Frigate - Telegram Notification

frigate_telegram_notification.yaml

blueprint:
  name: Frigate - Telegram Notification with Generative AI comments
  description: Create automations to receive Snapshots and Clips from Frigate
  domain: automation
  input:
    camera:
      name: Frigate Camera
      description: "Select the camera entity that will trigger notifications. \nIf
        you do not see cameras listed in the drop down, check you have the frigate
        integration installed.\nNote: The automation relies on this matching your
        frigate config (by default it does).\n"
      selector:
        entity:
          integration: frigate
          domain:
          - camera
          device_class:
          - camera
          multiple: false
    target_chat:
      name: Target
      description: 'The chat_id to be used by the Telegram bot. !secret chat_id is
        not allowed on Blueprint, you will need the chat_id code.
        '
    notification:
      name: Notification
      description: 'Select "true" to disable notification, leave "false" to receive
        notification. '
      selector:
        select:
          options:
          - 'true'
          - 'false'
          multiple: false
          sort: false
          custom_value: false
      default: 'false'
    prompt: 
      name: Gemini Prompt
      description: Customise your AI prompt for https://www.home-assistant.io/integrations/google_generative_ai_conversation/
      default: Very briefly describe what you see in this image from my cctv camera.
        Your message needs to be short to fit in a phone notification. Don't 
        describe stationary objects or buildings. Only describe animals, vehicles or persons.
    base_url:
      name: (Optional) Base URL
      description: The external url for your Home Assistant instance. This will default
        to a relative URL and will open the clips in the app instead of the browser,
        which may cause issues on some devices.
      default: ''
    zone_filter:
      name: (Optional) Zone Filter
      description: Only notify if object has entered a defined zone.
      default: false
      selector:
        boolean: {}
    zones:
      name: (Optional) Trigger Zones
      description: A list (-) of zones you wish to recieve notifications for.
      default: []
      selector:
        object: {}
    labels:
      name: (Optional) Trigger Objects
      description: A list (-) of objects you wish to recieve notifications for.
      default: []
      selector:
        object: {}
    presence_filter:
      name: (Optional) Presence Filter
      description: Only notify if selected presence entity is not "home".
      default: ''
      selector:
        entity:
          multiple: false
  source_url: https://gist.github.com/jakesOneFourSeven/cc8fab1e8712a34dbd596ebd355d76ea
mode: single
max_exceeded: silent
trigger_variables:
  input_camera: !input camera
  camera: '{{ input_camera | replace(''camera.'', '''') }}'
trigger:
- platform: mqtt
  topic: frigate/events
  payload: '{{ camera }}/new'
  value_template: '{{ value_json[''after''][''camera''] | lower | replace(''-'',''_'')
    }}/{{ value_json[''type'']}}'
  id: frigate-event
variables:
  id: '{{ trigger.payload_json[''after''][''id''] }}'
  input_camera: !input camera
  camera: '{{ input_camera | replace(''camera.'', '''') }}'
  camera_name: '{{ camera | replace(''_'', '' '') | title }}'
  target_chat: !input target_chat
  object: '{{ trigger.payload_json[''after''][''label''] }}'
  label: '{{ object | title }}'
  entered_zones: '{{ trigger.payload_json[''after''][''entered_zones''] }}'
  type: '{{ trigger.payload_json[''type''] }}'
  base_url: !input base_url
  zone_only: !input zone_filter
  input_zones: !input zones
  zones: '{{ input_zones | list }}'
  input_labels: !input labels
  labels: '{{ input_labels | list }}'
  presence_entity: !input presence_filter
  notification: !input notification
condition:
- '{{ type != ''end'' }}'
- '{{ not zone_only or entered_zones|length > 0 }}'
- '{{ not zones|length or zones|select(''in'', entered_zones)|list|length > 0 }}'
- '{{ not labels|length or object in labels }}'
- '{{ not presence_entity or not is_state(presence_entity, ''home'') }}'
action:
- service: downloader.download_file
  data:
    overwrite: true
    url: '{{base_url}}/api/frigate/notifications/{{id}}/snapshot.jpg'
    filename: snapshot1.jpg
- alias: Extra delay before getting snapshot file
  delay: 1
- service: downloader.download_file
  data:
    overwrite: true
    url: '{{base_url}}/api/frigate/notifications/{{id}}/snapshot.jpg'
    filename: snapshot2.jpg
- alias: Extra delay before getting snapshot file
  delay: 1
- service: downloader.download_file
  data:
    overwrite: true
    url: '{{base_url}}/api/frigate/notifications/{{id}}/snapshot.jpg'
    filename: snapshot3.jpg
- alias: Extra delay before getting snapshot file
  delay: 1
- service: google_generative_ai_conversation.generate_content
  data:
    filenames: 
      - www/downloads/snapshot1.jpg
      - www/downloads/snapshot2.jpg
      - www/downloads/snapshot3.jpg
    prompt: !input prompt        
  response_variable: generated_content
- service: telegram_bot.send_photo
  data:
    target: '{{ target_chat }}'
    disable_notification: '{{ notification }}'
    caption: ' {{ generated_content.text }} '
    url: '{{base_url}}/api/frigate/notifications/{{id}}/snapshot.jpg'
- repeat:
    sequence:
    - wait_for_trigger:
      - platform: mqtt
        topic: frigate/events
        payload: '{{ id }}'
        value_template: '{{ value_json[''after''][''id''] }}'
      timeout:
        minutes: 2
      continue_on_timeout: false
    - condition: template
      value_template: '{{ wait.trigger.payload_json[''type''] == ''end'' }}'
    - service: telegram_bot.send_video
      data:
        target: '{{ target_chat }}'
        disable_notification: '{{ notification }}'
        caption: ' {{ generated_content.text }} '
        url: '{{base_url}}/api/frigate/notifications/{{id}}/{{camera}}/clip.mp4'
    until: '{{ wait.trigger.payload_json[''type''] == ''end'' }}'

Hello friend, I really liked it and I use it!
Is there any way to adapt this code for ollama and use it locally?

Hello friend, I really liked it and I use it! Is there any way to adapt this code for ollama and use it locally?

Hi, glad to hear you are enjoying the changes. I haven't tried using Ollama, not hearing great things at the moment.

You could replace the following action that calls the Gemini service with the equivalent service for Ollama:

"- service: google_generative_ai_conversation.generate_content
data:
prompt: '>-{{ prompt }}'
image_filename: www/downloads/snapshot.jpg
response_variable: generated_content"

I will try using Ollama once it matures a little bit more.

I've added an option to send the response from gemini to a google speaker, do you want me to update the gist to include that option?

"- action: tts.google_say
data:
cache: false
entity_id: '{{ speaker_target }}'
message: ' {{ generated_content.text }} ' "

Hello friend, I really liked it and I use it! Is there any way to adapt this code for ollama and use it locally?

Hi, glad to hear you are enjoying the changes. I haven't tried using Ollama, not hearing great things at the moment.

You could replace the following action that calls the Gemini service with the equivalent service for Ollama:

"- service: google_generative_ai_conversation.generate_content data: prompt: '>-{{ prompt }}' image_filename: www/downloads/snapshot.jpg response_variable: generated_content"

I will try using Ollama once it matures a little bit more.

I understand, my friend, I'll wait too!
Even though Gemini is "free", I don't feel comfortable with so much information being given to Google.

I've added an option to send the response from gemini to a google speaker, do you want me to update the gist to include that option?

"- action: tts.google_say data: cache: false entity_id: '{{ speaker_target }}' message: ' {{ generated_content.text }} ' "

For me, it wouldn't work, because there are 8 cameras and 2 are pointed at the street. I would never stop talking lol.
I would really like to get answers in my primary language, pt-BR. I tried to modify the prompt and even asking in pt-BR or asking to answer in pt-BR, it keeps answering in en-US.

I created an automation to alert cameras that were offline for more than 1 minute and report again which ones are still online. The interval for resending messages is every 5 minutes. If there are no more, it reports "All cameras are online". I tried using the "initial_delay" variable in the loop but it always gave an error... :S with this exception, could you check and give your opinion @jakesOneFourSeven ?
Code:

Automation_Home_Assistant_Telegram_Notify_and_Check_Offline_Cameras_Frigate

This doesn't work for me. Error while running automation.
The automation "Living_Room_Camera_Notify_with_AI" (automation.living_room_camera_notify_with_ai) has an unknown action: downloader.download_file.

This error prevents the automation from running correctly. Maybe this action is no longer available, or perhaps a typo caused it.

Executed: November 11, 2024 at 7:27:44 PM
Error: Action downloader.download_file not found
Result:
params:
domain: downloader
service: download_file
service_data:
overwrite: true
url: >-
https://my-web-address/api/frigate/notifications/1731371263.671046-ohz2no/snapshot.jpg
filename: snapshot.jpg
target: {}
running_script: false

Hey @mrfixitpa , look and adjust your config of HA to add:

downloader: download_dir: /config/www/downloads

View first comment :)

Changes

2024-03-30

• service: google_generative_ai_conversation.generate_content incorporated into telegram messages.

• Added 2 second delay after downloader because the AI was sometimes using the old snapshot.jpg

• Need to add the following to configuration.yaml:

downloader: download_dir: /config/www/downloads

2024-03-31

• Added input field for AI prompt in order to allow per automation prompt.
• Added camera selector dropdown in blueprint

Thanks for this. Apparently that is also not the fix. I did what your reply suggested and I received and immediate error. It seems now the downloader is now within the integrations. I have made the changes and no errors. Hope this works :)

I had the same problem as you, I just didn't remember how I solved it, if I'm not mistaken, just download the service in settings/Devices and services - "Downloader" and in the configuration you enter the path. I checked my .yaml and I don't have anything from the first comment. Remember configure the Google Gemini for HA.