[BUG] Incorrect payload sent to OpenAI compatible endpoint after clustering

[nathn123]

Describe the bug
When using the OpenAI env vars and pointing to llamacpp-server which uses openAI chat completions endpoint, it incorrectly sends ollama post structure

To Reproduce
Steps to reproduce the behavior:

1. Set up OpenAI env vars
2. Connect to llamacpp-server
3. Run Clustering

Expected behavior
It sends correctly formatted openAI chat completions payload and the LLM returns a playlist name
Screenshots
If applicable, add screenshots to help explain your problem.

2026-03-10T22:57:00.060442898Z srv  log_server_r: done request: POST /v1/chat/completions 192.168.1.102 400
2026-03-10T22:57:00.060453708Z srv  log_server_r: request:  {"model": "Qwen3.5-9B-Q4_K_M.gguf", "prompt": "You are an expert music collector and MUST give a title to this playlist

Environment (please complete the following information):

• OS: Linux
• Deployment: Docker-compose
• AudioMuse-AI Version: v0.8.14
• Jellyfin/Navidrome Version: N/A
• Browser: Brave
• CPU: xeon E5
• RAM: 128GB
• Disk: a few :D

Additional context
this appears to be this line causing the issue

AudioMuse-AI/ai.py

Line 64 in 8fc2492

is_openai_format = api_key != "no-key-needed" or "openai" in server_url.lower() or "openrouter" in server_url.lower()

it looks like it uses the url to select between openAI/openrouter and ollama maybe two funcs one ollama the other OpenAICompatible

[NeptuneHub]

Hi ant thanks for the feedback. Thanks for point to it, that part of code was overcomplicated and need to be semplified just by checking the env variable AI_MODEL_PROVIDER.

Just to be sure that I understood the problem:

• you set AI_MODEL_PROVIDER to OPENAI;
• you are using llamacpp-server locally, that don't need API

Is that correct?

Thanks!

[nathn123]

so the llamacpp-server endpoint is http://llamacpp-server-url/v1/chat/completions which is the same as openAI but because of how that func detects the difference between ollam/openAI/openRouter it is using the ollama post body not the openAI post body and then i'm getting a bad request since the server isn't expecting that format

[NeptuneHub]

Hi,
I added the fix in devel and the new image should be ready in 30-60 minute.

Important: in this devel version I'm releasing a different machine learning model in place of clap that need a fresh installation an a new rescan. So I suggest for now to use it in a different deployment.

[NeptuneHub]

Released in version 0.9.0 also.