local llm setup with docker sandbox + lm studio + pi + qwen

model - qwen/qwen3.6-35b-a3b

model - qwen2.5-coder-3b

---

open terminal pwsh

check existing docker sandbox with :

sbx

or

sbx ls

if none :

then to make a new docker sandbox with pi as agent with all the skills setup from my GitHub run :

sbx run pi --kit "git+https://github.com/ArifAziz99/docker-sandbox.git#dir=pi" --kit "git+https://github.com/ArifAziz99/docker-sandbox.git#dir=skills"

if exists :

then to get back into sandbox run :

sbx exec -it pi-AI /bin/bash

note: instead of pi-AI add your sandbox name shown with sbx ls

then inside the sandbox to start/continue pi run :

pi -c

then in another terminal:

to start lm studio server run :

lms server start

to load a model from lm studio run :

lms load