Description
Mistral model inputs can be preprocessed with the official inference support library from a git snapshot. AI engineers use this package for tokenization, chat templates, tool-call formatting, and model-compatible request preparation. Prompts, tool arguments, and generated metadata can contain sensitive data.