Camera Notification V0.8 with Actionable Buttons Android IOS - Frigate LLM Vision AI

Looks good, thanks for the update

for me it seems to all work as expected.

EDIT: yes, latest DEV version

Thank you for the update.

Frigate LLM Notification v0.7 [iOS-aware] - Intelligent Security Camera Automation

Update December 2025

Is there a way I can adapt this with ollama vision?

I’m pretty new to this but I’ve been able to make a basic ollama based image analyser… I use it to check bins and it works but I can’t figure out how to adapt it to enable this frigate environment

alias: Snapshot Driveway for Analysis
description: ""
triggers:
  - trigger: time_pattern
    minutes: /10
actions:
  - data:
      entity_id: camera.driveway
      filename: /config/www/driveway.jpg
    action: camera.snapshot
  - delay:
      seconds: 2
  - data:
      image_url: /config/www/driveway.jpg
      image_name: driveway
      prompt: >-
        How many bins are visible in the left side of the image? Note that they
        might be partially obscured. Provide just a number
      device_id: xxx
    action: ollama_vision.analyze_image

Hi the whole automation is based on frigate and LLM vision so it is probably not the base to start with to be honest there is many blueprints for using the HA built in snapshot . What exactly are you looking for

Some other blueprints allow you to customize the icon that is displayed in lieu of the default home assistant one. This is helpful in differentiating at a glance if this is a CCTV zone notification or another home assistant related notification. I can help integrate this if you would like?

Yes certainly that would be a nice feature…

In LLMVision 1.6.0 ā€žRememberā€œ changed to ā€žStore in timelineā€œ. Seems like this feature is no more working with this blueprint V0.7.

yes i need to look at this shortly. Thanks for the reminder

1 Like

Hi Jack, Update made to latest.yaml

Rename ā€˜remember’ to ā€˜store_in_timeline’ in Latest.yaml

Hopefully thats the only breaking change worth noting any others let me know or add a an issue on git.

Cheers for the heads up.

I can confirm this fixed it.

thanks a lot

Im running an external frigate server and was completely unable to get this automation to work, i think its because to many parts of the config expect that frigate is running with the same IP as the Homeassistant Instance. This should work for anyone using an external frigate Server, it also includes a switch to switch off ai video analysis altogether if your model doesnt support it. GitHub - lenne0815/Frigate-LLM-Vision Ā· GitHub

Hi there thanks for the update and options, it may be better to merge these updates on git for the next update

I beleve your version is using the Frigate API directly rather than the Home Assistant Frigate integration api if I am correct, it would be nice to add another option or maybe a simple toggle to enable external frigate and just have a single host input with a different description. If external frigate server was enabled for example this would then change the api calls like you have used.

Yes thats what it does, it basically only changed the link generation to use frigate directly but im not good enough to pack that into a pr / toggle so it could be merged. If i understand frigate correctly the frigate proxy app should allow your app to work with an external frigate server too but i just couldnt get it to work, there were heaps of permission related errors when trying that.

Okay well il take another look through the yaml you have and see what we need to do to update this to allow an external server for a start.

1 Like

Frigate LLM Notification – v0.8 update

Update to the blueprint — mostly bug fixes and cleanup, no new inputs to configure:

Fixed iOS clip attachments – the iOS-only clip option wasn’t matching correctly before, so it could silently fall back to nothing. Works now.
Fixed LLM Vision timeline storage – was using an invalid field name, so ā€œUse Rememberā€ silently did nothing. Now correctly stores to the timeline.
Retry loops are now capped – no more risk of an endless loop if a clip/snapshot never finishes downloading.
Refreshed the AI model list – removed models that providers have since discontinued, added current ones. Default is now gemini-3.1-flash-lite. Custom model IDs still work if your provider adds something new.
Simplified clip handling internally – no functional change for you, just less redundant data sent to LLM Vision.
Grab it here: https://raw.githubusercontent.com/willhaggan/HA_Frigate_VLLM_Notification/main/Latest.yaml

If you’re upgrading from an older version, just re-import — your existing inputs/settings carry over.

Open your Home Assistant instance and show the blueprint import dialog with a specific blueprint pre-filled.