v6: bump to Gradio 5 (fixes Starlette 0.40+ TemplateResponse signature)

v3: install llama-cpp-python from prebuilt wheel (HF dataset, no source build)

v2: pin py3.12 + add inference tab on TinyLlama Q4_K_M

initial: TurboQuant visualizer (rotation effect on quantization)

initial commit