fix: address Phase 2 review findings from #1819 #2296

	name: Benchmark Regression Gate

	on:
	pull_request:
	types: [synchronize, opened, reopened]

	jobs:
	benchmark:
	if: github.event.pull_request.head.repo.full_name == github.repository \|\| github.event_name != 'pull_request'
	runs-on: ubuntu-latest
	timeout-minutes: 30
	concurrency:
	group: benchmark-${{ github.event.pull_request.number }}
	cancel-in-progress: true
	steps:
	- name: Checkout
	uses: actions/checkout@v4

	- name: Setup Node.js
	uses: actions/setup-node@v6
	with:
	node-version: '24'
	cache: 'npm'
	cache-dependency-path: \|
	package-lock.json
	package-lock.json

	# Route through the shared wrapper so the MCP install inherits the
	# Playwright browser-download skip (promptfoo's optional
	# @playwright/browser-chromium otherwise wedges `npm ci` past the cap on a
	# cold runner) plus retry/timeout resilience. A plain `npm ci` here
	# bypassed both. See docs/rca/2026-05-31-npm-ci-playwright-browser-wedge.md.
	- name: Install dependencies
	run: bash "$GITHUB_WORKSPACE/tools/release/npm-ci-retry.sh"

	- name: Run benchmarks
	run: npm run bench -- --outputJson benchmark-results.json
	env:
	RUN_BENCHMARKS: 'true'

	- name: Upload benchmark results
	uses: actions/upload-artifact@v4
	if: always()
	with:
	name: benchmark-results
	path: benchmark-results.json
	retention-days: 7

	- name: Check for regressions
	continue-on-error: true
	run: bash tools/audit/gates/check-benchmark-regression.sh --results benchmark-results.json --baselines src/projections/telemetry/benchmarks/baselines.json

Navigation Menu