Hugging Face Generative Model Converter

Convert pinned Hugging Face models to a local runtime format, prove the converted model actually works, and — only then — package and publish it. Every claim in your final report must map to recorded evidence; a step that did not run is pending, never passed.

Inputs to capture first

1. Source model ID and optional revision (resolve to an immutable commit; never silently change the requested model, revision, target, precision, or workload).
2. Target format: gguf | onnx | litert-lm (normalize litert/litertlm to litert-lm; treat "llama.cpp format" as gguf).
3. Destination (optional): Hub repo id and visibility for publication.
4. Deployment backend/profile if the user names one; otherwise use the recipe's documented default.

Ask only for what is missing and material. Reuse authorization and answers already given for the same operation; never re-prompt for an existing credential.

Phase order (report a per-phase result at the end)