With a self-hosted LLM, that loop happens locally. The model is downloaded to your machine, loaded into memory, and runs ...
If you run Docker long enough on a home server or NAS, this is an inevitable problem. Images pile up, old versions might stay ...