docker/unified: add ik_llama.cpp to CUDA container (#620)

docker/unified/Dockerfile

@@ -12,7 +12,7 @@ ARG BACKEND=cuda
 
 # ── Builder bases ──────────────────────────────────────────────────────
 
-FROM nvidia/cuda:12.4.0-devel-ubuntu22.04 AS builder-base-cuda
+FROM nvidia/cuda:12.9.1-devel-ubuntu24.04 AS builder-base-cuda
 
 ENV DEBIAN_FRONTEND=noninteractive
 ENV CMAKE_CUDA_ARCHITECTURES="60;61;75;86;89"
@@ -29,7 +29,7 @@ WORKDIR /build
 
 # ──
 
-FROM ubuntu:26.04 AS builder-base-vulkan
+FROM ubuntu:24.04 AS builder-base-vulkan
 
 ENV DEBIAN_FRONTEND=noninteractive
 ENV CCACHE_DIR=/ccache
@@ -78,6 +78,27 @@ RUN --mount=type=cache,id=ccache-${BACKEND},target=/ccache \
     --mount=type=cache,id=llama-${BACKEND},target=/src/llama.cpp/build \
     BACKEND=${BACKEND} bash /build/install-llama.sh "${LLAMA_COMMIT_HASH}"
 
+# ── Build ik_llama.cpp (CUDA only) ────────────────────────────────────
+#
+# Two named stages allow ARG BACKEND to select at build time:
+#   - ik-llama-cuda  : real build (from builder-base-cuda)
+#   - ik-llama-vulkan: no-op (empty /install/bin, skips CUDA pull entirely)
+# BuildKit only evaluates the selected branch, so vulkan builds never
+# pull nvidia/cuda:*-devel or compile ik_llama.cpp.
+
+FROM builder-base-vulkan AS ik-llama-vulkan
+RUN mkdir -p /install/bin
+
+FROM builder-base-cuda AS ik-llama-cuda
+ARG IK_LLAMA_COMMIT_HASH=main
+COPY install-ik-llama.sh /build/
+RUN --mount=type=cache,id=ccache-cuda,target=/ccache \
+    --mount=type=cache,id=ik-llama-cuda,target=/src/ik_llama.cpp/build \
+    bash /build/install-ik-llama.sh "${IK_LLAMA_COMMIT_HASH}"
+
+ARG BACKEND=cuda
+FROM ik-llama-${BACKEND} AS ik-llama-build
+
 # ── Download llama-swap release binary ────────────────────────────────
 
 FROM builder-base AS llama-swap-download
@@ -87,14 +108,14 @@ RUN bash /build/install-llama-swap.sh "${LS_VERSION}"
 
 # ── Runtime bases ─────────────────────────────────────────────────────
 
-FROM nvidia/cuda:12.4.0-runtime-ubuntu22.04 AS runtime-cuda
+FROM nvidia/cuda:12.9.1-runtime-ubuntu24.04 AS runtime-cuda
 
 ENV DEBIAN_FRONTEND=noninteractive
 ENV LD_LIBRARY_PATH="/usr/local/cuda/lib64:${LD_LIBRARY_PATH}"
 ENV PATH="/usr/local/bin:${PATH}"
 
 RUN apt-get update && apt-get install -y --no-install-recommends \
-    libgomp1 python3 curl ca-certificates git \
+    libgomp1 python3 curl ca-certificates \
     && rm -rf /var/lib/apt/lists/*
 
 # CUDA stub drivers for container compatibility
@@ -103,14 +124,14 @@ COPY --from=builder-base-cuda /usr/local/cuda/lib64/stubs/libcuda.so /usr/local/
 
 # ──
 
-FROM ubuntu:26.04 AS runtime-vulkan
+FROM ubuntu:24.04 AS runtime-vulkan
 
 ENV DEBIAN_FRONTEND=noninteractive
 ENV PATH="/usr/local/bin:${PATH}"
 
 RUN apt-get update && apt-get install -y --no-install-recommends \
     libgomp1 libvulkan1 mesa-vulkan-drivers \
-    python3 curl ca-certificates git \
+    python3 curl ca-certificates \
     && rm -rf /var/lib/apt/lists/*
 
 # ── Select runtime base by BACKEND ────────────────────────────────────
@@ -121,6 +142,7 @@ ARG BACKEND=cuda
 ARG LLAMA_COMMIT_HASH=unknown
 ARG WHISPER_COMMIT_HASH=unknown
 ARG SD_COMMIT_HASH=unknown
+ARG IK_LLAMA_COMMIT_HASH=unknown
 
 RUN apt-get update && apt-get install -y --no-install-recommends \
     python3-numpy python3-sentencepiece \
@@ -147,6 +169,9 @@ COPY --from=sd-build /install/lib/ /usr/local/lib/
 COPY --from=llama-build /install/bin/llama-server /usr/local/bin/
 COPY --from=llama-build /install/bin/llama-cli /usr/local/bin/
 
+# Copy ik-llama-server (CUDA only; empty copy for vulkan)
+COPY --from=ik-llama-build /install/bin/ /usr/local/bin/
+
 # Copy llama-swap binary
 COPY --from=llama-swap-download /install/bin/llama-swap /usr/local/bin/
 COPY --from=llama-swap-download /install/llama-swap-version /tmp/
@@ -159,6 +184,7 @@ COPY config.example.yaml /etc/llama-swap/config/config.yaml
 RUN echo "llama.cpp: ${LLAMA_COMMIT_HASH}" > /versions.txt && \
     echo "whisper.cpp: ${WHISPER_COMMIT_HASH}" >> /versions.txt && \
     echo "stable-diffusion.cpp: ${SD_COMMIT_HASH}" >> /versions.txt && \
+    echo "ik_llama.cpp: ${IK_LLAMA_COMMIT_HASH}" >> /versions.txt && \
     echo "llama-swap: $(cat /tmp/llama-swap-version)" >> /versions.txt && \
     echo "backend: ${BACKEND}" >> /versions.txt && \
     echo "build_timestamp: $(date -u +%Y-%m-%dT%H:%M:%SZ)" >> /versions.txt