# Multi-stage build for phostt with CUDA support
# Build: docker build -f Dockerfile.cuda -t phostt-cuda .
# Run: docker run --gpus all -p 9876:9876 phostt-cuda
#
# CUDA binary falls back to CPU if no GPU is available at runtime.
# --- Builder stage ---
FROM nvidia/cuda:12.6.3-devel-ubuntu22.04 AS builder
RUN apt-get update && \
apt-get install -y --no-install-recommends curl build-essential pkg-config ca-certificates && \
curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh -s -- -y --default-toolchain 1.85.0 && \
rm -rf /var/lib/apt/lists/*
ENV PATH="/root/.cargo/bin:${PATH}"
WORKDIR /build
# Dependency-compilation cache (see Dockerfile for rationale).
COPY Cargo.toml Cargo.lock ./
RUN mkdir -p src && \
echo 'fn main() {}' > src/main.rs && \
touch src/lib.rs && \
cargo build --release --features cuda && \
rm -rf src target/release/deps/phostt-* target/release/phostt*
COPY src/ src/
# ort's download-binaries feature fetches CUDA-enabled ONNX Runtime.
# copy-dylibs copies libonnxruntime*.so to target/release/.
RUN cargo build --release --features cuda && \
strip target/release/phostt
# --- Model bake stage (runs only when PHOSTT_BAKE_MODEL=1) ---
FROM debian:bookworm-slim AS model-fetcher
ARG PHOSTT_BAKE_MODEL=0
RUN apt-get update && \
apt-get install -y --no-install-recommends ca-certificates && \
rm -rf /var/lib/apt/lists/*
COPY --from=builder /build/target/release/phostt /usr/local/bin/phostt
RUN mkdir -p /models && \
if [ "$PHOSTT_BAKE_MODEL" = "1" ]; then \
phostt download --model-dir /models; \
fi
# --- Runtime stage ---
FROM nvidia/cuda:12.6.3-cudnn-runtime-ubuntu22.04
ARG PHOSTT_BAKE_MODEL=0
RUN apt-get update && \
apt-get install -y --no-install-recommends ca-certificates curl && \
rm -rf /var/lib/apt/lists/*
# Copy binary AND ort's dynamic libraries (.so files copied by copy-dylibs)
COPY --from=builder /build/target/release/phostt /usr/local/bin/phostt
COPY --from=builder /build/target/release/libonnxruntime*.so* /usr/local/lib/
RUN ldconfig
RUN groupadd -r phostt && useradd -r -g phostt phostt && \
mkdir -p /home/phostt/.phostt/models && chown -R phostt:phostt /home/phostt
# Copy baked model files (only present when PHOSTT_BAKE_MODEL=1)
COPY --from=model-fetcher --chown=phostt:phostt /models/. /home/phostt/.phostt/models/
USER phostt
ENV RUST_LOG=phostt=info
ENV NVIDIA_VISIBLE_DEVICES=all
ENV LD_LIBRARY_PATH="/usr/local/lib:${LD_LIBRARY_PATH}"
EXPOSE 9876
HEALTHCHECK --interval=30s --timeout=5s --retries=3 \
CMD curl -f http://localhost:9876/health || exit 1
ENTRYPOINT ["phostt"]
CMD ["serve", "--port", "9876", "--host", "0.0.0.0", "--bind-all"]