# llama-cpp-turboquant → llama-server (CUDA 13).
# Build context: root of https://github.com/TheTom/llama-cpp-turboquant
#   docker build -f docker/server-cuda13.Dockerfile -t llama-turboquant:server-cuda13 ../llama-cpp-turboquant
#
# Для Blackwell (RTX 50xx, sm_120). Toolkit 13.1.x — под драйвер с nvidia-smi CUDA 13.1.
# Только 120a-real: быстрее сборка, FP4 tensor cores; не для карт старше Blackwell.

ARG UBUNTU_VERSION=24.04
ARG CUDA_VERSION=13.1.2
ARG GCC_VERSION=14
ARG BASE_CUDA_DEV_CONTAINER=docker.io/nvidia/cuda:${CUDA_VERSION}-devel-ubuntu${UBUNTU_VERSION}
ARG BASE_CUDA_RUN_CONTAINER=docker.io/nvidia/cuda:${CUDA_VERSION}-runtime-ubuntu${UBUNTU_VERSION}

ARG BUILD_DATE=N/A
ARG APP_VERSION=N/A
ARG APP_REVISION=N/A

ARG NODE_VERSION=24

FROM docker.io/node:$NODE_VERSION AS web

ARG APP_VERSION

WORKDIR /app/tools/ui

COPY tools/ui/package.json tools/ui/package-lock.json ./
RUN npm ci

COPY tools/ui/ ./
RUN LLAMA_BUILD_NUMBER="$APP_VERSION" npm run build

FROM ${BASE_CUDA_DEV_CONTAINER} AS build

ARG GCC_VERSION
# 120 → CMake подменит на 120a; явно 120a-real под Blackwell (не forwards-compatible).
ARG CUDA_DOCKER_ARCH=120a-real

RUN apt-get update && \
    apt-get install -y gcc-${GCC_VERSION} g++-${GCC_VERSION} build-essential cmake python3 python3-pip git libssl-dev libgomp1

ENV CC=gcc-${GCC_VERSION} CXX=g++-${GCC_VERSION} CUDAHOSTCXX=g++-${GCC_VERSION}

WORKDIR /app

COPY . .

COPY --from=web /app/tools/ui/dist tools/ui/dist

RUN if [ "${CUDA_DOCKER_ARCH}" != "default" ]; then \
    export CMAKE_ARGS="-DCMAKE_CUDA_ARCHITECTURES=${CUDA_DOCKER_ARCH}"; \
    fi && \
    cmake -B build -DGGML_NATIVE=OFF -DGGML_CUDA=ON -DGGML_BACKEND_DL=ON -DGGML_CPU_ALL_VARIANTS=ON -DLLAMA_BUILD_TESTS=OFF ${CMAKE_ARGS} -DCMAKE_EXE_LINKER_FLAGS=-Wl,--allow-shlib-undefined . && \
    cmake --build build --config Release -j"$(nproc)" --target llama-server

RUN mkdir -p /app/lib && \
    find build -name "*.so*" -exec cp -P {} /app/lib \;

RUN mkdir -p /app/full \
    && cp build/bin/llama-server /app/full/

FROM ${BASE_CUDA_RUN_CONTAINER} AS server

ARG BUILD_DATE=N/A
ARG APP_VERSION=N/A
ARG APP_REVISION=N/A
LABEL org.opencontainers.image.created=$BUILD_DATE \
      org.opencontainers.image.version=$APP_VERSION \
      org.opencontainers.image.revision=$APP_REVISION \
      org.opencontainers.image.title="llama-cpp-turboquant" \
      org.opencontainers.image.description="llama-server with TurboQuant KV (CUDA 13 / sm_120)" \
      org.opencontainers.image.source="https://github.com/TheTom/llama-cpp-turboquant"

RUN apt-get update \
    && apt-get install -y libgomp1 curl \
    && apt autoremove -y \
    && apt clean -y \
    && rm -rf /tmp/* /var/tmp/* \
    && find /var/cache/apt/archives /var/lib/apt/lists -not -name lock -type f -delete \
    && find /var/cache -type f -delete

ENV LLAMA_ARG_HOST=0.0.0.0

COPY --from=build /app/lib/ /app
COPY --from=build /app/full/llama-server /app/

WORKDIR /app

HEALTHCHECK CMD [ "curl", "-f", "http://localhost:8080/health" ]

ENTRYPOINT [ "/app/llama-server" ]