Allow timeout to be configurable

[firstof9]

Please allow the timeout in providers.py for the API call to be configurable.

Thanks.

[BlueMahib]

Hi, thanks a lot for this project!

I can confirm the same behavior with a local Ollama setup using vision models: the Ollama provider uses a 60s timeout, which can cause image analysis to fail on slower or currently loaded hardware (e.g., J4012 Jetson Orin).

Symptoms (LLM Vision logs)

• “Provider Ollama failed: Request failed:”
• With debug logging enabled, the request starts (“Posting to …/api/chat”) and then fails almost exactly ~60 seconds later.

Root cause: In custom_components/llmvision/providers.py the Ollama request uses aiohttp with:
timeout=ClientTimeout(total=60)

For local vision workloads (e.g., gemma3:12b), the response time can exceed 60 seconds depending on model load and system utilization, which triggers the failure.

Workaround (worked immediately for me)

1. Backup the file:
   cp /config/custom_components/llmvision/providers.py /config/custom_components/llmvision/providers.py.bak

2. Increase the timeout (I used 180s first, later 300s):
   sed -i 's/ClientTimeout(total=60)/ClientTimeout(total=180)/' /config/custom_components/llmvision/providers.py
   (or use 300)

3. Restart Home Assistant.

After this change, llmvision.image_analyzer works reliably with gemma3:12b (vision) via my Ollama HTTPS endpoint (using NGNIX custom configuration to increase timeout also there).

Suggestion: If possible, it would be great to make this timeout configurable as proposed above (integration option / provider option / service parameter), or to increase the default for local providers (especially for vision models).

BR
Tim