GingerLabsPlatform/qwen38-prompt-enhancer-q8-runtime

Gingerlabs Qwen3.8 prompt-enhancer runtime mirror

This immutable deployment mirror contains only the GGUF artifacts needed by the Gingerlabs Runpod prompt-enhancer worker:

• —Q8KP target model
• —HauhauCS FastMTP 32K draft-vocabulary sidecar
• —BF16 vision projector, retained for optional future image captioning

The artifacts are byte-identical copies from upstream revision 993a5971fda8f30dd1b7eb2654792ba4415c7460. Signed upstream provenance and the FastMTP runtime patch are included. This repository intentionally omits all other quantizations so Runpod Cached Models does not prepare unused files.

The worker is text-only. The 32K in the FastMTP filename refers to the draft vocabulary, not the serving context length.

See THIRD_PARTY_NOTICES.md and LICENSE before redistribution.