[CI] Replace the Lark queue-digest card with a daily queue-timeline chart (#38380)

.github/workflows/ci-lark-notify.yml

@@ -9,14 +9,13 @@ on:
     branches: [main]
   schedule:
     - cron: '*/15 * * * *'   # runner-health
-    - cron: '30 */8 * * *'   # queue-digest
   workflow_dispatch:
     inputs:
       task:
         description: 'Which report to run'
         required: true
         type: choice
-        options: [ci-status, runner-health, queue-digest]
+        options: [ci-status, runner-health]
       run_id:
         description: 'Workflow run id (ci-status only)'
         required: false
@@ -94,33 +93,11 @@ jobs:
           LARK_WEBHOOK: ${{ secrets.LARK_WEBHOOK }}
         run: |
           python scripts/ci_monitor/lark_notify.py $DRY_RUN_FLAG runner-health \
-            --state-file runner-health-state.json
+            --state-file runner-health-state.json \
+            --remind-hours 6
       - name: Save runner-health state
         if: ${{ !inputs.dry_run }}
         uses: actions/cache/save@v4
         with:
           path: runner-health-state.json
           key: lark-runner-health-${{ github.run_id }}
-
-  queue-digest:
-    if: >-
-      github.repository == 'sgl-project/sglang' && (
-        (github.event_name == 'schedule' && github.event.schedule == '30 */8 * * *') ||
-        (github.event_name == 'workflow_dispatch' && inputs.task == 'queue-digest')
-      )
-    runs-on: ubuntu-latest
-    timeout-minutes: 30
-    steps:
-      - uses: actions/checkout@v4
-        with:
-          sparse-checkout: scripts/ci_monitor
-      - uses: actions/setup-python@v5
-        with:
-          python-version: '3.12'
-      - name: Post queue digest
-        env:
-          # PAT: an 8h window is ~400-500 API calls, too close to GITHUB_TOKEN's 1000/h
-          GITHUB_TOKEN: ${{ secrets.GH_PAT_FOR_NIGHTLY_CI_DATA }}
-          LARK_WEBHOOK: ${{ secrets.LARK_WEBHOOK }}
-        run: |
-          python scripts/ci_monitor/lark_notify.py $DRY_RUN_FLAG queue-digest --hours 8 --only-if-slow

.github/workflows/runner-utilization.yml

@@ -68,4 +68,18 @@ jobs:
           python scripts/ci/utils/runner_utilization_report.py \
             --repo ${{ github.repository }} \
             --hours ${{ (github.event_name == 'pull_request' && '0.34') || inputs.hours || '24' }} \
-            ${{ inputs.filter && format('--filter {0}', inputs.filter) || '' }}
+            ${{ inputs.filter && format('--filter {0}', inputs.filter) || '' }} \
+            --queue-series-out queue-series.json
+
+      # The card rides on the scan above rather than running its own: a 24h
+      # window is ~5000 API calls, the whole hourly budget of ci-lark-notify's PAT.
+      - name: Post queue timeline card
+        if: >-
+          github.repository == 'sgl-project/sglang' &&
+          github.event_name != 'pull_request'
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+          LARK_WEBHOOK: ${{ secrets.LARK_WEBHOOK }}
+        run: |
+          python scripts/ci_monitor/lark_notify.py queue-timeline \
+            --series-file queue-series.json