Hi,
I have a problem with AI-notification. In developer tools it works (AI description) but when my cameras recognise objects I have small notify with “person”, “dog”, “car” etc. I have Frigate (and notification from other blueprint works good). I have Google Generative AI intalled and added API (the same for Google Provider). And it is my notification on IOS (iPhone):
Traces: Traces - Pastebin.com
And my YAML for blueprint:
“alias: AI Event Summary (LLM Vision v1.3.5)
description: “”
use_blueprint:
path: valentinfrlch/event_summary.yaml
input:
mode: Frigate
notify_device:
- baee6028aad0ca03df8ad0dcf7a1ca1b
camera_entities:
- camera.garaz
- camera.kanciapa
- camera.kuchnia
- camera.ogrod
preview_mode: Live Preview
provider: 01JH86KC6YATGA5KPC6G2G7CVY
object_type:
cooldown: 1
message: >-
Describe what you see in this image from my security camera. Focus only on
dynamic objects like cars, people and animals. Rules for the response: 1.
Response must be in Polish 2. Keep it short (max 200 characters) to fit in
a mobile notification 3. Ignore all static objects like buildings, gates,
letterboxes etc. 4. If you see a person: mention their main
characteristics 5. If you see a car: mention its type and color 6. If
you see an animal: mention its species and behavior 7. Add a funny
comment or emoji occasionally 8. If you see multiple objects, focus on the
most important one - this will be most probablu surrounded by a bounding
box from the camer 8. Use present tense
max_tokens: 90
temperature: 0.5
detail: high
model: gemini1.5-flash”