Commit History

  1. Fix: download GGUF eagerly at container boot, outside the @spaces.GPU lease (was burning GPU-seconds on a 6-7min network download, causing the lease to be reclaimed mid-download with no error surfaced)

    Verified signature
    db2d88c
  2. Real fix: preload nvidia-*-cu12 pip packages' shared libraries via ctypes before llama_cpp import (libcudart.so.12 not found on ZeroGPU worker, confirmed via real runtime logs)

    Verified signature
    312640e
  3. Real fix: restore demo.launch() as entrypoint (required for ZeroGPU detection), expose /generate via Gradio's native api_name mechanism instead of a custom FastAPI route

    Verified signature
    89b4e16
  4. Fix /generate route being shadowed by Gradio's own catch-all -- mount Gradio at /ui instead of adding routes post-hoc to demo.app

    Verified signature
    83e90ad
  5. Switch to Gradio SDK -- ZeroGPU only works with Gradio SDK, not Docker/FastAPI

    Verified signature
    00bec90
  6. chatPDB API: FastAPI + llama-cpp-python ZeroGPU backend, Q4_K_M GGUF

    Verified signature
    5be713f
  7. initial commit

    Verified signature
    26cc93e