Merge branch 'main' into ehu/cve-bump-deps

Author: leseb

.github/actions/launch-gpu-runner/action.yml

@@ -0,0 +1,82 @@
+name: 'Launch GPU EC2 Runner'
+description: 'Launch GPU-enabled EC2 instance as GitHub Actions self-hosted runner (wrapper for machulav/ec2-github-runner)'
+
+on:
+  workflow_dispatch:
+inputs:
+  mode:
+    description: 'Mode: start or stop'
+    required: true
+  github-token:
+    description: 'GitHub Personal Access Token with repo scope for runner registration'
+    required: true
+  instance-type:
+    description: 'EC2 instance type (e.g., g6.2xlarge, g5.2xlarge, g6.8xlarge)'
+    required: false
+    default: 'g6.2xlarge'
+  aws-region:
+    description: 'AWS region'
+    required: true
+  availability-zones-config:
+    description: 'JSON array of AZ configs with imageId, subnetId, securityGroupId for fallback'
+    required: false
+    default: ''
+  # For stop mode
+  label:
+    description: 'Runner label (for stop mode)'
+    required: false
+  ec2-instance-id:
+    description: 'EC2 instance ID (for stop mode)'
+    required: false
+  # Optional
+  ec2-instance-tags:
+    description: 'JSON array of tags to apply to EC2 instance'
+    required: false
+    default: '[]'
+  runner-home-dir:
+    description: 'Home directory for the runner'
+    required: false
+    default: ''
+  iam-role-name:
+    description: 'IAM role name to attach to the instance (optional, for enhanced security)'
+    required: false
+    default: ''
+
+outputs:
+  label:
+    description: 'Unique label for the launched runner'
+    value: ${{ steps.ec2-runner.outputs.label }}
+  ec2-instance-id:
+    description: 'EC2 instance ID'
+    value: ${{ steps.ec2-runner.outputs.ec2-instance-id }}
+
+runs:
+  using: 'composite'
+  steps:
+    - name: Start EC2 runner
+      if: inputs.mode == 'start'
+      id: ec2-runner
+      uses: machulav/ec2-github-runner@343a1b2ae682e681c3cec9a235d882da17ff04ef # v2.6.1
+      env:
+        AWS_DEFAULT_REGION: ${{ inputs.aws-region }}
+        AWS_REGION: ${{ inputs.aws-region }}
+      with:
+        mode: start
+        github-token: ${{ inputs.github-token }}
+        ec2-instance-type: ${{ inputs.instance-type }}
+        aws-resource-tags: ${{ inputs.ec2-instance-tags }}
+        runner-home-dir: ${{ inputs.runner-home-dir }}
+        iam-role-name: ${{ inputs.iam-role-name }}
+        availability-zones-config: ${{ inputs.availability-zones-config }}
+
+    - name: Stop EC2 runner
+      if: inputs.mode == 'stop'
+      uses: machulav/ec2-github-runner@343a1b2ae682e681c3cec9a235d882da17ff04ef # v2.6.1
+      env:
+        AWS_DEFAULT_REGION: ${{ inputs.aws-region }}
+        AWS_REGION: ${{ inputs.aws-region }}
+      with:
+        mode: stop
+        github-token: ${{ inputs.github-token }}
+        label: ${{ inputs.label }}
+        ec2-instance-id: ${{ inputs.ec2-instance-id }}

.github/actions/run-and-record-tests/action.yml

@@ -62,30 +62,35 @@ runs:
         INPUT_PATTERN: ${{ inputs.pattern }}
         INPUT_TEXT_MODEL: ${{ inputs.text-model }}
       run: |
-        SCRIPT_ARGS="--stack-config ${INPUT_STACK_CONFIG} --inference-mode ${INPUT_INFERENCE_MODE}"
+        SCRIPT_ARGS=(
+          --stack-config "${INPUT_STACK_CONFIG}"
+          --inference-mode "${INPUT_INFERENCE_MODE}"
+        )
 
         # Add optional arguments only if they are provided
         if [ -n "${INPUT_SETUP}" ]; then
-          SCRIPT_ARGS="${SCRIPT_ARGS} --setup ${INPUT_SETUP}"
+          SCRIPT_ARGS+=(--setup "${INPUT_SETUP}")
         fi
         if [ -n "${INPUT_SUITE}" ]; then
-          SCRIPT_ARGS="${SCRIPT_ARGS} --suite ${INPUT_SUITE}"
+          SCRIPT_ARGS+=(--suite "${INPUT_SUITE}")
         fi
         if [ -n "${INPUT_SUBDIRS}" ]; then
-          SCRIPT_ARGS="${SCRIPT_ARGS} --subdirs ${INPUT_SUBDIRS}"
+          SCRIPT_ARGS+=(--subdirs "${INPUT_SUBDIRS}")
         fi
         if [ -n "${INPUT_PATTERN}" ]; then
-          SCRIPT_ARGS="${SCRIPT_ARGS} --pattern ${INPUT_PATTERN}"
+          SCRIPT_ARGS+=(--pattern "${INPUT_PATTERN}")
         fi
         if [ -n "${INPUT_TEXT_MODEL}" ]; then
-          SCRIPT_ARGS="${SCRIPT_ARGS} --text-model ${INPUT_TEXT_MODEL}"
+          SCRIPT_ARGS+=(--text-model "${INPUT_TEXT_MODEL}")
         fi
 
         echo "=== Running command ==="
-        echo "uv run --no-sync ./scripts/integration-tests.sh $SCRIPT_ARGS"
+        printf 'uv run --no-sync ./scripts/integration-tests.sh'
+        printf ' %q' "${SCRIPT_ARGS[@]}"
+        printf '\n'
         echo ""
 
-        uv run --no-sync ./scripts/integration-tests.sh $SCRIPT_ARGS | tee "pytest-${INPUT_INFERENCE_MODE}.log"
+        uv run --no-sync ./scripts/integration-tests.sh "${SCRIPT_ARGS[@]}" | tee "pytest-${INPUT_INFERENCE_MODE}.log"
 
 
     - name: Commit and push recordings

.github/actions/setup-test-environment/action.yml

@@ -27,6 +27,10 @@ inputs:
     description: 'Explicit branch override (used by scheduled CI to pass the matrix branch)'
     required: false
     default: ''
+  enable-hf-cache:
+    description: 'Whether to cache HuggingFace models and datasets'
+    required: false
+    default: 'true'
 
 runs:
   using: 'composite'
@@ -40,6 +44,7 @@ runs:
         branch: ${{ inputs.branch }}
 
     - name: Cache HuggingFace models and datasets
+      if: ${{ inputs.enable-hf-cache == 'true' }}
       uses: actions/cache@v4
       with:
         path: ~/.cache/huggingface
@@ -48,6 +53,7 @@ runs:
           hf-cache-${{ runner.os }}-
 
     - name: Pre-download HuggingFace assets for offline use
+      if: ${{ inputs.enable-hf-cache == 'true' }}
       shell: bash
       env:
         HF_HUB_ENABLE_HF_TRANSFER: "1"
@@ -74,7 +80,7 @@ runs:
         setup: ${{ inputs.setup }}
 
     - name: Setup vllm
-      if: ${{ startsWith(inputs.setup, 'vllm') && inputs.inference-mode != 'replay' }}
+      if: ${{ startsWith(inputs.setup, 'vllm') && !startsWith(inputs.setup, 'vllm-gpu-') && inputs.inference-mode != 'replay' }}
       uses: ./.github/actions/setup-vllm
 
     - name: Start Postgres service