requirements.txt

fastapi==0.111.0
uvicorn[standard]==0.29.0
pydantic==2.7.1
# Must be >=0.3.9 -- that's the release that added Qwen3 architecture
# support to llama.cpp. Anything older will fail to load a Qwen3 GGUF
# with "Failed to load model from file".
# Installed by the Dockerfile from abetlen's prebuilt CPU wheel index
# (not compiled from source -- compiling timed out/OOMed the build).
llama-cpp-python==0.3.30