diff --git a/.github/actions/build-image/action.yml b/.github/actions/build-image/action.yml
index ad2031dd7601..42b1d7b69f62 100644
--- a/.github/actions/build-image/action.yml
+++ b/.github/actions/build-image/action.yml
@@ -29,6 +29,46 @@ inputs:
   dockerfile-path:
     description: 'Path to the Dockerfile'
     required: true
+  arch-type:
+    description: 'Architecture type (e.g., x86, arm64)'
+    required: false
+    default: 'x86'
+  device-type:
+    description: 'Device type (e.g., gpu, cpu)'
+    required: false
+    default: 'gpu'
+  cuda-version:
+    description: 'CUDA version (e.g., cu129)'
+    required: false
+    default: ''
+  python-version:
+    description: 'Python version (e.g., py312)'
+    required: false
+    default: ''
+  os-version:
+    description: 'OS version (e.g., ubuntu24.04)'
+    required: false
+    default: ''
+  contributor:
+    description: 'Contributor name'
+    required: false
+    default: 'None'
+  customer-type:
+    description: 'Customer type (e.g., sagemaker, ec2, rayserve_ec2)'
+    required: false
+    default: ''
+  inference-toolkit-version:
+    description: 'Inference toolkit version (e.g., 2.0.25)'
+    required: false
+    default: ''
+  torchserve-version:
+    description: 'TorchServe version (e.g., 0.12.0)'
+    required: false
+    default: ''
+  transformers-version:
+    description: 'Transformers library version (e.g., 4.28.1)'
+    required: false
+    default: ''
 
 outputs:
   image-uri:
@@ -60,16 +100,90 @@ runs:
     - name: Build image
       shell: bash
       run: |
-        docker buildx build --progress plain \
-          --build-arg CACHE_REFRESH="$(date +"%Y-%m-%d")" \
-          --build-arg BASE_IMAGE="${{ inputs.base-image }}" \
-          --build-arg CONTAINER_TYPE="${{ inputs.container-type }}" \
-          --build-arg FRAMEWORK="${{ inputs.framework }}" \
-          --build-arg FRAMEWORK_VERSION="${{ inputs.framework-version }}" \
+        # Process label values
+        LABEL_FRAMEWORK=$(echo "${{ inputs.framework }}" | tr '_' '-')
+        LABEL_FRAMEWORK_VERSION=$(echo "${{ inputs.framework-version }}" | tr '.' '-')
+        LABEL_ARCH="${{ inputs.arch-type }}"
+        LABEL_PYTHON_VERSION=$(echo "${{ inputs.python-version }}" | tr '.' '-')
+        LABEL_OS_VERSION=$(echo "${{ inputs.os-version }}" | tr '.' '-')
+        LABEL_CONTRIBUTOR="${{ inputs.contributor }}"
+
+        # Build transformers label if version provided
+        LABEL_TRANSFORMERS=""
+        if [[ -n "${{ inputs.transformers-version }}" ]]; then
+          LABEL_TRANSFORMERS=$(echo "${{ inputs.transformers-version }}" | tr '.' '-')
+        fi
+
+        # Build inference toolkit label if both versions provided
+        LABEL_INFERENCE_TOOLKIT=""
+        if [[ -n "${{ inputs.inference-toolkit-version }}" && -n "${{ inputs.torchserve-version }}" ]]; then
+          TOOLKIT_VER=$(echo "${{ inputs.inference-toolkit-version }}" | tr '.' '-')
+          TORCHSERVE_VER=$(echo "${{ inputs.torchserve-version }}" | tr '.' '-')
+          LABEL_INFERENCE_TOOLKIT="${TOOLKIT_VER}.torchserve.${TORCHSERVE_VER}"
+        fi
+
+        # Construct device type label
+        LABEL_DEVICE_TYPE="${{ inputs.device-type }}"
+        if [[ "${{ inputs.device-type }}" == "gpu" && -n "${{ inputs.cuda-version }}" ]]; then
+          LABEL_DEVICE_TYPE="${{ inputs.device-type }}.${{ inputs.cuda-version }}"
+        fi
+
+        # Build base command
+        BUILD_CMD="docker buildx build --progress plain \
+          --build-arg CACHE_REFRESH=\"\$(date +\"%Y-%m-%d\")\" \
+          --build-arg BASE_IMAGE=\"${{ inputs.base-image }}\" \
+          --build-arg CONTAINER_TYPE=\"${{ inputs.container-type }}\" \
+          --build-arg FRAMEWORK=\"${{ inputs.framework }}\" \
+          --build-arg FRAMEWORK_VERSION=\"${{ inputs.framework-version }}\""
+
+        # Add SageMaker labels if customer-type is 'sagemaker'
+        if [[ "${{ inputs.customer-type }}" == "sagemaker" ]]; then
+          BUILD_CMD="${BUILD_CMD} \
+          --label \"com.amazonaws.ml.engines.sagemaker.dlc.arch.${LABEL_ARCH}=true\" \
+          --label \"com.amazonaws.ml.engines.sagemaker.dlc.device.${LABEL_DEVICE_TYPE}=true\" \
+          --label \"com.amazonaws.ml.engines.sagemaker.dlc.framework.${LABEL_FRAMEWORK}.${LABEL_FRAMEWORK_VERSION}=true\" \
+          --label \"com.amazonaws.ml.engines.sagemaker.dlc.job.${{ inputs.container-type }}=true\""
+
+          # Add OS version label if provided
+          if [[ -n "${{ inputs.os-version }}" ]]; then
+            BUILD_CMD="${BUILD_CMD} \
+          --label \"com.amazonaws.ml.engines.sagemaker.dlc.os.${LABEL_OS_VERSION}=true\""
+          fi
+
+          # Add Python version label if provided
+          if [[ -n "${{ inputs.python-version }}" ]]; then
+            BUILD_CMD="${BUILD_CMD} \
+          --label \"com.amazonaws.ml.engines.sagemaker.dlc.python.${LABEL_PYTHON_VERSION}=true\""
+          fi
+
+          # Add contributor label if provided
+          if [[ -n "${LABEL_CONTRIBUTOR}" ]]; then
+            BUILD_CMD="${BUILD_CMD} \
+          --label \"com.amazonaws.ml.engines.sagemaker.dlc.contributor.${LABEL_CONTRIBUTOR}=true\""
+          fi
+
+          # Add transformers library label if provided
+          if [[ -n "${LABEL_TRANSFORMERS}" ]]; then
+            BUILD_CMD="${BUILD_CMD} \
+          --label \"com.amazonaws.ml.engines.sagemaker.dlc.lib.transformers.${LABEL_TRANSFORMERS}=true\""
+          fi
+
+          # Add inference toolkit label if provided
+          if [[ -n "${LABEL_INFERENCE_TOOLKIT}" ]]; then
+            BUILD_CMD="${BUILD_CMD} \
+          --label \"com.amazonaws.ml.engines.sagemaker.dlc.inference-toolkit.${LABEL_INFERENCE_TOOLKIT}=true\""
+          fi
+        fi
+
+        # Complete the build command
+        BUILD_CMD="${BUILD_CMD} \
           --cache-to=type=inline \
           --cache-from=type=registry,ref=${CI_IMAGE_URI} \
           --tag ${CI_IMAGE_URI} \
           --push \
           --target ${{ inputs.target }} \
-          -f ${{ inputs.dockerfile-path }} .
+          -f ${{ inputs.dockerfile-path }} ."
+
+        # Execute build
+        eval ${BUILD_CMD}
         docker rmi ${CI_IMAGE_URI}
diff --git a/.github/actions/load-config/action.yml b/.github/actions/load-config/action.yml
new file mode 100644
index 000000000000..6c321d0b27e7
--- /dev/null
+++ b/.github/actions/load-config/action.yml
@@ -0,0 +1,51 @@
+name: 'Load Configuration Files'
+description: 'Load and parse YAML configuration files into JSON format'
+
+inputs:
+  config-files:
+    description: 'Comma-separated list of config file paths to load'
+    required: true
+
+outputs:
+  ec2-config:
+    description: 'EC2 configuration as JSON'
+    value: ${{ steps.load-configs.outputs.ec2-config }}
+  sagemaker-config:
+    description: 'SageMaker configuration as JSON'
+    value: ${{ steps.load-configs.outputs.sagemaker-config }}
+  rayserve-config:
+    description: 'RayServe configuration as JSON'
+    value: ${{ steps.load-configs.outputs.rayserve-config }}
+
+runs:
+  using: 'composite'
+  steps:
+    - name: Install yq
+      shell: bash
+      run: |
+        if ! command -v yq &> /dev/null; then
+          sudo wget -qO /usr/local/bin/yq https://github.com/mikefarah/yq/releases/latest/download/yq_linux_amd64
+          sudo chmod +x /usr/local/bin/yq
+        fi
+
+    - name: Load configuration files
+      id: load-configs
+      shell: bash
+      run: |
+        # Split comma-separated config files
+        IFS=',' read -ra CONFIG_FILES <<< "${{ inputs.config-files }}"
+
+        # Load each config file and output individually
+        for CONFIG_FILE in "${CONFIG_FILES[@]}"; do
+          # Extract config name from file path (e.g., "ec2" from "sglang-ec2.yml")
+          CONFIG_NAME=$(basename "$CONFIG_FILE" .yml | sed 's/.*-//')
+
+          echo "Loading config: $CONFIG_FILE as '${CONFIG_NAME}-config'"
+
+          # Output individual config
+          {
+            echo "${CONFIG_NAME}-config<<EOF"
+            yq eval -o=json '.' $CONFIG_FILE
+            echo "EOF"
+          } >> $GITHUB_OUTPUT
+        done
diff --git a/.github/config/sglang-ec2.yml b/.github/config/sglang-ec2.yml
new file mode 100644
index 000000000000..8162d97d4a73
--- /dev/null
+++ b/.github/config/sglang-ec2.yml
@@ -0,0 +1,30 @@
+# SGLang EC2 Image Configuration
+# This file contains all configuration for building, testing, and releasing the SGLang EC2 image
+
+# Image identification
+image:
+  name: "sglang-ec2"
+  description: "SGLang for EC2 instances"
+
+# Build configuration
+common:
+  framework: "sglang"
+  framework_version: "0.5.8"
+  job_type: "general"
+  python_version: "py312"
+  cuda_version: "cu129"
+  os_version: "ubuntu24.04"
+  customer_type: "ec2"
+  arch_type: "x86"
+  prod_image: "sglang:0.5-gpu-py312-ec2"
+  device_type: "gpu"
+  contributor: "None"
+
+# Release configuration
+release:
+  release: false
+  force_release: false
+  public_registry: true
+  private_registry: true
+  enable_soci: true
+  environment: gamma  # gamma or production
diff --git a/.github/config/sglang-sagemaker.yml b/.github/config/sglang-sagemaker.yml
new file mode 100644
index 000000000000..9430ec68db1a
--- /dev/null
+++ b/.github/config/sglang-sagemaker.yml
@@ -0,0 +1,30 @@
+# SGLang SageMaker Image Configuration
+# This file contains all configuration for building, testing, and releasing the SGLang SageMaker image
+
+# Image identification
+image:
+  name: "sglang-sagemaker"
+  description: "SGLang for SageMaker"
+
+# Build configuration
+common:
+  framework: "sglang"
+  framework_version: "0.5.8"
+  job_type: "general"
+  python_version: "py312"
+  cuda_version: "cu129"
+  os_version: "ubuntu24.04"
+  customer_type: "sagemaker"
+  arch_type: "x86"
+  prod_image: "sglang:0.5-gpu-py312"
+  device_type: "gpu"
+  contributor: "None"
+
+# Release configuration
+release:
+  release: false
+  force_release: false
+  public_registry: true
+  private_registry: true
+  enable_soci: true
+  environment: gamma  # gamma or production
diff --git a/.github/config/vllm-ec2.yml b/.github/config/vllm-ec2.yml
new file mode 100644
index 000000000000..5c94048f9487
--- /dev/null
+++ b/.github/config/vllm-ec2.yml
@@ -0,0 +1,30 @@
+# vLLM EC2 Image Configuration
+# This file contains all configuration for building, testing, and releasing the vLLM EC2 image
+
+# Image identification
+image:
+  name: "vllm-ec2"
+  description: "vLLM for EC2 instances"
+
+# Build configuration
+common:
+  framework: "vllm"
+  framework_version: "0.15.1"
+  job_type: "general"
+  python_version: "py312"
+  cuda_version: "cu129"
+  os_version: "ubuntu22.04"
+  customer_type: "ec2"
+  arch_type: "x86"
+  prod_image: "vllm:0.15-gpu-py312-ec2"
+  device_type: "gpu"
+  contributor: "None"
+
+# Release configuration
+release:
+  release: false
+  force_release: false
+  public_registry: true
+  private_registry: true
+  enable_soci: true
+  environment: gamma  # gamma or production
diff --git a/.github/config/vllm-0.10.2-rayserve.yml b/.github/config/vllm-rayserve.yml
similarity index 87%
rename from .github/config/vllm-0.10.2-rayserve.yml
rename to .github/config/vllm-rayserve.yml
index 49f09d785900..d806d20d7430 100644
--- a/.github/config/vllm-0.10.2-rayserve.yml
+++ b/.github/config/vllm-rayserve.yml
@@ -16,7 +16,7 @@ common:
   os_version: "ubuntu22.04"
   customer_type: "rayserve_ec2"
   arch_type: "x86"
-  prod_image: "vllm:0.10-gpu-py312-rayserve"
+  prod_image: "vllm:0.10-gpu-py312-rayserve-ec2"
   device_type: "gpu"
 
 # Release configuration
@@ -26,4 +26,4 @@ release:
   public_registry: true
   private_registry: false
   enable_soci: true
-  environment: production  # gamma or production
+  environment: gamma  # gamma or production
diff --git a/.github/config/vllm-sagemaker.yml b/.github/config/vllm-sagemaker.yml
new file mode 100644
index 000000000000..32710bd1dcb5
--- /dev/null
+++ b/.github/config/vllm-sagemaker.yml
@@ -0,0 +1,30 @@
+# vLLM SageMaker Image Configuration
+# This file contains all configuration for building, testing, and releasing the vLLM SageMaker image
+
+# Image identification
+image:
+  name: "vllm-sagemaker"
+  description: "vLLM for SageMaker"
+
+# Build configuration
+common:
+  framework: "vllm"
+  framework_version: "0.15.1"
+  job_type: "general"
+  python_version: "py312"
+  cuda_version: "cu129"
+  os_version: "ubuntu22.04"
+  customer_type: "sagemaker"
+  arch_type: "x86"
+  prod_image: "vllm:0.15-gpu-py312"
+  device_type: "gpu"
+  contributor: "None"
+
+# Release configuration
+release:
+  release: false
+  force_release: false
+  public_registry: true
+  private_registry: true
+  enable_soci: true
+  environment: gamma  # gamma or production
diff --git a/.github/workflows/pr-sglang.yml b/.github/workflows/pr-sglang.yml
index e364bcf24578..5474137e7834 100644
--- a/.github/workflows/pr-sglang.yml
+++ b/.github/workflows/pr-sglang.yml
@@ -13,24 +13,14 @@ permissions:
   pull-requests: read
 
 env:
-  # Common
-  CONTAINER_TYPE: "general"
-  FRAMEWORK: "sglang"
-  SGLANG_VERSION: "0.5.8"
-  PYTHON_VERSION: "py312"
-  CUDA_VERSION: "cu129"
-  OS_VERSION: "ubuntu24.04"
-
-  # SGLang EC2
-  PROD_EC2_IMAGE: sglang:0.5-gpu-py312-ec2
-
-  # SGLang SageMaker
-  PROD_SAGEMAKER_IMAGE: sglang:0.5-gpu-py312
-
   # CI environment configuration
   FORCE_COLOR: "1"
   TEST_ARTIFACTS_DIRECTORY: "/test_artifacts/sglang"
 
+  # Config file paths
+  EC2_CONFIG: ".github/config/sglang-ec2.yml"
+  SAGEMAKER_CONFIG: ".github/config/sglang-sagemaker.yml"
+
 jobs:
   gatekeeper:
     runs-on: ubuntu-latest
@@ -47,10 +37,27 @@ jobs:
       - name: Run permission gate (from base)
         uses: ./.github/actions/pr-permission-gate
 
-  check-changes:
+  load-config:
     needs: [gatekeeper]
     if: success()
     runs-on: ubuntu-latest
+    outputs:
+      ec2-config: ${{ steps.load.outputs.ec2-config }}
+      sagemaker-config: ${{ steps.load.outputs.sagemaker-config }}
+    steps:
+      - name: Checkout code
+        uses: actions/checkout@v5
+
+      - name: Load configurations
+        id: load
+        uses: ./.github/actions/load-config
+        with:
+          config-files: ${{ env.EC2_CONFIG }},${{ env.SAGEMAKER_CONFIG }}
+
+  check-changes:
+    needs: [load-config]
+    if: success()
+    runs-on: ubuntu-latest
     concurrency:
       group: ${{ github.workflow }}-check-changes-${{ github.event.pull_request.number }}
       cancel-in-progress: true
@@ -89,7 +96,7 @@ jobs:
   # =============== SGLang EC2 jobs ======================
   # ======================================================
   build-sglang-ec2-image:
-    needs: [check-changes]
+    needs: [check-changes, load-config]
     if: needs.check-changes.outputs.build-change == 'true'
     runs-on:
       - codebuild-runner-${{ github.run_id }}-${{ github.run_attempt }}
@@ -103,22 +110,44 @@ jobs:
     steps:
       - uses: actions/checkout@v5
 
+      - name: Parse EC2 config
+        id: config
+        run: |
+          echo '${{ needs.load-config.outputs.ec2-config }}' > config.json
+          echo "framework=$(jq -r '.common.framework' config.json)" >> $GITHUB_OUTPUT
+          echo "framework-version=$(jq -r '.common.framework_version' config.json)" >> $GITHUB_OUTPUT
+          echo "container-type=$(jq -r '.common.job_type' config.json)" >> $GITHUB_OUTPUT
+          echo "python-version=$(jq -r '.common.python_version' config.json)" >> $GITHUB_OUTPUT
+          echo "cuda-version=$(jq -r '.common.cuda_version' config.json)" >> $GITHUB_OUTPUT
+          echo "os-version=$(jq -r '.common.os_version' config.json)" >> $GITHUB_OUTPUT
+          echo "device-type=$(jq -r '.common.device_type // "gpu"' config.json)" >> $GITHUB_OUTPUT
+          echo "arch-type=$(jq -r '.common.arch_type // "x86"' config.json)" >> $GITHUB_OUTPUT
+          echo "contributor=$(jq -r '.common.contributor // "None"' config.json)" >> $GITHUB_OUTPUT
+          echo "customer-type=$(jq -r '.common.customer_type // ""' config.json)" >> $GITHUB_OUTPUT
+
       - name: Build image
         id: build
         uses: ./.github/actions/build-image
         with:
-          framework: ${{ env.FRAMEWORK }}
+          framework: ${{ steps.config.outputs.framework }}
           target: sglang-ec2
-          base-image: lmsysorg/sglang:v${{ env.SGLANG_VERSION }}-${{ env.CUDA_VERSION }}-amd64
-          framework-version: ${{ env.SGLANG_VERSION }}
-          container-type: ${{ env.CONTAINER_TYPE }}
+          base-image: lmsysorg/sglang:v${{ steps.config.outputs.framework-version }}-${{ steps.config.outputs.cuda-version }}-amd64
+          framework-version: ${{ steps.config.outputs.framework-version }}
+          container-type: ${{ steps.config.outputs.container-type }}
           aws-account-id: ${{ vars.CI_AWS_ACCOUNT_ID }}
           aws-region: ${{ vars.AWS_REGION }}
-          tag-pr: ${{ env.FRAMEWORK }}-${{ env.SGLANG_VERSION }}-gpu-${{ env.PYTHON_VERSION }}-${{ env.CUDA_VERSION }}-${{ env.OS_VERSION }}-ec2-pr-${{ github.event.pull_request.number }}
-          dockerfile-path: docker/${{ env.FRAMEWORK }}/Dockerfile
+          tag-pr: ${{ steps.config.outputs.framework }}-${{ steps.config.outputs.framework-version }}-gpu-${{ steps.config.outputs.python-version }}-${{ steps.config.outputs.cuda-version }}-${{ steps.config.outputs.os-version }}-ec2-pr-${{ github.event.pull_request.number }}
+          dockerfile-path: docker/${{ steps.config.outputs.framework }}/Dockerfile
+          arch-type: ${{ steps.config.outputs.arch-type }}
+          device-type: ${{ steps.config.outputs.device-type }}
+          cuda-version: ${{ steps.config.outputs.cuda-version }}
+          python-version: ${{ steps.config.outputs.python-version }}
+          os-version: ${{ steps.config.outputs.os-version }}
+          contributor: ${{ steps.config.outputs.contributor }}
+          customer-type: ${{ steps.config.outputs.customer-type }}
 
   set-ec2-test-environment:
-    needs: [check-changes, build-sglang-ec2-image]
+    needs: [check-changes, build-sglang-ec2-image, load-config]
     if: |
       always() && !failure() && !cancelled() &&
       (needs.check-changes.outputs.build-change == 'true' || needs.check-changes.outputs.test-change == 'true')
@@ -129,10 +158,18 @@ jobs:
     outputs:
       aws-account-id: ${{ steps.set-env.outputs.AWS_ACCOUNT_ID }}
       image-uri: ${{ steps.set-env.outputs.IMAGE_URI }}
+      framework-version: ${{ steps.config.outputs.framework-version }}
     steps:
       - name: Checkout code
         uses: actions/checkout@v4
 
+      - name: Parse EC2 config
+        id: config
+        run: |
+          echo '${{ needs.load-config.outputs.ec2-config }}' > config.json
+          echo "framework-version=$(jq -r '.common.framework_version' config.json)" >> $GITHUB_OUTPUT
+          echo "prod-image=$(jq -r '.common.prod_image' config.json)" >> $GITHUB_OUTPUT
+
       - name: Set test environment
         id: set-env
         run: |
@@ -141,7 +178,7 @@ jobs:
             IMAGE_URI=${{ needs.build-sglang-ec2-image.outputs.ci-image }}
           else
             AWS_ACCOUNT_ID=${{ vars.PROD_AWS_ACCOUNT_ID }}
-            IMAGE_URI=${{ vars.PROD_AWS_ACCOUNT_ID }}.dkr.ecr.${{ vars.AWS_REGION }}.amazonaws.com/${{ env.PROD_EC2_IMAGE }}
+            IMAGE_URI=${{ vars.PROD_AWS_ACCOUNT_ID }}.dkr.ecr.${{ vars.AWS_REGION }}.amazonaws.com/${{ steps.config.outputs.prod-image }}
           fi
 
           echo "Image URI to test: ${IMAGE_URI}"
@@ -231,7 +268,7 @@ jobs:
         uses: actions/checkout@v5
         with:
           repository: sgl-project/sglang
-          ref: v${{ env.SGLANG_VERSION }}
+          ref: v${{ needs.set-ec2-test-environment.outputs.framework-version }}
           path: sglang_source
 
       - name: Start container
@@ -269,7 +306,7 @@ jobs:
   # =============== SGLang SageMaker jobs ================
   # ======================================================
   build-sglang-sagemaker-image:
-    needs: [check-changes]
+    needs: [check-changes, load-config]
     if: needs.check-changes.outputs.build-change == 'true'
     runs-on:
       - codebuild-runner-${{ github.run_id }}-${{ github.run_attempt }}
@@ -283,22 +320,44 @@ jobs:
     steps:
       - uses: actions/checkout@v5
 
+      - name: Parse SageMaker config
+        id: config
+        run: |
+          echo '${{ needs.load-config.outputs.sagemaker-config }}' > config.json
+          echo "framework=$(jq -r '.common.framework' config.json)" >> $GITHUB_OUTPUT
+          echo "framework-version=$(jq -r '.common.framework_version' config.json)" >> $GITHUB_OUTPUT
+          echo "container-type=$(jq -r '.common.job_type' config.json)" >> $GITHUB_OUTPUT
+          echo "python-version=$(jq -r '.common.python_version' config.json)" >> $GITHUB_OUTPUT
+          echo "cuda-version=$(jq -r '.common.cuda_version' config.json)" >> $GITHUB_OUTPUT
+          echo "os-version=$(jq -r '.common.os_version' config.json)" >> $GITHUB_OUTPUT
+          echo "device-type=$(jq -r '.common.device_type // "gpu"' config.json)" >> $GITHUB_OUTPUT
+          echo "arch-type=$(jq -r '.common.arch_type // "x86"' config.json)" >> $GITHUB_OUTPUT
+          echo "contributor=$(jq -r '.common.contributor // "None"' config.json)" >> $GITHUB_OUTPUT
+          echo "customer-type=$(jq -r '.common.customer_type // ""' config.json)" >> $GITHUB_OUTPUT
+
       - name: Build image
         id: build
         uses: ./.github/actions/build-image
         with:
-          framework: ${{ env.FRAMEWORK }}
+          framework: ${{ steps.config.outputs.framework }}
           target: sglang-sagemaker
-          base-image: lmsysorg/sglang:v${{ env.SGLANG_VERSION }}-${{ env.CUDA_VERSION }}-amd64
-          framework-version: ${{ env.SGLANG_VERSION }}
-          container-type: ${{ env.CONTAINER_TYPE }}
+          base-image: lmsysorg/sglang:v${{ steps.config.outputs.framework-version }}-${{ steps.config.outputs.cuda-version }}-amd64
+          framework-version: ${{ steps.config.outputs.framework-version }}
+          container-type: ${{ steps.config.outputs.container-type }}
           aws-account-id: ${{ vars.CI_AWS_ACCOUNT_ID }}
           aws-region: ${{ vars.AWS_REGION }}
-          tag-pr: ${{ env.FRAMEWORK }}-${{ env.SGLANG_VERSION }}-gpu-${{ env.PYTHON_VERSION }}-${{ env.CUDA_VERSION }}-${{ env.OS_VERSION }}-sagemaker-pr-${{ github.event.pull_request.number }}
-          dockerfile-path: docker/${{ env.FRAMEWORK }}/Dockerfile
+          tag-pr: ${{ steps.config.outputs.framework }}-${{ steps.config.outputs.framework-version }}-gpu-${{ steps.config.outputs.python-version }}-${{ steps.config.outputs.cuda-version }}-${{ steps.config.outputs.os-version }}-sagemaker-pr-${{ github.event.pull_request.number }}
+          dockerfile-path: docker/${{ steps.config.outputs.framework }}/Dockerfile
+          arch-type: ${{ steps.config.outputs.arch-type }}
+          device-type: ${{ steps.config.outputs.device-type }}
+          cuda-version: ${{ steps.config.outputs.cuda-version }}
+          python-version: ${{ steps.config.outputs.python-version }}
+          os-version: ${{ steps.config.outputs.os-version }}
+          contributor: ${{ steps.config.outputs.contributor }}
+          customer-type: ${{ steps.config.outputs.customer-type }}
 
   set-sagemaker-test-environment:
-    needs: [check-changes, build-sglang-sagemaker-image]
+    needs: [check-changes, build-sglang-sagemaker-image, load-config]
     if: |
       always() && !failure() && !cancelled() &&
       (needs.check-changes.outputs.build-change == 'true' || needs.check-changes.outputs.test-change == 'true')
@@ -309,10 +368,18 @@ jobs:
     outputs:
       aws-account-id: ${{ steps.set-env.outputs.AWS_ACCOUNT_ID }}
       image-uri: ${{ steps.set-env.outputs.IMAGE_URI }}
+      framework-version: ${{ steps.config.outputs.framework-version }}
     steps:
       - name: Checkout code
         uses: actions/checkout@v4
 
+      - name: Parse SageMaker config
+        id: config
+        run: |
+          echo '${{ needs.load-config.outputs.sagemaker-config }}' > config.json
+          echo "framework-version=$(jq -r '.common.framework_version' config.json)" >> $GITHUB_OUTPUT
+          echo "prod-image=$(jq -r '.common.prod_image' config.json)" >> $GITHUB_OUTPUT
+
       - name: Set test environment
         id: set-env
         run: |
@@ -321,7 +388,7 @@ jobs:
             IMAGE_URI=${{ needs.build-sglang-sagemaker-image.outputs.ci-image }}
           else
             AWS_ACCOUNT_ID=${{ vars.PROD_AWS_ACCOUNT_ID }}
-            IMAGE_URI=${{ vars.PROD_AWS_ACCOUNT_ID }}.dkr.ecr.${{ vars.AWS_REGION }}.amazonaws.com/${{ env.PROD_SAGEMAKER_IMAGE }}
+            IMAGE_URI=${{ vars.PROD_AWS_ACCOUNT_ID }}.dkr.ecr.${{ vars.AWS_REGION }}.amazonaws.com/${{ steps.config.outputs.prod-image }}
           fi
 
           echo "Image URI to test: ${IMAGE_URI}"
@@ -411,7 +478,7 @@ jobs:
         uses: actions/checkout@v5
         with:
           repository: sgl-project/sglang
-          ref: v${{ env.SGLANG_VERSION }}
+          ref: v${{ needs.set-sagemaker-test-environment.outputs.framework-version }}
           path: sglang_source
 
       - name: Start container
diff --git a/.github/workflows/pr-vllm.yml b/.github/workflows/pr-vllm.yml
index fbf5aae2ea5d..48f9418afab7 100644
--- a/.github/workflows/pr-vllm.yml
+++ b/.github/workflows/pr-vllm.yml
@@ -13,27 +13,14 @@ permissions:
   pull-requests: read
 
 env:
-  # Common
-  CONTAINER_TYPE: "general"
-  FRAMEWORK: "vllm"
-  PYTHON_VERSION: "py312"
-  CUDA_VERSION: "cu129"
-  OS_VERSION: "ubuntu22.04"
-
-  # vLLM EC2
-  VLLM_VERSION: 0.15.1
-  PROD_EC2_IMAGE: vllm:0.15-gpu-py312-ec2
-
-  # vLLM SageMaker
-  PROD_SAGEMAKER_IMAGE: vllm:0.15-gpu-py312
-
-  # vLLM RayServe
-  VLLM_RAYSERVE_VERSION: 0.10.2
-  PROD_RAYSERVE_IMAGE: vllm:0.10-gpu-py312-rayserve
-
   # CI environment configuration
   FORCE_COLOR: "1"
 
+  # Config file paths
+  EC2_CONFIG: ".github/config/vllm-ec2.yml"
+  SAGEMAKER_CONFIG: ".github/config/vllm-sagemaker.yml"
+  RAYSERVE_CONFIG: ".github/config/vllm-rayserve.yml"
+
 
 jobs:
   gatekeeper:
@@ -51,10 +38,28 @@ jobs:
       - name: Run permission gate (from base)
         uses: ./.github/actions/pr-permission-gate
 
-  check-changes:
+  load-config:
     needs: [gatekeeper]
     if: success()
     runs-on: ubuntu-latest
+    outputs:
+      ec2-config: ${{ steps.load.outputs.ec2-config }}
+      sagemaker-config: ${{ steps.load.outputs.sagemaker-config }}
+      rayserve-config: ${{ steps.load.outputs.rayserve-config }}
+    steps:
+      - name: Checkout code
+        uses: actions/checkout@v5
+
+      - name: Load configurations
+        id: load
+        uses: ./.github/actions/load-config
+        with:
+          config-files: ${{ env.EC2_CONFIG }},${{ env.SAGEMAKER_CONFIG }},${{ env.RAYSERVE_CONFIG }}
+
+  check-changes:
+    needs: [load-config]
+    if: success()
+    runs-on: ubuntu-latest
     concurrency:
       group: ${{ github.workflow }}-check-changes-${{ github.event.pull_request.number }}
       cancel-in-progress: true
@@ -93,7 +98,7 @@ jobs:
   # =============== vLLM EC2 jobs ================
   # ==============================================
   build-vllm-ec2-image:
-    needs: [check-changes]
+    needs: [check-changes, load-config]
     if: needs.check-changes.outputs.build-change == 'true'
     runs-on:
       - codebuild-runner-${{ github.run_id }}-${{ github.run_attempt }}
@@ -107,22 +112,44 @@ jobs:
     steps:
       - uses: actions/checkout@v5
 
+      - name: Parse EC2 config
+        id: config
+        run: |
+          echo '${{ needs.load-config.outputs.ec2-config }}' > config.json
+          echo "framework=$(jq -r '.common.framework' config.json)" >> $GITHUB_OUTPUT
+          echo "framework-version=$(jq -r '.common.framework_version' config.json)" >> $GITHUB_OUTPUT
+          echo "container-type=$(jq -r '.common.job_type' config.json)" >> $GITHUB_OUTPUT
+          echo "python-version=$(jq -r '.common.python_version' config.json)" >> $GITHUB_OUTPUT
+          echo "cuda-version=$(jq -r '.common.cuda_version' config.json)" >> $GITHUB_OUTPUT
+          echo "os-version=$(jq -r '.common.os_version' config.json)" >> $GITHUB_OUTPUT
+          echo "device-type=$(jq -r '.common.device_type // "gpu"' config.json)" >> $GITHUB_OUTPUT
+          echo "arch-type=$(jq -r '.common.arch_type // "x86"' config.json)" >> $GITHUB_OUTPUT
+          echo "contributor=$(jq -r '.common.contributor // "None"' config.json)" >> $GITHUB_OUTPUT
+          echo "customer-type=$(jq -r '.common.customer_type // ""' config.json)" >> $GITHUB_OUTPUT
+
       - name: Build image
         id: build
         uses: ./.github/actions/build-image
         with:
-          framework: ${{ env.FRAMEWORK }}
+          framework: ${{ steps.config.outputs.framework }}
           target: vllm-ec2
-          base-image: vllm/vllm-openai:v${{ env.VLLM_VERSION }}
-          framework-version: ${{ env.VLLM_VERSION }}
-          container-type: ${{ env.CONTAINER_TYPE }}
+          base-image: vllm/vllm-openai:v${{ steps.config.outputs.framework-version }}
+          framework-version: ${{ steps.config.outputs.framework-version }}
+          container-type: ${{ steps.config.outputs.container-type }}
           aws-account-id: ${{ vars.CI_AWS_ACCOUNT_ID }}
           aws-region: ${{ vars.AWS_REGION }}
-          tag-pr: ${{ env.FRAMEWORK }}-${{ env.VLLM_VERSION }}-gpu-${{ env.PYTHON_VERSION }}-${{ env.CUDA_VERSION }}-${{ env.OS_VERSION }}-ec2-pr-${{ github.event.pull_request.number }}
-          dockerfile-path: docker/${{ env.FRAMEWORK }}/Dockerfile
+          tag-pr: ${{ steps.config.outputs.framework }}-${{ steps.config.outputs.framework-version }}-gpu-${{ steps.config.outputs.python-version }}-${{ steps.config.outputs.cuda-version }}-${{ steps.config.outputs.os-version }}-ec2-pr-${{ github.event.pull_request.number }}
+          dockerfile-path: docker/${{ steps.config.outputs.framework }}/Dockerfile
+          arch-type: ${{ steps.config.outputs.arch-type }}
+          device-type: ${{ steps.config.outputs.device-type }}
+          cuda-version: ${{ steps.config.outputs.cuda-version }}
+          python-version: ${{ steps.config.outputs.python-version }}
+          os-version: ${{ steps.config.outputs.os-version }}
+          contributor: ${{ steps.config.outputs.contributor }}
+          customer-type: ${{ steps.config.outputs.customer-type }}
 
   set-ec2-test-environment:
-    needs: [check-changes, build-vllm-ec2-image]
+    needs: [check-changes, build-vllm-ec2-image, load-config]
     if: |
       always() && !failure() && !cancelled() &&
       (needs.check-changes.outputs.build-change == 'true' || needs.check-changes.outputs.test-change == 'true')
@@ -133,10 +160,18 @@ jobs:
     outputs:
       aws-account-id: ${{ steps.set-env.outputs.AWS_ACCOUNT_ID }}
       image-uri: ${{ steps.set-env.outputs.IMAGE_URI }}
+      framework-version: ${{ steps.config.outputs.framework-version }}
     steps:
       - name: Checkout code
         uses: actions/checkout@v4
 
+      - name: Parse EC2 config
+        id: config
+        run: |
+          echo '${{ needs.load-config.outputs.ec2-config }}' > config.json
+          echo "framework-version=$(jq -r '.common.framework_version' config.json)" >> $GITHUB_OUTPUT
+          echo "prod-image=$(jq -r '.common.prod_image' config.json)" >> $GITHUB_OUTPUT
+
       - name: Set test environment
         id: set-env
         run: |
@@ -145,7 +180,7 @@ jobs:
             IMAGE_URI=${{ needs.build-vllm-ec2-image.outputs.ci-image }}
           else
             AWS_ACCOUNT_ID=${{ vars.PROD_AWS_ACCOUNT_ID }}
-            IMAGE_URI=${{ vars.PROD_AWS_ACCOUNT_ID }}.dkr.ecr.${{ vars.AWS_REGION }}.amazonaws.com/${{ env.PROD_EC2_IMAGE }}
+            IMAGE_URI=${{ vars.PROD_AWS_ACCOUNT_ID }}.dkr.ecr.${{ vars.AWS_REGION }}.amazonaws.com/${{ steps.config.outputs.prod-image }}
           fi
 
           echo "Image URI to test: ${IMAGE_URI}"
@@ -177,7 +212,7 @@ jobs:
         uses: actions/checkout@v5
         with:
           repository: vllm-project/vllm
-          ref: v${{ env.VLLM_VERSION }}
+          ref: v${{ needs.set-ec2-test-environment.outputs.framework-version }}
           path: vllm_source
 
       - name: Start container
@@ -223,7 +258,7 @@ jobs:
         uses: actions/checkout@v5
         with:
           repository: vllm-project/vllm
-          ref: v${{ env.VLLM_VERSION }}
+          ref: v${{ needs.set-ec2-test-environment.outputs.framework-version }}
           path: vllm_source
 
       - name: Start container
@@ -269,7 +304,7 @@ jobs:
         uses: actions/checkout@v5
         with:
           repository: vllm-project/vllm
-          ref: v${{ env.VLLM_VERSION }}
+          ref: v${{ needs.set-ec2-test-environment.outputs.framework-version }}
           path: vllm_source
 
       - name: Start container
@@ -290,11 +325,12 @@ jobs:
         run: |
           docker exec ${CONTAINER_ID} scripts/vllm/vllm_ec2_examples_test.sh
 
+
   # ===================================================
   # =============== vLLM RayServe jobs ================
   # ===================================================
   build-vllm-rayserve-image:
-    needs: [check-changes]
+    needs: [check-changes, load-config]
     if: needs.check-changes.outputs.build-change == 'true'
     runs-on:
       - codebuild-runner-${{ github.run_id }}-${{ github.run_attempt }}
@@ -308,22 +344,44 @@ jobs:
     steps:
       - uses: actions/checkout@v5
 
+      - name: Parse RayServe config
+        id: config
+        run: |
+          echo '${{ needs.load-config.outputs.rayserve-config }}' > config.json
+          echo "framework=$(jq -r '.common.framework' config.json)" >> $GITHUB_OUTPUT
+          echo "framework-version=$(jq -r '.common.framework_version' config.json)" >> $GITHUB_OUTPUT
+          echo "container-type=$(jq -r '.common.job_type' config.json)" >> $GITHUB_OUTPUT
+          echo "python-version=$(jq -r '.common.python_version' config.json)" >> $GITHUB_OUTPUT
+          echo "cuda-version=$(jq -r '.common.cuda_version' config.json)" >> $GITHUB_OUTPUT
+          echo "os-version=$(jq -r '.common.os_version' config.json)" >> $GITHUB_OUTPUT
+          echo "device-type=$(jq -r '.common.device_type // "gpu"' config.json)" >> $GITHUB_OUTPUT
+          echo "arch-type=$(jq -r '.common.arch_type // "x86"' config.json)" >> $GITHUB_OUTPUT
+          echo "contributor=$(jq -r '.common.contributor // "None"' config.json)" >> $GITHUB_OUTPUT
+          echo "customer-type=$(jq -r '.common.customer_type // ""' config.json)" >> $GITHUB_OUTPUT
+
       - name: Build image
         id: build
         uses: ./.github/actions/build-image
         with:
-          framework: ${{ env.FRAMEWORK }}
+          framework: ${{ steps.config.outputs.framework }}
           target: vllm-rayserve-ec2
-          base-image: vllm/vllm-openai:v${{ env.VLLM_RAYSERVE_VERSION }}
-          framework-version: ${{ env.VLLM_RAYSERVE_VERSION }}
-          container-type: ${{ env.CONTAINER_TYPE }}
+          base-image: vllm/vllm-openai:v${{ steps.config.outputs.framework-version }}
+          framework-version: ${{ steps.config.outputs.framework-version }}
+          container-type: ${{ steps.config.outputs.container-type }}
           aws-account-id: ${{ vars.CI_AWS_ACCOUNT_ID }}
           aws-region: ${{ vars.AWS_REGION }}
-          tag-pr: ${{ env.FRAMEWORK }}-${{ env.VLLM_VERSION }}-gpu-${{ env.PYTHON_VERSION }}-${{ env.CUDA_VERSION }}-${{ env.OS_VERSION }}-rayserve-ec2-pr-${{ github.event.pull_request.number }}
-          dockerfile-path: docker/${{ env.FRAMEWORK }}/Dockerfile
+          tag-pr: ${{ steps.config.outputs.framework }}-${{ steps.config.outputs.framework-version }}-gpu-${{ steps.config.outputs.python-version }}-${{ steps.config.outputs.cuda-version }}-${{ steps.config.outputs.os-version }}-rayserve-ec2-pr-${{ github.event.pull_request.number }}
+          dockerfile-path: docker/${{ steps.config.outputs.framework }}/Dockerfile
+          arch-type: ${{ steps.config.outputs.arch-type }}
+          device-type: ${{ steps.config.outputs.device-type }}
+          cuda-version: ${{ steps.config.outputs.cuda-version }}
+          python-version: ${{ steps.config.outputs.python-version }}
+          os-version: ${{ steps.config.outputs.os-version }}
+          contributor: ${{ steps.config.outputs.contributor }}
+          customer-type: ${{ steps.config.outputs.customer-type }}
 
   set-rayserve-test-environment:
-    needs: [check-changes, build-vllm-rayserve-image]
+    needs: [check-changes, build-vllm-rayserve-image, load-config]
     if: |
       always() && !failure() && !cancelled() &&
       (needs.check-changes.outputs.build-change == 'true' || needs.check-changes.outputs.test-change == 'true')
@@ -334,10 +392,18 @@ jobs:
     outputs:
       aws-account-id: ${{ steps.set-env.outputs.AWS_ACCOUNT_ID }}
       image-uri: ${{ steps.set-env.outputs.IMAGE_URI }}
+      framework-version: ${{ steps.config.outputs.framework-version }}
     steps:
       - name: Checkout code
         uses: actions/checkout@v4
 
+      - name: Parse RayServe config
+        id: config
+        run: |
+          echo '${{ needs.load-config.outputs.rayserve-config }}' > config.json
+          echo "framework-version=$(jq -r '.common.framework_version' config.json)" >> $GITHUB_OUTPUT
+          echo "prod-image=$(jq -r '.common.prod_image' config.json)" >> $GITHUB_OUTPUT
+
       - name: Set test environment
         id: set-env
         run: |
@@ -346,7 +412,7 @@ jobs:
             IMAGE_URI=${{ needs.build-vllm-rayserve-image.outputs.ci-image }}
           else
             AWS_ACCOUNT_ID=${{ vars.PROD_AWS_ACCOUNT_ID }}
-            IMAGE_URI=${{ vars.PROD_AWS_ACCOUNT_ID }}.dkr.ecr.${{ vars.AWS_REGION }}.amazonaws.com/${{ env.PROD_RAYSERVE_IMAGE }}
+            IMAGE_URI=${{ vars.PROD_AWS_ACCOUNT_ID }}.dkr.ecr.${{ vars.AWS_REGION }}.amazonaws.com/${{ steps.config.outputs.prod-image }}
           fi
 
           echo "Image URI to test: ${IMAGE_URI}"
@@ -378,7 +444,7 @@ jobs:
         uses: actions/checkout@v5
         with:
           repository: vllm-project/vllm
-          ref: v${{ env.VLLM_RAYSERVE_VERSION }}
+          ref: v${{ needs.set-rayserve-test-environment.outputs.framework-version }}
           path: vllm_source
 
       - name: Start container
@@ -424,7 +490,7 @@ jobs:
         uses: actions/checkout@v5
         with:
           repository: vllm-project/vllm
-          ref: v${{ env.VLLM_RAYSERVE_VERSION }}
+          ref: v${{ needs.set-rayserve-test-environment.outputs.framework-version }}
           path: vllm_source
 
       - name: Start container
@@ -470,7 +536,7 @@ jobs:
         uses: actions/checkout@v5
         with:
           repository: vllm-project/vllm
-          ref: v${{ env.VLLM_RAYSERVE_VERSION }}
+          ref: v${{ needs.set-rayserve-test-environment.outputs.framework-version }}
           path: vllm_source
 
       - name: Start container
@@ -495,7 +561,7 @@ jobs:
   # =============== vLLM SageMaker jobs ================
   # ====================================================
   build-vllm-sagemaker-image:
-    needs: [check-changes]
+    needs: [check-changes, load-config]
     if: needs.check-changes.outputs.build-change == 'true'
     runs-on:
       - codebuild-runner-${{ github.run_id }}-${{ github.run_attempt }}
@@ -509,22 +575,44 @@ jobs:
     steps:
       - uses: actions/checkout@v5
 
+      - name: Parse SageMaker config
+        id: config
+        run: |
+          echo '${{ needs.load-config.outputs.sagemaker-config }}' > config.json
+          echo "framework=$(jq -r '.common.framework' config.json)" >> $GITHUB_OUTPUT
+          echo "framework-version=$(jq -r '.common.framework_version' config.json)" >> $GITHUB_OUTPUT
+          echo "container-type=$(jq -r '.common.job_type' config.json)" >> $GITHUB_OUTPUT
+          echo "python-version=$(jq -r '.common.python_version' config.json)" >> $GITHUB_OUTPUT
+          echo "cuda-version=$(jq -r '.common.cuda_version' config.json)" >> $GITHUB_OUTPUT
+          echo "os-version=$(jq -r '.common.os_version' config.json)" >> $GITHUB_OUTPUT
+          echo "device-type=$(jq -r '.common.device_type // "gpu"' config.json)" >> $GITHUB_OUTPUT
+          echo "arch-type=$(jq -r '.common.arch_type // "x86"' config.json)" >> $GITHUB_OUTPUT
+          echo "contributor=$(jq -r '.common.contributor // "None"' config.json)" >> $GITHUB_OUTPUT
+          echo "customer-type=$(jq -r '.common.customer_type // ""' config.json)" >> $GITHUB_OUTPUT
+
       - name: Build image
         id: build
         uses: ./.github/actions/build-image
         with:
-          framework: ${{ env.FRAMEWORK }}
+          framework: ${{ steps.config.outputs.framework }}
           target: vllm-sagemaker
-          base-image: vllm/vllm-openai:v${{ env.VLLM_VERSION }}
-          framework-version: ${{ env.VLLM_VERSION }}
-          container-type: ${{ env.CONTAINER_TYPE }}
+          base-image: vllm/vllm-openai:v${{ steps.config.outputs.framework-version }}
+          framework-version: ${{ steps.config.outputs.framework-version }}
+          container-type: ${{ steps.config.outputs.container-type }}
           aws-account-id: ${{ vars.CI_AWS_ACCOUNT_ID }}
           aws-region: ${{ vars.AWS_REGION }}
-          tag-pr: ${{ env.FRAMEWORK }}-${{ env.VLLM_VERSION }}-gpu-${{ env.PYTHON_VERSION }}-${{ env.CUDA_VERSION }}-${{ env.OS_VERSION }}-sagemaker-pr-${{ github.event.pull_request.number }}
-          dockerfile-path: docker/${{ env.FRAMEWORK }}/Dockerfile
+          tag-pr: ${{ steps.config.outputs.framework }}-${{ steps.config.outputs.framework-version }}-gpu-${{ steps.config.outputs.python-version }}-${{ steps.config.outputs.cuda-version }}-${{ steps.config.outputs.os-version }}-sagemaker-pr-${{ github.event.pull_request.number }}
+          dockerfile-path: docker/${{ steps.config.outputs.framework }}/Dockerfile
+          arch-type: ${{ steps.config.outputs.arch-type }}
+          device-type: ${{ steps.config.outputs.device-type }}
+          cuda-version: ${{ steps.config.outputs.cuda-version }}
+          python-version: ${{ steps.config.outputs.python-version }}
+          os-version: ${{ steps.config.outputs.os-version }}
+          contributor: ${{ steps.config.outputs.contributor }}
+          customer-type: ${{ steps.config.outputs.customer-type }}
 
   set-sagemaker-test-environment:
-    needs: [check-changes, build-vllm-sagemaker-image]
+    needs: [check-changes, build-vllm-sagemaker-image, load-config]
     if: |
       always() && !failure() && !cancelled() &&
       (needs.check-changes.outputs.build-change == 'true' || needs.check-changes.outputs.test-change == 'true')
@@ -535,10 +623,18 @@ jobs:
     outputs:
       aws-account-id: ${{ steps.set-env.outputs.AWS_ACCOUNT_ID }}
       image-uri: ${{ steps.set-env.outputs.IMAGE_URI }}
+      framework-version: ${{ steps.config.outputs.framework-version }}
     steps:
       - name: Checkout code
         uses: actions/checkout@v4
 
+      - name: Parse SageMaker config
+        id: config
+        run: |
+          echo '${{ needs.load-config.outputs.sagemaker-config }}' > config.json
+          echo "framework-version=$(jq -r '.common.framework_version' config.json)" >> $GITHUB_OUTPUT
+          echo "prod-image=$(jq -r '.common.prod_image' config.json)" >> $GITHUB_OUTPUT
+
       - name: Set test environment
         id: set-env
         run: |
@@ -547,7 +643,7 @@ jobs:
             IMAGE_URI=${{ needs.build-vllm-sagemaker-image.outputs.ci-image }}
           else
             AWS_ACCOUNT_ID=${{ vars.PROD_AWS_ACCOUNT_ID }}
-            IMAGE_URI=${{ vars.PROD_AWS_ACCOUNT_ID }}.dkr.ecr.${{ vars.AWS_REGION }}.amazonaws.com/${{ env.PROD_SAGEMAKER_IMAGE }}
+            IMAGE_URI=${{ vars.PROD_AWS_ACCOUNT_ID }}.dkr.ecr.${{ vars.AWS_REGION }}.amazonaws.com/${{ steps.config.outputs.prod-image }}
           fi
 
           echo "Image URI to test: ${IMAGE_URI}"
@@ -579,7 +675,7 @@ jobs:
         uses: actions/checkout@v5
         with:
           repository: vllm-project/vllm
-          ref: v${{ env.VLLM_VERSION }}
+          ref: v${{ needs.set-sagemaker-test-environment.outputs.framework-version }}
           path: vllm_source
 
       - name: Start container
@@ -625,7 +721,7 @@ jobs:
         uses: actions/checkout@v5
         with:
           repository: vllm-project/vllm
-          ref: v${{ env.VLLM_VERSION }}
+          ref: v${{ needs.set-sagemaker-test-environment.outputs.framework-version }}
           path: vllm_source
 
       - name: Start container
@@ -671,7 +767,7 @@ jobs:
         uses: actions/checkout@v5
         with:
           repository: vllm-project/vllm
-          ref: v${{ env.VLLM_VERSION }}
+          ref: v${{ needs.set-sagemaker-test-environment.outputs.framework-version }}
           path: vllm_source
 
       - name: Start container
diff --git a/.github/workflows/vllm-rayserver-auto-release.yml b/.github/workflows/vllm-rayserver-auto-release.yml
index 12e1beb5849f..10473a7dac66 100644
--- a/.github/workflows/vllm-rayserver-auto-release.yml
+++ b/.github/workflows/vllm-rayserver-auto-release.yml
@@ -3,11 +3,18 @@ name: vLLM_RAYSERVE_AUTO_RELEASE
 on:
   # schedule run must be on the default branch
   # https://docs.github.com/en/actions/reference/workflows-and-actions/events-that-trigger-workflows#schedule
-  schedule:
-    # Runs at 10:00 AM PST/PDT on Monday and Wednesday
-    # Note: GitHub Actions uses UTC time. PST is UTC-8, PDT is UTC-7
-    # Using 17:00 UTC = 10:00 AM PDT (most of the year) / 9:00 AM PST (winter)
-    - cron: '00 17 * * 1,3' # Monday and Wednesday at 10:00 AM PDT / 9:00 AM PST
+  # schedule:
+  #   # Runs at 10:00 AM PST/PDT on Monday and Wednesday
+  #   # Note: GitHub Actions uses UTC time. PST is UTC-8, PDT is UTC-7
+  #   # Using 17:00 UTC = 10:00 AM PDT (most of the year) / 9:00 AM PST (winter)
+  #   - cron: '00 17 * * 1,3' # Monday and Wednesday at 10:00 AM PDT / 9:00 AM PST
+
+  pull_request:
+    branches: [main]
+    types: [opened, reopened, synchronize]
+    paths:
+      - "**vllm**"
+      - "!docs/**"
 
 permissions:
   contents: read
@@ -18,7 +25,7 @@ env:
   FORCE_COLOR: "1"
 
   # Config file paths
-  RAYSERVE_CONFIG: ".github/config/vllm-0.10.2-rayserve.yml"
+  RAYSERVE_CONFIG: ".github/config/vllm-rayserve.yml"
 
 
 jobs:
@@ -42,25 +49,16 @@ jobs:
     if: success()
     runs-on: ubuntu-latest
     outputs:
-      rayserve-config: ${{ steps.load-configs.outputs.rayserve-config }}
+      rayserve-config: ${{ steps.load.outputs.rayserve-config }}
     steps:
       - name: Checkout code
         uses: actions/checkout@v5
 
-      - name: Load configuration files
-        id: load-configs
-        run: |
-          # Install yq for YAML parsing
-          sudo wget -qO /usr/local/bin/yq https://github.com/mikefarah/yq/releases/latest/download/yq_linux_amd64
-          sudo chmod +x /usr/local/bin/yq
-
-          # Load and output configs as JSON for easy parsing in other jobs
-          # Using multiline output format to handle special characters
-          {
-            echo "rayserve-config<<EOF"
-            yq eval -o=json '.' ${{ env.RAYSERVE_CONFIG }}
-            echo "EOF"
-          } >> $GITHUB_OUTPUT
+      - name: Load configurations
+        id: load
+        uses: ./.github/actions/load-config
+        with:
+          config-files: ${{ env.RAYSERVE_CONFIG }}
 
   # ===================================================
   # =============== vLLM RayServe jobs ================
@@ -90,6 +88,9 @@ jobs:
           echo "cuda-version=$(jq -r '.common.cuda_version' config.json)" >> $GITHUB_OUTPUT
           echo "os-version=$(jq -r '.common.os_version' config.json)" >> $GITHUB_OUTPUT
           echo "device-type=$(jq -r '.common.device_type // "gpu"' config.json)" >> $GITHUB_OUTPUT
+          echo "arch-type=$(jq -r '.common.arch_type // "x86"' config.json)" >> $GITHUB_OUTPUT
+          echo "contributor=$(jq -r '.common.contributor // "None"' config.json)" >> $GITHUB_OUTPUT
+          echo "customer-type=$(jq -r '.common.customer_type // ""' config.json)" >> $GITHUB_OUTPUT
 
       - name: Build image
         id: build
@@ -104,6 +105,13 @@ jobs:
           aws-region: ${{ vars.AWS_REGION }}
           tag-pr: ${{ steps.config.outputs.framework }}-${{ steps.config.outputs.framework-version }}-${{ steps.config.outputs.device-type }}-${{ steps.config.outputs.python-version }}-${{ steps.config.outputs.cuda-version }}-${{ steps.config.outputs.os-version }}-rayserve-ec2-pr-${{ github.event.pull_request.number }}
           dockerfile-path: docker/${{ steps.config.outputs.framework }}/Dockerfile
+          arch-type: ${{ steps.config.outputs.arch-type }}
+          device-type: ${{ steps.config.outputs.device-type }}
+          cuda-version: ${{ steps.config.outputs.cuda-version }}
+          python-version: ${{ steps.config.outputs.python-version }}
+          os-version: ${{ steps.config.outputs.os-version }}
+          contributor: ${{ steps.config.outputs.contributor }}
+          customer-type: ${{ steps.config.outputs.customer-type }}
 
   set-rayserve-test-environment:
     needs: [build-vllm-rayserve-image, load-config]