scursel/qwen38-27b-obliterated

Restore working ZeroGPU app and OpenAI bridge contract

Fix clean mount_gradio_app integration

Restore native 4-bit NF4 with app_kwargs middleware

Switch to Q5_K_M GGUF and app_kwargs middleware

Restore clean working ZeroGPU Gradio app

Add FastAPI routes on top of working ZeroGPU demo

Mirror known working ZeroGPU implementation from profgabrielramos demo

Remove type parameter from ChatInterface

Register routes with FastAPI and mount Gradio app properly with top-level @spaces.GPU

Fix literal newlines in Markdown string

Align with ZeroGPU requirements: sdk 6.25.0 and accurate deps

Fix syntax error

Fix type annotations

Fix ZeroGPU initialization: use gr.Blocks with fn=gradio_chat and device_map auto

Fix route middleware for FastAPI + Gradio on ZeroGPU

Add app.py and requirements.txt

Initial deploy: Qwen3.8-27B OBLITERATED on ZeroGPU with OpenAI compatibility

initial commit