Support / Provide include files to simplify advance configuration

Right now there are a lot of FFMPEG settings to support server acceleration and to FIX specific camera issues however many of these will be common to large groups of users.

Home assistant and a lot of other YAML configured projects use the idea of importing include files for simplifying the main config and also for storing secrets etc.

What I would propose is to:

1. support user made include files or document it if already possible.
2. Bundle / include a library of common include files (with each frigate release) for Server HW / Camera HW specific settings…

IE instead of depending on documentation for Intel support

Change this:

ffmpeg:
  hwaccel_args:
    - -hwaccel
    - vaapi
    - -hwaccel_device
    - /dev/dri/renderD128
    - -hwaccel_output_format
    - yuv420p

to

ffmpeg:
  hwaccel_args:  !include global_hwaccel_args_intel_7thGen_or_less.yaml

Instead of a large FFMPEG block in a camera section.

 reolink:
    ffmpeg:
      hwaccel_args:
      input_args:
        - -avoid_negative_ts
        - make_zero
        - -fflags
        - nobuffer+genpts+discardcorrupt
        - -flags
        - low_delay
        - -strict
        - experimental
        - -analyzeduration
        - 1000M
        - -probesize
        - 1000M
        - -rw_timeout
        - "5000000"

to

 reolink:
    ffmpeg: !include camera_ffmpeg_reolink_generic.yaml

If the include files where maintained with the releases it would help make upgrades easier going forward, and would make the config file a bit less lengthy.

Documentation would also possibly need less updating if it just generally referred to looking at the include library for matching configurations.

Using include files in the camera section in general would be very helpful. Many users will have the same over all settings for several cameras so being able to use includes would clean up these sections.

[bagobones]

More config files could be exposed but it gets complicated quickly because we either have to give the container access to the entire /config directory (security implications) or like now we only have access to /config/frigate.yml

Maybe I am not understanding this but since this is all based on docker is this not easily fixed by moving the config file or the includes into their own path / folder on the host and mappings that folder for that use?/config/includes but only includes is mapped with folder access?

Also going back to presets instead of !include you go !preset and use a named preset and !preset just maps to a folder in the image.

End of the day these are just ideas.

Looking at my current config I would love to take this:

cameras:
  doorbell:
    ffmpeg:
      inputs:
        - path: rtsp://admin:password@192.168.103.8:554
          roles:
            - record
            - rtmp
        - path: rtsp://admin:password@192.168.103.8:554/cam/realmonitor?channel=1&subtype=1&authbasic=64
          roles:
            - detect
    rtmp:
      enabled: True
    detect:
      width: 640
      height: 480
      fps: 5
    motion:
      threshold: 30
      mask:
        - 0,30,640,31,640,0,0,0
        - 0,442,162,403,254,409,294,451,412,451,531,428,598,406,594,248,0,249
        - 640,296,640,133,568,131,566,300
    zones:
      front_steps:
        coordinates: 227,480,222,402,158,396,0,439,0,480
        objects:
          - person
      front_deck:
        coordinates: 640,480,640,418,602,410,447,444,289,458,295,480
        objects:
          - person
    snapshots:
      enabled: True
    record:
      enabled: True
      events:
        required_zones:
          - front_steps
          - front_deck
  garage_front:
    ffmpeg:
      inputs:
        - path: rtsp://admin:passwor@192.168.103.3:554/
          roles:
            - record
            - rtmp
        - path: rtsp://admin:passwor@192.168.103.3:554/h264Preview_01_sub
          roles:
            - detect
    rtmp:
      enabled: True
    detect:
      width: 640
      height: 360
      fps: 7
    objects:
      track:
        - person
        - car
    motion:
      threshold: 35
      mask:
        - 640,360,400,360,398,334,640,328
        - 149,83,21,0,640,0,640,300
    zones:
      front_steps:
        coordinates: 124,70,127,135,163,135,213,162,215,179,154,193,0,124,0,0,30,0
        objects:
          - person
      parking_spot:
        coordinates: 0,360,0,262,262,159,640,291,640,360
        objects:
          - person
          - car
      mailbox:
        coordinates: 66,158,72,221,0,244,0,133
        objects:
          - person
    snapshots:
      enabled: True
    record:
      enabled: True
      events:
        required_zones:
          - front_steps
          - parking_spot

And turn it into

cameras:
  doorbell: !include camera_doorbell.yaml
  garage_front: !include camera_garage_front.yaml

Having each camera as a file would greatly simplify version control and make cloning, swapping back and forth between versions of a config, much easier than tracking a DIFF in the main config file.

[bagobones]

Still think this is a good idea but will let this go stale next time around. The more I tune detection min_size etc per camera the longer my config file gets.