gx10-orchestrator

name	gx10-orchestrator
description	Manage local AI runtimes (Ollama, ComfyUI, Stable Diffusion WebUI) on the ASUS GX10 workstation. Use when starting, stopping, or checking status of local LLMs and image generators.

GX10 AI Runtime Orchestration

Hardware

Item	Value
SoC	NVIDIA Grace Blackwell GB10 (ARM aarch64)
RAM	128 GB LPDDR5x Coherent Unified Memory (CPU+GPU shared)
OS	Ubuntu 24.04.4 LTS, CUDA 13.0
Storage	916 GB NVMe, ~598 GB free
Tailscale	100.108.147.12
Domains	ai.hjs.space → ComfyUI :8188

Ollama

Item	Value
Binary	/home/hjs/bin/ollama
API	http://127.0.0.1:11434
Restart	systemctl --user restart ollama

Models

Model	Size	Role
qwen3.5-queen-27b	28 GB	Primary (drafts, review, char gen)
gemma-4-31b-uncensored	32 GB	Creative fallback, uncensored
golddiamondgold-70b	42 GB	High-quality heavy fallback
glm-4-9b-abliterated	10 GB	Fast lightweight tasks
hermes3	4.7 GB	Tool-use / function calling
nemotron-mini	2.7 GB	Ultra-fast triage

Ollama Python API

import urllib.request, json

def ollama_generate(model: str, prompt: str) -> str:
    payload = json.dumps({"model": model, "prompt": prompt, "stream": False}).encode()
    req = urllib.request.Request("http://127.0.0.1:11434/api/generate",
                                  data=payload, headers={"Content-Type": "application/json"})
    with urllib.request.urlopen(req, timeout=120) as r:
        return json.loads(r.read())["response"].strip()

ComfyUI

Item	Value
Location	/home/hjs/ComfyUI
Venv	/home/hjs/comfyui_env
Port	8188
Model	animagine-xl-3.1.safetensors (symlink)
Speed	~67s/image

# Start
source /home/hjs/comfyui_env/bin/activate
nohup python3 /home/hjs/ComfyUI/main.py --listen 127.0.0.1 --port 8188 > /tmp/comfyui.log 2>&1 &

# Health check
curl -s http://127.0.0.1:8188/system_stats | python3 -c "import sys,json; d=json.load(sys.stdin); print('OK v'+d['system']['comfyui_version'])"

SD WebUI (A1111)

• Location: /home/hjs/stable-diffusion-webui | Port: 7860
• Start: cd ~/stable-diffusion-webui && ./webui.sh

Model Symlinking (save disk space)

# Symlink SD WebUI checkpoints → ComfyUI (already done for animagine)
ln -sf ~/stable-diffusion-webui/models/Stable-diffusion/animagine-xl-3.1.safetensors \
       ~/ComfyUI/models/checkpoints/animagine-xl-3.1.safetensors

Memory Budget

• qwen3.5-queen-27b: 28 GB
• animagine-xl-3.1 SDXL: ~7 GB VRAM
• Both run simultaneously in 128 GB unified pool
• Free at runtime: ~83 GB

Fallback Chain (LLM)

1. qwen3.5-queen-27b (local primary)
2. Gemini CLI (cloud, rate-limited)
3. gemma-4-31b-uncensored (local fallback)
4. Claude CLI (cloud escalation)
5. golddiamondgold-70b (heavy local last resort)