* Ignore numpy get limits warning
* Add function wrapper to redirect stdout and stderr to logpipe
* Save stderr too
* Add more to catch
* run logpipe
* Use other logging redirect class
* Use other logging redirect class
* add decorator for redirecting c/c++ level output to logger
* fix typing
---------
Co-authored-by: Josh Hawkins <32435876+hawkeye217@users.noreply.github.com>
* Move log level initialization to log
* Use logger config
* Formatting
* Fix config order
* Set process names
---------
Co-authored-by: Nicolas Mowen <nickmowen213@gmail.com>
* Set runtime
* Use count correctly
* Don't assume camera sizes
* Use separate zmq proxy for object detection
* Correct order
* Use forkserver
* Only store PID instead of entire process reference
* Cleanup
* Catch correct errors
* Fix typing
* Remove before_run from process util
The before_run never actually ran because:
You're right to suspect an issue with before_run not being called and a potential deadlock. The way you've implemented the run_wrapper using __getattribute__ for the run method of BaseProcess is a common pitfall in Python's multiprocessing, especially when combined with how multiprocessing.Process works internally.
Here's a breakdown of why before_run isn't being called and why you might be experiencing a deadlock:
The Problem: __getattribute__ and Process Serialization
When you create a multiprocessing.Process object and call start(), the multiprocessing module needs to serialize the process object (or at least enough of it to re-create the process in the new interpreter). It then pickles this serialized object and sends it to the newly spawned process.
The issue with your __getattribute__ implementation for run is that:
run is retrieved during serialization: When multiprocessing tries to pickle your Process object to send to the new process, it will likely access the run attribute. This triggers your __getattribute__ wrapper, which then tries to bind run_wrapper to self.
run_wrapper is bound to the parent process's self: The run_wrapper closure, when created in the parent process, captures the self (the Process instance) from the parent's memory space.
Deserialization creates a new object: In the child process, a new Process object is created by deserializing the pickled data. However, the run_wrapper method that was pickled still holds a reference to the self from the parent process. This is a subtle but critical distinction.
The child's run is not your wrapped run: When the child process starts, it internally calls its own run method. Because of the serialization and deserialization process, the run method that's ultimately executed in the child process is the original multiprocessing.Process.run or the Process.run if you had directly overridden it. Your __getattribute__ magic, which wraps run, isn't correctly applied to the Process object within the child's context.
* Cleanup
* Logging bugfix (#18465)
* use mp Manager to handle logging queues
A Python bug (https://github.com/python/cpython/issues/91555) was preventing logs from the embeddings maintainer process from printing. The bug is fixed in Python 3.14, but a viable workaround is to use the multiprocessing Manager, which better manages mp queues and causes the logging to work correctly.
* consolidate
* fix typing
* Fix typing
* Use global log queue
* Move to using process for logging
* Convert camera tracking to process
* Add more processes
* Finalize process
* Cleanup
* Cleanup typing
* Formatting
* Remove daemon
---------
Co-authored-by: Josh Hawkins <32435876+hawkeye217@users.noreply.github.com>
* Add base class for global config updates
* Add or remove camera states
* Move camera process management to separate thread
* Move camera management fully to separate class
* Cleanup
* Stop camera processes when stop command is sent
* Start processes dynamically when needed
* Adjust
* Leave extra room in tracked object queue for two cameras
* Dynamically set extra config pieces
* Add some TODOs
* Fix type check
* Simplify config updates
* Improve typing
* Correctly handle indexed entries
* Cleanup
* Create out SHM
* Use ZMQ for signaling object detectoin is completed
* Get camera correctly created
* Cleanup for updating the cameras config
* Cleanup
* Don't enable audio if no cameras have audio transcription
* Use exact string so similar camera names don't interfere
* Add ability to update config via json body to config/set endpoint
Additionally, update the config in a single rather than multiple calls for each updated key
* fix autotracking calibration to support new config updater function
---------
Co-authored-by: Josh Hawkins <32435876+hawkeye217@users.noreply.github.com>
* install new packages for transcription support
* add config options
* audio maintainer modifications to support transcription
* pass main config to audio process
* embeddings support
* api and transcription post processor
* embeddings maintainer support for post processor
* live audio transcription with sherpa and faster-whisper
* update dispatcher with live transcription topic
* frontend websocket
* frontend live transcription
* frontend changes for speech events
* i18n changes
* docs
* mqtt docs
* fix linter
* use float16 and small model on gpu for real-time
* fix return value and use requestor to embed description instead of passing embeddings
* run real-time transcription in its own thread
* tweaks
* publish live transcriptions on their own topic instead of tracked_object_update
* config validator and docs
* clarify docs
* Fix the `Any` typing hint treewide
There has been confusion between the Any type[1] and the any function[2]
in typing hints.
[1] https://docs.python.org/3/library/typing.html#typing.Any
[2] https://docs.python.org/3/library/functions.html#any
* Fix typing for various frame_shape members
Frame shapes are most likely defined by height and width, so a single int
cannot express that.
* Wrap gpu stats functions in Optional[]
These can return `None`, so they need to be `Type | None`, which is what
`Optional` expresses very nicely.
* Fix return type in get_latest_segment_datetime
Returns a datetime object, not an integer.
* Make the return type of FrameManager.write optional
This is necessary since the SharedMemoryFrameManager.write function can
return None.
* Fix total_seconds() return type in get_tz_modifiers
The function returns a float, not an int.
https://docs.python.org/3/library/datetime.html#datetime.timedelta.total_seconds
* Account for floating point results in to_relative_box
Because the function uses division the return types may either be int or
float.
* Resolve ruff deprecation warning
The config has been split into formatter and linter, and the global
options are deprecated.
* Fix audio events in explore section
Make sure that audio events are listed in the explore section
* Update audio.py
* Hide other submit options
Only allow submits for objects only
* Add service manager infrastructure
The changes are (This will be a bit long):
- A ServiceManager class that spawns a background thread and deals with
service lifecycle management. The idea is that service lifecycle code
will run in async functions, so a single thread is enough to manage
any (reasonable) amount of services.
- A Service class, that offers start(), stop() and restart() methods
that simply notify the service manager to... well. Start, stop or
restart a service.
(!) Warning: Note that this differs from mp.Process.start/stop in that
the service commands are sent asynchronously and will complete
"eventually". This is good because it means that business logic is
fast when booting up and shutting down, but we need to make sure
that code does not rely on start() and stop() being instant
(Mainly pid assignments).
Subclasses of the Service class should use the on_start and on_stop
methods to monitor for service events. These will be run by the
service manager thread, so we need to be careful not to block
execution here. Standard async stuff.
(!) Note on service names: Service names should be unique within a
ServiceManager. Make sure that you pass the name you want to
super().__init__(name="...") if you plan to spawn multiple instances
of a service.
- A ServiceProcess class: A Service that wraps a multiprocessing.Process
into a Service. It offers a run() method subclasses can override and
can support in-place restarting using the service manager.
And finally, I lied a bit about this whole thing using a single thread.
I can't find any way to run python multiprocessing in async, so there is
a MultiprocessingWaiter thread that waits for multiprocessing events and
notifies any pending futures. This was uhhh... fun? No, not really.
But it works. Using this part of the code just involves calling the
provided wait method. See the implementation of ServiceProcess for more
details.
Mirror util.Process hooks onto service process
Remove Service.__name attribute
Do not serialize process object on ServiceProcess start.
asd
* Update frigate dictionary
* Convert AudioProcessor to service process
* Added stop_event to util.Process
util.Process will take care of receiving signals when the stop_event is
accessed in the subclass. If it never is, SystemExit is raised instead.
This has the effect of still behaving like multiprocessing.Process when
stop_event is not accessed, while still allowing subclasses to not deal
with the hassle of setting it up.
* Give each util.Process their own logger
This will help to reduce boilerplate in subclasses.
* Give explicit types to util.Process.__init__
This gives better type hinting in the editor.
* Use util.Process facilities in AudioProcessor
Boilerplate begone!
* Removed pointless check in util.Process
The log_listener.queue should never be None, unless something has gone
extremely wrong in the log setup code. If we're that far gone, crashing
is better.
* Make sure faulthandler is enabled in all processes
This has no effect currently since we're using the fork start_method.
However, when we inevidably switch to forkserver (either by choice, or
by upgrading to python 3.14+) not having this makes for some really fun
failure modes :D
* Subclass Process for audio_process
* Introduce custom mp.Process subclass
In preparation to switch the multiprocessing startup method away from
"fork", we cannot rely on os.fork cloning the log state at fork time.
Instead, we have to set up logging before we run the business logic of
each process.
* Make camera_metrics into a class
* Make ptz_metrics into a class
* Fixed PtzMotionEstimator.ptz_metrics type annotation
* Removed pointless variables
* Do not start audio processor when no audio cameras are configured
* Moved FrigateApp.init_config() into FrigateConfig.load()
* Move frigate config loading into main
* Store PlusApi in FrigateConfig
* Register SIGTERM handler in main
* Ensure logging is setup during config parsing
* Removed pointless try
* Moved config initialization out of FrigateApp
* Made FrigateApp.shm_frame_count into a function
* Removed log calls from signal handlers
python's logging calls are not re-entrant, which caused at least one of
these to deadlock randomly.
* Reopen stdout/err on process fork
This helps avoid deadlocks (https://github.com/python/cpython/issues/91776).
* Make mypy happy
* Whoops. I might have forgotten to save.
Truly an amateur mistake.
* Always call FrigateApp.stop()
* Install multiple ffmpeg versions and add config to make it configurable
* Update docs
* Run ffprobe too
* Cleanup
* Apply config to go2rtc as well
* Fix ffmpeg bin
* Docs
* Restore path
* Cleanup env var
* Fix ffmpeg path for encoding
* Fix export
* Formatting
* Initial re-implementation of semantic search
* put docker-compose back and make reindex match docs
* remove debug code and fix import
* fix docs
* manually build pysqlite3 as binaries are only available for x86-64
* update comment in build_pysqlite3.sh
* only embed objects
* better error handling when genai fails
* ask ollama to pull requested model at startup
* update ollama docs
* address some PR review comments
* fix lint
* use IPC to write description, update docs for reindex
* remove gemini-pro-vision from docs as it will be unavailable soon
* fix OpenAI doc available models
* fix api error in gemini and metadata for embeddings
* intentionally handle queues during shutdown and carefully manage shutdown order
* more carefully manage shutdown to avoid threadlocks
* use debug for signal logging
* ensure disabled cameras dont break shutdown
* typo
* Restrict nginx to 4 processes if more are available
* Fix bash
* Different sed structure
* Limit ffmpeg thread counts for secondary ffmpeg processes
* Add up / down keyboard shortcut
* Add config pub / sub pattern
* remove recording from feature metrics
* remove audio and feature metrics
* Check for updates from all cameras
* remove birdseye from camera metrics
* remove motion and detection camera metrics
* Ensure that all processes are stopped
* Stop communicators
* Detections
* Cleanup video output queue
* Use select for time sensitive polls
* Use ipc instead of tcp
* Use zmq for inter process communication
* Use localhost for reply and request
* Use pyobj instead of json and Need to use separate requestors for each audio listener
* Cleanup port defining
* Send mqtt message when audio is detected
* Fix value
* Add audio topics to mqtt docs and add mqtt headers
* Use existing standard for values
* Update mqtt.md
