ci: collapse power-agent into one job and skip test collection on top level

.github/workflows/post-merge-ci.yml

@@ -141,27 +141,28 @@ jobs:
           azure_acr_password: ${{ secrets.AZURE_ACR_PASSWORD }}
           extra_tags: |
             ${{ secrets.AZURE_ACR_HOSTNAME }}/ai-dynamo/dynamo:${{ github.sha }}-power-agent
-
-  power-agent-test:
-    name: Power Agent  # group under the same UI umbrella as the build job
-    if: github.event_name != 'workflow_dispatch'
-    runs-on: ubuntu-latest
-    steps:
-      - name: Checkout code
-        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd  # v6.0.2
+      - name: Refresh BuildKit builder
+        # The runtime build above can leave the remote BuildKit connection stale;
+        # re-establish it (re-routing only if unhealthy) before the test build.
+        uses: ./.github/actions/builder-refresher
+        with:
+          builder_name: b-${{ github.run_id }}-${{ github.run_attempt }}
+          flavor: general
+          arch: linux/amd64
       - name: Run unit tests inside the container image
         # Build the Dockerfile `test` stage (FROM runtime): its RUN pytest runs
         # the suite against the exact shipped image (Python 3.12 + baked deps),
         # validating the container itself rather than a pip-installed runner env.
-        run: docker build --target test -f deploy/power-agent/Dockerfile deploy/power-agent
-
-  power-agent-helm-tests:
-    name: Power Agent  # group under the same UI umbrella as the build job
-    if: github.event_name != 'workflow_dispatch'
-    runs-on: ubuntu-latest
-    steps:
-      - name: Checkout code
-        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd  # v6.0.2
+        # Reuses the builder + ECR Docker Hub mirror the build step set up above
+        # (runtime layers are a cache hit); no --push/--load runs pytest without
+        # exporting the never-shipped test image.
+        env:
+          ECR_HOSTNAME: ${{ secrets.AWS_ACCOUNT_ID }}.dkr.ecr.${{ secrets.AWS_DEFAULT_REGION }}.amazonaws.com
+        run: |
+          docker buildx build --builder b-${{ github.run_id }}-${{ github.run_attempt }} \
+            --target test \
+            --build-arg DOCKER_PROXY=${ECR_HOSTNAME}/dockerhub/ \
+            -f deploy/power-agent/Dockerfile deploy/power-agent
       - name: Set up Helm
         uses: azure/setup-helm@b9e51907a09c216f16ebe8536097933489208112  # v4.3.0
         with:

.github/workflows/pr.yaml

@@ -61,8 +61,6 @@ jobs:
       - snapshot-placeholder-vllm
       - snapshot-placeholder-sglang
       - power-agent
-      - power-agent-test
-      - power-agent-helm-tests
       - vllm-build
       - vllm-dev-build
       - vllm-test
@@ -229,29 +227,28 @@ jobs:
           builder_name: ${{ needs.changed-files.outputs.builder_name }}
           aws_default_region: ${{ secrets.AWS_DEFAULT_REGION }}
           aws_account_id: ${{ secrets.AWS_ACCOUNT_ID }}
-
-  power-agent-test:
-    needs: changed-files
-    if: needs.changed-files.outputs.power_agent == 'true'
-    name: Power Agent  # group under the same UI umbrella as the build job
-    runs-on: ubuntu-latest
-    steps:
-      - name: Checkout code
-        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd  # v6.0.2
+      - name: Refresh BuildKit builder
+        # The runtime build above can leave the remote BuildKit connection stale;
+        # re-establish it (re-routing only if unhealthy) before the test build.
+        uses: ./.github/actions/builder-refresher
+        with:
+          builder_name: ${{ needs.changed-files.outputs.builder_name }}
+          flavor: general
+          arch: linux/amd64
       - name: Run unit tests inside the container image
         # Build the Dockerfile `test` stage (FROM runtime): its RUN pytest runs
         # the suite against the exact shipped image (Python 3.12 + baked deps),
         # validating the container itself rather than a pip-installed runner env.
-        run: docker build --target test -f deploy/power-agent/Dockerfile deploy/power-agent
-
-  power-agent-helm-tests:
-    needs: changed-files
-    if: needs.changed-files.outputs.power_agent == 'true'
-    name: Power Agent  # group under the same UI umbrella as the build job
-    runs-on: ubuntu-latest
-    steps:
-      - name: Checkout code
-        uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd  # v6.0.2
+        # Reuses the builder + ECR Docker Hub mirror the build step set up above
+        # (runtime layers are a cache hit); no --push/--load runs pytest without
+        # exporting the never-shipped test image.
+        env:
+          ECR_HOSTNAME: ${{ secrets.AWS_ACCOUNT_ID }}.dkr.ecr.${{ secrets.AWS_DEFAULT_REGION }}.amazonaws.com
+        run: |
+          docker buildx build --builder ${{ needs.changed-files.outputs.builder_name }} \
+            --target test \
+            --build-arg DOCKER_PROXY=${ECR_HOSTNAME}/dockerhub/ \
+            -f deploy/power-agent/Dockerfile deploy/power-agent
       - name: Set up Helm
         uses: azure/setup-helm@b9e51907a09c216f16ebe8536097933489208112  # v4.3.0
         with:

deploy/power-agent/pytest.ini

@@ -7,15 +7,11 @@
 # filterwarnings=error with TRTLLM/pydantic-specific filters (and extra plugins)
 # that are absent from the power-agent CI image, which installs only
 # pytest + pynvml + kubernetes + prometheus-client (see .github/workflows/pr.yaml
-# and post-merge-ci.yml power-agent-test job).
+# and post-merge-ci.yml power-agent job).
 [pytest]
 testpaths = tests
 # rootdir is deploy/power-agent (this file); "." puts it on sys.path so
 # `from power_agent import …` resolves both locally and inside the container
 # image's test stage (WORKDIR /app, where power_agent.py + this file live).
 pythonpath = .
 addopts = -ra --strict-markers --strict-config
-markers =
-    pre_merge: marks tests to run before merging
-    gpu_0: marks tests that don't require GPU
-    unit: marks tests as unit tests

deploy/power-agent/tests/test_annotation_scope.py

@@ -18,11 +18,8 @@
 from unittest.mock import MagicMock, patch
 
 import power_agent
-import pytest
 from power_agent import POWER_ANNOTATION_KEY, PowerAgent
 
-pytestmark = [pytest.mark.pre_merge, pytest.mark.gpu_0, pytest.mark.unit]
-
 SAFE_DEFAULT = 500
 
 

deploy/power-agent/tests/test_reconcile_failsafe.py

@@ -19,11 +19,8 @@
 import unittest
 from unittest.mock import MagicMock
 
-import pytest
 from power_agent import PowerAgent
 
-pytestmark = [pytest.mark.pre_merge, pytest.mark.gpu_0, pytest.mark.unit]
-
 
 def _make_agent(core_v1, device_count: int = 2) -> PowerAgent:
     """Build a PowerAgent without touching NVML / K8s (bypass __init__)."""

pyproject.toml

@@ -182,6 +182,10 @@ addopts = [
     "--ignore-glob=components/src/dynamo/sglang/multimodal_utils/*",
     "--ignore-glob=components/src/dynamo/vllm/multimodal_utils/*",
     "--ignore-glob=examples/backends/sglang/slurm_jobs/*",
+    # power-agent is a self-contained deployable with its own pytest.ini; its
+    # `tests` package is tested inside its container, and collecting it here
+    # collides with the repo-root `tests` package.
+    "--ignore-glob=deploy/power-agent/*",
     # FIXME: Get relative/generic blob paths to work here
 ]
 xfail_strict = true