No suggestions

[Joly0]

Hey guys, i just installed this and configured it with openrouter. For the model i used "deepseek/deepseek-r1-0528". When i test through developer tools and aktions i run it like this:

I can send that and in openrouter i see that this was received and processed, but i do not gget anything in home assistant. Nothing. It just shows this all time:

and no notifications or anything as per the documentation. What ma i missing/doing wrong?

[Joly0]

Also getting some warnings:

2025-06-30 19:32:00.722 WARNING (MainThread) [homeassistant.components.sensor] Setup of sensor platform ai_automation_suggester is taking over 10 seconds.
2025-06-30 19:38:33.486 WARNING (MainThread) [homeassistant.helpers.frame] Detected that custom integration 'ai_automation_suggester' sets option flow config_entry explicitly, which is deprecated at custom_components/ai_automation_suggester/config_flow.py, line 529: self.config_entry = config_entry. This will stop working in Home Assistant 2025.12, please create a bug report at https://github.com/ITSpecialist111/ai_automation_suggester/issues

[RmG152]

@Joly0 These suggestions are stored in attributes.

Below is an example of a card for viewing AI messages.

type: markdown
content: |-
  Status: {{ states('sensor.ai_automation_suggestions_openrouter')
  }}
  Model: {{ states('sensor.ai_model_openrouter') }}
  Last update: {{
  state_attr('sensor.ai_automation_suggestions_openrouter',
  'last_update') }}

  {# Full AI response #}

  {{state_attr('sensor.ai_automation_suggestions_openrouter',
  'suggestions') }}
grid_options:
  columns: full
  rows: auto
title: 🤖 AI Suggestion — Openrouter

or

• Displaying the Description:

  {{ state_attr('sensor.ai_automation_suggestions_<provider_name>', 'description') }}

• Displaying the YAML Block:

  {{ state_attr('sensor.ai_automation_suggestions_<provider_name>', 'yaml_block') }}
  

[Joly0]

Just tried that, nothing:

Also had to adjust the sensor for the model

[Joly0]

Displaying the yaml block fully with
{{ state_attr('sensor.ai_automation_suggestions_openrouter', 'yaml_block') }} just shows "None"

[Joly0]

I can run this as often as I want, the "last update" field updates, but nothing more. It keeps saying "No Suggestions" although i can see in open router, that the request was received and processed

[RmG152]

To ensure the request is handled correctly, check the attributes:

[Joly0]

Ye, I checked that like 30 times already. There is nothing

[RmG152]

Try to delete the AI provider and set it up again.

[Joly0]

Try to delete the AI provider and set it up again.

I trid already, and reinstalling the Plugin. Doesnt work

[ITSpecialist111]

v1.5.3 is now released with provider validation and endpoint fixes, but this report also touches the no-suggestions/output path, so I’m leaving it open. Please retest after updating and share the latest provider response/log details if notifications or stored suggestions are still empty.

[ITSpecialist111]

Recent releases may have addressed this:

• v1.5.0 added structured parsing, YAML fallback, and truncation detection.
• v1.5.1 improved best-effort recovery when providers return malformed or partial structured output.
• v1.5.2 made warning text clearer in persistent notifications.

Please retest on the latest release and share:

1. Exact provider/model that produced missing or truncated YAML
2. Whether YAML now appears, or whether a truncation warning is shown
3. Any relevant integration logs

If this is fixed for you, I can close it; otherwise I’ll use the new details for the next parser hardening pass.

[HexRebuilt]

Hi, by using opencode i was able to get it working on my system.
This is what it fixed to enable me to use my llamaswap istance (custopm openai api)

Stuff that i was doing unkownly wrong:

• wrong endpoint URL, my LlamaSwap URL was missing /v1/chat/completions. i fixed it via the UI. NOTE: i was not able to modify the integration entities without a UI crash. eacht ime i had to remove the created ewntities and re-add them.

Fix for reasoning models returning empty content (Qwen, DeepSeek R1, etc.)
The AI provider responds with 200 OK but content: "" because reasoning models (like Qwen3.5-9B) output their response in reasoning_content instead of content. The integration only reads content, so suggestions are silently discarded.
Three changes needed:
1. coordinator.py — _extract_chat_content(): fall back to reasoning_content when content is empty
2. Config max_output_tokens — reasoning models need more tokens (burn many on thinking before producing output); default 500 is too low
3. suggestions.py — _validate_yaml(): auto-fix YAML indentation (AI reasoning models often indent 4 spaces instead of 2)
The reasoning_content fallback is the key fix — it makes the integration compatible with any reasoning model.

i hope it helps

[ITSpecialist111]

@HexRebuilt — thanks for the excellent diagnosis, that was exactly the right call. Applied the first of your three fixes in v1.5.8 (d8fd997):

coordinator._extract_chat_content now falls back to message.reasoning_content (and message.reasoning) when content is empty. That should unblock Qwen3 / DeepSeek R1 and other OpenAI-compatible reasoning-model deployments where the answer arrives in reasoning_content instead of content.

The other two items you flagged are still on the list:

• Raising max_output_tokens guidance/default for reasoning models — worth documenting rather than silently bumping since it costs users tokens.
• YAML indentation auto-fix in suggestions._validate_yaml — I want a repro sample before touching the parser; if you still have a saved LLM response that produced 4-space YAML, please paste it.

Please update to v1.5.8 via HACS and confirm reasoning-model suggestions now come through. Leaving open for the two follow-ups.