v1 API on real Ray Serve (v0.6.0) (#46) #104

Workflow file for this run

.github/workflows/test_back.yml at 04c973b

	# =================================
	# Copyright: CEA-LIST/DIASI/SIALV
	# Author : pixano@cea.fr
	# License: CECILL-C
	# =================================

	name: Test Backend

	on:
	push:
	paths:
	- "src/**"
	- "packages/**"
	- "tests/**"
	- ".github/workflows/test_back.yml"
	- "pyproject.toml"
	- ".pre-commit-config.yaml"
	branches:
	- "main"
	pull_request:
	paths:
	- "src/**"
	- "packages/**"
	- "tests/**"
	- ".github/workflows/test_back.yml"
	- "pyproject.toml"
	- ".pre-commit-config.yaml"
	branches:
	- "main"

	permissions:
	contents: read

	jobs:
	python_test:
	name: Backend
	runs-on: ubuntu-latest
	strategy:
	fail-fast: false
	matrix:
	python-version: ["3.10", "3.11", "3.12", "3.13"]
	steps:
	- name: Checkout repository
	uses: actions/checkout@v4

	- name: Set up Python ${{ matrix.python-version }}
	uses: actions/setup-python@v5
	with:
	python-version: ${{ matrix.python-version }}
	cache: "pip"

	# Install the CPU torch build so the model code paths (sam2/transformers) are
	# actually exercised instead of silently skipped via importorskip.
	- name: Install dependencies
	run: \|
	python -m pip install --upgrade pip
	# Install torch AND torchvision from the same CPU index so their compiled ops stay
	# ABI-coherent (open_clip / transformers pull torchvision; a PyPI torchvision against
	# a CPU-index torch breaks with "operator torchvision::nms does not exist").
	python -m pip install torch torchvision --index-url https://download.pytorch.org/whl/cpu
	# The client is a hard core dep not yet on PyPI, so install it editable BEFORE `pip install .`.
	python -m pip install -e packages/pixano-inference-client
	python -m pip install .[test,transformers]
	python -m pip install -e examples/numpy_detector # example entry-point plugin
	python -m pip install -e packages/pixano-inference-sam # SAM2 plugin (fake predictor in tests; sam-2 lib not needed)
	python -m pip install -e packages/pixano-inference-clip # CLIP embedding plugin (open_clip only needed for the integration test)

	# Fast unit tests (the Ray Serve integration suite runs in its own job below).
	- name: Test with pytest
	run: \|
	pytest --cov=pixano_inference --cov-report=xml -m "not integration" tests/
	- name: Upload coverage reports to Codecov with GitHub Action
	uses: codecov/codecov-action@v4.2.0
	env:
	CODECOV_TOKEN: ${{ secrets.CODECOV_TOKEN }}

	integration:
	name: Ray Serve integration (CPU)
	runs-on: ubuntu-latest
	steps:
	- name: Checkout repository
	uses: actions/checkout@v4

	- name: Set up Python
	uses: actions/setup-python@v5
	with:
	python-version: "3.12"
	cache: "pip"

	# Integration tests boot a real local Ray + Serve runtime with a numpy-only stub
	# model (no torch, no GPU) and exercise deploy -> RUNNING -> predict -> undeploy.
	- name: Install dependencies
	run: \|
	python -m pip install --upgrade pip
	# torch + torchvision from the same index (open_clip needs matching torchvision ops).
	python -m pip install torch torchvision --index-url https://download.pytorch.org/whl/cpu
	python -m pip install -e packages/pixano-inference-client # hard core dep, not yet on PyPI
	python -m pip install .[test]
	python -m pip install -e examples/numpy_detector # example entry-point plugin
	python -m pip install -e packages/pixano-inference-clip # CLIP plugin: real MobileCLIP2 deploy+embed test

	- name: Run integration tests
	run: \|
	pytest -m integration tests/integration/

	docker_smoke:
	name: Docker build & smoke (CPU, framework-free)
	runs-on: ubuntu-latest
	steps:
	- name: Checkout repository
	uses: actions/checkout@v4

	# Build a minimal, framework-free image (no torch, no SAM) with the numpy example
	# plugin, so the Dockerfile mechanics + entrypoint + healthcheck + Serve deploy path
	# are validated fast, without pulling multi-GB torch/sam-2.
	- name: Build minimal image
	run: \|
	docker build \
	--build-arg TORCH_INDEX_URL= \
	--build-arg PIXANO_EXTRAS= \
	--build-arg INSTALL_SAM=false \
	--build-arg INSTALL_EXAMPLE=true \
	-t pixano-inference:smoke .

	- name: Run container and smoke-test /v1/ready
	run: \|
	docker run -d --name pi -p 7463:7463 --shm-size=2g \
	-v "$PWD/docker/models.numpy.py:/config/models.py:ro" \
	pixano-inference:smoke --host 0.0.0.0 --port 7463 --config /config/models.py
	for i in $(seq 1 40); do
	if curl -fsS http://localhost:7463/health >/dev/null 2>&1; then echo "healthy"; break; fi
	sleep 3
	done
	curl -fsS http://localhost:7463/v1/ready \
	\| python3 -c "import sys, json; d = json.load(sys.stdin); print(d); sys.exit(0 if d['ready'] else 1)" \
	\|\| (docker logs pi; exit 1)
	docker rm -f pi

	framework_free_core:
	name: Framework-free core
	runs-on: ubuntu-latest
	strategy:
	fail-fast: false
	matrix:
	python-version: ["3.10", "3.13"]
	steps:
	- name: Checkout repository
	uses: actions/checkout@v4

	- name: Set up Python ${{ matrix.python-version }}
	uses: actions/setup-python@v5
	with:
	python-version: ${{ matrix.python-version }}
	cache: "pip"

	# Deliberately install NO ML framework (no torch/jax/tensorflow/mlx). The core
	# must import and the guardrail tests must pass with numpy only.
	- name: Install core only
	run: \|
	python -m pip install --upgrade pip
	python -m pip install -e packages/pixano-inference-client # hard core dep, not yet on PyPI
	python -m pip install .[test]

	- name: Assert core imports pull no ML framework
	run: \|
	pytest tests/test_core_framework_free.py -v

	# A framework-free custom model, installed as a plugin package and discovered via its
	# entry point — the recommended extension path, exercised here with no ML framework.
	- name: Install and discover the example plugin (no ML framework)
	run: \|
	python -m pip install -e examples/numpy_detector
	pytest tests/test_plugins.py -v

	openapi:
	name: OpenAPI schema in sync
	runs-on: ubuntu-latest
	steps:
	- name: Checkout repository
	uses: actions/checkout@v4

	# Generate the schema in the LOCKED environment (uv.lock) so the check is deterministic:
	# a fresh `pip install` resolves newer FastAPI/pydantic that emit a subtly different schema
	# than the committed docs/openapi.json (which is generated against the lockfile).
	- name: Install uv
	uses: astral-sh/setup-uv@v5

	- name: Set up Python
	uses: actions/setup-python@v5
	with:
	python-version: "3.12"

	- name: Check the committed OpenAPI schema is in sync
	run: \|
	uv sync --no-default-groups
	uv run python scripts/gen_openapi.py --check

	client_standalone:
	name: Standalone client (no server deps)
	runs-on: ubuntu-latest
	strategy:
	fail-fast: false
	matrix:
	python-version: ["3.10", "3.13"]
	steps:
	- name: Checkout repository
	uses: actions/checkout@v4

	- name: Set up Python ${{ matrix.python-version }}
	uses: actions/setup-python@v5
	with:
	python-version: ${{ matrix.python-version }}
	cache: "pip"

	# Install ONLY the standalone client (no core, no ray/fastapi/torch). Its tests assert the
	# import pulls in no heavy modules — the acceptance proof that third-party callers get a
	# small install. Also confirm the whole server stack is genuinely absent.
	- name: Install the client only
	run: \|
	python -m pip install --upgrade pip
	python -m pip install -e "packages/pixano-inference-client[masks]"
	python -m pip install pytest

	- name: Assert the server stack is not installed
	run: \|
	python - <<'PY'
	import importlib.util
	for mod in ["ray", "fastapi", "uvicorn", "torch", "pixano_inference"]:
	assert importlib.util.find_spec(mod) is None, f"{mod} should not be installed"
	import pixano_inference_client
	print("standalone client OK:", pixano_inference_client.PixanoInferenceClient)
	PY

	- name: Run the client package tests
	run: \|
	pytest packages/pixano-inference-client/tests -v

Navigation Menu

Uh oh!

v1 API on real Ray Serve (v0.6.0) (#46) #104

Workflow file

v1 API on real Ray Serve (v0.6.0) (#46) #104

Uh oh!

Workflow file for this run