Error when processing

[Viggeborn]

Installed this (what i hop) awesome docker on my server today.

When it processes i get error on everything exept 5 images(Not fully done). No info why the error come but the logs say

2026-05-12 13:14:09,047 ERROR backend.pipeline: AI processing failed for /photos/appdata_oc17x2tlq5t2/preview/f/a/e/f/e/c/4/5398/64-64-crop.jpg: Server error '500 Internal Server Error' for url 'http://IP-REMOVED/api/generate'

Any suggestions why this comes?

[vonhex]

try giving my fork a try I think I fixed this issue, he has yet to pull my changes. https://github.com/vonhex/a-eye its also on Unraid in the community apps.

[Viggeborn]

try giving my fork a try I think I fixed this issue, he has yet to pull my changes. https://github.com/vonhex/a-eye its also on Unraid in the community apps.

I saw now that it was the Vonhex one i hade installed

[vonhex]

so the issue you are having is on my fork?

[vonhex]

Are you using Nextcloud for this?

[vonhex]

Ok I pushed a fix, if using Unraid please force update and let me know if its working now.

[Viggeborn]

Ok I pushed a fix, if using Unraid please force update and let me know if its working now.

Dont know how you Force update in Unraid but uninstalled and installed again and still same

[vonhex]

can you close this here and make an issue on my fork please.

[vonhex]

● Now the error logged will say something like Ollama error: model 'llava:13b' not found or Ollama error: CUDA out of memory
instead of the useless HTTP status message.

Root cause in this report: The Ollama server at 192.168.1.214:11434 returned a 500. The most common causes on Unraid are:

1. Model not found — the vision model configured in A-EYE isn't pulled on that Ollama instance (ollama pull )
2. OOM — GPU/RAM ran out when loading a large model; try a smaller quantization (e.g. q4_K_M instead of q8)
3. Model timeout — Ollama unloaded the model between the /api/tags check and the generate call (unlikely but possible under
   memory pressure)

The user's next step should be to check the Ollama container logs on their Unraid (docker logs ) — with this
fix, A-EYE will also now log the specific Ollama error message so they won't have to dig into Ollama's logs at all.