Google_generative_ai_conversation.generate_content deprecated

Configuration
1

Hello,

I have a whole bunch of automations that use the deprecated google_generative_ai_conversation.generate_content action… I want to switch over to ai_task.generate_data action, but I`m confused how to attach a file to it like I did before. Few of my automations use this type of action. How can I substitute for the new way ?

action: google_generative_ai_conversation.generate_content
metadata: {}
data:
filenames: /config/www/proxy/XXXXXXX.jpeg
prompt: >-
Very briefly describe what you see in this image from my XXXXX
camera.Don’t describe stationary objects or buildings.
response_variable: doorbellimagechanged
continue_on_error: true

I`m a little confused on how to add the /config/www/proxy/XXXXXXX.jpeg file, I looked at the docs and it seems only media is able to attach from the build in media HA interface ?

2

@stev3 this is really easy fix.

1) Map your folder in configuration.yaml

If your image is in /config/www/proxy/…, map that folder (or the whole /config/www) under media_dirs:

# configuration.yaml
homeassistant:
  media_dirs:
    local: /media                       # keep default
    www: /config/www                    # exposes ALL of /config/www
    # or, if you prefer just the proxy subfolder:
    # proxy: /config/www/proxy

Then Restart Home Assistant (or Developer Tools → YAML → Reload core configuration if offered). This makes your folder appear in the Media Browser and gives it a media-source name (www or proxy). Home Assistant

2) Reference the file in your AI Task

Now attach the file with a Media Source URI:

image
image1098×1365 20.3 KB

- action: ai_task.generate_data
  response_variable: doorbellimagechanged
  data:
    task_name: "doorbell snapshot description"
    instructions: >
      Very briefly describe what you see in this image.
      Focus on people/vehicles; skip background objects.
    attachments:
      # If you mapped "www: /config/www"
      - media_content_id: media-source://media_source/www/proxy/XXXXXXX.jpeg
        media_content_type: image/jpeg

      # If you mapped "proxy: /config/www/proxy" instead, use:
      # - media_content_id: media-source://media_source/proxy/XXXXXXX.jpeg
      #   media_content_type: image/jpeg

However id simply adjust my automation to snap a picture directly from the camera as its easy to do now. (Correct me if i wrong you use to capture snapshot then save it to www under the same file name every time then reference that in google generative ai?)

image
image1769×1328 163 KB

image
image1010×1278 172 KB

image
image736×1301 53.2 KB

1 Like
3

wow you are great ! Yes I take a snapshot and save it to the folder and then reference it in the AI task. I did see that you can take a snapshot directly from the camera using media now, but I want to send the same original snapshot via notification and email, does taking the snapshot from the ai task will save it somewhere and I can reference it in the email/notificaion ?

4

Thanks, and not that i know of haven’t toyed with it enough yet.
-Take care

5

got it. Thank you again!