integration.yml

name: Integration Tests

on:
  push:
    branches: [main]
  pull_request:
    branches: [main]
  workflow_dispatch: {}
  # Weekly drift catch
  schedule:
    - cron: '0 6 * * 2'  # Tuesday 06:00 UTC — failures land in EU/IN working hours

permissions:
  contents: read

# Avoid spawning parallel docker-compose stacks for back-to-back pushes;
# also cancels stale PR runs when a new commit lands. Push to main keys on SHA
# so main runs never queue or cancel each other.
concurrency:
  group: integration-${{ github.event.pull_request.number || github.sha }}
  cancel-in-progress: ${{ github.event_name == 'pull_request' }}

env:
  AXONFLOW_TELEMETRY: 'off'

jobs:
  # Contract tests run on every PR and push. No network, no stack.
  contract-tests:
    name: Contract Tests
    runs-on: ubuntu-latest
    timeout-minutes: 10
    steps:
      - uses: actions/checkout@v4

      - name: Setup Node.js
        uses: actions/setup-node@v4
        with:
          node-version: '20.x'
          cache: 'npm'

      - name: Install dependencies
        run: npm ci

      - name: Run contract tests
        run: npm run test:contract

  # Example + integration-test type-check on every PR and push. Catches API
  # drift at PR time without needing a live stack. Mirrors Python SDK
  # demo-scripts job.
  example-typecheck:
    name: Example Type-Check
    runs-on: ubuntu-latest
    timeout-minutes: 10
    needs: contract-tests
    steps:
      - uses: actions/checkout@v4

      - name: Setup Node.js
        uses: actions/setup-node@v4
        with:
          node-version: '20.x'
          cache: 'npm'

      - name: Install dependencies
        run: npm ci

      - name: Type-check examples against local source
        run: npx tsc -p tsconfig.examples.json

      # Integration test files live behind jest's testPathIgnorePatterns
      # (only run on push-to-main), so jest never type-checks them at PR
      # time. Drift in those files (e.g. method renames on AxonFlow) only
      # surfaced after merge. Catch it on every PR. The dedicated tsconfig
      # inherits strict + esModuleInterop + types from tsconfig.json so
      # the gate matches what ts-jest applies at runtime.
      - name: Type-check integration test files
        run: npx tsc -p tsconfig.tests.json

  # Integration tests run against a live community stack via docker compose.
  # Runs on main merges, scheduled, and manual dispatch. Skipped on PRs to keep
  # latency reasonable — PR-level live coverage is added in QF-13.
  # Mirrors axonflow-sdk-go/.github/workflows/integration.yml.
  integration:
    name: Integration Tests (Community Stack)
    runs-on: ubuntu-latest
    timeout-minutes: 20
    needs: [contract-tests, example-typecheck]
    if: github.event_name != 'pull_request'
    steps:
      - name: Checkout SDK
        uses: actions/checkout@v4

      - name: Setup Node.js
        uses: actions/setup-node@v4
        with:
          node-version: '20.x'
          cache: 'npm'

      - name: Install dependencies
        run: npm ci

      - name: Build SDK
        run: npm run build

      - name: Clone community stack
        run: git clone --depth 1 https://github.com/getaxonflow/axonflow.git ../axonflow

      - name: Start community stack
        run: |
          cd ../axonflow
          docker compose up -d --wait --wait-timeout 120

          # Belt-and-suspenders: also poll /health, since not every compose
          # service has a healthcheck wired.
          echo "Waiting for agent to be healthy..."
          timeout 120 bash -c 'until curl -sf http://localhost:8080/health; do sleep 2; done'
          echo "Agent is healthy"

          echo "Waiting for orchestrator to be healthy..."
          timeout 60 bash -c 'until curl -sf http://localhost:8081/health; do sleep 2; done'
          echo "Orchestrator is healthy"

      # tests/integration.test.ts targets a fully-featured stack (LLM
      # provider configured, planning engine initialised, PII enforcement
      # blocking) — community boots without those, so several assertions
      # fail at runtime even when the wire shape is correct. Tracked in
      # axonflow-sdk-typescript#198 — refactor that suite to community-
      # vs enterprise-tier groupings, then re-enable here. PR-time
      # typecheck against the integration files still runs in
      # example-typecheck above, so wire-shape drift is still caught.
      #
      # The basic example below remains the live community-stack smoke.

      - name: Run basic example against live stack
        env:
          AXONFLOW_AGENT_URL: http://localhost:8080
          AXONFLOW_CLIENT_ID: demo-client
          AXONFLOW_CLIENT_SECRET: demo-secret
        run: |
          set -e
          # Clean tarball dir so a stale .tgz from a re-run can't shadow the
          # fresh one. `npm pack --json` gives us the exact filename — don't
          # rely on alphabetical glob.
          rm -rf /tmp/sdk-tarball /tmp/example-runner
          mkdir -p /tmp/sdk-tarball /tmp/example-runner
          PACK_JSON=$(npm pack --pack-destination /tmp/sdk-tarball --json)
          TARBALL_NAME=$(printf '%s' "$PACK_JSON" | node -e 'process.stdin.on("data",d=>{const o=JSON.parse(d);console.log(o[0].filename||o[0].name)})')
          TARBALL="/tmp/sdk-tarball/$TARBALL_NAME"
          test -f "$TARBALL" || { echo "Tarball not found at $TARBALL"; ls -la /tmp/sdk-tarball; exit 1; }
          cp examples/basic/index.ts /tmp/example-runner/
          cd /tmp/example-runner
          npm init -y >/dev/null
          npm install --prefer-offline --no-audit --no-fund \
            "$TARBALL" tsx@4 typescript@5 @types/node@20
          timeout 60 npx tsx index.ts

      # Logs MUST be captured before `Stop community stack` runs — `compose
      # down` destroys the containers and `compose logs` then returns
      # nothing. `if: failure()` here, `if: always()` on teardown.
      - name: Show docker logs on failure
        if: failure()
        run: |
          if [ -d "../axonflow" ]; then
            cd ../axonflow
            docker compose logs --tail=200 || true
          fi

      - name: Upload docker logs on failure
        if: failure()
        uses: actions/upload-artifact@v4
        with:
          name: docker-compose-logs
          path: ../axonflow/docker-compose-logs.txt
          if-no-files-found: ignore

      - name: Stop community stack
        if: always()
        run: |
          if [ -d "../axonflow" ]; then
            cd ../axonflow
            # Persist logs to disk so the upload step can grab them even after teardown.
            docker compose logs --tail=500 > docker-compose-logs.txt 2>/dev/null || true
            docker compose down --volumes --remove-orphans || true
          fi