Fix start.py: use subprocess instead of execv to keep training thread alive

Add startup script that runs sandbox server + training in parallel

Add training deps and startup script to run SADC training in sandbox

Add FSD model code and training script for SADC training

Update metadata with huggingface_hub

Setup sandbox server

update skill and index for persistence

dd tmux to docker

add provision warning

back to debian

use apk

add uv python

hash space ids

remove gpu from skill

add index.html

fix docker

fix dockerfile

add uv in docker

skill md

scripts

add gitignore

make it runnable

Create Dockerfile

initial commit