diff --git a/docker/Dockerfile b/docker/Dockerfile
index 0b79785c9..15e492504 100644
--- a/docker/Dockerfile
+++ b/docker/Dockerfile
@@ -523,7 +523,8 @@ RUN --mount=type=cache,target=/root/.cache/pip \
     pandas \
     matplotlib \
     tabulate \
-    termplotlib
+    termplotlib \
+    "runai-model-streamer[s3,gcs,azure]>=0.15.7"
 
 RUN --mount=type=cache,target=/root/.cache/pip \
    python3 -m pip install "nvidia-cutlass-dsl>=4.4.1" "nvidia-cutlass-dsl-libs-base>=4.4.1" --force-reinstall --no-deps;
@@ -615,15 +616,15 @@ RUN --mount=type=cache,target=/root/.cache/pip \
     && python3 -m pip install --no-deps -e "python[${BUILD_TYPE}]" \
     && kernels lock python \
     && ( success=0; for i in 1 2 3; do \
-        echo "Attempt $i/3: downloading sgl-kernel cubins..." && \
-        kernels download python && \
-        success=1 && break; \
-        echo "sgl-kernel cubin download failed, retrying in 30s..." && sleep 30; \
+    echo "Attempt $i/3: downloading sgl-kernel cubins..." && \
+    kernels download python && \
+    success=1 && break; \
+    echo "sgl-kernel cubin download failed, retrying in 30s..." && sleep 30; \
     done; [ "$success" = "1" ] ) \
-    && mv python/kernels.lock /root/.cache/sglang \
+    && mkdir -p /root/.cache/sglang \
+    && mv python/kernels.lock /root/.cache/sglang/ \
     && find /usr/local/lib/python3.12/dist-packages -type d -name "__pycache__" -exec rm -rf {} + 2>/dev/null || true \
-    && find /usr/local/lib/python3.12/dist-packages -type d -name "tests" -exec rm -rf {} + 2>/dev/null || true \
-    && find /usr/local/lib/python3.12/dist-packages -name "*.pyc" -delete 2>/dev/null || true
+
 
 # Install pre-built gateway artifacts from parallel builder
 COPY --from=gateway_builder /build/sgl-model-gateway-bin /usr/local/bin/sgl-model-gateway