ha-llmvision: Blueprint not working

Bug Description

Since update the blueprint to 1.3.7 nothing trigger the automation, but if i look into frigate, i got event

Version: 1.3.7

Service Call

alias: Notif AI Cam Sonette
description: ""
use_blueprint:
  path: valentinfrlch/event_summary.yaml
  input:
    mode: Frigate
    notify_device:
      - d0d9e52e51bf00ca693a6161e29b1b6a
    camera_entities:
      - camera.cam_sonette
    required_zones:
      - cam sonette motion
    object_type:
      - person
      - car
    tap_navigate: "{{video if video != '''' else image}}"
    provider: 01JHNP36WMNRXSF2WYRTKT704X
    temperature: 0.3

Logs

Paste relevant logs here

Additional context

About this issue

  • Original URL
  • State: closed
  • Created 2 years ago
  • Comments: 59 (12 by maintainers)

Commits related to this issue

Most upvoted comments

The blueprint has become complicated, and many people have had problems with it. I haven’t upgraded Frigate to 0.15, but I think having the summaries in Frigate directly is easier to set up and use.

I don’t think it’ll make the blueprint obsolete, but perhaps the “Frigate” mode is unnecessary now. Removing the Frigate mode and focussing on Camera mode for the people who don’t have Frigate would make maintaining the blueprint much easier and less prone to failure.

That’s a few of us then. I’m happy to test anything needed if I can help in any way.

Automation is triggered: Image

but: Image

The issue is likely the updated frigate trigger:

trigger: mqtt topic: frigate/events id: frigate_trigger enabled: “{{mode == ‘Frigate’}}” The enabled template should evaluate to true only if mode is set to “Frigate”. Previously there was an issue that would throw an error if mqtt wasn’t installed even though camera mode doesn’t need mqtt.

Confirmed, in part. By removing any condition based on mode == 'Frigate', the automation fires and listens on frigate/events. Note that line 306 in the blueprint sets mode: single, which probably conflicts. However, I’ve broken it somehow in my config because the trace is failing at the value template when I go walk in front of the camera. Not sure what’s going on there, but will investigate more as I have time.

Here’s what I changed it to:

trigger:
  - platform: mqtt
    topic: frigate/events
    id: frigate_trigger
  - platform: state
    entity_id: !input camera_entities
    to: !input trigger_state
    id: camera_trigger
  - platform: state
    entity_id: !input motion_sensors
    to: "on"
    id: motion_sensor_trigger
condition:
  - condition: template
    value_template: >
      {{ trigger.payload_json["type"] == "new"
        and ('camera.' + trigger.payload_json['after']['camera']|lower) in camera_entities_list
        and ((object_types_list|length) == 0 or ((trigger.payload_json['after']['label']|lower) in object_types_list))
        and (not required_zones_list or ((set(trigger.payload_json['after']['current_zones']).intersection(required_zones_list))))
        and not (ignore_stationary and trigger.payload_json['after']['stationary'])
      }}

Note that it also wasn’t triggering when type was set to end.

Im not sure how i can do, but can we test a basic automation without the blueprint to see if is working ?

You can do that under developer tools > actions https://llm-vision.gitbook.io/examples/examples/scripts