Switch to Gradio SDK -- ZeroGPU only works with Gradio SDK, not Docker/FastAPI

chatPDB API: FastAPI + llama-cpp-python ZeroGPU backend, Q4_K_M GGUF

initial commit