fix(tests): close four suite-hygiene loose ends, and say why CI was green on the failing one #2482

Workflow file for this run

	name: CI

	on:
	push:
	branches: [main]
	pull_request:
	branches: [main]

	jobs:
	changes:
	runs-on: ubuntu-latest
	outputs:
	brain_bar: ${{ steps.filter.outputs.brain_bar }}
	launchd: ${{ steps.filter.outputs.launchd }}
	steps:
	- uses: actions/checkout@v4

	- id: filter
	uses: dorny/paths-filter@v3
	with:
	filters: \|
	brain_bar:
	- 'brain-bar/**'
	launchd:
	- 'scripts/launchd/**'
	- 'src/brainlayer/pause.py'
	- 'tests/test_fleet_watchdog.py'
	- '.github/workflows/ci.yml'

	swift:
	name: swift (macos-15)
	needs: changes
	runs-on: macos-15
	steps:
	- uses: actions/checkout@v4

	- name: Select Xcode (Swift 6)
	if: needs.changes.outputs.brain_bar == 'true'
	uses: maxim-lobanov/setup-xcode@v1
	with:
	xcode-version: latest-stable

	- name: Swift version
	run: swift --version

	- name: Cache SwiftPM
	if: needs.changes.outputs.brain_bar == 'true'
	uses: actions/cache@v4
	with:
	path: \|
	~/.swiftpm
	brain-bar/.build
	key: ${{ runner.os }}-swiftpm-${{ hashFiles('brain-bar/Package.resolved') }}
	restore-keys: ${{ runner.os }}-swiftpm-

	- name: Build BrainBar
	if: needs.changes.outputs.brain_bar == 'true'
	run: cd brain-bar && swift build -c release

	- name: Test BrainBar
	if: needs.changes.outputs.brain_bar == 'true'
	run: cd brain-bar && swift test

	- name: Skip BrainBar Swift build
	if: needs.changes.outputs.brain_bar != 'true'
	run: echo "no brain-bar changes — swift build/test skipped"

	launchd:
	# fleet-watchdog.sh reads the pause sentinel with plutil and BSD `date -j`, so those drills
	# skip themselves on the ubuntu matrix. This is the runner that actually executes them;
	# tests/test_fleet_watchdog.py asserts this job exists and runs that file.
	name: launchd (macos-15)
	needs: changes
	if: needs.changes.outputs.launchd == 'true'
	runs-on: macos-15
	steps:
	- uses: actions/checkout@v4

	- uses: actions/setup-python@v5
	with:
	python-version: "3.13"

	# The drills need pytest, pyyaml and stdlib-only brainlayer.pause off PYTHONPATH -- not
	# the full dev extra, which would pull torch onto a macOS runner for no benefit.
	- name: Install test deps
	run: pip install pytest pyyaml

	- name: Fleet-watchdog and pause-sentinel drills
	env:
	PYTHONPATH: src
	run: pytest tests/test_fleet_watchdog.py -v --tb=short

	test:
	runs-on: ubuntu-latest
	strategy:
	matrix:
	python-version: ["3.11", "3.12", "3.13"]
	steps:
	- uses: actions/checkout@v4

	- uses: actions/setup-python@v5
	with:
	python-version: ${{ matrix.python-version }}

	- name: Free runner disk
	run: \|
	df -h
	sudo rm -rf /usr/local/lib/android /usr/share/dotnet /opt/ghc /opt/hostedtoolcache/CodeQL \|\| true
	df -h

	- name: Cache pip
	uses: actions/cache@v4
	with:
	path: ~/.cache/pip
	key: ${{ runner.os }}-pip-${{ matrix.python-version }}-${{ hashFiles('pyproject.toml') }}
	restore-keys: ${{ runner.os }}-pip-${{ matrix.python-version }}-

	- name: Cache Hugging Face models
	uses: actions/cache@v4
	with:
	path: ~/.cache/huggingface
	key: ${{ runner.os }}-hf-bge-large-en-v1.5-bge-m3

	- name: Install
	run: pip install -e ".[dev]"

	- name: Warm Hugging Face model cache
	run: \|
	python - <<'PY'
	import time

	from huggingface_hub import snapshot_download

	model_ids = ["BAAI/bge-large-en-v1.5", "BAAI/bge-m3"]
	delays = [5, 15, 30]

	for model_id in model_ids:
	for attempt in range(1, len(delays) + 2):
	try:
	snapshot_download(repo_id=model_id)
	print(f"Warmed Hugging Face cache for {model_id}")
	break
	except Exception as exc:
	if attempt == len(delays) + 1:
	print(f"::warning::Unable to warm Hugging Face cache for {model_id}: {exc!r}")
	break
	delay = delays[attempt - 1]
	print(f"Unable to warm Hugging Face cache for {model_id} on attempt {attempt}; retrying in {delay}s: {exc!r}")
	time.sleep(delay)
	PY

	- name: Unit tests
	env:
	HF_HUB_OFFLINE: "1"
	TRANSFORMERS_OFFLINE: "1"
	run: >
	pytest tests/ -v --tb=short -m "not integration and not live" -x
	--ignore=tests/test_eval_framework.py
	--ignore=tests/test_follow_up_rewrite.py
	--ignore=tests/test_prompt_classification.py

	- name: Doctor fixture gate
	env:
	HF_HUB_OFFLINE: "1"
	TRANSFORMERS_OFFLINE: "1"
	run: pytest tests/test_doctor.py -v --tb=short

	- name: Isolated eval and hook routing tests
	env:
	HF_HUB_OFFLINE: "1"
	TRANSFORMERS_OFFLINE: "1"
	run: >
	pytest
	tests/test_eval_framework.py
	tests/test_follow_up_rewrite.py
	tests/test_prompt_classification.py
	-v --tb=short -x

	- name: MCP tool registration
	env:
	HF_HUB_OFFLINE: "1"
	TRANSFORMERS_OFFLINE: "1"
	run: pytest tests/test_think_recall_integration.py::TestMCPToolCount -v --tb=short

	lint:
	runs-on: ubuntu-latest
	steps:
	- uses: actions/checkout@v4

	- name: Reject tracked docs.local files
	shell: bash
	run: \|
	tracked_docs_local="$(git ls-files docs.local)"
	if [[ -n "$tracked_docs_local" ]]; then
	echo "::error::docs.local is ignored but contains tracked files:"
	printf '%s\n' "$tracked_docs_local"
	exit 1
	fi
	echo "No tracked files under docs.local/."

	- uses: actions/setup-python@v5
	with:
	python-version: "3.13"

	- name: Cache pip
	uses: actions/cache@v4
	with:
	path: ~/.cache/pip
	key: ${{ runner.os }}-pip-lint-${{ hashFiles('pyproject.toml') }}

	- run: pip install ruff
	- run: ruff check src/ tests/
	- run: ruff format --check src/ tests/

Navigation Menu

fix(tests): close four suite-hygiene loose ends, and say why CI was green on the failing one #2482

Workflow file

fix(tests): close four suite-hygiene loose ends, and say why CI was green on the failing one #2482

Uh oh!

Workflow file for this run