MQTT/Remote logs possible

[matknowles]

Hi everyone, is it possible to get logs via MQTT or some other remote method? I'm having some silent errors happening occaisonally which are necessitating a container restart, but if left un-restarted are causing loss of operation/recording/autotracking/alerts.
I'm hoping to be able to monitor through home assistant so if there's an error I can be notified to restart or ideally restart the container automatically.

The issues include dropped detections (error on camera feed), too many unprocessed segments (resulting in loss of recording - see discussion 15683) and a memory error which may be related to the fact im running on linux mint (planning on migrating to debian soon to see if that fixes it)

I am using Dockge to manage containers if it matters/helps.

[NickM-27]

This can be done using the api https://docs.frigate.video/integrations/api/logs-logs-service-get

[matknowles]

That's great thanks for that. For anyone else here's how I setup a rest sensor to get the frigate logs into home assistant (this goes in configuration.yaml and the logs are in the attributes of the created sensor):

sensor:
  - platform: rest
    name: "Frigate Raw Logs"
    resource: http://<frigate-IP>:5000/api/logs/frigate
    scan_interval: 300
    json_attributes:
      - lines
    value_template: "OK" # Simple state value

validate, reboot and the sensor appears as sensor.frigate_raw_logs

I have an automation that's supposed to notify by TTS when an error appears in the logs, but I haven't validated it yet. Below if anyone's interested:

alias: (Anytime but only Notify between 0630-2200) Frigate Log Error TTS Alerts
description: Notify on Frigate errors from raw logs
triggers:
  - entity_id: sensor.frigate_raw_logs
    for: "00:01:00"
    trigger: state
conditions:
  - condition: time
    after: "06:30:00"
    before: "22:00:00"
actions:
  - choose:
      - conditions:
          - condition: template
            value_template: "{{ error_count > 0 }}"
        sequence:
          - action: tts.speak
            metadata: {}
            data:
              cache: false
              media_player_entity_id: media_player
              message: "    Frigate logs error:  {% if 'WARNING' in logs %}Warning - {% elif 'ERROR' in logs     %}Error - {% endif %} {{ latest_errors | join('... ') | truncate(100) }}"
            target:
              entity_id: tts.piper
variables:
  error_pattern: (ERROR|WARNING|CRITICAL|failed|exception|maintenance)
  logs: |-
    {{ state_attr('sensor.frigate_raw_logs', 'lines') | join('
    ') }}
  error_count: "{{ logs | regex_findall(error_pattern) | count }}"
  latest_errors: >-
    {% set matches = logs | regex_findall('.*' ~ error_pattern ~ '.*',
    multiline=True, ignorecase=True) %} {{ matches[-3:] if matches | length > 0
    else [] }}  # Last 3 errors

[bagobones]

It can also be done in docker via the logging driver

services:
  frigate:
    logging:
      driver: "syslog"
      options:
        syslog-address: "udp://192.168.102.11:514"