Examples

Six reference models share the identical folder layout and CLI, from a one-command support-ticket triage to a vLLM-servable vision-language model. Each is a standalone folder under examples/ with its own model.yml — install maatml from PyPI, clone the repo for the seed data and plugins, and point the CLI at the folder.

Model	Task	Architecture	Base
Support Ticket Triage	triage → JSON	causal_sft (LoRA)	Qwen3-0.6B
Vision VLM	describe a scene image	vlm_sft (vLLM-servable)	SmolVLM-256M-Instruct
Vision	scene + detect + pose	vision_multitask	MobileNetV3-Large
Vision Describer	caption from vision JSON	seq2seq	flan-t5-small
JCL Validator	jcl_validation	classifier (4-head)	ModernBERT-base
Spool Interpreter	spool_interpretation	seq2seq	flan-t5-base

New to MaatML? Support Ticket Triage is the shortest path end to end — see the 5-minute quickstart. For the multimodal path, start with Vision (trains a checkpoint) and Vision Describer (reads its output and captions it). Vision VLM folds both into one vision-language model, servable directly by vLLM.

Any directory with a valid model.yml works the same way. Scaffold a new one with:

maatml scaffold ~/models/my-task --architecture causal_sft --name my-task

See the plugin author guide for wiring up a validator, metrics, and a predictor for your own task.