ALICE + CLIO Integration

CLIO automates ALICE deployments, manages GPU fleets, and orchestrates generation infrastructure from your terminal.

Overview

CLIO's remote execution and multi-agent capabilities make it ideal for managing ALICE at scale. From a single Steam Deck to a fleet of GPU servers, CLIO handles the infrastructure so you can focus on generating.

Deployment Automation

Single Server Deployment

clio --new
> SSH into gpu-server, detect GPU type, install Docker, deploy ALICE with appropriate profile, verify health

Fleet Deployment (Puppeteer Mode)

clio --new
> Deploy ALICE to all hosts in inventory: [gpu-1, gpu-2, gpu-3, gpu-4] using Puppeteer mode with rate limiting

CLIO spawns parallel agents for each target with file/git locking to prevent conflicts.

Steam Deck Deployment

clio --new
> SSH into Steam Deck (deck@steamdeck.local), run ALICE SteamOS installer, configure ROCm, start user service, verify web UI

Model Management at Scale

Bulk Model Download

clio --new
> On all ALICE servers: download top 5 SDXL models from CivitAI, verify SHA256, update model registry

Model Synchronization

clio --new
> Sync model library from primary ALICE server to all replicas, deduplicate, verify integrity

Monitoring & Benchmarking

Automated Health Checks

clio --new
> Every 5 minutes: check ALICE health endpoint on all servers, alert if any unhealthy, collect GPU metrics

Performance Benchmarks

clio --new
> Run standardized benchmark on all GPUs: 20 generations each, collect timing/VRAM/temperature, generate comparison report

Configuration Management

Config Synchronization

clio --new
> Push updated config.yaml to all ALICE instances, rolling restart, verify zero-downtime

Automatic Backend Selection

CLIO detects hardware and selects the appropriate ALICE backend:

Hardware	Backend	Docker Profile
NVIDIA GPU	CUDA + PyTorch	`--profile cuda`
AMD GPU (Linux)	ROCm + PyTorch	`--profile rocm`
Apple Silicon	MPS + PyTorch	Native (launchd)
Steam Deck	ROCm/Vulkan	SteamOS installer
CPU Only	CPU fallback	`--profile cpu`

CLIO Calling ALICE API

CLIO can call ALICE's API for testing and validation:

clio --new
> POST http://alice-server:8080/v1/images/generations with test prompt, verify response structure, download and validate image