Podman Quadlets¶

Bazzite AI workloads run as systemd user services using Podman Quadlets. The ujust <workload> config command automatically generates quadlet files.

How Quadlets Work¶

When you run ujust <workload> config, a quadlet file is generated in ~/.config/containers/systemd/. When started, systemd manages the container:

Auto-start on login - Containers start automatically
Restart policies - Automatic restart on failure
Service management - Standard systemctl commands work
GPU passthrough - Automatically configured

Quick Start¶

Step	Command	Description
1	`ujust ollama config`	Generate quadlet
2	`ujust ollama start`	Start service
3	`ujust ollama status`	Check status

Workload Lifecycle¶

All workloads follow the config → start pattern:

Command	Description
`ujust <workload> config`	Configure and generate quadlet
`ujust <workload> start`	Start via systemctl
`ujust <workload> status`	Check service status
`ujust <workload> logs`	View container logs
`ujust <workload> stop`	Stop service
`ujust <workload> restart`	Restart service
`ujust <workload> delete`	Remove quadlet and config

Multiple Instances¶

Run multiple instances with separate quadlets:

# First instance (default)
ujust ollama config
ujust ollama start

# Second instance (different port)
ujust ollama config -n 2 --port=11435
ujust ollama start -n 2

Each instance has:

Separate quadlet file: ~/.config/containers/systemd/ollama-<n>.container
Separate config: ~/.config/ollama/<n>/config
Separate container name: ollama-<n>

Manual Systemctl Commands¶

After configuration, standard systemctl commands work:

# Check status
systemctl --user status ollama-1.service

# View logs
journalctl --user -u ollama-1.service -f

# Stop/Start
systemctl --user stop ollama-1.service
systemctl --user start ollama-1.service

# Disable auto-start
systemctl --user disable ollama-1.service

Generated Quadlet Location¶

Quadlet files are stored in:

~/.config/containers/systemd/<workload>-<instance>.container

Example ollama-1.container:

[Unit]
Description=Ollama LLM Server (Instance 1)
After=network-online.target

[Container]
Image=ghcr.io/atrawog/bazzite-ai-pod-ollama:stable
PublishPort=11434:11434
Volume=%h/.config/ollama/1:/home/jovian/.ollama
AddDevice=nvidia.com/gpu=all
Network=bazzite-ai.network

[Service]
Restart=on-failure

[Install]
WantedBy=default.target

Available Workloads¶

Workload	Config Command	Recording
Ollama	`ujust ollama config`
Jupyter	`ujust jupyter config`
ComfyUI	`ujust comfyui config`
OpenWebUI	`ujust openwebui config`
FiftyOne	`ujust fiftyone config`
Jellyfin	`ujust jellyfin config`
Portainer	`ujust portainer config`