.provider.chat.completions.create(**request_params)  # type: ignore[call-overload]
 
             for chunk in stream:
                 if not chunk or not chunk.choices:

frigate/jobs/media_sync.py

@@ -5,7 +5,7 @@ import os
 import threading
 from dataclasses import dataclass, field
 from datetime import datetime
-from typing import Optional
+from typing import Optional, cast
 
 from frigate.comms.inter_process import InterProcessRequestor
 from frigate.const import CONFIG_DIR, UPDATE_JOB_STATE
@@ -122,7 +122,7 @@ def start_media_sync_job(
     if job_is_running("media_sync"):
         current = get_current_job("media_sync")
         logger.warning(
-            f"Media sync job {current.id} is already running. Rejecting new request."
+            f"Media sync job {current.id if current else 'unknown'} is already running. Rejecting new request."
         )
         return None
 
@@ -146,9 +146,9 @@ def start_media_sync_job(
 
 def get_current_media_sync_job() -> Optional[MediaSyncJob]:
     """Get the current running/queued media sync job, if any."""
-    return get_current_job("media_sync")
+    return cast(Optional[MediaSyncJob], get_current_job("media_sync"))
 
 
 def get_media_sync_job_by_id(job_id: str) -> Optional[MediaSyncJob]:
     """Get media sync job by ID. Currently only tracks the current job."""
-    return get_job_by_id("media_sync", job_id)
+    return cast(Optional[MediaSyncJob], get_job_by_id("media_sync", job_id))

frigate/jobs/motion_search.py

@@ -6,7 +6,7 @@ import threading
 from concurrent.futures import Future, ThreadPoolExecutor, as_completed
 from dataclasses import asdict, dataclass, field
 from datetime import datetime
-from typing import Any, Optional
+from typing import Any, Optional, cast
 
 import cv2
 import numpy as np
@@ -96,7 +96,7 @@ def create_polygon_mask(
         dtype=np.int32,
     )
     mask = np.zeros((frame_height, frame_width), dtype=np.uint8)
-    cv2.fillPoly(mask, [motion_points], 255)
+    cv2.fillPoly(mask, [motion_points], (255,))
     return mask
 
 
@@ -116,7 +116,7 @@ def compute_roi_bbox_normalized(
 
 
 def heatmap_overlaps_roi(
-    heatmap: dict[str, int], roi_bbox: tuple[float, float, float, float]
+    heatmap: object, roi_bbox: tuple[float, float, float, float]
 ) -> bool:
     """Check if a sparse motion heatmap has any overlap with the ROI bounding box.
 
@@ -155,9 +155,9 @@ def segment_passes_activity_gate(recording: Recordings) -> bool:
     Returns True if any of motion, objects, or regions is non-zero/non-null.
     Returns True if all are null (old segments without data).
     """
-    motion = recording.motion
-    objects = recording.objects
-    regions = recording.regions
+    motion: Any = recording.motion
+    objects: Any = recording.objects
+    regions: Any = recording.regions
 
     # Old segments without metadata - pass through (conservative)
     if motion is None and objects is None and regions is None:
@@ -278,6 +278,9 @@ class MotionSearchRunner(threading.Thread):
         frame_width = camera_config.detect.width
         frame_height = camera_config.detect.height
 
+        if frame_width is None or frame_height is None:
+            raise ValueError(f"Camera {camera_name} detect dimensions not configured")
+
         # Create polygon mask
         polygon_mask = create_polygon_mask(
             self.job.polygon_points, frame_width, frame_height
@@ -415,11 +418,13 @@ class MotionSearchRunner(threading.Thread):
                 if self._should_stop():
                     break
 
+                rec_start: float = recording.start_time  # type: ignore[assignment]
+                rec_end: float = recording.end_time  # type: ignore[assignment]
                 future = executor.submit(
                     self._process_recording_for_motion,
-                    recording.path,
-                    recording.start_time,
-                    recording.end_time,
+                    str(recording.path),
+                    rec_start,
+                    rec_end,
                     self.job.start_time_range,
                     self.job.end_time_range,
                     polygon_mask,
@@ -524,10 +529,12 @@ class MotionSearchRunner(threading.Thread):
                 break
 
             try:
+                rec_start: float = recording.start_time  # type: ignore[assignment]
+                rec_end: float = recording.end_time  # type: ignore[assignment]
                 results, frames = self._process_recording_for_motion(
-                    recording.path,
-                    recording.start_time,
-                    recording.end_time,
+                    str(recording.path),
+                    rec_start,
+                    rec_end,
                     self.job.start_time_range,
                     self.job.end_time_range,
                     polygon_mask,
@@ -672,7 +679,9 @@ class MotionSearchRunner(threading.Thread):
                 # Handle frame dimension changes
                 if gray.shape != polygon_mask.shape:
                     resized_mask = cv2.resize(
-                        polygon_mask, (gray.shape[1], gray.shape[0]), cv2.INTER_NEAREST
+                        polygon_mask,
+                        (gray.shape[1], gray.shape[0]),
+                        interpolation=cv2.INTER_NEAREST,
                     )
                     current_bbox = cv2.boundingRect(resized_mask)
                 else:
@@ -698,7 +707,7 @@ class MotionSearchRunner(threading.Thread):
                 )
 
                 if prev_frame_gray is not None:
-                    diff = cv2.absdiff(prev_frame_gray, masked_gray)
+                    diff = cv2.absdiff(prev_frame_gray, masked_gray)  # type: ignore[unreachable]
                     diff_blurred = cv2.GaussianBlur(diff, (3, 3), 0)
                     _, thresh = cv2.threshold(
                         diff_blurred, threshold, 255, cv2.THRESH_BINARY
@@ -825,7 +834,7 @@ def get_motion_search_job(job_id: str) -> Optional[MotionSearchJob]:
         if job_entry:
             return job_entry[0]
     # Check completed jobs via manager
-    return get_job_by_id("motion_search", job_id)
+    return cast(Optional[MotionSearchJob], get_job_by_id("motion_search", job_id))
 
 
 def cancel_motion_search_job(job_id: str) -> bool:

frigate/jobs/vlm_watch.py

@@ -25,6 +25,9 @@ logger = logging.getLogger(__name__)
 _MIN_INTERVAL = 1
 _MAX_INTERVAL = 300
 
+# Minimum seconds between VLM iterations when woken by detections (no zone filter)
+_DETECTION_COOLDOWN_WITHOUT_ZONE = 10
+
 # Max user/assistant turn pairs to keep in conversation history
 _MAX_HISTORY = 10
 
@@ -40,6 +43,7 @@ class VLMWatchJob(Job):
     labels: list = field(default_factory=list)
     zones: list = field(default_factory=list)
     last_reasoning: str = ""
+    notification_message: str = ""
     iteration_count: int = 0
 
     def to_dict(self) -> dict[str, Any]:
@@ -54,9 +58,9 @@ class VLMWatchRunner(threading.Thread):
         job: VLMWatchJob,
         config: FrigateConfig,
         cancel_event: threading.Event,
-        frame_processor,
-        genai_manager,
-        dispatcher,
+        frame_processor: Any,
+        genai_manager: Any,
+        dispatcher: Any,
     ) -> None:
         super().__init__(daemon=True, name=f"vlm_watch_{job.id}")
         self.job = job
@@ -196,6 +200,7 @@ class VLMWatchRunner(threading.Thread):
                 min(_MAX_INTERVAL, int(parsed.get("next_run_in", 30))),
             )
             reasoning = str(parsed.get("reasoning", ""))
+            notification_message = str(parsed.get("notification_message", ""))
         except (json.JSONDecodeError, ValueError, TypeError) as e:
             logger.warning(
                 "VLM watch job %s: failed to parse VLM response: %s", self.job.id, e
@@ -203,6 +208,7 @@ class VLMWatchRunner(threading.Thread):
             return 30
 
         self.job.last_reasoning = reasoning
+        self.job.notification_message = notification_message
         self.job.iteration_count += 1
         self._broadcast_status()
 
@@ -213,22 +219,41 @@ class VLMWatchRunner(threading.Thread):
                 self.job.camera,
                 reasoning,
             )
-            self._send_notification(reasoning)
+            self._send_notification(notification_message or reasoning)
             self.job.status = JobStatusTypesEnum.success
             return 0
 
         return next_run_in
 
     def _wait_for_trigger(self, max_wait: float) -> None:
-        """Wait up to max_wait seconds, returning early if a relevant detection fires on the target camera."""
-        deadline = time.time() + max_wait
+        """Wait up to max_wait seconds, returning early if a relevant detection fires on the target camera.
+
+        With zones configured, a matching detection wakes immediately (events
+        are already filtered).  Without zones, detections are frequent so a
+        cooldown is enforced: messages are continuously drained to prevent
+        queue backup, but the loop only exits once a match has been seen
+        *and* the cooldown period has elapsed.
+        """
+        now = time.time()
+        deadline = now + max_wait
+        use_cooldown = not self.job.zones
+        earliest_wake = now + _DETECTION_COOLDOWN_WITHOUT_ZONE if use_cooldown else 0
+        triggered = False
+
         while not self.cancel_event.is_set():
             remaining = deadline - time.time()
             if remaining <= 0:
                 break
-            topic, payload = self.detection_subscriber.check_for_update(
+
+            if triggered and time.time() >= earliest_wake:
+                break
+
+            result = self.detection_subscriber.check_for_update(
                 timeout=min(1.0, remaining)
             )
+            if result is None:
+                continue
+            topic, payload = result
             if topic is None or payload is None:
                 continue
             # payload = (camera, frame_name, frame_time, tracked_objects, motion_boxes, regions)
@@ -247,12 +272,22 @@ class VLMWatchRunner(threading.Thread):
             if cam != self.job.camera or not tracked_objects:
                 continue
             if self._detection_matches_filters(tracked_objects):
-                logger.debug(
-                    "VLM watch job %s: woken early by detection event on %s",
-                    self.job.id,
-                    self.job.camera,
-                )
-                break
+                if not use_cooldown:
+                    logger.debug(
+                        "VLM watch job %s: woken early by detection event on %s",
+                        self.job.id,
+                        self.job.camera,
+                    )
+                    break
+
+                if not triggered:
+                    logger.debug(
+                        "VLM watch job %s: detection match on %s, draining for %.0fs",
+                        self.job.id,
+                        self.job.camera,
+                        max(0, earliest_wake - time.time()),
+                    )
+                    triggered = True
 
     def _detection_matches_filters(self, tracked_objects: list) -> bool:
         """Return True if any tracked object passes the label and zone filters."""
@@ -281,7 +316,11 @@ class VLMWatchRunner(threading.Thread):
             f"You will receive a sequence of frames over time. Use the conversation history to understand "
             f"what is stationary vs. actively changing.\n\n"
             f"For each frame respond with JSON only:\n"
-            f'{{"condition_met": <true/false>, "next_run_in": <integer seconds 1-300>, "reasoning": "<brief explanation>"}}\n\n'
+            f'{{"condition_met": <true/false>, "next_run_in": <integer seconds 1-300>, "reasoning": "<brief explanation>", "notification_message": "<natural language notification>"}}\n\n'
+            f"Guidelines for notification_message:\n"
+            f"- Only required when condition_met is true.\n"
+            f"- Write a short, natural notification a user would want to receive on their phone.\n"
+            f'- Example: "Your package has been delivered to the front porch."\n\n'
             f"Guidelines for next_run_in:\n"
             f"- Scene is empty / nothing of interest visible: 60-300.\n"
             f"- Relevant object(s) visible anywhere in frame (even outside the target zone): 3-10. "
@@ -291,12 +330,13 @@ class VLMWatchRunner(threading.Thread):
             f"- Keep reasoning to 1-2 sentences."
         )
 
-    def _send_notification(self, reasoning: str) -> None:
+    def _send_notification(self, message: str) -> None:
         """Publish a camera_monitoring event so downstream handlers (web push, MQTT) can notify users."""
         payload = {
             "camera": self.job.camera,
             "condition": self.job.condition,
-            "reasoning": reasoning,
+            "message": message,
+            "reasoning": self.job.last_reasoning,
             "job_id": self.job.id,
         }
 
@@ -328,9 +368,9 @@ def start_vlm_watch_job(
     condition: str,
     max_duration_minutes: int,
     config: FrigateConfig,
-    frame_processor,
-    genai_manager,
-    dispatcher,
+    frame_processor: Any,
+    genai_manager: Any,
+    dispatcher: Any,
     labels: list[str] | None = None,
     zones: list[str] | None = None,
 ) -> str:

frigate/motion/__init__.py

@@ -13,10 +13,10 @@ class MotionDetector(ABC):
         frame_shape: Tuple[int, int, int],
         config: MotionConfig,
         fps: int,
-        improve_contrast,
-        threshold,
-        contour_area,
-    ):
+        improve_contrast: bool,
+        threshold: int,
+        contour_area: int | None,
+    ) -> None:
         pass
 
     @abstractmethod
@@ -25,7 +25,7 @@ class MotionDetector(ABC):
         pass
 
     @abstractmethod
-    def is_calibrating(self):
+    def is_calibrating(self) -> bool:
         """Return if motion is recalibrating."""
         pass
 
@@ -35,6 +35,6 @@ class MotionDetector(ABC):
         pass
 
     @abstractmethod
-    def stop(self):
+    def stop(self) -> None:
         """Stop any ongoing work and processes."""
         pass

frigate/motion/frigate_motion.py

@@ -1,7 +1,9 @@
+from typing import Any
+
 import cv2
 import numpy as np
 
-from frigate.config import MotionConfig
+from frigate.config.config import RuntimeMotionConfig
 from frigate.motion import MotionDetector
 from frigate.util.image import grab_cv2_contours
 
@@ -9,19 +11,20 @@ from frigate.util.image import grab_cv2_contours
 class FrigateMotionDetector(MotionDetector):
     def __init__(
         self,
-        frame_shape,
-        config: MotionConfig,
+        frame_shape: tuple[int, ...],
+        config: RuntimeMotionConfig,
         fps: int,
-        improve_contrast,
-        threshold,
-        contour_area,
-    ):
+        improve_contrast: Any,
+        threshold: Any,
+        contour_area: Any,
+    ) -> None:
         self.config = config
         self.frame_shape = frame_shape
-        self.resize_factor = frame_shape[0] / config.frame_height
+        frame_height = config.frame_height or frame_shape[0]
+        self.resize_factor = frame_shape[0] / frame_height
         self.motion_frame_size = (
-            config.frame_height,
-            config.frame_height * frame_shape[1] // frame_shape[0],
+            frame_height,
+            frame_height * frame_shape[1] // frame_shape[0],
         )
         self.avg_frame = np.zeros(self.motion_frame_size, np.float32)
         self.avg_delta = np.zeros(self.motion_frame_size, np.float32)
@@ -38,10 +41,10 @@ class FrigateMotionDetector(MotionDetector):
         self.threshold = threshold
         self.contour_area = contour_area
 
-    def is_calibrating(self):
+    def is_calibrating(self) -> bool:
         return False
 
-    def detect(self, frame):
+    def detect(self, frame: np.ndarray) -> list:
         motion_boxes = []
 
         gray = frame[0 : self.frame_shape[0], 0 : self.frame_shape[1]]
@@ -99,7 +102,7 @@ class FrigateMotionDetector(MotionDetector):
 
             # dilate the thresholded image to fill in holes, then find contours
             # on thresholded image
-            thresh_dilated = cv2.dilate(thresh, None, iterations=2)
+            thresh_dilated = cv2.dilate(thresh, None, iterations=2)  # type: ignore[call-overload]
             contours = cv2.findContours(
                 thresh_dilated, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE
             )

frigate/motion/improved_motion.py

@@ -1,11 +1,12 @@
 import logging
+from typing import Optional
 
 import cv2
 import numpy as np
 from scipy.ndimage import gaussian_filter
 
 from frigate.camera import PTZMetrics
-from frigate.config import MotionConfig
+from frigate.config.config import RuntimeMotionConfig
 from frigate.motion import MotionDetector
 from frigate.util.image import grab_cv2_contours
 
@@ -15,22 +16,23 @@ logger = logging.getLogger(__name__)
 class ImprovedMotionDetector(MotionDetector):
     def __init__(
         self,
-        frame_shape,
-        config: MotionConfig,
+        frame_shape: tuple[int, ...],
+        config: RuntimeMotionConfig,
         fps: int,
-        ptz_metrics: PTZMetrics = None,
-        name="improved",
-        blur_radius=1,
-        interpolation=cv2.INTER_NEAREST,
-        contrast_frame_history=50,
-    ):
+        ptz_metrics: Optional[PTZMetrics] = None,
+        name: str = "improved",
+        blur_radius: int = 1,
+        interpolation: int = cv2.INTER_NEAREST,
+        contrast_frame_history: int = 50,
+    ) -> None:
         self.name = name
         self.config = config
         self.frame_shape = frame_shape
-        self.resize_factor = frame_shape[0] / config.frame_height
+        frame_height = config.frame_height or frame_shape[0]
+        self.resize_factor = frame_shape[0] / frame_height
         self.motion_frame_size = (
-            config.frame_height,
-            config.frame_height * frame_shape[1] // frame_shape[0],
+            frame_height,
+            frame_height * frame_shape[1] // frame_shape[0],
         )
         self.avg_frame = np.zeros(self.motion_frame_size, np.float32)
         self.motion_frame_count = 0
@@ -44,20 +46,20 @@ class ImprovedMotionDetector(MotionDetector):
         self.contrast_values[:, 1:2] = 255
         self.contrast_values_index = 0
         self.ptz_metrics = ptz_metrics
-        self.last_stop_time = None
+        self.last_stop_time: float | None = None
 
-    def is_calibrating(self):
+    def is_calibrating(self) -> bool:
         return self.calibrating
 
-    def detect(self, frame):
-        motion_boxes = []
+    def detect(self, frame: np.ndarray) -> list[tuple[int, int, int, int]]:
+        motion_boxes: list[tuple[int, int, int, int]] = []
 
         if not self.config.enabled:
             return motion_boxes
 
         # if ptz motor is moving from autotracking, quickly return
         # a single box that is 80% of the frame
-        if (
+        if self.ptz_metrics is not None and (
             self.ptz_metrics.autotracker_enabled.value
             and not self.ptz_metrics.motor_stopped.is_set()
         ):
@@ -130,19 +132,19 @@ class ImprovedMotionDetector(MotionDetector):
 
         # dilate the thresholded image to fill in holes, then find contours
         # on thresholded image
-        thresh_dilated = cv2.dilate(thresh, None, iterations=1)
+        thresh_dilated = cv2.dilate(thresh, None, iterations=1)  # type: ignore[call-overload]
         contours = cv2.findContours(
             thresh_dilated, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE
         )
         contours = grab_cv2_contours(contours)
 
         # loop over the contours
-        total_contour_area = 0
+        total_contour_area: float = 0
         for c in contours:
             # if the contour is big enough, count it as motion
             contour_area = cv2.contourArea(c)
             total_contour_area += contour_area
-            if contour_area > self.config.contour_area:
+            if contour_area > (self.config.contour_area or 0):
                 x, y, w, h = cv2.boundingRect(c)
                 motion_boxes.append(
                     (
@@ -159,7 +161,7 @@ class ImprovedMotionDetector(MotionDetector):
 
         # check if the motor has just stopped from autotracking
         # if so, reassign the average to the current frame so we begin with a new baseline
-        if (
+        if self.ptz_metrics is not None and (
             # ensure we only do this for cameras with autotracking enabled
             self.ptz_metrics.autotracker_enabled.value
             and self.ptz_metrics.motor_stopped.is_set()

frigate/mypy.ini

@@ -22,50 +22,43 @@ warn_unreachable = true
 no_implicit_reexport = true
 
 [mypy-frigate.*]
+ignore_errors = false
+
+# Third-party code imported from https://github.com/ufal/whisper_streaming
+[mypy-frigate.data_processing.real_time.whisper_online]
 ignore_errors = true
 
-[mypy-frigate.__main__]
-ignore_errors = false
-disallow_untyped_calls = false
+# TODO: Remove ignores for these modules as they are updated with type annotations.
 
-[mypy-frigate.app]
-ignore_errors = false
-disallow_untyped_calls = false
+[mypy-frigate.api.*]
+ignore_errors = true
 
-[mypy-frigate.const]
-ignore_errors = false
+[mypy-frigate.config.*]
+ignore_errors = true
 
-[mypy-frigate.comms.*]
-ignore_errors = false
+[mypy-frigate.debug_replay]
+ignore_errors = true
 
-[mypy-frigate.events]
-ignore_errors = false
+[mypy-frigate.detectors.*]
+ignore_errors = true
 
-[mypy-frigate.log]
-ignore_errors = false
+[mypy-frigate.embeddings.*]
+ignore_errors = true
 
-[mypy-frigate.models]
-ignore_errors = false
+[mypy-frigate.http]
+ignore_errors = true
 
-[mypy-frigate.plus]
-ignore_errors = false
+[mypy-frigate.ptz.*]
+ignore_errors = true
 
-[mypy-frigate.stats]
-ignore_errors = false
+[mypy-frigate.stats.*]
+ignore_errors = true
 
-[mypy-frigate.track.*]
-ignore_errors = false
+[mypy-frigate.test.*]
+ignore_errors = true
 
-[mypy-frigate.types]
-ignore_errors = false
+[mypy-frigate.util.*]
+ignore_errors = true
 
-[mypy-frigate.version]
-ignore_errors = false
-
-[mypy-frigate.watchdog]
-ignore_errors = false
-disallow_untyped_calls = false
-
-
-[mypy-frigate.service_manager.*]
-ignore_errors = false
+[mypy-frigate.video.*]
+ignore_errors = true

frigate/object_detection/base.py

@@ -7,6 +7,7 @@ from abc import ABC, abstractmethod
 from collections import deque
 from multiprocessing import Queue, Value
 from multiprocessing.synchronize import Event as MpEvent
+from typing import Any, Optional
 
 import numpy as np
 import zmq
@@ -34,26 +35,25 @@ logger = logging.getLogger(__name__)
 
 class ObjectDetector(ABC):
     @abstractmethod
-    def detect(self, tensor_input, threshold: float = 0.4):
+    def detect(self, tensor_input: np.ndarray, threshold: float = 0.4) -> list:
         pass
 
 
 class BaseLocalDetector(ObjectDetector):
     def __init__(
         self,
-        detector_config: BaseDetectorConfig = None,
-        labels: str = None,
-        stop_event: MpEvent = None,
-    ):
+        detector_config: Optional[BaseDetectorConfig] = None,
+        labels: Optional[str] = None,
+        stop_event: Optional[MpEvent] = None,
+    ) -> None:
         self.fps = EventsPerSecond()
         if labels is None:
-            self.labels = {}
+            self.labels: dict[int, str] = {}
         else:
             self.labels = load_labels(labels)
 
-        if detector_config:
+        if detector_config and detector_config.model:
             self.input_transform = tensor_transform(detector_config.model.input_tensor)
-
             self.dtype = detector_config.model.input_dtype
         else:
             self.input_transform = None
@@ -77,10 +77,10 @@ class BaseLocalDetector(ObjectDetector):
 
         return tensor_input
 
-    def detect(self, tensor_input: np.ndarray, threshold=0.4):
+    def detect(self, tensor_input: np.ndarray, threshold: float = 0.4) -> list:
         detections = []
 
-        raw_detections = self.detect_raw(tensor_input)
+        raw_detections = self.detect_raw(tensor_input)  # type: ignore[attr-defined]
 
         for d in raw_detections:
             if int(d[0]) < 0 or int(d[0]) >= len(self.labels):
@@ -96,28 +96,28 @@ class BaseLocalDetector(ObjectDetector):
 
 
 class LocalObjectDetector(BaseLocalDetector):
-    def detect_raw(self, tensor_input: np.ndarray):
+    def detect_raw(self, tensor_input: np.ndarray) -> np.ndarray:
         tensor_input = self._transform_input(tensor_input)
-        return self.detect_api.detect_raw(tensor_input=tensor_input)
+        return self.detect_api.detect_raw(tensor_input=tensor_input)  # type: ignore[no-any-return]
 
 
 class AsyncLocalObjectDetector(BaseLocalDetector):
-    def async_send_input(self, tensor_input: np.ndarray, connection_id: str):
+    def async_send_input(self, tensor_input: np.ndarray, connection_id: str) -> None:
         tensor_input = self._transform_input(tensor_input)
-        return self.detect_api.send_input(connection_id, tensor_input)
+        self.detect_api.send_input(connection_id, tensor_input)
 
-    def async_receive_output(self):
+    def async_receive_output(self) -> Any:
         return self.detect_api.receive_output()
 
 
 class DetectorRunner(FrigateProcess):
     def __init__(
         self,
-        name,
+        name: str,
         detection_queue: Queue,
         cameras: list[str],
-        avg_speed: Value,
-        start_time: Value,
+        avg_speed: Any,
+        start_time: Any,
         config: FrigateConfig,
         detector_config: BaseDetectorConfig,
         stop_event: MpEvent,
@@ -129,11 +129,11 @@ class DetectorRunner(FrigateProcess):
         self.start_time = start_time
         self.config = config
         self.detector_config = detector_config
-        self.outputs: dict = {}
+        self.outputs: dict[str, Any] = {}
 
-    def create_output_shm(self, name: str):
+    def create_output_shm(self, name: str) -> None:
         out_shm = UntrackedSharedMemory(name=f"out-{name}", create=False)
-        out_np = np.ndarray((20, 6), dtype=np.float32, buffer=out_shm.buf)
+        out_np: np.ndarray = np.ndarray((20, 6), dtype=np.float32, buffer=out_shm.buf)
         self.outputs[name] = {"shm": out_shm, "np": out_np}
 
     def run(self) -> None:
@@ -155,8 +155,8 @@ class DetectorRunner(FrigateProcess):
                 connection_id,
                 (
                     1,
-                    self.detector_config.model.height,
-                    self.detector_config.model.width,
+                    self.detector_config.model.height,  # type: ignore[union-attr]
+                    self.detector_config.model.width,  # type: ignore[union-attr]
                     3,
                 ),
             )
@@ -187,11 +187,11 @@ class DetectorRunner(FrigateProcess):
 class AsyncDetectorRunner(FrigateProcess):
     def __init__(
         self,
-        name,
+        name: str,
         detection_queue: Queue,
         cameras: list[str],
-        avg_speed: Value,
-        start_time: Value,
+        avg_speed: Any,
+        start_time: Any,
         config: FrigateConfig,
         detector_config: BaseDetectorConfig,
         stop_event: MpEvent,
@@ -203,15 +203,15 @@ class AsyncDetectorRunner(FrigateProcess):
         self.start_time = start_time
         self.config = config
         self.detector_config = detector_config
-        self.outputs: dict = {}
+        self.outputs: dict[str, Any] = {}
         self._frame_manager: SharedMemoryFrameManager | None = None
         self._publisher: ObjectDetectorPublisher | None = None
         self._detector: AsyncLocalObjectDetector | None = None
-        self.send_times = deque()
+        self.send_times: deque[float] = deque()
 
-    def create_output_shm(self, name: str):
+    def create_output_shm(self, name: str) -> None:
         out_shm = UntrackedSharedMemory(name=f"out-{name}", create=False)
-        out_np = np.ndarray((20, 6), dtype=np.float32, buffer=out_shm.buf)
+        out_np: np.ndarray = np.ndarray((20, 6), dtype=np.float32, buffer=out_shm.buf)
         self.outputs[name] = {"shm": out_shm, "np": out_np}
 
     def _detect_worker(self) -> None:
@@ -222,12 +222,13 @@ class AsyncDetectorRunner(FrigateProcess):
             except queue.Empty:
                 continue
 
+            assert self._frame_manager is not None
             input_frame = self._frame_manager.get(
                 connection_id,
                 (
                     1,
-                    self.detector_config.model.height,
-                    self.detector_config.model.width,
+                    self.detector_config.model.height,  # type: ignore[union-attr]
+                    self.detector_config.model.width,  # type: ignore[union-attr]
                     3,
                 ),
             )
@@ -238,11 +239,13 @@ class AsyncDetectorRunner(FrigateProcess):
 
             # mark start time and send to accelerator
             self.send_times.append(time.perf_counter())
+            assert self._detector is not None
             self._detector.async_send_input(input_frame, connection_id)
 
     def _result_worker(self) -> None:
         logger.info("Starting Result Worker Thread")
         while not self.stop_event.is_set():
+            assert self._detector is not None
             connection_id, detections = self._detector.async_receive_output()
 
             # Handle timeout case (queue.Empty) - just continue
@@ -256,6 +259,7 @@ class AsyncDetectorRunner(FrigateProcess):
             duration = time.perf_counter() - ts
 
             # release input buffer
+            assert self._frame_manager is not None
             self._frame_manager.close(connection_id)
 
             if connection_id not in self.outputs:
@@ -264,6 +268,7 @@ class AsyncDetectorRunner(FrigateProcess):
             # write results and publish
             if detections is not None:
                 self.outputs[connection_id]["np"][:] = detections[:]
+            assert self._publisher is not None
             self._publisher.publish(connection_id)
 
             # update timers
@@ -330,11 +335,14 @@ class ObjectDetectProcess:
         self.stop_event = stop_event
         self.start_or_restart()
 
-    def stop(self):
+    def stop(self) -> None:
         # if the process has already exited on its own, just return
         if self.detect_process and self.detect_process.exitcode:
             return
 
+        if self.detect_process is None:
+            return
+
         logging.info("Waiting for detection process to exit gracefully...")
         self.detect_process.join(timeout=30)
         if self.detect_process.exitcode is None:
@@ -343,8 +351,8 @@ class ObjectDetectProcess:
             self.detect_process.join()
         logging.info("Detection process has exited...")
 
-    def start_or_restart(self):
-        self.detection_start.value = 0.0
+    def start_or_restart(self) -> None:
+        self.detection_start.value = 0.0  # type: ignore[attr-defined]
         if (self.detect_process is not None) and self.detect_process.is_alive():
             self.stop()
 
@@ -389,17 +397,19 @@ class RemoteObjectDetector:
         self.detection_queue = detection_queue
         self.stop_event = stop_event
         self.shm = UntrackedSharedMemory(name=self.name, create=False)
-        self.np_shm = np.ndarray(
+        self.np_shm: np.ndarray = np.ndarray(
             (1, model_config.height, model_config.width, 3),
             dtype=np.uint8,
             buffer=self.shm.buf,
         )
         self.out_shm = UntrackedSharedMemory(name=f"out-{self.name}", create=False)
-        self.out_np_shm = np.ndarray((20, 6), dtype=np.float32, buffer=self.out_shm.buf)
+        self.out_np_shm: np.ndarray = np.ndarray(
+            (20, 6), dtype=np.float32, buffer=self.out_shm.buf
+        )
         self.detector_subscriber = ObjectDetectorSubscriber(name)
 
-    def detect(self, tensor_input, threshold=0.4):
-        detections = []
+    def detect(self, tensor_input: np.ndarray, threshold: float = 0.4) -> list:
+        detections: list = []
 
         if self.stop_event.is_set():
             return detections
@@ -431,7 +441,7 @@ class RemoteObjectDetector:
         self.fps.update()
         return detections
 
-    def cleanup(self):
+    def cleanup(self) -> None:
         self.detector_subscriber.stop()
         self.shm.unlink()
         self.out_shm.unlink()

frigate/object_detection/util.py

@@ -13,10 +13,10 @@ class RequestStore:
     A thread-safe hash-based response store that handles creating requests.
     """
 
-    def __init__(self):
+    def __init__(self) -> None:
         self.request_counter = 0
         self.request_counter_lock = threading.Lock()
-        self.input_queue = queue.Queue()
+        self.input_queue: queue.Queue[tuple[int, ndarray]] = queue.Queue()
 
     def __get_request_id(self) -> int:
         with self.request_counter_lock:
@@ -45,17 +45,19 @@ class ResponseStore:
     their request's result appears.
     """
 
-    def __init__(self):
-        self.responses = {}  # Maps request_id -> (original_input, infer_results)
+    def __init__(self) -> None:
+        self.responses: dict[
+            int, ndarray
+        ] = {}  # Maps request_id -> (original_input, infer_results)
         self.lock = threading.Lock()
         self.cond = threading.Condition(self.lock)
 
-    def put(self, request_id: int, response: ndarray):
+    def put(self, request_id: int, response: ndarray) -> None:
         with self.cond:
             self.responses[request_id] = response
             self.cond.notify_all()
 
-    def get(self, request_id: int, timeout=None) -> ndarray:
+    def get(self, request_id: int, timeout: float | None = None) -> ndarray:
         with self.cond:
             if not self.cond.wait_for(
                 lambda: request_id in self.responses, timeout=timeout
@@ -65,7 +67,9 @@ class ResponseStore:
             return self.responses.pop(request_id)
 
 
-def tensor_transform(desired_shape: InputTensorEnum):
+def tensor_transform(
+    desired_shape: InputTensorEnum,
+) -> tuple[int, int, int, int] | None:
     # Currently this function only supports BHWC permutations
     if desired_shape == InputTensorEnum.nhwc:
         return None

frigate/output/birdseye.py

@@ -4,13 +4,13 @@ import datetime
 import glob
 import logging
 import math
-import multiprocessing as mp
 import os
 import queue
 import subprocess as sp
 import threading
 import time
 import traceback
+from multiprocessing.synchronize import Event as MpEvent
 from typing import Any, Optional
 
 import cv2
@@ -74,25 +74,25 @@ class Canvas:
         self,
         canvas_width: int,
         canvas_height: int,
-        scaling_factor: int,
+        scaling_factor: float,
     ) -> None:
         self.scaling_factor = scaling_factor
         gcd = math.gcd(canvas_width, canvas_height)
         self.aspect = get_standard_aspect_ratio(
-            (canvas_width / gcd), (canvas_height / gcd)
+            int(canvas_width / gcd), int(canvas_height / gcd)
         )
         self.width = canvas_width
-        self.height = (self.width * self.aspect[1]) / self.aspect[0]
-        self.coefficient_cache: dict[int, int] = {}
+        self.height: float = (self.width * self.aspect[1]) / self.aspect[0]
+        self.coefficient_cache: dict[int, float] = {}
         self.aspect_cache: dict[str, tuple[int, int]] = {}
 
-    def get_aspect(self, coefficient: int) -> tuple[int, int]:
+    def get_aspect(self, coefficient: float) -> tuple[float, float]:
         return (self.aspect[0] * coefficient, self.aspect[1] * coefficient)
 
-    def get_coefficient(self, camera_count: int) -> int:
+    def get_coefficient(self, camera_count: int) -> float:
         return self.coefficient_cache.get(camera_count, self.scaling_factor)
 
-    def set_coefficient(self, camera_count: int, coefficient: int) -> None:
+    def set_coefficient(self, camera_count: int, coefficient: float) -> None:
         self.coefficient_cache[camera_count] = coefficient
 
     def get_camera_aspect(
@@ -105,7 +105,7 @@ class Canvas:
 
         gcd = math.gcd(camera_width, camera_height)
         camera_aspect = get_standard_aspect_ratio(
-            camera_width / gcd, camera_height / gcd
+            int(camera_width / gcd), int(camera_height / gcd)
         )
         self.aspect_cache[cam_name] = camera_aspect
         return camera_aspect
@@ -116,7 +116,7 @@ class FFMpegConverter(threading.Thread):
         self,
         ffmpeg: FfmpegConfig,
         input_queue: queue.Queue,
-        stop_event: mp.Event,
+        stop_event: MpEvent,
         in_width: int,
         in_height: int,
         out_width: int,
@@ -128,7 +128,7 @@ class FFMpegConverter(threading.Thread):
         self.camera = "birdseye"
         self.input_queue = input_queue
         self.stop_event = stop_event
-        self.bd_pipe = None
+        self.bd_pipe: int | None = None
 
         if birdseye_rtsp:
             self.recreate_birdseye_pipe()
@@ -181,7 +181,8 @@ class FFMpegConverter(threading.Thread):
         os.close(stdin)
         self.reading_birdseye = False
 
-    def __write(self, b) -> None:
+    def __write(self, b: bytes) -> None:
+        assert self.process.stdin is not None
         self.process.stdin.write(b)
 
         if self.bd_pipe:
@@ -200,13 +201,13 @@ class FFMpegConverter(threading.Thread):
 
                 return
 
-    def read(self, length):
+    def read(self, length: int) -> Any:
         try:
-            return self.process.stdout.read1(length)
+            return self.process.stdout.read1(length)  # type: ignore[union-attr]
         except ValueError:
             return False
 
-    def exit(self):
+    def exit(self) -> None:
         if self.bd_pipe:
             os.close(self.bd_pipe)
 
@@ -233,8 +234,8 @@ class BroadcastThread(threading.Thread):
         self,
         camera: str,
         converter: FFMpegConverter,
-        websocket_server,
-        stop_event: mp.Event,
+        websocket_server: Any,
+        stop_event: MpEvent,
     ):
         super().__init__()
         self.camera = camera
@@ -242,7 +243,7 @@ class BroadcastThread(threading.Thread):
         self.websocket_server = websocket_server
         self.stop_event = stop_event
 
-    def run(self):
+    def run(self) -> None:
         while not self.stop_event.is_set():
             buf = self.converter.read(65536)
             if buf:
@@ -270,16 +271,16 @@ class BirdsEyeFrameManager:
     def __init__(
         self,
         config: FrigateConfig,
-        stop_event: mp.Event,
+        stop_event: MpEvent,
     ):
         self.config = config
         width, height = get_canvas_shape(config.birdseye.width, config.birdseye.height)
         self.frame_shape = (height, width)
         self.yuv_shape = (height * 3 // 2, width)
-        self.frame = np.ndarray(self.yuv_shape, dtype=np.uint8)
+        self.frame: np.ndarray = np.ndarray(self.yuv_shape, dtype=np.uint8)
         self.canvas = Canvas(width, height, config.birdseye.layout.scaling_factor)
         self.stop_event = stop_event
-        self.last_refresh_time = 0
+        self.last_refresh_time: float = 0
 
         # initialize the frame as black and with the Frigate logo
         self.blank_frame = np.zeros(self.yuv_shape, np.uint8)
@@ -323,15 +324,15 @@ class BirdsEyeFrameManager:
 
         self.frame[:] = self.blank_frame
 
-        self.cameras = {}
+        self.cameras: dict[str, Any] = {}
         for camera in self.config.cameras.keys():
             self.add_camera(camera)
 
-        self.camera_layout = []
-        self.active_cameras = set()
+        self.camera_layout: list[Any] = []
+        self.active_cameras: set[str] = set()
         self.last_output_time = 0.0
 
-    def add_camera(self, cam: str):
+    def add_camera(self, cam: str) -> None:
         """Add a camera to self.cameras with the correct structure."""
         settings = self.config.cameras[cam]
         # precalculate the coordinates for all the channels
@@ -361,16 +362,21 @@ class BirdsEyeFrameManager:
             },
         }
 
-    def remove_camera(self, cam: str):
+    def remove_camera(self, cam: str) -> None:
         """Remove a camera from self.cameras."""
         if cam in self.cameras:
             del self.cameras[cam]
 
-    def clear_frame(self):
+    def clear_frame(self) -> None:
         logger.debug("Clearing the birdseye frame")
         self.frame[:] = self.blank_frame
 
-    def copy_to_position(self, position, camera=None, frame: np.ndarray = None):
+    def copy_to_position(
+        self,
+        position: Any,
+        camera: Optional[str] = None,
+        frame: Optional[np.ndarray] = None,
+    ) -> None:
         if camera is None:
             frame = None
             channel_dims = None
@@ -389,7 +395,9 @@ class BirdsEyeFrameManager:
             channel_dims,
         )
 
-    def camera_active(self, mode, object_box_count, motion_box_count):
+    def camera_active(
+        self, mode: Any, object_box_count: int, motion_box_count: int
+    ) -> bool:
         if mode == BirdseyeModeEnum.continuous:
             return True
 
@@ -399,6 +407,8 @@ class BirdsEyeFrameManager:
         if mode == BirdseyeModeEnum.objects and object_box_count > 0:
             return True
 
+        return False
+
     def get_camera_coordinates(self) -> dict[str, dict[str, int]]:
         """Return the coordinates of each camera in the current layout."""
         coordinates = {}
@@ -451,7 +461,7 @@ class BirdsEyeFrameManager:
                         - self.cameras[active_camera]["last_active_frame"]
                     ),
                 )
-                active_cameras = limited_active_cameras[:max_cameras]
+                active_cameras = set(limited_active_cameras[:max_cameras])
                 max_camera_refresh = True
                 self.last_refresh_time = now
 
@@ -510,7 +520,7 @@ class BirdsEyeFrameManager:
 
                     # center camera view in canvas and ensure that it fits
                     if scaled_width < self.canvas.width:
-                        coefficient = 1
+                        coefficient: float = 1
                         x_offset = int((self.canvas.width - scaled_width) / 2)
                     else:
                         coefficient = self.canvas.width / scaled_width
@@ -557,7 +567,7 @@ class BirdsEyeFrameManager:
                         calculating = False
                         self.canvas.set_coefficient(len(active_cameras), coefficient)
 
-                    self.camera_layout = layout_candidate
+                    self.camera_layout = layout_candidate or []
                 frame_changed = True
 
             # Draw the layout
@@ -577,10 +587,12 @@ class BirdsEyeFrameManager:
         self,
         cameras_to_add: list[str],
         coefficient: float,
-    ) -> tuple[Any]:
+    ) -> Optional[list[list[Any]]]:
         """Calculate the optimal layout for 2+ cameras."""
 
-        def map_layout(camera_layout: list[list[Any]], row_height: int):
+        def map_layout(
+            camera_layout: list[list[Any]], row_height: int
+        ) -> tuple[int, int, Optional[list[list[Any]]]]:
             """Map the calculated layout."""
             candidate_layout = []
             starting_x = 0
@@ -777,11 +789,11 @@ class Birdseye:
     def __init__(
         self,
         config: FrigateConfig,
-        stop_event: mp.Event,
-        websocket_server,
+        stop_event: MpEvent,
+        websocket_server: Any,
     ) -> None:
         self.config = config
-        self.input = queue.Queue(maxsize=10)
+        self.input: queue.Queue[bytes] = queue.Queue(maxsize=10)
         self.converter = FFMpegConverter(
             config.ffmpeg,
             self.input,
@@ -806,7 +818,7 @@ class Birdseye:
         )
 
         if config.birdseye.restream:
-            self.birdseye_buffer = self.frame_manager.create(
+            self.birdseye_buffer: Any = self.frame_manager.create(
                 "birdseye",
                 self.birdseye_manager.yuv_shape[0] * self.birdseye_manager.yuv_shape[1],
             )

frigate/output/camera.py

@@ -1,10 +1,11 @@
 """Handle outputting individual cameras via jsmpeg."""
 
 import logging
-import multiprocessing as mp
 import queue
 import subprocess as sp
 import threading
+from multiprocessing.synchronize import Event as MpEvent
+from typing import Any
 
 from frigate.config import CameraConfig, FfmpegConfig
 
@@ -17,7 +18,7 @@ class FFMpegConverter(threading.Thread):
         camera: str,
         ffmpeg: FfmpegConfig,
         input_queue: queue.Queue,
-        stop_event: mp.Event,
+        stop_event: MpEvent,
         in_width: int,
         in_height: int,
         out_width: int,
@@ -64,16 +65,17 @@ class FFMpegConverter(threading.Thread):
             start_new_session=True,
         )
 
-    def __write(self, b) -> None:
+    def __write(self, b: bytes) -> None:
+        assert self.process.stdin is not None
         self.process.stdin.write(b)
 
-    def read(self, length):
+    def read(self, length: int) -> Any:
         try:
-            return self.process.stdout.read1(length)
+            return self.process.stdout.read1(length)  # type: ignore[union-attr]
         except ValueError:
             return False
 
-    def exit(self):
+    def exit(self) -> None:
         self.process.terminate()
 
         try:
@@ -98,8 +100,8 @@ class BroadcastThread(threading.Thread):
         self,
         camera: str,
         converter: FFMpegConverter,
-        websocket_server,
-        stop_event: mp.Event,
+        websocket_server: Any,
+        stop_event: MpEvent,
     ):
         super().__init__()
         self.camera = camera
@@ -107,7 +109,7 @@ class BroadcastThread(threading.Thread):
         self.websocket_server = websocket_server
         self.stop_event = stop_event
 
-    def run(self):
+    def run(self) -> None:
         while not self.stop_event.is_set():
             buf = self.converter.read(65536)
             if buf:
@@ -133,15 +135,15 @@ class BroadcastThread(threading.Thread):
 
 class JsmpegCamera:
     def __init__(
-        self, config: CameraConfig, stop_event: mp.Event, websocket_server
+        self, config: CameraConfig, stop_event: MpEvent, websocket_server: Any
     ) -> None:
         self.config = config
-        self.input = queue.Queue(maxsize=config.detect.fps)
+        self.input: queue.Queue[bytes] = queue.Queue(maxsize=config.detect.fps)
         width = int(
             config.live.height * (config.frame_shape[1] / config.frame_shape[0])
         )
         self.converter = FFMpegConverter(
-            config.name,
+            config.name or "",
             config.ffmpeg,
             self.input,
             stop_event,
@@ -152,13 +154,13 @@ class JsmpegCamera:
             config.live.quality,
         )
         self.broadcaster = BroadcastThread(
-            config.name, self.converter, websocket_server, stop_event
+            config.name or "", self.converter, websocket_server, stop_event
         )
 
         self.converter.start()
         self.broadcaster.start()
 
-    def write_frame(self, frame_bytes) -> None:
+    def write_frame(self, frame_bytes: bytes) -> None:
         try:
             self.input.put_nowait(frame_bytes)
         except queue.Full:

frigate/output/output.py

@@ -61,6 +61,12 @@ def check_disabled_camera_update(
             # last camera update was more than 1 second ago
             # need to send empty data to birdseye because current
             # frame is now out of date
+            cam_width = config.cameras[camera].detect.width
+            cam_height = config.cameras[camera].detect.height
+
+            if cam_width is None or cam_height is None:
+                raise ValueError(f"Camera {camera} detect dimensions not configured")
+
             if birdseye and offline_time < 10:
                 # we only need to send blank frames to birdseye at the beginning of a camera being offline
                 birdseye.write_data(
@@ -68,10 +74,7 @@ def check_disabled_camera_update(
                     [],
                     [],
                     now,
-                    get_blank_yuv_frame(
-                        config.cameras[camera].detect.width,
-                        config.cameras[camera].detect.height,
-                    ),
+                    get_blank_yuv_frame(cam_width, cam_height),
                 )
 
     if not has_enabled_camera and birdseye:
@@ -173,7 +176,7 @@ class OutputProcess(FrigateProcess):
                 birdseye_config_subscriber.check_for_update()
             )
 
-            if update_topic is not None:
+            if update_topic is not None and birdseye_config is not None:
                 previous_global_mode = self.config.birdseye.mode
                 self.config.birdseye = birdseye_config
 
@@ -198,7 +201,10 @@ class OutputProcess(FrigateProcess):
                             birdseye,
                         )
 
-            (topic, data) = detection_subscriber.check_for_update(timeout=1)
+            _result = detection_subscriber.check_for_update(timeout=1)
+            if _result is None:
+                continue
+            (topic, data) = _result
             now = datetime.datetime.now().timestamp()
 
             if now - last_disabled_cam_check > 5:
@@ -208,7 +214,7 @@ class OutputProcess(FrigateProcess):
                     self.config, birdseye, preview_recorders, preview_write_times
                 )
 
-            if not topic:
+            if not topic or data is None:
                 continue
 
             (
@@ -262,11 +268,15 @@ class OutputProcess(FrigateProcess):
                 jsmpeg_cameras[camera].write_frame(frame.tobytes())
 
             # send output data to birdseye if websocket is connected or restreaming
-            if self.config.birdseye.enabled and (
-                self.config.birdseye.restream
-                or any(
-                    ws.environ["PATH_INFO"].endswith("birdseye")
-                    for ws in websocket_server.manager
+            if (
+                self.config.birdseye.enabled
+                and birdseye is not None
+                and (
+                    self.config.birdseye.restream
+                    or any(
+                        ws.environ["PATH_INFO"].endswith("birdseye")
+                        for ws in websocket_server.manager
+                    )
                 )
             ):
                 birdseye.write_data(
@@ -282,9 +292,12 @@ class OutputProcess(FrigateProcess):
         move_preview_frames("clips")
 
         while True:
-            (topic, data) = detection_subscriber.check_for_update(timeout=0)
+            _cleanup_result = detection_subscriber.check_for_update(timeout=0)
+            if _cleanup_result is None:
+                break
+            (topic, data) = _cleanup_result
 
-            if not topic:
+            if not topic or data is None:
                 break
 
             (
@@ -322,7 +335,7 @@ class OutputProcess(FrigateProcess):
         logger.info("exiting output process...")
 
 
-def move_preview_frames(loc: str):
+def move_preview_frames(loc: str) -> None:
     preview_holdover = os.path.join(CLIPS_DIR, "preview_restart_cache")
     preview_cache = os.path.join(CACHE_DIR, "preview_frames")
 

frigate/output/preview.py

@@ -22,7 +22,6 @@ from frigate.ffmpeg_presets import (
     parse_preset_hardware_acceleration_encode,
 )
 from frigate.models import Previews
-from frigate.track.object_processing import TrackedObject
 from frigate.util.image import copy_yuv_to_position, get_blank_yuv_frame, get_yuv_crop
 
 logger = logging.getLogger(__name__)
@@ -66,7 +65,9 @@ def get_cache_image_name(camera: str, frame_time: float) -> str:
     )
 
 
-def get_most_recent_preview_frame(camera: str, before: float = None) -> str | None:
+def get_most_recent_preview_frame(
+    camera: str, before: float | None = None
+) -> str | None:
     """Get the most recent preview frame for a camera."""
     if not os.path.exists(PREVIEW_CACHE_DIR):
         return None
@@ -147,12 +148,12 @@ class FFMpegConverter(threading.Thread):
             if t_idx == item_count - 1:
                 # last frame does not get a duration
                 playlist.append(
-                    f"file '{get_cache_image_name(self.config.name, self.frame_times[t_idx])}'"
+                    f"file '{get_cache_image_name(self.config.name, self.frame_times[t_idx])}'"  # type: ignore[arg-type]
                 )
                 continue
 
             playlist.append(
-                f"file '{get_cache_image_name(self.config.name, self.frame_times[t_idx])}'"
+                f"file '{get_cache_image_name(self.config.name, self.frame_times[t_idx])}'"  # type: ignore[arg-type]
             )
             playlist.append(
                 f"duration {self.frame_times[t_idx + 1] - self.frame_times[t_idx]}"
@@ -199,30 +200,33 @@ class FFMpegConverter(threading.Thread):
         # unlink files from cache
         # don't delete last frame as it will be used as first frame in next segment
         for t in self.frame_times[0:-1]:
-            Path(get_cache_image_name(self.config.name, t)).unlink(missing_ok=True)
+            Path(get_cache_image_name(self.config.name, t)).unlink(missing_ok=True)  # type: ignore[arg-type]
 
 
 class PreviewRecorder:
     def __init__(self, config: CameraConfig) -> None:
         self.config = config
-        self.start_time = 0
-        self.last_output_time = 0
+        self.camera_name: str = config.name or ""
+        self.start_time: float = 0
+        self.last_output_time: float = 0
         self.offline = False
-        self.output_frames = []
+        self.output_frames: list[float] = []
 
-        if config.detect.width > config.detect.height:
+        if config.detect.width is None or config.detect.height is None:
+            raise ValueError("Detect width and height must be set for previews.")
+
+        self.detect_width: int = config.detect.width
+        self.detect_height: int = config.detect.height
+
+        if self.detect_width > self.detect_height:
             self.out_height = PREVIEW_HEIGHT
             self.out_width = (
-                int((config.detect.width / config.detect.height) * self.out_height)
-                // 4
-                * 4
+                int((self.detect_width / self.detect_height) * self.out_height) // 4 * 4
             )
         else:
             self.out_width = PREVIEW_HEIGHT
             self.out_height = (
-                int((config.detect.height / config.detect.width) * self.out_width)
-                // 4
-                * 4
+                int((self.detect_height / self.detect_width) * self.out_width) // 4 * 4
             )
 
         # create communication for finished previews
@@ -302,7 +306,7 @@ class PreviewRecorder:
         )
         self.start_time = frame_time
         self.last_output_time = frame_time
-        self.output_frames: list[float] = []
+        self.output_frames = []
 
     def should_write_frame(
         self,
@@ -342,7 +346,9 @@ class PreviewRecorder:
 
     def write_frame_to_cache(self, frame_time: float, frame: np.ndarray) -> None:
         # resize yuv frame
-        small_frame = np.zeros((self.out_height * 3 // 2, self.out_width), np.uint8)
+        small_frame: np.ndarray = np.zeros(
+            (self.out_height * 3 // 2, self.out_width), np.uint8
+        )
         copy_yuv_to_position(
             small_frame,
             (0, 0),
@@ -356,7 +362,7 @@ class PreviewRecorder:
             cv2.COLOR_YUV2BGR_I420,
         )
         cv2.imwrite(
-            get_cache_image_name(self.config.name, frame_time),
+            get_cache_image_name(self.camera_name, frame_time),
             small_frame,
             [
                 int(cv2.IMWRITE_WEBP_QUALITY),
@@ -396,7 +402,7 @@ class PreviewRecorder:
                 ).start()
             else:
                 logger.debug(
-                    f"Not saving preview for {self.config.name} because there are no saved frames."
+                    f"Not saving preview for {self.camera_name} because there are no saved frames."
                 )
 
             self.reset_frame_cache(frame_time)
@@ -416,9 +422,7 @@ class PreviewRecorder:
         if not self.offline:
             self.write_frame_to_cache(
                 frame_time,
-                get_blank_yuv_frame(
-                    self.config.detect.width, self.config.detect.height
-                ),
+                get_blank_yuv_frame(self.detect_width, self.detect_height),
             )
             self.offline = True
 
@@ -431,9 +435,9 @@ class PreviewRecorder:
                 return
 
             old_frame_path = get_cache_image_name(
-                self.config.name, self.output_frames[-1]
+                self.camera_name, self.output_frames[-1]
             )
-            new_frame_path = get_cache_image_name(self.config.name, frame_time)
+            new_frame_path = get_cache_image_name(self.camera_name, frame_time)
             shutil.copy(old_frame_path, new_frame_path)
 
             # save last frame to ensure consistent duration
@@ -447,13 +451,12 @@ class PreviewRecorder:
             self.reset_frame_cache(frame_time)
 
     def stop(self) -> None:
-        self.config_subscriber.stop()
         self.requestor.stop()
 
 
 def get_active_objects(
-    frame_time: float, camera_config: CameraConfig, all_objects: list[TrackedObject]
-) -> list[TrackedObject]:
+    frame_time: float, camera_config: CameraConfig, all_objects: list[dict[str, Any]]
+) -> list[dict[str, Any]]:
     """get active objects for detection."""
     return [
         o

frigate/record/cleanup.py

@@ -7,6 +7,7 @@ import os
 import threading
 from multiprocessing.synchronize import Event as MpEvent
 from pathlib import Path
+from typing import Any
 
 from playhouse.sqlite_ext import SqliteExtDatabase
 
@@ -60,7 +61,9 @@ class RecordingCleanup(threading.Thread):
             db.execute_sql("PRAGMA wal_checkpoint(TRUNCATE);")
             db.close()
 
-    def expire_review_segments(self, config: CameraConfig, now: datetime) -> set[Path]:
+    def expire_review_segments(
+        self, config: CameraConfig, now: datetime.datetime
+    ) -> set[Path]:
         """Delete review segments that are expired"""
         alert_expire_date = (
             now - datetime.timedelta(days=config.record.alerts.retain.days)
@@ -68,7 +71,7 @@ class RecordingCleanup(threading.Thread):
         detection_expire_date = (
             now - datetime.timedelta(days=config.record.detections.retain.days)
         ).timestamp()
-        expired_reviews: ReviewSegment = (
+        expired_reviews = (
             ReviewSegment.select(ReviewSegment.id, ReviewSegment.thumb_path)
             .where(ReviewSegment.camera == config.name)
             .where(
@@ -109,13 +112,13 @@ class RecordingCleanup(threading.Thread):
         continuous_expire_date: float,
         motion_expire_date: float,
         config: CameraConfig,
-        reviews: ReviewSegment,
+        reviews: list[Any],
     ) -> set[Path]:
         """Delete recordings for existing camera based on retention config."""
         # Get the timestamp for cutoff of retained days
 
         # Get recordings to check for expiration
-        recordings: Recordings = (
+        recordings = (
             Recordings.select(
                 Recordings.id,
                 Recordings.start_time,
@@ -148,13 +151,12 @@ class RecordingCleanup(threading.Thread):
         review_start = 0
         deleted_recordings = set()
         kept_recordings: list[tuple[float, float]] = []
-        recording: Recordings
         for recording in recordings:
             keep = False
             mode = None
             # Now look for a reason to keep this recording segment
             for idx in range(review_start, len(reviews)):
-                review: ReviewSegment = reviews[idx]
+                review = reviews[idx]
                 severity = review.severity
                 pre_capture = config.record.get_review_pre_capture(severity)
                 post_capture = config.record.get_review_post_capture(severity)
@@ -214,7 +216,7 @@ class RecordingCleanup(threading.Thread):
                 Recordings.id << deleted_recordings_list[i : i + max_deletes]
             ).execute()
 
-        previews: list[Previews] = (
+        previews = (
             Previews.select(
                 Previews.id,
                 Previews.start_time,
@@ -290,13 +292,13 @@ class RecordingCleanup(threading.Thread):
         expire_before = (
             datetime.datetime.now() - datetime.timedelta(days=expire_days)
         ).timestamp()
-        no_camera_recordings: Recordings = (
+        no_camera_recordings = (
             Recordings.select(
                 Recordings.id,
                 Recordings.path,
             )
             .where(
-                Recordings.camera.not_in(list(self.config.cameras.keys())),
+                Recordings.camera.not_in(list(self.config.cameras.keys())),  # type: ignore[call-arg, arg-type, misc]
                 Recordings.end_time < expire_before,
             )
             .namedtuples()
@@ -341,7 +343,7 @@ class RecordingCleanup(threading.Thread):
             ).timestamp()
 
             # Get all the reviews to check against
-            reviews: ReviewSegment = (
+            reviews = (
                 ReviewSegment.select(
                     ReviewSegment.start_time,
                     ReviewSegment.end_time,

frigate/record/export.py

@@ -85,10 +85,6 @@ def validate_ffmpeg_args(args: str) -> tuple[bool, str]:
     return True, ""
 
 
-def lower_priority():
-    os.nice(PROCESS_PRIORITY_LOW)
-
-
 class PlaybackSourceEnum(str, Enum):
     recordings = "recordings"
     preview = "preview"
@@ -150,7 +146,7 @@ class RecordingExporter(threading.Thread):
         ):
             # has preview mp4
             try:
-                preview: Previews = (
+                preview = (
                     Previews.select(
                         Previews.camera,
                         Previews.path,
@@ -231,20 +227,19 @@ class RecordingExporter(threading.Thread):
 
     def get_record_export_command(
         self, video_path: str, use_hwaccel: bool = True
-    ) -> list[str]:
+    ) -> tuple[list[str], str | list[str]]:
         # handle case where internal port is a string with ip:port
         internal_port = self.config.networking.listen.internal
         if type(internal_port) is str:
             internal_port = int(internal_port.split(":")[-1])
 
+        playlist_lines: list[str] = []
         if (self.end_time - self.start_time) <= MAX_PLAYLIST_SECONDS:
-            playlist_lines = f"http://127.0.0.1:{internal_port}/vod/{self.camera}/start/{self.start_time}/end/{self.end_time}/index.m3u8"
+            playlist_url = f"http://127.0.0.1:{internal_port}/vod/{self.camera}/start/{self.start_time}/end/{self.end_time}/index.m3u8"
             ffmpeg_input = (
-                f"-y -protocol_whitelist pipe,file,http,tcp -i {playlist_lines}"
+                f"-y -protocol_whitelist pipe,file,http,tcp -i {playlist_url}"
             )
         else:
-            playlist_lines = []
-
             # get full set of recordings
             export_recordings = (
                 Recordings.select(
@@ -305,7 +300,7 @@ class RecordingExporter(threading.Thread):
 
     def get_preview_export_command(
         self, video_path: str, use_hwaccel: bool = True
-    ) -> list[str]:
+    ) -> tuple[list[str], list[str]]:
         playlist_lines = []
         codec = "-c copy"
 
@@ -355,7 +350,6 @@ class RecordingExporter(threading.Thread):
             .iterator()
         )
 
-        preview: Previews
         for preview in export_previews:
             playlist_lines.append(f"file '{preview.path}'")
 
@@ -441,10 +435,9 @@ class RecordingExporter(threading.Thread):
             return
 
         p = sp.run(
-            ffmpeg_cmd,
+            ["nice", "-n", str(PROCESS_PRIORITY_LOW)] + ffmpeg_cmd,
             input="\n".join(playlist_lines),
             encoding="ascii",
-            preexec_fn=lower_priority,
             capture_output=True,
         )
 
@@ -469,10 +462,9 @@ class RecordingExporter(threading.Thread):
                 )
 
             p = sp.run(
-                ffmpeg_cmd,
+                ["nice", "-n", str(PROCESS_PRIORITY_LOW)] + ffmpeg_cmd,
                 input="\n".join(playlist_lines),
                 encoding="ascii",
-                preexec_fn=lower_priority,
                 capture_output=True,
             )
 
@@ -493,7 +485,7 @@ class RecordingExporter(threading.Thread):
         logger.debug(f"Finished exporting {video_path}")
 
 
-def migrate_exports(ffmpeg: FfmpegConfig, camera_names: list[str]):
+def migrate_exports(ffmpeg: FfmpegConfig, camera_names: list[str]) -> None:
     Path(os.path.join(CLIPS_DIR, "export")).mkdir(exist_ok=True)
 
     exports = []

frigate/record/maintainer.py

@@ -266,7 +266,7 @@ class RecordingMaintainer(threading.Thread):
 
             # get all reviews with the end time after the start of the oldest cache file
             # or with end_time None
-            reviews: ReviewSegment = (
+            reviews = (
                 ReviewSegment.select(
                     ReviewSegment.start_time,
                     ReviewSegment.end_time,
@@ -301,7 +301,9 @@ class RecordingMaintainer(threading.Thread):
                 RecordingsDataTypeEnum.saved.value,
             )
 
-        recordings_to_insert: list[Optional[Recordings]] = await asyncio.gather(*tasks)
+        recordings_to_insert: list[Optional[dict[str, Any]]] = await asyncio.gather(
+            *tasks
+        )
 
         # fire and forget recordings entries
         self.requestor.send_data(
@@ -314,8 +316,8 @@ class RecordingMaintainer(threading.Thread):
         self.end_time_cache.pop(cache_path, None)
 
     async def validate_and_move_segment(
-        self, camera: str, reviews: list[ReviewSegment], recording: dict[str, Any]
-    ) -> Optional[Recordings]:
+        self, camera: str, reviews: Any, recording: dict[str, Any]
+    ) -> Optional[dict[str, Any]]:
         cache_path: str = recording["cache_path"]
         start_time: datetime.datetime = recording["start_time"]
 
@@ -456,6 +458,8 @@ class RecordingMaintainer(threading.Thread):
             if end_time < retain_cutoff:
                 self.drop_segment(cache_path)
 
+        return None
+
     def _compute_motion_heatmap(
         self, camera: str, motion_boxes: list[tuple[int, int, int, int]]
     ) -> dict[str, int] | None:
@@ -481,7 +485,7 @@ class RecordingMaintainer(threading.Thread):
         frame_width = camera_config.detect.width
         frame_height = camera_config.detect.height
 
-        if frame_width <= 0 or frame_height <= 0:
+        if not frame_width or frame_width <= 0 or not frame_height or frame_height <= 0:
             return None
 
         GRID_SIZE = 16
@@ -575,13 +579,13 @@ class RecordingMaintainer(threading.Thread):
         duration: float,
         cache_path: str,
         store_mode: RetainModeEnum,
-    ) -> Optional[Recordings]:
+    ) -> Optional[dict[str, Any]]:
         segment_info = self.segment_stats(camera, start_time, end_time)
 
         # check if the segment shouldn't be stored
         if segment_info.should_discard_segment(store_mode):
             self.drop_segment(cache_path)
-            return
+            return None
 
         # directory will be in utc due to start_time being in utc
         directory = os.path.join(
@@ -620,7 +624,8 @@ class RecordingMaintainer(threading.Thread):
 
                 if p.returncode != 0:
                     logger.error(f"Unable to convert {cache_path} to {file_path}")
-                    logger.error((await p.stderr.read()).decode("ascii"))
+                    if p.stderr:
+                        logger.error((await p.stderr.read()).decode("ascii"))
                     return None
                 else:
                     logger.debug(
@@ -684,11 +689,16 @@ class RecordingMaintainer(threading.Thread):
             stale_frame_count_threshold = 10
             # empty the object recordings info queue
             while True:
-                (topic, data) = self.detection_subscriber.check_for_update(
+                result = self.detection_subscriber.check_for_update(
                     timeout=FAST_QUEUE_TIMEOUT
                 )
 
-                if not topic:
+                if not result:
+                    break
+
+                topic, data = result
+
+                if not topic or not data:
                     break
 
                 if topic == DetectionTypeEnum.video.value:

frigate/review/maintainer.py

@@ -31,7 +31,7 @@ from frigate.const import (
 )
 from frigate.models import ReviewSegment
 from frigate.review.types import SeverityEnum
-from frigate.track.object_processing import ManualEventState, TrackedObject
+from frigate.track.object_processing import ManualEventState
 from frigate.util.image import SharedMemoryFrameManager, calculate_16_9_crop
 
 logger = logging.getLogger(__name__)
@@ -69,7 +69,9 @@ class PendingReviewSegment:
             self.last_alert_time = frame_time
 
         # thumbnail
-        self._frame = np.zeros((THUMB_HEIGHT * 3 // 2, THUMB_WIDTH), np.uint8)
+        self._frame: np.ndarray[Any, Any] = np.zeros(
+            (THUMB_HEIGHT * 3 // 2, THUMB_WIDTH), np.uint8
+        )
         self.has_frame = False
         self.frame_active_count = 0
         self.frame_path = os.path.join(
@@ -77,8 +79,11 @@ class PendingReviewSegment:
         )
 
     def update_frame(
-        self, camera_config: CameraConfig, frame, objects: list[TrackedObject]
-    ):
+        self,
+        camera_config: CameraConfig,
+        frame: np.ndarray,
+        objects: list[dict[str, Any]],
+    ) -> None:
         min_x = camera_config.frame_shape[1]
         min_y = camera_config.frame_shape[0]
         max_x = 0
@@ -114,7 +119,7 @@ class PendingReviewSegment:
                 self.frame_path, self._frame, [int(cv2.IMWRITE_WEBP_QUALITY), 60]
             )
 
-    def save_full_frame(self, camera_config: CameraConfig, frame):
+    def save_full_frame(self, camera_config: CameraConfig, frame: np.ndarray) -> None:
         color_frame = cv2.cvtColor(frame, cv2.COLOR_YUV2BGR_I420)
         width = int(THUMB_HEIGHT * color_frame.shape[1] / color_frame.shape[0])
         self._frame = cv2.resize(
@@ -165,13 +170,13 @@ class ActiveObjects:
         self,
         frame_time: float,
         camera_config: CameraConfig,
-        all_objects: list[TrackedObject],
+        all_objects: list[dict[str, Any]],
     ):
         self.camera_config = camera_config
 
         # get current categorization of objects to know if
         # these objects are currently being categorized
-        self.categorized_objects = {
+        self.categorized_objects: dict[str, list[dict[str, Any]]] = {
             "alerts": [],
             "detections": [],
         }
@@ -250,7 +255,7 @@ class ActiveObjects:
 
         return False
 
-    def get_all_objects(self) -> list[TrackedObject]:
+    def get_all_objects(self) -> list[dict[str, Any]]:
         return (
             self.categorized_objects["alerts"] + self.categorized_objects["detections"]
         )
@@ -309,7 +314,7 @@ class ReviewSegmentMaintainer(threading.Thread):
             "reviews",
             json.dumps(review_update),
         )
-        self.review_publisher.publish(review_update, segment.camera)
+        self.review_publisher.publish(review_update, segment.camera)  # type: ignore[arg-type]
         self.requestor.send_data(
             f"{segment.camera}/review_status", segment.severity.value.upper()
         )
@@ -318,8 +323,8 @@ class ReviewSegmentMaintainer(threading.Thread):
         self,
         segment: PendingReviewSegment,
         camera_config: CameraConfig,
-        frame,
-        objects: list[TrackedObject],
+        frame: Optional[np.ndarray],
+        objects: list[dict[str, Any]],
         prev_data: dict[str, Any],
     ) -> None:
         """Update segment."""
@@ -337,7 +342,7 @@ class ReviewSegmentMaintainer(threading.Thread):
             "reviews",
             json.dumps(review_update),
         )
-        self.review_publisher.publish(review_update, segment.camera)
+        self.review_publisher.publish(review_update, segment.camera)  # type: ignore[arg-type]
         self.requestor.send_data(
             f"{segment.camera}/review_status", segment.severity.value.upper()
         )
@@ -346,7 +351,7 @@ class ReviewSegmentMaintainer(threading.Thread):
         self,
         segment: PendingReviewSegment,
         prev_data: dict[str, Any],
-    ) -> float:
+    ) -> Any:
         """End segment."""
         final_data = segment.get_data(ended=True)
         end_time = final_data[ReviewSegment.end_time.name]
@@ -360,24 +365,25 @@ class ReviewSegmentMaintainer(threading.Thread):
             "reviews",
             json.dumps(review_update),
         )
-        self.review_publisher.publish(review_update, segment.camera)
+        self.review_publisher.publish(review_update, segment.camera)  # type: ignore[arg-type]
         self.requestor.send_data(f"{segment.camera}/review_status", "NONE")
         self.active_review_segments[segment.camera] = None
         return end_time
 
-    def forcibly_end_segment(self, camera: str) -> float:
+    def forcibly_end_segment(self, camera: str) -> Any:
         """Forcibly end the pending segment for a camera."""
         segment = self.active_review_segments.get(camera)
         if segment:
             prev_data = segment.get_data(False)
             return self._publish_segment_end(segment, prev_data)
+        return None
 
     def update_existing_segment(
         self,
         segment: PendingReviewSegment,
         frame_name: str,
         frame_time: float,
-        objects: list[TrackedObject],
+        objects: list[dict[str, Any]],
     ) -> None:
         """Validate if existing review segment should continue."""
         camera_config = self.config.cameras[segment.camera]
@@ -492,8 +498,11 @@ class ReviewSegmentMaintainer(threading.Thread):
                 except FileNotFoundError:
                     return
 
-            if segment.severity == SeverityEnum.alert and frame_time > (
-                segment.last_alert_time + camera_config.review.alerts.cutoff_time
+            if (
+                segment.severity == SeverityEnum.alert
+                and segment.last_alert_time is not None
+                and frame_time
+                > (segment.last_alert_time + camera_config.review.alerts.cutoff_time)
             ):
                 needs_new_detection = (
                     segment.last_detection_time > segment.last_alert_time
@@ -516,23 +525,18 @@ class ReviewSegmentMaintainer(threading.Thread):
                         new_zones.update(o["current_zones"])
 
                     if new_detections:
-                        self.active_review_segments[activity.camera_config.name] = (
-                            PendingReviewSegment(
-                                activity.camera_config.name,
-                                end_time,
-                                SeverityEnum.detection,
-                                new_detections,
-                                sub_labels={},
-                                audio=set(),
-                                zones=list(new_zones),
-                            )
+                        new_segment = PendingReviewSegment(
+                            segment.camera,
+                            end_time,
+                            SeverityEnum.detection,
+                            new_detections,
+                            sub_labels={},
+                            audio=set(),
+                            zones=list(new_zones),
                         )
-                        self._publish_segment_start(
-                            self.active_review_segments[activity.camera_config.name]
-                        )
-                        self.active_review_segments[
-                            activity.camera_config.name
-                        ].last_detection_time = last_detection_time
+                        self.active_review_segments[segment.camera] = new_segment
+                        self._publish_segment_start(new_segment)
+                        new_segment.last_detection_time = last_detection_time
             elif segment.severity == SeverityEnum.detection and frame_time > (
                 segment.last_detection_time
                 + camera_config.review.detections.cutoff_time
@@ -544,7 +548,7 @@ class ReviewSegmentMaintainer(threading.Thread):
         camera: str,
         frame_name: str,
         frame_time: float,
-        objects: list[TrackedObject],
+        objects: list[dict[str, Any]],
     ) -> None:
         """Check if a new review segment should be created."""
         camera_config = self.config.cameras[camera]
@@ -581,7 +585,7 @@ class ReviewSegmentMaintainer(threading.Thread):
                         zones.append(zone)
 
             if severity:
-                self.active_review_segments[camera] = PendingReviewSegment(
+                new_segment = PendingReviewSegment(
                     camera,
                     frame_time,
                     severity,
@@ -590,6 +594,7 @@ class ReviewSegmentMaintainer(threading.Thread):
                     audio=set(),
                     zones=zones,
                 )
+                self.active_review_segments[camera] = new_segment
 
                 try:
                     yuv_frame = self.frame_manager.get(
@@ -600,11 +605,11 @@ class ReviewSegmentMaintainer(threading.Thread):
                         logger.debug(f"Failed to get frame {frame_name} from SHM")
                         return
 
-                    self.active_review_segments[camera].update_frame(
+                    new_segment.update_frame(
                         camera_config, yuv_frame, activity.get_all_objects()
                     )
                     self.frame_manager.close(frame_name)
-                    self._publish_segment_start(self.active_review_segments[camera])
+                    self._publish_segment_start(new_segment)
                 except FileNotFoundError:
                     return
 
@@ -621,9 +626,14 @@ class ReviewSegmentMaintainer(threading.Thread):
                 for camera in updated_topics["enabled"]:
                     self.forcibly_end_segment(camera)
 
-            (topic, data) = self.detection_subscriber.check_for_update(timeout=1)
+            result = self.detection_subscriber.check_for_update(timeout=1)
 
-            if not topic:
+            if not result:
+                continue
+
+            topic, data = result
+
+            if not topic or not data:
                 continue
 
             if topic == DetectionTypeEnum.video.value:
@@ -712,10 +722,13 @@ class ReviewSegmentMaintainer(threading.Thread):
                         if topic == DetectionTypeEnum.api:
                             # manual_info["label"] contains 'label: sub_label'
                             # so split out the label without modifying manual_info
+                            det_labels = self.config.cameras[
+                                camera
+                            ].review.detections.labels
                             if (
                                 self.config.cameras[camera].review.detections.enabled
-                                and manual_info["label"].split(": ")[0]
-                                in self.config.cameras[camera].review.detections.labels
+                                and det_labels is not None
+                                and manual_info["label"].split(": ")[0] in det_labels
                             ):
                                 current_segment.last_detection_time = manual_info[
                                     "end_time"
@@ -744,14 +757,15 @@ class ReviewSegmentMaintainer(threading.Thread):
                         ):
                             # manual_info["label"] contains 'label: sub_label'
                             # so split out the label without modifying manual_info
+                            det_labels = self.config.cameras[
+                                camera
+                            ].review.detections.labels
                             if (
                                 not self.config.cameras[
                                     camera
                                 ].review.detections.enabled
-                                or manual_info["label"].split(": ")[0]
-                                not in self.config.cameras[
-                                    camera
-                                ].review.detections.labels
+                                or det_labels is None
+                                or manual_info["label"].split(": ")[0] not in det_labels
                             ):
                                 current_segment.severity = SeverityEnum.alert
                         elif (
@@ -828,17 +842,18 @@ class ReviewSegmentMaintainer(threading.Thread):
                     severity = None
                     # manual_info["label"] contains 'label: sub_label'
                     # so split out the label without modifying manual_info
+                    det_labels = self.config.cameras[camera].review.detections.labels
                     if (
                         self.config.cameras[camera].review.detections.enabled
-                        and manual_info["label"].split(": ")[0]
-                        in self.config.cameras[camera].review.detections.labels
+                        and det_labels is not None
+                        and manual_info["label"].split(": ")[0] in det_labels
                     ):
                         severity = SeverityEnum.detection
                     elif self.config.cameras[camera].review.alerts.enabled:
                         severity = SeverityEnum.alert
 
                     if severity:
-                        self.active_review_segments[camera] = PendingReviewSegment(
+                        api_segment = PendingReviewSegment(
                             camera,
                             frame_time,
                             severity,
@@ -847,32 +862,25 @@ class ReviewSegmentMaintainer(threading.Thread):
                             [],
                             set(),
                         )
+                        self.active_review_segments[camera] = api_segment
 
                         if manual_info["state"] == ManualEventState.start:
                             self.indefinite_events[camera][manual_info["event_id"]] = (
                                 manual_info["label"]
                             )
                             # temporarily make it so this event can not end
-                            self.active_review_segments[
-                                camera
-                            ].last_alert_time = sys.maxsize
-                            self.active_review_segments[
-                                camera
-                            ].last_detection_time = sys.maxsize
+                            api_segment.last_alert_time = sys.maxsize
+                            api_segment.last_detection_time = sys.maxsize
                         elif manual_info["state"] == ManualEventState.complete:
-                            self.active_review_segments[
-                                camera
-                            ].last_alert_time = manual_info["end_time"]
-                            self.active_review_segments[
-                                camera
-                            ].last_detection_time = manual_info["end_time"]
+                            api_segment.last_alert_time = manual_info["end_time"]
+                            api_segment.last_detection_time = manual_info["end_time"]
                     else:
                         logger.warning(
                             f"Manual event API has been called for {camera}, but alerts and detections are disabled. This manual event will not appear as an alert or detection."
                         )
                 elif topic == DetectionTypeEnum.lpr:
                     if self.config.cameras[camera].review.detections.enabled:
-                        self.active_review_segments[camera] = PendingReviewSegment(
+                        lpr_segment = PendingReviewSegment(
                             camera,
                             frame_time,
                             SeverityEnum.detection,
@@ -881,25 +889,18 @@ class ReviewSegmentMaintainer(threading.Thread):
                             [],
                             set(),
                         )
+                        self.active_review_segments[camera] = lpr_segment
 
                         if manual_info["state"] == ManualEventState.start:
                             self.indefinite_events[camera][manual_info["event_id"]] = (
                                 manual_info["label"]
                             )
                             # temporarily make it so this event can not end
-                            self.active_review_segments[
-                                camera
-                            ].last_alert_time = sys.maxsize
-                            self.active_review_segments[
-                                camera
-                            ].last_detection_time = sys.maxsize
+                            lpr_segment.last_alert_time = sys.maxsize
+                            lpr_segment.last_detection_time = sys.maxsize
                         elif manual_info["state"] == ManualEventState.complete:
-                            self.active_review_segments[
-                                camera
-                            ].last_alert_time = manual_info["end_time"]
-                            self.active_review_segments[
-                                camera
-                            ].last_detection_time = manual_info["end_time"]
+                            lpr_segment.last_alert_time = manual_info["end_time"]
+                            lpr_segment.last_detection_time = manual_info["end_time"]
                     else:
                         logger.warning(
                             f"Dedicated LPR camera API has been called for {camera}, but detections are disabled. LPR events will not appear as a detection."

frigate/stats/util.py

@@ -19,6 +19,7 @@ from frigate.types import StatsTrackingTypes
 from frigate.util.services import (
     calculate_shm_requirements,
     get_amd_gpu_stats,
+    get_axcl_npu_stats,
     get_bandwidth_stats,
     get_cpu_stats,
     get_fs_type,
@@ -324,6 +325,10 @@ async def set_npu_usages(config: FrigateConfig, all_stats: dict[str, Any]) -> No
             # OpenVINO NPU usage
             ov_usage = get_openvino_npu_stats()
             stats["openvino"] = ov_usage
+        elif detector.type == "axengine":
+            # AXERA NPU usage
+            axcl_usage = get_axcl_npu_stats()
+            stats["axengine"] = axcl_usage
 
     if stats:
         all_stats["npu_usages"] = stats

frigate/storage.py

@@ -3,6 +3,7 @@
 import logging
 import shutil
 import threading
+from multiprocessing.synchronize import Event as MpEvent
 from pathlib import Path
 
 from peewee import SQL, fn
@@ -23,7 +24,7 @@ MAX_CALCULATED_BANDWIDTH = 10000  # 10Gb/hr
 class StorageMaintainer(threading.Thread):
     """Maintain frigates recording storage."""
 
-    def __init__(self, config: FrigateConfig, stop_event) -> None:
+    def __init__(self, config: FrigateConfig, stop_event: MpEvent) -> None:
         super().__init__(name="storage_maintainer")
         self.config = config
         self.stop_event = stop_event
@@ -114,7 +115,7 @@ class StorageMaintainer(threading.Thread):
         logger.debug(
             f"Storage cleanup check: {hourly_bandwidth} hourly with remaining storage: {remaining_storage}."
         )
-        return remaining_storage < hourly_bandwidth
+        return remaining_storage < float(hourly_bandwidth)
 
     def reduce_storage_consumption(self) -> None:
         """Remove oldest hour of recordings."""
@@ -124,7 +125,7 @@ class StorageMaintainer(threading.Thread):
             [b["bandwidth"] for b in self.camera_storage_stats.values()]
         )
 
-        recordings: Recordings = (
+        recordings = (
             Recordings.select(
                 Recordings.id,
                 Recordings.camera,
@@ -138,7 +139,7 @@ class StorageMaintainer(threading.Thread):
             .iterator()
         )
 
-        retained_events: Event = (
+        retained_events = (
             Event.select(
                 Event.start_time,
                 Event.end_time,
@@ -278,7 +279,7 @@ class StorageMaintainer(threading.Thread):
                 Recordings.id << deleted_recordings_list[i : i + max_deletes]
             ).execute()
 
-    def run(self):
+    def run(self) -> None:
         """Check every 5 minutes if storage needs to be cleaned up."""
         if self.config.safe_mode:
             logger.info("Safe mode enabled, skipping storage maintenance")

frigate/test/test_config.py

@@ -10,7 +10,7 @@ from ruamel.yaml.constructor import DuplicateKeyError
 from frigate.config import BirdseyeModeEnum, FrigateConfig
 from frigate.const import MODEL_CACHE_DIR
 from frigate.detectors import DetectorTypeEnum
-from frigate.util.builtin import deep_merge
+from frigate.util.builtin import deep_merge, load_labels
 
 
 class TestConfig(unittest.TestCase):
@@ -288,6 +288,65 @@ class TestConfig(unittest.TestCase):
         frigate_config = FrigateConfig(**config)
         assert "dog" in frigate_config.cameras["back"].objects.filters
 
+    def test_default_audio_filters(self):
+        config = {
+            "mqtt": {"host": "mqtt"},
+            "audio": {"listen": ["speech", "yell"]},
+            "cameras": {
+                "back": {
+                    "ffmpeg": {
+                        "inputs": [
+                            {"path": "rtsp://10.0.0.1:554/video", "roles": ["detect"]}
+                        ]
+                    },
+                    "detect": {
+                        "height": 1080,
+                        "width": 1920,
+                        "fps": 5,
+                    },
+                }
+            },
+        }
+
+        frigate_config = FrigateConfig(**config)
+        all_audio_labels = {
+            label
+            for label in load_labels("/audio-labelmap.txt", prefill=521).values()
+            if label
+        }
+
+        assert all_audio_labels.issubset(
+            set(frigate_config.cameras["back"].audio.filters.keys())
+        )
+
+    def test_override_audio_filters(self):
+        config = {
+            "mqtt": {"host": "mqtt"},
+            "cameras": {
+                "back": {
+                    "ffmpeg": {
+                        "inputs": [
+                            {"path": "rtsp://10.0.0.1:554/video", "roles": ["detect"]}
+                        ]
+                    },
+                    "detect": {
+                        "height": 1080,
+                        "width": 1920,
+                        "fps": 5,
+                    },
+                    "audio": {
+                        "listen": ["speech", "yell"],
+                        "filters": {"speech": {"threshold": 0.9}},
+                    },
+                }
+            },
+        }
+
+        frigate_config = FrigateConfig(**config)
+        assert "speech" in frigate_config.cameras["back"].audio.filters
+        assert frigate_config.cameras["back"].audio.filters["speech"].threshold == 0.9
+        assert "babbling" in frigate_config.cameras["back"].audio.filters
+
     def test_inherit_object_filters(self):
         config = {
             "mqtt": {"host": "mqtt"},

frigate/timeline.py

@@ -8,7 +8,7 @@ from multiprocessing.synchronize import Event as MpEvent
 from typing import Any
 
 from frigate.config import FrigateConfig
-from frigate.events.maintainer import EventStateEnum, EventTypeEnum
+from frigate.events.types import EventStateEnum, EventTypeEnum
 from frigate.models import Timeline
 from frigate.util.builtin import to_relative_box
 
@@ -28,7 +28,7 @@ class TimelineProcessor(threading.Thread):
         self.config = config
         self.queue = queue
         self.stop_event = stop_event
-        self.pre_event_cache: dict[str, list[dict[str, Any]]] = {}
+        self.pre_event_cache: dict[str, list[dict[Any, Any]]] = {}
 
     def run(self) -> None:
         while not self.stop_event.is_set():
@@ -56,7 +56,7 @@ class TimelineProcessor(threading.Thread):
 
     def insert_or_save(
         self,
-        entry: dict[str, Any],
+        entry: dict[Any, Any],
         prev_event_data: dict[Any, Any],
         event_data: dict[Any, Any],
     ) -> None:
@@ -84,11 +84,15 @@ class TimelineProcessor(threading.Thread):
         event_type: str,
         prev_event_data: dict[Any, Any],
         event_data: dict[Any, Any],
-    ) -> bool:
+    ) -> None:
         """Handle object detection."""
         camera_config = self.config.cameras.get(camera)
-        if camera_config is None:
-            return False
+        if (
+            camera_config is None
+            or camera_config.detect.width is None
+            or camera_config.detect.height is None
+        ):
+            return
         event_id = event_data["id"]
 
         # Base timeline entry data that all entries will share

frigate/track/object_processing.py

@@ -81,6 +81,7 @@ class TrackedObjectProcessor(threading.Thread):
                 CameraConfigUpdateEnum.motion,
                 CameraConfigUpdateEnum.objects,
                 CameraConfigUpdateEnum.remove,
+                CameraConfigUpdateEnum.timestamp_style,
                 CameraConfigUpdateEnum.zones,
             ],
         )

frigate/track/tracked_object.py

@@ -67,8 +67,8 @@ class TrackedObject:
         self.has_snapshot = False
         self.top_score = self.computed_score = 0.0
         self.thumbnail_data: dict[str, Any] | None = None
-        self.last_updated = 0
-        self.last_published = 0
+        self.last_updated: float = 0
+        self.last_published: float = 0
         self.frame = None
         self.active = True
         self.pending_loitering = False

frigate/util/builtin.py

@@ -12,7 +12,7 @@ import shlex
 import struct
 import urllib.parse
 from collections.abc import Mapping
-from multiprocessing.sharedctypes import Synchronized
+from multiprocessing.managers import ValueProxy
 from pathlib import Path
 from typing import Any, Dict, Optional, Tuple, Union
 
@@ -64,7 +64,7 @@ class EventsPerSecond:
 
 
 class InferenceSpeed:
-    def __init__(self, metric: Synchronized) -> None:
+    def __init__(self, metric: ValueProxy[float]) -> None:
         self.__metric = metric
         self.__initialized = False
 

frigate/util/services.py

@@ -488,6 +488,43 @@ def get_rockchip_npu_stats() -> Optional[dict[str, float | str]]:
     return stats
 
 
+def get_axcl_npu_stats() -> Optional[dict[str, str | float]]:
+    """Get NPU stats using axcl."""
+    # Check if axcl-smi exists
+    axcl_smi_path = "/usr/bin/axcl/axcl-smi"
+    if not os.path.exists(axcl_smi_path):
+        return None
+
+    try:
+        # Run axcl-smi command to get NPU stats
+        axcl_command = [axcl_smi_path, "sh", "cat", "/proc/ax_proc/npu/top"]
+        p = sp.run(
+            axcl_command,
+            capture_output=True,
+            text=True,
+        )
+
+        if p.returncode != 0:
+            pass
+        else:
+            utilization = None
+
+            for line in p.stdout.strip().splitlines():
+                line = line.strip()
+                if line.startswith("utilization:"):
+                    match = re.search(r"utilization:(\d+)%", line)
+                    if match:
+                        utilization = float(match.group(1))
+
+            if utilization is not None:
+                stats: dict[str, str | float] = {"npu": utilization, "mem": "-%"}
+                return stats
+    except Exception:
+        pass
+
+    return None
+
+
 def try_get_info(f, h, default="N/A", sensor=None):
     try:
         if h:

web/public/locales/en/config/cameras.json

@@ -529,7 +529,7 @@
     },
     "detections": {
       "label": "Detections config",
-      "description": "Settings for creating detection events (non-alert) and how long to keep them.",
+      "description": "Settings for which tracked objects generate detections (non-alert) and how detections are retained.",
       "enabled": {
         "label": "Enable detections",
         "description": "Enable or disable detection events for this camera."

web/public/locales/en/config/global.json

@@ -293,7 +293,7 @@
       "label": "Detector specific model configuration",
       "description": "Detector-specific model configuration options (path, input size, etc.).",
       "path": {
-        "label": "Custom Object detection model path",
+        "label": "Custom object detector model path",
         "description": "Path to a custom detection model file (or plus://<model_id> for Frigate+ models)."
       },
       "labelmap_path": {
@@ -466,7 +466,7 @@
     "label": "Detection model",
     "description": "Settings to configure a custom object detection model and its input shape.",
     "path": {
-      "label": "Custom Object detection model path",
+      "label": "Custom object detector model path",
       "description": "Path to a custom detection model file (or plus://<model_id> for Frigate+ models)."
     },
     "labelmap_path": {
@@ -752,7 +752,7 @@
   },
   "live": {
     "label": "Live playback",
-    "description": "Settings used by the Web UI to control live stream resolution and quality.",
+    "description": "Settings to control the jsmpeg live stream resolution and quality. This does not affect restreamed cameras that use go2rtc for live view.",
     "streams": {
       "label": "Live stream names",
       "description": "Mapping of configured stream names to restream/go2rtc names used for live playback."
@@ -1044,7 +1044,7 @@
     },
     "detections": {
       "label": "Detections config",
-      "description": "Settings for creating detection events (non-alert) and how long to keep them.",
+      "description": "Settings for which tracked objects generate detections (non-alert) and how detections are retained.",
       "enabled": {
         "label": "Enable detections",
         "description": "Enable or disable detection events for all cameras; can be overridden per-camera."

web/public/locales/en/views/settings.json

@@ -825,6 +825,12 @@
       "area": "Area"
     }
   },
+  "timestampPosition": {
+    "tl": "Top left",
+    "tr": "Top right",
+    "bl": "Bottom left",
+    "br": "Bottom right"
+  },
   "users": {
     "title": "Users",
     "management": {
@@ -1342,7 +1348,22 @@
         "preset-nvidia": "NVIDIA GPU",
         "preset-jetson-h264": "NVIDIA Jetson (H.264)",
         "preset-jetson-h265": "NVIDIA Jetson (H.265)",
-        "preset-rkmpp": "Rockchip RKMPP"
+        "preset-rkmpp": "Rockchip RKMPP",
+        "preset-http-jpeg-generic": "HTTP JPEG (Generic)",
+        "preset-http-mjpeg-generic": "HTTP MJPEG (Generic)",
+        "preset-http-reolink": "HTTP - Reolink Cameras",
+        "preset-rtmp-generic": "RTMP (Generic)",
+        "preset-rtsp-generic": "RTSP (Generic)",
+        "preset-rtsp-restream": "RTSP - Restream from go2rtc",
+        "preset-rtsp-restream-low-latency": "RTSP - Restream from go2rtc (Low Latency)",
+        "preset-rtsp-udp": "RTSP - UDP",
+        "preset-rtsp-blue-iris": "RTSP - Blue Iris",
+        "preset-record-generic": "Record (Generic, no audio)",
+        "preset-record-generic-audio-copy": "Record (Generic + Copy Audio)",
+        "preset-record-generic-audio-aac": "Record (Generic + Audio to AAC)",
+        "preset-record-mjpeg": "Record - MJPEG Cameras",
+        "preset-record-jpeg": "Record - JPEG Cameras",
+        "preset-record-ubiquiti": "Record - Ubiquiti Cameras"
       }
     },
     "cameraInputs": {
@@ -1410,6 +1431,11 @@
       "summary": "{{count}} object types selected",
       "empty": "No object labels available"
     },
+    "reviewLabels": {
+      "summary": "{{count}} labels selected",
+      "empty": "No labels available",
+      "allNonAlertDetections": "All non-alert activity will be included as detections."
+    },
     "filters": {
       "objectFieldLabel": "{{field}} for {{label}}"
     },
@@ -1453,7 +1479,8 @@
     "timestamp_style": {
       "title": "Timestamp Settings"
     },
-    "searchPlaceholder": "Search..."
+    "searchPlaceholder": "Search...",
+    "addCustomLabel": "Add custom label..."
   },
   "globalConfig": {
     "title": "Global Configuration",
@@ -1500,6 +1527,8 @@
     "noOverrides": "No overrides",
     "cameraCount_one": "{{count}} camera",
     "cameraCount_other": "{{count}} cameras",
+    "columnCamera": "Camera",
+    "columnOverrides": "Profile Overrides",
     "baseConfig": "Base Config",
     "addProfile": "Add Profile",
     "newProfile": "New Profile",

web/src/components/config-form/section-configs/audio.ts

@@ -19,6 +19,16 @@ const audio: SectionConfigOverrides = {
     hiddenFields: ["enabled_in_config"],
     advancedFields: ["min_volume", "max_not_heard", "num_threads"],
     uiSchema: {
+      filters: {
+        "ui:options": {
+          expandable: false,
+        },
+      },
+      "filters.*": {
+        "ui:options": {
+          additionalPropertyKeyReadonly: true,
+        },
+      },
       listen: {
         "ui:widget": "audioLabels",
       },

web/src/components/config-form/section-configs/objects.ts

@@ -29,6 +29,11 @@ const objects: SectionConfigOverrides = {
     ],
     advancedFields: ["genai"],
     uiSchema: {
+      filters: {
+        "ui:options": {
+          expandable: false,
+        },
+      },
       "filters.*.min_area": {
         "ui:options": {
           suppressMultiSchema: true,

web/src/components/config-form/section-configs/record.ts

@@ -23,7 +23,7 @@ const record: SectionConfigOverrides = {
     uiSchema: {
       export: {
         hwaccel_args: {
-          "ui:options": { size: "lg" },
+          "ui:options": { suppressMultiSchema: true, size: "lg" },
         },
       },
     },

web/src/components/config-form/section-configs/review.ts

@@ -3,14 +3,16 @@ import type { SectionConfigOverrides } from "./types";
 const review: SectionConfigOverrides = {
   base: {
     sectionDocs: "/configuration/review",
+    fieldDocs: {
+      "alerts.labels": "/configuration/review/#alerts-and-detections",
+      "detections.labels": "/configuration/review/#alerts-and-detections",
+    },
     restartRequired: [],
     fieldOrder: ["alerts", "detections", "genai"],
     fieldGroups: {},
     hiddenFields: [
       "enabled_in_config",
-      "alerts.labels",
       "alerts.enabled_in_config",
-      "detections.labels",
       "detections.enabled_in_config",
       "genai.enabled_in_config",
     ],
@@ -18,20 +20,35 @@ const review: SectionConfigOverrides = {
     uiSchema: {
       alerts: {
         "ui:before": { render: "CameraReviewStatusToggles" },
+        labels: {
+          "ui:widget": "reviewLabels",
+          "ui:options": {
+            suppressMultiSchema: true,
+          },
+        },
         required_zones: {
           "ui:widget": "hidden",
         },
       },
       detections: {
+        labels: {
+          "ui:widget": "reviewLabels",
+          "ui:options": {
+            suppressMultiSchema: true,
+            emptySelectionHintKey:
+              "configForm.reviewLabels.allNonAlertDetections",
+          },
+        },
         required_zones: {
           "ui:widget": "hidden",
         },
       },
       genai: {
         additional_concerns: {
-          "ui:widget": "textarea",
+          "ui:widget": "ArrayAsTextWidget",
           "ui:options": {
             size: "full",
+            multiline: true,
           },
         },
         activity_context_prompt: {

web/src/components/config-form/section-configs/timestamp_style.ts

@@ -4,12 +4,13 @@ const timestampStyle: SectionConfigOverrides = {
   base: {
     sectionDocs: "/configuration/reference",
     restartRequired: [],
-    fieldOrder: ["position", "format", "color", "thickness"],
+    fieldOrder: ["position", "format", "thickness", "color"],
     hiddenFields: ["effect", "enabled_in_config"],
     advancedFields: [],
     uiSchema: {
       position: {
         "ui:size": "xs",
+        "ui:options": { enumI18nPrefix: "timestampPosition" },
       },
       format: {
         "ui:size": "xs",
@@ -17,7 +18,7 @@ const timestampStyle: SectionConfigOverrides = {
     },
   },
   global: {
-    restartRequired: ["position", "format", "color", "thickness", "effect"],
+    restartRequired: [],
   },
   camera: {
     restartRequired: [],

web/src/components/config-form/sections/BaseSection.tsx

@@ -152,6 +152,10 @@ export interface BaseSectionProps {
   profileBorderColor?: string;
   /** Callback to delete the current profile's overrides for this section */
   onDeleteProfileSection?: () => void;
+  /** Whether a SaveAll operation is in progress (disables individual Save) */
+  isSavingAll?: boolean;
+  /** Callback when this section's saving state changes */
+  onSavingChange?: (isSaving: boolean) => void;
 }
 
 export interface CreateSectionOptions {
@@ -186,6 +190,8 @@ export function ConfigSection({
   profileFriendlyName,
   profileBorderColor,
   onDeleteProfileSection,
+  isSavingAll = false,
+  onSavingChange,
 }: ConfigSectionProps) {
   // For replay level, treat as camera-level config access
   const effectiveLevel = level === "replay" ? "camera" : level;
@@ -246,6 +252,7 @@ export function ConfigSection({
     [onPendingDataChange, effectiveSectionPath, cameraName],
   );
   const [isSaving, setIsSaving] = useState(false);
+  const [isResettingToDefault, setIsResettingToDefault] = useState(false);
   const [hasValidationErrors, setHasValidationErrors] = useState(false);
   const [extraHasChanges, setExtraHasChanges] = useState(false);
   const [formKey, setFormKey] = useState(0);
@@ -577,6 +584,7 @@ export function ConfigSection({
     if (!pendingData) return;
 
     setIsSaving(true);
+    onSavingChange?.(true);
     try {
       const basePath =
         effectiveLevel === "camera" && cameraName
@@ -699,6 +707,7 @@ export function ConfigSection({
       }
     } finally {
       setIsSaving(false);
+      onSavingChange?.(false);
     }
   }, [
     sectionPath,
@@ -718,12 +727,14 @@ export function ConfigSection({
     setPendingData,
     requiresRestartForOverrides,
     skipSave,
+    onSavingChange,
   ]);
 
   // Handle reset to global/defaults - removes camera-level override or resets global to defaults
   const handleResetToGlobal = useCallback(async () => {
     if (effectiveLevel === "camera" && !cameraName) return;
 
+    setIsResettingToDefault(true);
     try {
       const basePath =
         effectiveLevel === "camera" && cameraName
@@ -758,6 +769,8 @@ export function ConfigSection({
           defaultValue: "Failed to reset settings",
         }),
       );
+    } finally {
+      setIsResettingToDefault(false);
     }
   }, [
     effectiveSectionPath,
@@ -945,9 +958,12 @@ export function ConfigSection({
                 <Button
                   onClick={() => setIsResetDialogOpen(true)}
                   variant="outline"
-                  disabled={isSaving || disabled}
+                  disabled={isSaving || isResettingToDefault || disabled}
                   className="flex flex-1 gap-2"
                 >
+                  {isResettingToDefault && (
+                    <ActivityIndicator className="h-4 w-4" />
+                  )}
                   {effectiveLevel === "global"
                     ? t("button.resetToDefault", {
                         ns: "common",
@@ -980,7 +996,7 @@ export function ConfigSection({
               <Button
                 onClick={handleReset}
                 variant="outline"
-                disabled={isSaving || disabled}
+                disabled={isSaving || isSavingAll || disabled}
                 className="flex min-w-36 flex-1 gap-2"
               >
                 {t("button.undo", { ns: "common", defaultValue: "Undo" })}
@@ -990,7 +1006,11 @@ export function ConfigSection({
               onClick={handleSave}
               variant="select"
               disabled={
-                !hasChanges || hasValidationErrors || isSaving || disabled
+                !hasChanges ||
+                hasValidationErrors ||
+                isSaving ||
+                isSavingAll ||
+                disabled
               }
               className="flex min-w-36 flex-1 gap-2"
             >

web/src/components/config-form/theme/frigateTheme.ts

@@ -20,6 +20,7 @@ import { TextareaWidget } from "./widgets/TextareaWidget";
 import { SwitchesWidget } from "./widgets/SwitchesWidget";
 import { ObjectLabelSwitchesWidget } from "./widgets/ObjectLabelSwitchesWidget";
 import { AudioLabelSwitchesWidget } from "./widgets/AudioLabelSwitchesWidget";
+import { ReviewLabelSwitchesWidget } from "./widgets/ReviewLabelSwitchesWidget";
 import { ZoneSwitchesWidget } from "./widgets/ZoneSwitchesWidget";
 import { ArrayAsTextWidget } from "./widgets/ArrayAsTextWidget";
 import { FfmpegArgsWidget } from "./widgets/FfmpegArgsWidget";
@@ -76,6 +77,7 @@ export const frigateTheme: FrigateTheme = {
     switches: SwitchesWidget,
     objectLabels: ObjectLabelSwitchesWidget,
     audioLabels: AudioLabelSwitchesWidget,
+    reviewLabels: ReviewLabelSwitchesWidget,
     zoneNames: ZoneSwitchesWidget,
     timezoneSelect: TimezoneSelectWidget,
     optionalField: OptionalFieldWidget,

web/src/components/config-form/theme/widgets/ArrayAsTextWidget.tsx

@@ -1,33 +1,101 @@
-// Widget that displays an array as a concatenated text string
+// Widget that displays an array as editable text.
+// Single-line mode (default): space-separated in an Input.
+// Multiline mode (options.multiline): one item per line in a Textarea.
 import type { WidgetProps } from "@rjsf/utils";
 import { Input } from "@/components/ui/input";
-import { useCallback } from "react";
+import { Textarea } from "@/components/ui/textarea";
+import { cn } from "@/lib/utils";
+import { getSizedFieldClassName } from "../utils";
+import { useCallback, useEffect, useState } from "react";
+
+function arrayToText(value: unknown, multiline: boolean): string {
+  const sep = multiline ? "\n" : " ";
+  if (Array.isArray(value) && value.length > 0) {
+    return value.join(sep);
+  }
+  if (typeof value === "string") {
+    return value;
+  }
+  return "";
+}
+
+function textToArray(text: string, multiline: boolean): string[] {
+  if (text.trim() === "") {
+    return [];
+  }
+  return multiline
+    ? text.split("\n").filter((line) => line.trim() !== "")
+    : text.trim().split(/\s+/);
+}
 
 export function ArrayAsTextWidget(props: WidgetProps) {
-  const { value, onChange, disabled, readonly, placeholder } = props;
+  const {
+    id,
+    value,
+    disabled,
+    readonly,
+    onChange,
+    onBlur,
+    onFocus,
+    placeholder,
+    schema,
+    options,
+  } = props;
 
-  // Convert array or string to text
-  let textValue = "";
-  if (typeof value === "string" && value.length > 0) {
-    textValue = value;
-  } else if (Array.isArray(value) && value.length > 0) {
-    textValue = value.join(" ");
-  }
+  const multiline = !!(options.multiline as boolean);
 
-  const handleChange = useCallback(
-    (event: React.ChangeEvent<HTMLInputElement>) => {
-      const newText = event.target.value;
-      // Convert space-separated string back to array
-      const newArray = newText.trim() ? newText.trim().split(/\s+/) : [];
-      onChange(newArray);
+  // Local state keeps raw text so newlines aren't stripped mid-typing
+  const [text, setText] = useState(() => arrayToText(value, multiline));
+
+  useEffect(() => {
+    setText(arrayToText(value, multiline));
+  }, [value, multiline]);
+
+  const fieldClassName = multiline
+    ? getSizedFieldClassName(options, "md")
+    : undefined;
+
+  const handleInputChange = useCallback(
+    (e: React.ChangeEvent<HTMLInputElement | HTMLTextAreaElement>) => {
+      const raw = e.target.value;
+      setText(raw);
+      onChange(textToArray(raw, multiline));
     },
-    [onChange],
+    [onChange, multiline],
   );
 
+  const handleBlur = useCallback(
+    (e: React.FocusEvent<HTMLInputElement | HTMLTextAreaElement>) => {
+      // Clean up: strip empty entries and sync
+      const cleaned = textToArray(e.target.value, multiline);
+      onChange(cleaned);
+      setText(arrayToText(cleaned, multiline));
+      onBlur?.(id, e.target.value);
+    },
+    [id, onChange, onBlur, multiline],
+  );
+
+  if (multiline) {
+    return (
+      <Textarea
+        id={id}
+        className={cn("text-md", fieldClassName)}
+        value={text}
+        disabled={disabled || readonly}
+        rows={(options.rows as number) || 3}
+        onChange={handleInputChange}
+        onBlur={handleBlur}
+        onFocus={(e) => onFocus?.(id, e.target.value)}
+        aria-label={schema.title}
+      />
+    );
+  }
+
   return (
     <Input
-      value={textValue}
-      onChange={handleChange}
+      value={text}
+      onChange={handleInputChange}
+      onBlur={handleBlur}
       disabled={disabled}
       readOnly={readonly}
       placeholder={placeholder}

web/src/components/config-form/theme/widgets/ReviewLabelSwitchesWidget.tsx

@@ -0,0 +1,84 @@
+// Review Label Switches Widget - For selecting review alert/detection labels via switches.
+// Combines object labels (from objects.track) and audio labels (from audio.listen)
+// since review labels can include both types.
+import type { WidgetProps } from "@rjsf/utils";
+import { SwitchesWidget } from "./SwitchesWidget";
+import type { FormContext } from "./SwitchesWidget";
+import { getTranslatedLabel } from "@/utils/i18n";
+import type { FrigateConfig } from "@/types/frigateConfig";
+import type { JsonObject } from "@/types/configForm";
+
+function getReviewLabels(context: FormContext): string[] {
+  const labels = new Set<string>();
+  const fullConfig = context.fullConfig as FrigateConfig | undefined;
+  const fullCameraConfig = context.fullCameraConfig;
+
+  // Object labels from tracked objects (camera-level, falling back to global)
+  const trackLabels =
+    fullCameraConfig?.objects?.track ?? fullConfig?.objects?.track;
+  if (Array.isArray(trackLabels)) {
+    trackLabels.forEach((label: string) => labels.add(label));
+  }
+
+  // Audio labels from listen config, only if audio detection is enabled
+  const audioEnabled =
+    fullCameraConfig?.audio?.enabled_in_config ??
+    fullConfig?.audio?.enabled_in_config;
+  if (audioEnabled) {
+    const audioLabels =
+      fullCameraConfig?.audio?.listen ?? fullConfig?.audio?.listen;
+    if (Array.isArray(audioLabels)) {
+      audioLabels.forEach((label: string) => labels.add(label));
+    }
+  }
+
+  // Include any labels already in the review form data (alerts + detections)
+  // so that previously saved labels remain visible even if tracking config changed
+  if (context.formData && typeof context.formData === "object") {
+    const formData = context.formData as JsonObject;
+    for (const section of ["alerts", "detections"] as const) {
+      const sectionData = formData[section];
+      if (sectionData && typeof sectionData === "object") {
+        const sectionLabels = (sectionData as JsonObject).labels;
+        if (Array.isArray(sectionLabels)) {
+          sectionLabels.forEach((label) => {
+            if (typeof label === "string") {
+              labels.add(label);
+            }
+          });
+        }
+      }
+    }
+  }
+
+  return [...labels].sort();
+}
+
+function getReviewLabelDisplayName(
+  label: string,
+  context?: FormContext,
+): string {
+  const fullCameraConfig = context?.fullCameraConfig;
+  const fullConfig = context?.fullConfig as FrigateConfig | undefined;
+  const audioLabels =
+    fullCameraConfig?.audio?.listen ?? fullConfig?.audio?.listen;
+  const isAudio = Array.isArray(audioLabels) && audioLabels.includes(label);
+  return getTranslatedLabel(label, isAudio ? "audio" : "object");
+}
+
+export function ReviewLabelSwitchesWidget(props: WidgetProps) {
+  return (
+    <SwitchesWidget
+      {...props}
+      options={{
+        ...props.options,
+        getEntities: getReviewLabels,
+        getDisplayLabel: getReviewLabelDisplayName,
+        i18nKey: "reviewLabels",
+        allowCustomEntries: true,
+        listClassName:
+          "relative max-h-none overflow-visible md:max-h-64 md:overflow-y-auto md:overscroll-contain md:scrollbar-container",
+      }}
+    />
+  );
+}

web/src/components/config-form/theme/widgets/SelectWidget.tsx

@@ -1,5 +1,6 @@
 // Select Widget - maps to shadcn/ui Select
 import type { WidgetProps } from "@rjsf/utils";
+import { useTranslation } from "react-i18next";
 import {
   Select,
   SelectContent,
@@ -21,9 +22,18 @@ export function SelectWidget(props: WidgetProps) {
     schema,
   } = props;
 
+  const { t } = useTranslation(["views/settings"]);
   const { enumOptions = [] } = options;
+  const enumI18nPrefix = options["enumI18nPrefix"] as string | undefined;
   const fieldClassName = getSizedFieldClassName(options, "sm");
 
+  const getLabel = (option: { value: unknown; label: string }) => {
+    if (enumI18nPrefix) {
+      return t(`${enumI18nPrefix}.${option.value}`);
+    }
+    return option.label;
+  };
+
   return (
     <Select
       value={value?.toString() ?? ""}
@@ -42,7 +52,7 @@ export function SelectWidget(props: WidgetProps) {
       <SelectContent>
         {enumOptions.map((option: { value: unknown; label: string }) => (
           <SelectItem key={String(option.value)} value={String(option.value)}>
-            {option.label}
+            {getLabel(option)}
           </SelectItem>
         ))}
       </SelectContent>

web/src/components/config-form/theme/widgets/SwitchesWidget.tsx

@@ -1,6 +1,6 @@
 // Generic Switches Widget - Reusable component for selecting from any list of entities
 import { WidgetProps } from "@rjsf/utils";
-import { useMemo, useState } from "react";
+import { useCallback, useMemo, useState } from "react";
 import { Switch } from "@/components/ui/switch";
 import { Button } from "@/components/ui/button";
 import { Input } from "@/components/ui/input";
@@ -43,6 +43,10 @@ export type SwitchesWidgetOptions = {
   listClassName?: string;
   /** Enable search input to filter the list */
   enableSearch?: boolean;
+  /** Allow users to add custom entries not in the predefined list */
+  allowCustomEntries?: boolean;
+  /** i18n key for a hint shown when no entities are selected */
+  emptySelectionHintKey?: string;
 };
 
 function normalizeValue(value: unknown): string[] {
@@ -122,20 +126,51 @@ export function SwitchesWidget(props: WidgetProps) {
     [props.options],
   );
 
+  const allowCustomEntries = useMemo(
+    () => props.options?.allowCustomEntries as boolean | undefined,
+    [props.options],
+  );
+
+  const emptySelectionHintKey = useMemo(
+    () => props.options?.emptySelectionHintKey as string | undefined,
+    [props.options],
+  );
+
   const selectedEntities = useMemo(() => normalizeValue(value), [value]);
   const [isOpen, setIsOpen] = useState(selectedEntities.length > 0);
   const [searchTerm, setSearchTerm] = useState("");
+  const [customEntries, setCustomEntries] = useState<string[]>([]);
+  const [customInput, setCustomInput] = useState("");
+
+  const allEntities = useMemo(() => {
+    if (customEntries.length === 0) {
+      return availableEntities;
+    }
+    const merged = new Set([...availableEntities, ...customEntries]);
+    return [...merged].sort();
+  }, [availableEntities, customEntries]);
 
   const filteredEntities = useMemo(() => {
     if (!enableSearch || !searchTerm.trim()) {
-      return availableEntities;
+      return allEntities;
     }
     const term = searchTerm.toLowerCase();
-    return availableEntities.filter((entity) => {
+    return allEntities.filter((entity) => {
       const displayLabel = getDisplayLabel(entity, context);
       return displayLabel.toLowerCase().includes(term);
     });
-  }, [availableEntities, searchTerm, enableSearch, getDisplayLabel, context]);
+  }, [allEntities, searchTerm, enableSearch, getDisplayLabel, context]);
+
+  const addCustomEntry = useCallback(() => {
+    const trimmed = customInput.trim().toLowerCase();
+    if (!trimmed || allEntities.includes(trimmed)) {
+      setCustomInput("");
+      return;
+    }
+    setCustomEntries((prev) => [...prev, trimmed]);
+    onChange([...selectedEntities, trimmed]);
+    setCustomInput("");
+  }, [customInput, allEntities, selectedEntities, onChange]);
 
   const toggleEntity = (entity: string, enabled: boolean) => {
     if (enabled) {
@@ -163,7 +198,7 @@ export function SwitchesWidget(props: WidgetProps) {
 
   return (
     <Collapsible open={isOpen} onOpenChange={setIsOpen}>
-      <div className="space-y-3">
+      <div className="space-y-2">
         <CollapsibleTrigger asChild>
           <Button
             type="button"
@@ -180,8 +215,14 @@ export function SwitchesWidget(props: WidgetProps) {
           </Button>
         </CollapsibleTrigger>
 
-        <CollapsibleContent className="rounded-lg bg-secondary p-2 pr-0 md:max-w-md">
-          {availableEntities.length === 0 ? (
+        {emptySelectionHintKey && selectedEntities.length === 0 && t && (
+          <div className="mt-0 pb-2 text-sm text-success">
+            {t(emptySelectionHintKey, { ns: namespace })}
+          </div>
+        )}
+
+        <CollapsibleContent className="rounded-lg border border-input bg-secondary pb-1 pr-0 pt-2 md:max-w-md">
+          {allEntities.length === 0 && !allowCustomEntries ? (
             <div className="text-sm text-muted-foreground">{emptyMessage}</div>
           ) : (
             <>
@@ -223,6 +264,26 @@ export function SwitchesWidget(props: WidgetProps) {
                   );
                 })}
               </div>
+              {allowCustomEntries && !disabled && !readonly && (
+                <div className="mx-2 mt-2 pb-1">
+                  <Input
+                    type="text"
+                    placeholder={t?.("configForm.addCustomLabel", {
+                      ns: "views/settings",
+                      defaultValue: "Add custom label...",
+                    })}
+                    value={customInput}
+                    onChange={(e) => setCustomInput(e.target.value)}
+                    onKeyDown={(e) => {
+                      if (e.key === "Enter") {
+                        e.preventDefault();
+                        addCustomEntry();
+                      }
+                    }}
+                    onBlur={addCustomEntry}
+                  />
+                </div>
+              )}
             </>
           )}
         </CollapsibleContent>

web/src/components/overlay/detail/AnnotationOffsetSlider.tsx

@@ -9,15 +9,16 @@ import { toast } from "sonner";
 import { Trans, useTranslation } from "react-i18next";
 import { LuExternalLink, LuInfo, LuMinus, LuPlus } from "react-icons/lu";
 import { cn } from "@/lib/utils";
+import {
+  ANNOTATION_OFFSET_MAX,
+  ANNOTATION_OFFSET_MIN,
+  ANNOTATION_OFFSET_STEP,
+} from "@/lib/const";
 import { isMobile } from "react-device-detect";
 import { useIsAdmin } from "@/hooks/use-is-admin";
 import { useDocDomain } from "@/hooks/use-doc-domain";
 import { Link } from "react-router-dom";
 
-const OFFSET_MIN = -2500;
-const OFFSET_MAX = 2500;
-const OFFSET_STEP = 50;
-
 type Props = {
   className?: string;
 };
@@ -43,7 +44,10 @@ export default function AnnotationOffsetSlider({ className }: Props) {
     (delta: number) => {
       setAnnotationOffset((prev) => {
         const next = prev + delta;
-        return Math.max(OFFSET_MIN, Math.min(OFFSET_MAX, next));
+        return Math.max(
+          ANNOTATION_OFFSET_MIN,
+          Math.min(ANNOTATION_OFFSET_MAX, next),
+        );
       });
     },
     [setAnnotationOffset],
@@ -114,17 +118,17 @@ export default function AnnotationOffsetSlider({ className }: Props) {
           size="icon"
           className="size-8 shrink-0"
           aria-label="-50ms"
-          onClick={() => stepOffset(-OFFSET_STEP)}
-          disabled={annotationOffset <= OFFSET_MIN}
+          onClick={() => stepOffset(-ANNOTATION_OFFSET_STEP)}
+          disabled={annotationOffset <= ANNOTATION_OFFSET_MIN}
         >
           <LuMinus className="size-4" />
         </Button>
         <div className="w-full flex-1 landscape:flex">
           <Slider
             value={[annotationOffset]}
-            min={OFFSET_MIN}
-            max={OFFSET_MAX}
-            step={OFFSET_STEP}
+            min={ANNOTATION_OFFSET_MIN}
+            max={ANNOTATION_OFFSET_MAX}
+            step={ANNOTATION_OFFSET_STEP}
             onValueChange={handleChange}
           />
         </div>
@@ -134,8 +138,8 @@ export default function AnnotationOffsetSlider({ className }: Props) {
           size="icon"
           className="size-8 shrink-0"
           aria-label="+50ms"
-          onClick={() => stepOffset(OFFSET_STEP)}
-          disabled={annotationOffset >= OFFSET_MAX}
+          onClick={() => stepOffset(ANNOTATION_OFFSET_STEP)}
+          disabled={annotationOffset >= ANNOTATION_OFFSET_MAX}
         >
           <LuPlus className="size-4" />
         </Button>

web/src/components/overlay/detail/AnnotationSettingsPane.tsx

@@ -13,10 +13,11 @@ import { Slider } from "@/components/ui/slider";
 import { Trans, useTranslation } from "react-i18next";
 import { useDocDomain } from "@/hooks/use-doc-domain";
 import { useIsAdmin } from "@/hooks/use-is-admin";
-
-const OFFSET_MIN = -2500;
-const OFFSET_MAX = 2500;
-const OFFSET_STEP = 50;
+import {
+  ANNOTATION_OFFSET_MAX,
+  ANNOTATION_OFFSET_MIN,
+  ANNOTATION_OFFSET_STEP,
+} from "@/lib/const";
 
 type AnnotationSettingsPaneProps = {
   event: Event;
@@ -49,7 +50,10 @@ export function AnnotationSettingsPane({
     (delta: number) => {
       setAnnotationOffset((prev) => {
         const next = prev + delta;
-        return Math.max(OFFSET_MIN, Math.min(OFFSET_MAX, next));
+        return Math.max(
+          ANNOTATION_OFFSET_MIN,
+          Math.min(ANNOTATION_OFFSET_MAX, next),
+        );
       });
     },
     [setAnnotationOffset],
@@ -128,16 +132,16 @@ export function AnnotationSettingsPane({
             size="icon"
             className="size-8 shrink-0"
             aria-label="-50ms"
-            onClick={() => stepOffset(-OFFSET_STEP)}
-            disabled={annotationOffset <= OFFSET_MIN}
+            onClick={() => stepOffset(-ANNOTATION_OFFSET_STEP)}
+            disabled={annotationOffset <= ANNOTATION_OFFSET_MIN}
           >
             <LuMinus className="size-4" />
           </Button>
           <Slider
             value={[annotationOffset]}
-            min={OFFSET_MIN}
-            max={OFFSET_MAX}
-            step={OFFSET_STEP}
+            min={ANNOTATION_OFFSET_MIN}
+            max={ANNOTATION_OFFSET_MAX}
+            step={ANNOTATION_OFFSET_STEP}
             onValueChange={handleSliderChange}
             className="flex-1"
           />
@@ -147,8 +151,8 @@ export function AnnotationSettingsPane({
             size="icon"
             className="size-8 shrink-0"
             aria-label="+50ms"
-            onClick={() => stepOffset(OFFSET_STEP)}
-            disabled={annotationOffset >= OFFSET_MAX}
+            onClick={() => stepOffset(ANNOTATION_OFFSET_STEP)}
+            disabled={annotationOffset >= ANNOTATION_OFFSET_MAX}
           >
             <LuPlus className="size-4" />
           </Button>

web/src/lib/const.ts

@@ -1,6 +1,10 @@
 /** ONNX embedding models that require local model downloads. GenAI providers are not in this list. */
 export const JINA_EMBEDDING_MODELS = ["jinav1", "jinav2"] as const;
 
+export const ANNOTATION_OFFSET_MIN = -10000;
+export const ANNOTATION_OFFSET_MAX = 5000;
+export const ANNOTATION_OFFSET_STEP = 50;
+
 export const supportedLanguageKeys = [
   "en",
   "es",

web/src/pages/Settings.tsx

@@ -724,6 +724,7 @@ export default function Settings() {
 
   // Save All state
   const [isSavingAll, setIsSavingAll] = useState(false);
+  const [isAnySectionSaving, setIsAnySectionSaving] = useState(false);
   const [restartDialogOpen, setRestartDialogOpen] = useState(false);
   const { send: sendRestart } = useRestart();
   const { data: fullSchema } = useSWR<RJSFSchema>("config/schema.json");
@@ -1299,6 +1300,10 @@ export default function Settings() {
     [],
   );
 
+  const handleSectionSavingChange = useCallback((saving: boolean) => {
+    setIsAnySectionSaving(saving);
+  }, []);
+
   // The active profile being edited for the selected camera
   const activeEditingProfile = selectedCamera
     ? (editingProfile[selectedCamera] ?? null)
@@ -1508,7 +1513,7 @@ export default function Settings() {
                     onClick={handleUndoAll}
                     variant="outline"
                     size="sm"
-                    disabled={isSavingAll}
+                    disabled={isSavingAll || isAnySectionSaving}
                     className="flex w-full items-center justify-center gap-2"
                   >
                     {t("button.undoAll", {
@@ -1520,7 +1525,11 @@ export default function Settings() {
                     onClick={handleSaveAll}
                     variant="select"
                     size="sm"
-                    disabled={isSavingAll || hasPendingValidationErrors}
+                    disabled={
+                      isSavingAll ||
+                      isAnySectionSaving ||
+                      hasPendingValidationErrors
+                    }
                     className="flex w-full items-center justify-center gap-2"
                   >
                     {isSavingAll ? (
@@ -1606,6 +1615,8 @@ export default function Settings() {
                     }
                     profilesUIEnabled={profilesUIEnabled}
                     setProfilesUIEnabled={setProfilesUIEnabled}
+                    isSavingAll={isSavingAll}
+                    onSectionSavingChange={handleSectionSavingChange}
                   />
                 );
               })()}
@@ -1674,7 +1685,7 @@ export default function Settings() {
                 onClick={handleUndoAll}
                 variant="outline"
                 size="sm"
-                disabled={isSavingAll}
+                disabled={isSavingAll || isAnySectionSaving}
                 className="flex items-center justify-center gap-2"
               >
                 {t("button.undoAll", {
@@ -1686,7 +1697,11 @@ export default function Settings() {
                 variant="select"
                 size="sm"
                 onClick={handleSaveAll}
-                disabled={isSavingAll || hasPendingValidationErrors}
+                disabled={
+                  isSavingAll ||
+                  isAnySectionSaving ||
+                  hasPendingValidationErrors
+                }
                 className="flex items-center justify-center gap-2"
               >
                 {isSavingAll ? (
@@ -1843,6 +1858,8 @@ export default function Settings() {
                   onDeleteProfileSection={handleDeleteProfileForCurrentSection}
                   profilesUIEnabled={profilesUIEnabled}
                   setProfilesUIEnabled={setProfilesUIEnabled}
+                  isSavingAll={isSavingAll}
+                  onSectionSavingChange={handleSectionSavingChange}
                 />
               );
             })()}

web/src/views/live/LiveCameraView.tsx

@@ -707,14 +707,23 @@ export default function LiveCameraView({
               }}
             >
               <div
-                className={`relative flex flex-col items-center justify-center ${growClassName}`}
+                className={cn(
+                  "flex flex-col items-center justify-center",
+                  growClassName,
+                )}
                 ref={clickOverlayRef}
                 style={{
                   aspectRatio: constrainedAspectRatio,
                 }}
               >
                 {clickOverlay && overlaySize.width > 0 && (
-                  <div className="absolute inset-0 z-40 cursor-crosshair">
+                  <div
+                    className="absolute z-40 cursor-crosshair"
+                    style={{
+                      width: overlaySize.width,
+                      height: overlaySize.height,
+                    }}
+                  >
                     <Stage
                       width={overlaySize.width}
                       height={overlaySize.height}

web/src/views/settings/ProfilesView.tsx

@@ -542,6 +542,20 @@ export default function ProfilesView({
                   <CollapsibleContent>
                     {cameras.length > 0 ? (
                       <div className="mx-4 mb-3 ml-11 border-l border-border/50 pl-4">
+                        <div className="flex items-baseline gap-3 border-b border-border/30 pb-1.5">
+                          <span className="min-w-[120px] shrink-0 text-xs font-semibold uppercase text-muted-foreground">
+                            {t("profiles.columnCamera", {
+                              ns: "views/settings",
+                              defaultValue: "Camera",
+                            })}
+                          </span>
+                          <span className="text-xs font-semibold uppercase text-muted-foreground">
+                            {t("profiles.columnOverrides", {
+                              ns: "views/settings",
+                              defaultValue: "Profile Overrides",
+                            })}
+                          </span>
+                        </div>
                         {cameras.map((camera) => {
                           const sections = cameraData[camera];
                           return (

web/src/views/settings/SingleSectionPage.tsx

@@ -39,6 +39,10 @@ export type SettingsPageProps = {
   onDeleteProfileSection?: (profileName: string) => void;
   profilesUIEnabled?: boolean;
   setProfilesUIEnabled?: React.Dispatch<React.SetStateAction<boolean>>;
+  /** Whether a SaveAll operation is in progress */
+  isSavingAll?: boolean;
+  /** Callback when a section's saving state changes */
+  onSectionSavingChange?: (isSaving: boolean) => void;
 };
 
 export type SectionStatus = {
@@ -73,6 +77,8 @@ export function SingleSectionPage({
   onPendingDataChange,
   profileState,
   onDeleteProfileSection,
+  isSavingAll,
+  onSectionSavingChange,
 }: SingleSectionPageProps) {
   const sectionNamespace =
     level === "camera" ? "config/cameras" : "config/global";
@@ -95,9 +101,10 @@ export function SingleSectionPage({
     ? getLocaleDocUrl(resolvedSectionConfig.sectionDocs)
     : undefined;
 
-  const currentEditingProfile = selectedCamera
-    ? (profileState?.editingProfile[selectedCamera] ?? null)
-    : null;
+  const currentEditingProfile =
+    level === "camera" && selectedCamera
+      ? (profileState?.editingProfile[selectedCamera] ?? null)
+      : null;
 
   const profileColor = useMemo(
     () =>
@@ -273,6 +280,8 @@ export function SingleSectionPage({
         onDeleteProfileSection={
           currentEditingProfile ? handleDeleteProfileSection : undefined
         }
+        isSavingAll={isSavingAll}
+        onSavingChange={onSectionSavingChange}
       />
     </div>
   );