Augmentoolkit transforms user documents into training datasets that update an LLM’s knowledge cutoff, enabling domain expertise in any chosen area. It generates fine-tuning data, optionally creates RAG-ready datasets, and can spin up a local inference server for the resulting model. The pipeline works fully offline without external API keys for data generation on most hardware.