ci: replace backend-e2e with e2e-docker and introduce playwright-e2e workflows

.github/workflows/e2e-docker.yml

@@ -0,0 +1,117 @@
+# Full backend Docker E2E suite.
+# Runs every test registered with `LABELS e2e` in tests/CMakeLists.txt
+# (see `deliveryoptimizer_add_docker_*e2e_test` helpers). Each E2E test brings
+# up the full Compose stack (postgres + osrm + http-server) under a unique
+# project name, exercises a real routing/job/proxy/health flow, and tears the
+# stack down on exit.
+#
+# Triggers:
+#   - schedule: weekly, Mondays 06:00 UTC (cron: '0 6 * * 1')
+#   - workflow_dispatch: manual run (Actions tab -> Run workflow, or
+#     `gh workflow run e2e-docker.yml`)
+#
+# Not wired to pull_request on purpose. PR CI (pr-light.yml) keeps excluding
+# e2e/docker via `ctest -LE 'e2e|docker'` because each E2E run can take a long
+# time on cold caches (Docker build of OSRM + http-server image, OSRM PBF
+# preprocess for Monaco, then 8 serial bring-up/tear-down cycles).
+#
+# Runner: native ARM64 to match deploy/compose/docker-compose.arm64.yml and
+# tests/integration/e2e/e2e_stack.env (DELIVERYOPTIMIZER_PLATFORM=linux/arm64).
+# If `ubuntu-24.04-arm` is unavailable on this org/repo, switch `runs-on` to
+# `ubuntu-latest` and add a Buildx + QEMU setup step before the ctest call;
+# emulated arm64 builds will be noticeably slower.
+#
+# Disabled by CMake (DISABLED TRUE): DeliveryOptimizerRoutingSmokeTest.OsrmAndVroom.
+# CTest skips disabled tests automatically, so `ctest -L e2e` is safe to use.
+
+name: E2E Docker (weekly + manual)
+
+on:
+  schedule:
+    - cron: '0 6 * * 1'
+  workflow_dispatch:
+
+concurrency:
+  group: e2e-docker-${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: true
+
+jobs:
+  e2e:
+    name: ctest -L e2e (full suite)
+    runs-on: ubuntu-24.04-arm
+    timeout-minutes: 240
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Install build + test prerequisites
+        uses: awalsh128/cache-apt-pkgs-action@v1
+        with:
+          packages: ninja-build libjsoncpp-dev libdrogon-dev uuid-dev libmariadb-dev libbrotli-dev libhiredis-dev libyaml-cpp-dev libgtest-dev python3
+          version: 1.0
+
+      - name: Verify toolchain
+        run: |
+          cmake --version
+          docker --version
+          docker compose version
+          python3 --version
+          bash --version | head -n 1
+
+      - name: Validate compose config with E2E env
+        run: |
+          docker compose \
+            -f deploy/compose/docker-compose.arm64.yml \
+            --env-file tests/integration/e2e/e2e_stack.env \
+            config > /dev/null
+
+      - name: Configure (CMake dev preset)
+        run: cmake --preset dev
+
+      - name: Build (CMake dev preset)
+        run: cmake --build --preset dev --parallel
+
+      - name: List discovered E2E tests
+        run: ctest --preset dev -N -L e2e
+
+      - name: Run full E2E suite
+        run: ctest --preset dev --output-on-failure -L e2e
+
+      - name: Capture Docker diagnostics on failure
+        if: failure()
+        run: |
+          mkdir -p e2e-diagnostics
+          {
+            echo "=== docker version ==="
+            docker version || true
+            echo
+            echo "=== docker ps -a ==="
+            docker ps -a || true
+            echo
+            echo "=== docker network ls ==="
+            docker network ls || true
+            echo
+            echo "=== docker volume ls ==="
+            docker volume ls || true
+            echo
+            echo "=== docker images (deliveryoptimizer*) ==="
+            docker images "deliveryoptimizer*" || true
+          } > e2e-diagnostics/docker-state.txt 2>&1
+
+          # Each E2E test uses its own unique compose project name and cleans
+          # up on EXIT, so logs are usually unavailable post-mortem. Attempt a
+          # best-effort dump for any leftover containers.
+          for cid in $(docker ps -aq); do
+            (
+              echo "=== docker logs ${cid} ==="
+              docker logs "${cid}" 2>&1 || true
+            ) >> e2e-diagnostics/leftover-container-logs.txt
+          done
+
+      - name: Upload diagnostics artifact
+        if: failure()
+        uses: actions/upload-artifact@v4
+        with:
+          name: e2e-docker-diagnostics
+          path: e2e-diagnostics/
+          retention-days: 7
+          if-no-files-found: ignore

.github/workflows/playwright-e2e.yml

@@ -0,0 +1,114 @@
+# Playwright browser E2E suite.
+# Spins up the full Docker Compose backend stack (OSRM + VROOM + Postgres + C++ API),
+# builds the Next.js production server, and runs a Playwright test that exercises
+# the landing → edit → optimize → results flow end-to-end.
+#
+# OSRM routing data for San Francisco is cached between runs via actions/cache so
+# PBF downloading and processing (~5-8 min) only happens on first run or when
+# e2e_sf_playwright.env changes.
+#
+# No GitHub secrets required. Google Maps is omitted; the map panel logs a console
+# error but the route sidebar (what the test asserts) renders independently.
+#
+# Triggers:
+#   - schedule: weekly, Mondays 06:00 UTC
+#   - workflow_dispatch: manual run (Actions tab or `gh workflow run playwright-e2e.yml`)
+
+name: Playwright E2E (weekly + manual)
+
+on:
+  schedule:
+    - cron: "0 6 * * 1"
+  workflow_dispatch:
+
+concurrency:
+  group: playwright-e2e-${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: true
+
+jobs:
+  frontend-e2e:
+    runs-on: ubuntu-latest
+    timeout-minutes: 90
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Restore OSRM cache
+        uses: actions/cache@v4
+        with:
+          path: /tmp/osrm-sf-data
+          key: osrm-sf-${{ hashFiles('tests/integration/e2e/e2e_sf_playwright.env') }}
+
+      - name: Prepare OSRM cache directory
+        run: mkdir -p /tmp/osrm-sf-data
+
+      - name: Start backend stack
+        run: |
+          OSRM_CACHE_DIR=/tmp/osrm-sf-data \
+          docker compose \
+            -f deploy/compose/docker-compose.arm64.yml \
+            -f deploy/compose/docker-compose.e2e-override.yml \
+            --env-file tests/integration/e2e/e2e_sf_playwright.env \
+            up -d --build
+
+      # 120 × 10 s = 20 min ceiling.
+      # Cold run: OSRM downloads + processes the SF PBF (~5-8 min).
+      # Cache hit: files already exist, entrypoint skips processing, starts in <1 min.
+      - name: Wait for backend health
+        run: |
+          for i in $(seq 1 120); do
+            RESP=$(curl -fsS http://127.0.0.1:39080/health 2>/dev/null || true)
+            if echo "$RESP" | grep -q '"status":"ok"'; then
+              echo "Healthy after $((i * 10))s"; exit 0
+            fi
+            echo "Attempt $i/120 — waiting 10s..."; sleep 10
+          done
+          docker compose \
+            -f deploy/compose/docker-compose.arm64.yml \
+            -f deploy/compose/docker-compose.e2e-override.yml \
+            --env-file tests/integration/e2e/e2e_sf_playwright.env \
+            logs >&2 || true
+          exit 1
+
+      - uses: actions/setup-node@v4
+        with:
+          node-version: 20
+          cache: npm
+          cache-dependency-path: app/ui/package-lock.json
+
+      - run: npm ci
+        working-directory: app/ui
+
+      - run: npx playwright install --with-deps chromium
+        working-directory: app/ui
+
+      - name: Build Next.js
+        run: npm run build
+        working-directory: app/ui
+        env:
+          DELIVERYOPTIMIZER_API_URL: http://127.0.0.1:39080
+
+      - name: Start Next.js server
+        run: npm start > /tmp/nextjs-server.log 2>&1 &
+        working-directory: app/ui
+        env:
+          DELIVERYOPTIMIZER_API_URL: http://127.0.0.1:39080
+
+      - run: npx wait-on http://localhost:3000 --timeout 30000
+        working-directory: app/ui
+
+      - name: Run Playwright E2E tests
+        run: npm run test:e2e
+        working-directory: app/ui
+        env:
+          PLAYWRIGHT_BASE_URL: http://localhost:3000
+          CI: "true"
+
+      - name: Upload Playwright report
+        if: failure()
+        uses: actions/upload-artifact@v4
+        with:
+          name: playwright-report
+          path: |
+            app/ui/playwright-report/
+            /tmp/nextjs-server.log
+          retention-days: 7

.github/workflows/pr-light.yml

@@ -63,16 +63,3 @@ jobs:
       - run: cmake --build --preset dev --parallel
       - run: ctest --preset dev --output-on-failure --no-tests=error -LE 'e2e|docker'
       - run: docker compose -f deploy/compose/docker-compose.arm64.yml --env-file deploy/env/http-server.arm64.env config
-  backend-e2e:
-    if: github.event_name != 'pull_request'
-    runs-on: ubuntu-latest
-    steps:
-      - uses: actions/checkout@v4
-      - uses: docker/setup-qemu-action@v3
-      - uses: awalsh128/cache-apt-pkgs-action@v1
-        with:
-          packages: ninja-build libjsoncpp-dev libdrogon-dev uuid-dev libmariadb-dev libbrotli-dev libhiredis-dev libyaml-cpp-dev libgtest-dev postgresql postgresql-client
-          version: 1.0
-      - run: cmake --preset dev
-      - run: cmake --build --preset dev --parallel
-      - run: ctest --preset dev --output-on-failure --no-tests=error -L 'e2e|docker'

.gitignore

@@ -92,5 +92,7 @@ services/vroom/data/
 # macOS / editor noise
 .DS_Store
 
-# Claude
+# Claude and Cursor
 Claude.md
+.claude/
+.cursor/

app/ui/package.json

@@ -10,6 +10,7 @@
     "typecheck": "tsc --noEmit",
     "test": "vitest run",
     "test:watch": "vitest",
+    "test:e2e": "playwright test",
     "format": "prettier --write .",
     "format:check": "prettier --check ."
   },
@@ -40,6 +41,7 @@
     "prettier": "^3.8.3",
     "tailwindcss": "^4",
     "typescript": "^5",
-    "vitest": "^4.1.4"
+    "vitest": "^4.1.4",
+    "@playwright/test": "^1.49.1"
   }
 }