Skip to content

fix(cli): write the workflow for init --workflow without --ci #83

fix(cli): write the workflow for init --workflow without --ci

fix(cli): write the workflow for init --workflow without --ci #83

Workflow file for this run

name: Publish memory
# ClawHub runs publish-time inspection inside a 512 MiB Node action. This gate
# measures that path (packed OpenClaw target preparation, retirement
# qualification, author-facing plugin check) on the largest official channel
# plugins and compares it with the 0.4.5 release on the same runner, so a
# change that regresses memory on a large plugin fails before release.
on:
pull_request:
push:
branches:
- main
permissions:
contents: read
jobs:
publish-memory:
name: Publish-time peak RSS
runs-on: ubuntu-latest
timeout-minutes: 20
steps:
- name: Locate scratch directories
run: |
echo "FIXTURES=$RUNNER_TEMP/publish-memory" >> "$GITHUB_ENV"
echo "BASELINE=$RUNNER_TEMP/baseline-0.4.5" >> "$GITHUB_ENV"
- name: Checkout
uses: actions/checkout@v7
- name: Setup Node
uses: actions/setup-node@v7
with:
node-version: 24.22.1
- name: Install dependencies
run: npm ci
- name: Cache npm fixtures
uses: actions/cache@v4
with:
path: ${{ runner.temp }}/publish-memory
key: publish-memory-openclaw-2026.10.10-channels-2026.10.10
- name: Fetch OpenClaw target and channel plugins
run: node scripts/fetch-publish-memory-fixtures.mjs --out "$FIXTURES"
- name: Install the 0.4.5 baseline
run: |
mkdir -p "$BASELINE"
cd "$BASELINE"
npm init -y >/dev/null
npm install --no-audit --no-fund --ignore-scripts @openclaw/plugin-inspector@0.4.5
# Every target is answered the same way: there is no per-release data, so
# the published release and a modified snapshot exercise the same path.
- name: OpenClaw 2026.10.10, at most 50% of 0.4.5
run: >-
node scripts/publish-memory-benchmark.mjs
--target "$FIXTURES/openclaw/package"
--plugin @openclaw/discord="$FIXTURES/plugins/discord/package"
--plugin @openclaw/matrix="$FIXTURES/plugins/matrix/package"
--plugin @openclaw/whatsapp="$FIXTURES/plugins/whatsapp/package"
--baseline "$BASELINE/node_modules/@openclaw/plugin-inspector/src/index.js"
--runs 5 --max-ratio 0.5 --diagnose --json publish-memory-released.json
- name: Unseen OpenClaw snapshot, at most 50% of 0.4.5
if: success() || failure()
run: >-
node scripts/publish-memory-benchmark.mjs --uncached-target
--target "$FIXTURES/openclaw/package"
--plugin @openclaw/discord="$FIXTURES/plugins/discord/package"
--plugin @openclaw/matrix="$FIXTURES/plugins/matrix/package"
--plugin @openclaw/whatsapp="$FIXTURES/plugins/whatsapp/package"
--baseline "$BASELINE/node_modules/@openclaw/plugin-inspector/src/index.js"
--runs 3 --max-ratio 0.5 --diagnose --json publish-memory-unseen.json
# A later release adds new SDK entrypoints and declaration shapes; nothing
# about it is known in advance. 0.4.5 type-checks it fully, so identical
# outputs prove the certificates answered exactly.
- name: Future OpenClaw release with new SDK surface, at most 50% of 0.4.5
if: success() || failure()
run: |
node scripts/make-future-openclaw-fixture.mjs --from "$FIXTURES/openclaw/package" --out "$RUNNER_TEMP/future/package"
node scripts/publish-memory-benchmark.mjs --target "$RUNNER_TEMP/future/package" \
--plugin @openclaw/discord="$FIXTURES/plugins/discord/package" \
--plugin @openclaw/matrix="$FIXTURES/plugins/matrix/package" \
--plugin @openclaw/whatsapp="$FIXTURES/plugins/whatsapp/package" \
--baseline "$BASELINE/node_modules/@openclaw/plugin-inspector/src/index.js" \
--runs 3 --max-ratio 0.5 --diagnose --json publish-memory-future.json
# Replacement names computed from template literals over finite unions are
# enumerated exactly, so this release is held to the same budget.
- name: Future release with computed names, at most 50% of 0.4.5
if: success() || failure()
run: |
node scripts/make-future-openclaw-fixture.mjs --from "$FIXTURES/openclaw/package" --out "$RUNNER_TEMP/future-computed/package" --computed-names
node scripts/publish-memory-benchmark.mjs --target "$RUNNER_TEMP/future-computed/package" \
--plugin @openclaw/discord="$FIXTURES/plugins/discord/package" \
--plugin @openclaw/matrix="$FIXTURES/plugins/matrix/package" \
--plugin @openclaw/whatsapp="$FIXTURES/plugins/whatsapp/package" \
--baseline "$BASELINE/node_modules/@openclaw/plugin-inspector/src/index.js" \
--runs 3 --max-ratio 0.5 --diagnose --json publish-memory-future-computed.json
# Names whose availability needs TypeScript's assignability (a conditional
# type over object types) are answered by the exact type-checked fallback.
# Outputs must stay identical to 0.4.5. Its memory does NOT meet the 50%
# budget (an exact type check of the SDK closure cannot); this guards it
# against regressions while that trade-off is decided.
- name: Undecidable future names take the exact fallback (identical outputs, at most 85% of 0.4.5)
if: success() || failure()
run: |
node scripts/make-future-openclaw-fixture.mjs --from "$FIXTURES/openclaw/package" --out "$RUNNER_TEMP/future-undecidable/package" --undecidable
node scripts/publish-memory-benchmark.mjs --target "$RUNNER_TEMP/future-undecidable/package" \
--plugin @openclaw/discord="$FIXTURES/plugins/discord/package" \
--plugin @openclaw/matrix="$FIXTURES/plugins/matrix/package" \
--plugin @openclaw/whatsapp="$FIXTURES/plugins/whatsapp/package" \
--baseline "$BASELINE/node_modules/@openclaw/plugin-inspector/src/index.js" \
--runs 1 --max-ratio 0.85 --json publish-memory-future-undecidable.json
# ClawHub's Convex actions run the inspector from an esbuild bundle, so a
# dependency hidden from bundlers would only fail after deploy.
- name: Bundled like ClawHub's Convex actions, at most 50% of 0.4.5
if: success() || failure()
run: |
banner='import { createRequire as __createRequire } from "node:module"; import { fileURLToPath as __fileURLToPath } from "node:url"; import { dirname as __dirnameOf } from "node:path"; const require = __createRequire(import.meta.url); const __filename = __fileURLToPath(import.meta.url); const __dirname = __dirnameOf(__filename);'
for implementation in candidate baseline; do
entry=src/index.js
if [ "$implementation" = baseline ]; then entry="$BASELINE/node_modules/@openclaw/plugin-inspector/src/index.js"; fi
npx --yes esbuild@0.25.12 "$entry" --bundle --platform=node --format=esm --target=node24 \
--banner:js="$banner" --log-level=warning --outfile="$RUNNER_TEMP/bundled/$implementation/index.mjs"
done
node scripts/publish-memory-benchmark.mjs \
--target "$FIXTURES/openclaw/package" \
--plugin @openclaw/discord="$FIXTURES/plugins/discord/package" \
--plugin @openclaw/matrix="$FIXTURES/plugins/matrix/package" \
--plugin @openclaw/whatsapp="$FIXTURES/plugins/whatsapp/package" \
--inspector "$RUNNER_TEMP/bundled/candidate/index.mjs" \
--baseline "$RUNNER_TEMP/bundled/baseline/index.mjs" \
--runs 5 --max-ratio 0.5 --diagnose --json publish-memory-bundled.json
# Certified answers must match the full type check for every name the
# snapshot spells, not only the retirement inventory.
- name: Retirement certificates agree with the full type check
if: success() || failure()
run: >-
node --max-old-space-size=4096 scripts/verify-packed-sdk-certificates.mjs
--target "$FIXTURES/openclaw/package"
--target "$RUNNER_TEMP/future/package"
--target "$RUNNER_TEMP/future-computed/package"
--target "$RUNNER_TEMP/future-undecidable/package"
# The runtime-import analysis reads scope facts from a light scope manager
# driven by eslint-scope's Referencer; they must equal the full analysis on
# every file of the plugin packages, dependencies included.
- name: Scope facts agree with eslint-scope
if: success() || failure()
run: node scripts/verify-scope-facts.mjs "$FIXTURES/plugins"
- name: Upload measurements
if: always()
uses: actions/upload-artifact@v4
with:
name: publish-memory
path: publish-memory-*.json
if-no-files-found: ignore