Steven Song StevenSong

VSCode + Cline + Ollama + Docker + Qwen3 Coder 30B

For environments that need to be particularly locked down and code/data should not be sent to an external service, a locally served LLM can still be used as a backend to agentic AI coding tools. This gist details steps to use Cline AI coding agent in VSCode using a locally served LLM running in an Ollama Docker image (assuming you use VSCode ± RemoteSSH on the same machine that will serve the model):

start Ollama docker:
```
docker run -d --rm --gpus='"device=0"' -v ollama:/root/.ollama -p 11434:11434 --name ollama ollama/ollama
```
- requires having docker with nvidia container toolkit installed
- set gpu device index to control specific resource usage on multi-gpu systems
serve a capable agentic code model inside the container (at the time of writing, Cline suggests Qwen3 Coder 30B at 8-bit quantization):

	#!/usr/bin/env python

	# eog not bundled with your distro?
	# no sudo access?
	# do it all with python!
	# easiest to use with miniconda env with matplotlib

	import sys
	import matplotlib.pyplot as plt
	from PIL import Image

	import os
	import re
	import gzip
	import hashlib
	from tqdm import tqdm
	from bs4 import BeautifulSoup, SoupStrainer

	fnames = []
	for fname in os.listdir('PubMed'):
	if fname.endswith('.xml'):

	import tensorflow as tf
	b = tf.constant(21306806, dtype=tf.float32)
	print((1+b)-b)
	print(1+(b-b))

	# Launch jupyter notebook in a new tmux session for Colab to connect to
	# https://research.google.com/colaboratory/local-runtimes.html

	SESSION=colab

	tmux new -d -s $SESSION
	tmux send-keys -t $SESSION "jupyter notebook \
	--NotebookApp.allow_origin='https://colab.research.google.com' \
	--port=8888 \
	--NotebookApp.port_retries=0 \

	#!/bin/bash

	# Log files are stored in subdirectories of current directly: .//log
	# Log files contain output like: ROC AUC * {label} * = 0.NNN
	# Where the AUC for a label on the test set is the last ocurrence of the above fragment in the file
	#
	# For example, in an experiment to predict "death", our results are in a
	# folder "experiment" with 10 subfolders "trial_N", where N is the trial number.
	# Within each "trial_N" subfolder, there is a file "log" with output lines:
	#

	#!/bin/python
	#
	# usage:
	# python dispatch.py \
	# --gpus 0-3 \
	# --bootstraps 0-9 \
	# --scripts \
	# train-simple.sh \
	# train-varied.sh \
	# train-deeper.sh