Switch to transformers with Qwen2.5-7B-Instruct

Use pre-built llama-cpp-python wheel for cu124

Switch to Gradio + ZeroGPU with llama-cpp-python

Use llama.cpp server with OpenAI-compatible API

Switch to Docker SDK with CUDA for llama-cpp

Use pre-built llama-cpp-python wheel for faster build

Switch to ZeroGPU with llama-cpp for GGUF model

Initial commit: Qwen3.5-9B API interface

initial commit