eval

fix(capabilities): add openclaw.compat.pluginApi to manifest and surface in host_compat #25

	name: eval

	# Gate retrieval quality: score kb.context against the committed labeled set
	# and fail on a P@5 regression beyond tolerance vs eval/baseline.json. Runs
	# only when retrieval code changes.
	on:
	pull_request:
	paths:
	- "src/vouch/embeddings/**"
	- "src/vouch/context.py"
	- "src/vouch/eval/**"
	- "eval/**"

	jobs:
	recall:
	name: recall eval
	runs-on: ubuntu-latest
	steps:
	- uses: actions/checkout@v4

	- uses: actions/setup-python@v5
	with:
	python-version: "3.12"
	cache: pip

	- name: install
	run: \|
	python -m pip install --upgrade pip
	pip install -e '.[dev]'

	- name: build fixture index
	working-directory: eval/fixture-kb
	run: python -m vouch.cli reindex

	- name: recall eval (fail on P@5 regression > 5%)
	working-directory: eval/fixture-kb
	run: >-
	python -m vouch.cli eval recall ../queries.jsonl
	--k 5 --baseline ../baseline.json --max-regression 0.05

Provide feedback