ci: run the jobs side by side, and split the E2E tests of the CLI in four shards

.github/actions/setup/action.yml

@@ -0,0 +1,26 @@
+name: Set up the repository
+description: Node.js at the version of .nvmrc, the dependencies of the lockfile and, when asked, the Chromium of Playwright. Run it after the checkout.
+
+inputs:
+  playwright:
+    description: Install the Chromium of Playwright with its system dependencies
+    default: "false"
+
+runs:
+  using: composite
+  steps:
+    - name: Setup Node.js
+      uses: actions/setup-node@v7
+      with:
+        node-version-file: .nvmrc
+        cache: npm
+
+    - name: Install dependencies
+      shell: bash
+      run: npm ci
+
+    - name: Install Playwright Chromium
+      if: inputs.playwright == 'true'
+      shell: bash
+      working-directory: cli
+      run: npx playwright install --with-deps chromium

.github/workflows/ci.yml

@@ -12,14 +12,22 @@ concurrency:
   group: ci-${{ github.event.pull_request.number || github.ref }}
   cancel-in-progress: ${{ github.event_name == 'pull_request' }}
 
+permissions:
+  contents: read
+
+# The jobs run side by side, each on its own runner. The E2E tests of the CLI are most of the time
+# of a run, so they are split in shards: a shard has a whole runner, as the suite had when it ran
+# alone, and the tests that read timings keep their margin.
 jobs:
-  validate:
+  # A pull request that only changes documentation skips the jobs that build and launch Chromium
+  changes:
     runs-on: ubuntu-latest
-    # A run takes about 12 minutes. Without a limit, one that hangs holds the runner for 6 hours
-    timeout-minutes: 25
+    timeout-minutes: 5
     permissions:
       contents: read
       pull-requests: read
+    outputs:
+      skip: ${{ steps.slow.outputs.skip }}
 
     steps:
       - name: Checkout repository
@@ -31,9 +39,8 @@ jobs:
         uses: actions/setup-node@v7
         with:
           node-version-file: .nvmrc
-          cache: 'npm'
 
-      - name: Decide whether to skip the slow steps
+      - name: Decide whether to skip the slow jobs
         id: slow
         env:
           EVENT_NAME: ${{ github.event_name }}
@@ -42,19 +49,22 @@ jobs:
           GH_TOKEN: ${{ github.token }}
         run: node scripts/ci-skip-slow-steps.js
 
-      - name: Install dependencies
-        run: npm ci
+  # What needs neither a build nor a browser
+  checks:
+    runs-on: ubuntu-latest
+    timeout-minutes: 10
 
-      - name: Lint
-        run: npm run lint
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v7
+        with:
+          persist-credentials: false
 
-      - name: Build
-        if: steps.slow.outputs.skip != 'true'
-        run: npm run build
+      - name: Set up the repository
+        uses: ./.github/actions/setup
 
-      - name: Check the built snippets
-        if: steps.slow.outputs.skip != 'true'
-        run: npm run check:built-snippets
+      - name: Lint
+        run: npm run lint
 
       - name: Check consistency
         run: npm run check:consistency
@@ -68,12 +78,18 @@ jobs:
         if: github.event_name == 'push'
         run: node scripts/release-versions.js --untagged
 
-      - name: Run demo tests
-        run: npm run test:demos
-
       - name: Check emojis have icons
         run: npm run check:emoji
 
+      - name: Check generated skills are up to date
+        run: npm run generate-skills:check
+
+      - name: Check generated llms files are up to date
+        run: npm run generate-llms:check
+
+      - name: Run demo tests
+        run: npm run test:demos
+
       - name: Run icon plugin tests
         run: npm run test:icons
 
@@ -89,12 +105,6 @@ jobs:
       - name: Run extension tests
         run: npm run test:extension
 
-      - name: Check generated skills are up to date
-        run: npm run generate-skills:check
-
-      - name: Check generated llms files are up to date
-        run: npm run generate-llms:check
-
       - name: Run skill script tests
         run: npm run test:skills
 
@@ -110,21 +120,77 @@ jobs:
       - name: Run unit tests
         run: npm run test:unit --prefix cli
 
-      - name: Install Playwright Chromium
-        if: steps.slow.outputs.skip != 'true'
-        run: npx playwright install --with-deps chromium
-        working-directory: cli
+  build:
+    needs: changes
+    if: needs.changes.outputs.skip != 'true'
+    runs-on: ubuntu-latest
+    timeout-minutes: 15
+
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v7
+        with:
+          persist-credentials: false
+
+      - name: Set up the repository
+        uses: ./.github/actions/setup
+        with:
+          playwright: "true"
+
+      - name: Build
+        run: npm run build
+
+      - name: Check the built snippets
+        run: npm run check:built-snippets
 
       # Reports to the job summary and does not block: a failing audit marks this step, not the job.
       - name: Audit agentic browsing with Lighthouse
-        if: steps.slow.outputs.skip != 'true'
         continue-on-error: true
         run: npm run audit:agentic
 
+  e2e-cli:
+    name: E2E of the CLI (${{ matrix.shard }}/4)
+    needs: changes
+    if: needs.changes.outputs.skip != 'true'
+    runs-on: ubuntu-latest
+    timeout-minutes: 20
+    strategy:
+      # A shard that fails does not stop the others: the run says everything that fails
+      fail-fast: false
+      matrix:
+        shard: [1, 2, 3, 4]
+
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v7
+        with:
+          persist-credentials: false
+
+      - name: Set up the repository
+        uses: ./.github/actions/setup
+        with:
+          playwright: "true"
+
       - name: Run E2E tests
-        if: steps.slow.outputs.skip != 'true'
-        run: npm run test:e2e --prefix cli
+        run: npm run test:e2e --prefix cli -- --shard=${{ matrix.shard }}/4
+
+  e2e-extension:
+    name: E2E of the extension
+    needs: changes
+    if: needs.changes.outputs.skip != 'true'
+    runs-on: ubuntu-latest
+    timeout-minutes: 15
+
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v7
+        with:
+          persist-credentials: false
+
+      - name: Set up the repository
+        uses: ./.github/actions/setup
+        with:
+          playwright: "true"
 
       - name: Run extension E2E tests
-        if: steps.slow.outputs.skip != 'true'
         run: npm run test:extension:e2e

CLAUDE.md

@@ -78,7 +78,7 @@ npm run smoke:cli-package     # before a CLI release: the packed CLI installed i
 npm run test:e2e --prefix cli   # Playwright against local fixtures (cli/tests/fixtures)
 ```
 
-CI (`.github/workflows/ci.yml`) runs lint, build, all the checks above except `eval:prompts`, `generate-skills:check` and the CLI unit and e2e tests. The pre-commit hook (`.githooks/pre-commit`) runs `check:consistency`, `check:emoji` and `lint`.
+CI (`.github/workflows/ci.yml`) runs lint, build, all the checks above except `eval:prompts`, `generate-skills:check` and the CLI unit and e2e tests. Its jobs run side by side: `checks` (what needs neither a build nor a browser), `build`, the e2e tests of the extension, and the e2e tests of the CLI in four shards; `.github/actions/setup` installs Node, the dependencies and Chromium for all of them. The pre-commit hook (`.githooks/pre-commit`) runs `check:consistency`, `check:emoji` and `lint`.
 
 ## Snippets