| 7 | | File | Deployed at | Purpose | | 7 | | File | Deployed at | Purpose | |
| 8 | |---|---|---| | 8 | |---|---|---| |
| 9 | | `whisper-lazy-proxy.py` | `/mnt/user/appdata/whisper-lazy/proxy.py` (container `whisper-lazy`, :10300) | Wyoming lazy proxy: answers HA availability checks from cache, starts `faster-whisper` on demand, stops it after 15 min idle (frees GPU VRAM). | | 9 | | `whisper-lazy-proxy.py` | `/mnt/user/appdata/whisper-lazy/proxy.py` (container `whisper-lazy`, :10300) | Wyoming lazy proxy: answers HA availability checks from cache, starts `faster-whisper` on demand, stops it after 15 min idle (frees GPU VRAM). | |
| 10 | | `runpod-lazy-proxy.py` | `/mnt/user/appdata/runpod-lazy/proxy.py` (container `runpod-lazy`, :8189) | ComfyUI-compatible lazy proxy for a volumeless RunPod GPU pod: wake-on-generate with model-group detection, `/lazy/status|warmup|shutdown|ping` control API for the SillyTavern warmup UI, NAS-side 15-min idle terminate. | | 10 | | `runpod-lazy-proxy.py` | `/mnt/user/appdata/runpod-lazy/proxy.py` (container `runpod-lazy`, :8189) | ComfyUI-compatible lazy proxy for a volumeless RunPod GPU pod: wake-on-generate with model-group detection, `/lazy/status|warmup|shutdown|ping|catalog` control API for the SillyTavern warmup UI, NAS-side 15-min idle terminate. Model-set changes on a running pod download in place via the worker image's model-manager (port 8189, LRU disk eviction); pods are only recreated when none exists or the image predates the manager. | |
| 11 | | `runpod-pod.sh` | `/boot/config/runpod-pod.sh` | CLI for the runpod-lazy control API (`status|warmup|shutdown|ping|watch`). | | 11 | | `runpod-pod.sh` | `/boot/config/runpod-pod.sh` | CLI for the runpod-lazy control API (`status|warmup|shutdown|ping|watch`). | |
| 12 | | 12 | |
| 13 | Related: the RunPod worker image (ComfyUI + custom nodes + boot-time model | 13 | Related: the RunPod worker image (ComfyUI + custom nodes + boot-time model |