elizaos--eliza/.github/workflows/views-soak.yml

name: Views Soak

# Scheduled, HARD-FAIL run of the real-app view-lifecycle soak
# (`packages/app audit:views` → audit-views-soak.mjs). The soak is the strongest
# view-isolation gate in the repo — it drives the ACTUAL running app (renderer +
# API/agent), enumerates every registered view via /api/views, cycles each one
# through the real ViewRouter many times, and fails (non-zero) on a render-storm,
# an unbounded module cache, or unbounded heap growth. Before this workflow it
# ran manual-only (audit item 7: "audit:views is in NO CI workflow"), so a
# view-isolation regression could land unseen.
#
# It hard-fails — no `|| true` — so a regression reds this lane. It runs nightly
# (off the PR path so it never blocks a merge on flake) plus on demand.

on:
  schedule:
    # 08:00 UTC daily — after the app-live-e2e nightly (07:00) settles.
    - cron: "0 8 * * *"
  workflow_dispatch: {}

concurrency:
  group: views-soak-${{ github.ref }}
  cancel-in-progress: true

permissions:
  contents: read

env:
  CI: "true"
  BUN_VERSION: "canary"
  NODE_VERSION: "24.15.0"
  NODE_NO_WARNINGS: "1"
  PGLITE_WASM_MODE: node
  # The soak connects to the dev stack on the default dev ports.
  ELIZA_UI_PORT: "2138"
  ELIZA_API_PORT: "31337"

jobs:
  views-soak:
    name: real-app view-lifecycle soak (hard-fail)
    runs-on: ubuntu-24.04
    timeout-minutes: 45
    steps:
      - uses: actions/checkout@34e114876b0b11c390a56381ad16ebd13914f8d5
        with:
          submodules: false

      - name: Free disk space for browser soak
        run: |
          set -euxo pipefail
          df -h /
          sudo rm -rf /usr/share/dotnet /opt/ghc /usr/local/lib/android /opt/hostedtoolcache/CodeQL || true
          docker system prune -af || true
          df -h /

      - name: Setup Node.js
        uses: actions/setup-node@48b55a011bda9f5d6aeb4c2d9c7362e8dae4041e
        with:
          node-version: ${{ env.NODE_VERSION }}

      - name: Setup workspace dependencies
        uses: ./.github/actions/setup-bun-workspace
        with:
          bun-version: ${{ env.BUN_VERSION }}
          install-command: bun install
          install-native-deps: "false"
          skip-avatar-clone: "true"
          no-vision-deps: "true"

      - name: Install Playwright Chromium
        run: |
          bunx playwright install --with-deps chromium
          sudo apt-get update
          sudo apt-get install -y --no-install-recommends ffmpeg

      - name: Generate shared i18n assets
        run: bun run --cwd packages/shared build:i18n

      # Prebuild the renderer + core/shared bundles so the dev server serves the
      # compiled browser bundle instead of a cold ~12 min on-demand vite graph
      # (the composer/views must mount inside the soak's per-view timeouts).
      - name: Build @elizaos/core + @elizaos/shared browser bundles
        run: bunx turbo run build --filter=@elizaos/core --filter=@elizaos/shared

      - name: Build app renderer bundle
        run: bun run --cwd packages/app build:web

      # Boot the real dev stack (renderer + API/agent) in the background, then
      # wait for BOTH ports before running the soak against it. The soak reads
      # UI=127.0.0.1:2138 / API=127.0.0.1:31337 (its defaults).
      - name: Boot dev stack + run views soak (hard-fail)
        run: |
          set -euo pipefail
          bun run dev > /tmp/views-soak-dev.log 2>&1 &
          DEV_PID=$!
          echo "dev stack PID=$DEV_PID"
          cleanup() { kill "$DEV_PID" 2>/dev/null || true; }
          trap cleanup EXIT

          wait_for() {
            local url="$1" name="$2" tries=0
            until curl -sf -o /dev/null "$url"; do
              tries=$((tries + 1))
              if [ "$tries" -gt 180 ]; then
                echo "::error::$name did not come up at $url within 6 minutes"
                echo "---- dev stack log (tail) ----"
                tail -n 200 /tmp/views-soak-dev.log || true
                exit 1
              fi
              sleep 2
            done
            echo "$name is up ($url)"
          }
          wait_for "http://127.0.0.1:31337/api/views" "API"
          wait_for "http://127.0.0.1:2138" "UI"

          # Hard-fail: no `|| true`. A render-storm / unbounded cache / heap growth
          # reds this lane.
          UI=http://127.0.0.1:2138 API=http://127.0.0.1:31337 OUT="${{ runner.temp }}/evidence/10196-views-state" \
            bun run --cwd packages/app audit:views

      - name: Upload views-soak artifacts
        if: ${{ always() }}
        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a
        with:
          name: views-soak
          path: |
            ${{ runner.temp }}/evidence/10196-views-state/
            /tmp/views-soak-dev.log
          if-no-files-found: ignore
          retention-days: 7

  # A hard-fail nightly that only reds in the Actions tab is indistinguishable
  # from no lane: without an alert nobody notices the view-isolation soak went
  # red and regressions accumulate silently. This turns a red scheduled run into
  # an actionable signal by commenting on the testing-loop umbrella (#13562, the
  # issue this soak lane closes) with the result and a link to the run. Scheduled
  # runs only — manual workflow_dispatch runs stay quiet to avoid noise. Mirrors
  # app-live-e2e.yml's notify-on-failure idiom.
  notify-on-failure:
    name: notify on red nightly
    needs: [views-soak]
    if: ${{ always() && github.event_name == 'schedule' && contains(needs.*.result, 'failure') }}
    runs-on: ubuntu-24.04
    permissions:
      issues: write
    steps:
      - name: Comment red-nightly diagnostic on umbrella issue #13562
        env:
          GH_TOKEN: ${{ github.token }}
          REPO: ${{ github.repository }}
          RUN_URL: ${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }}
          R_SOAK: ${{ needs.views-soak.result }}
        run: |
          set -euo pipefail
          {
            printf '## 🌙 Nightly Views Soak failed (scheduled run)\n\n'
            printf 'The hard-fail view-lifecycle soak (`packages/app audit:views`) went red — a render-storm, unbounded module cache, or unbounded heap growth in the real ViewRouter. Result:\n\n'
            printf -- '- real-app view-lifecycle soak: `%s`\n' "$R_SOAK"
            printf '\nRun log + failure artifacts: %s\n\n' "$RUN_URL"
            printf '_Auto-posted by views-soak.yml so a red nightly notifies someone (#13562)._\n'
          } > "$RUNNER_TEMP/nightly-red.md"
          gh issue comment 13562 --repo "$REPO" --body-file "$RUNNER_TEMP/nightly-red.md"