MMIRAGE Documentation¶
MMIRAGE — Modular Multimodal Intelligent Reformatting and Augmentation Generation Engine — is an advanced platform for large-scale dataset processing using generative models, including vision-language models (VLMs).
Install MMIRAGE and run your first pipeline in minutes.
Full YAML configuration reference for all parameters.
All mmirage subcommands, flags, and examples.
Auto-generated documentation for every public module.
Key Features¶
Multimodal support — process text and images with vision-language models.
YAML-driven — configure every aspect of a pipeline via a single file using Jinja2 templating and JMESPath queries.
Scalable — native sharding with multi-node SLURM support.
Modular — pluggable processors, loaders, and writers.
Automatic retry — configurable shard-level retry with budget tracking.
Structured output — produce plain text or validated JSON.