aitbc/.gitea/workflows/production-tests.yml

name: Production Tests

on:
  push:
    branches: [main, develop]
    paths:
      - 'tests/production/**'
      - 'apps/agent-coordinator/**'
      - '.gitea/workflows/production-tests.yml'
  pull_request:
    branches: [main, develop]
  workflow_dispatch:

concurrency:
  group: production-tests-${{ github.ref }}
  cancel-in-progress: true

jobs:
  test-production:
    name: Production Integration Tests
    runs-on: debian
    timeout-minutes: 20

    steps:
      - name: Clone repository
        run: |
          WORKSPACE="/var/lib/aitbc-workspaces/production-tests"
          rm -rf "$WORKSPACE"
          mkdir -p "$WORKSPACE"
          cd "$WORKSPACE"
          git clone --depth 1 http://gitea.bubuit.net:3000/oib/aitbc.git repo

      - name: Initialize job logging
        run: |
          cd /var/lib/aitbc-workspaces/production-tests/repo
          bash scripts/ci/setup-job-logging.sh

      - name: Setup test environment
        run: |
          cd /var/lib/aitbc-workspaces/production-tests/repo

          # Remove any existing venv to avoid cache corruption issues
          rm -rf venv

          bash scripts/ci/setup-python-venv.sh \
            --repo-dir "$PWD" \
            --venv-dir "$PWD/venv" \
            --skip-requirements \
            --extra-packages "pytest pytest-asyncio pytest-timeout requests httpx pyjwt fastapi uvicorn[standard] redis bcrypt websockets numpy psutil prometheus-client celery aiohttp pydantic pydantic-settings python-dotenv cryptography"

          # Ensure standard directories exist
          mkdir -p /var/lib/aitbc/data /var/lib/aitbc/keystore /etc/aitbc /var/log/aitbc

      - name: Ensure Redis server
        run: |
          if command -v redis-server >/dev/null 2>&1 && command -v redis-cli >/dev/null 2>&1; then
            echo "✅ Redis binaries already available"
            exit 0
          fi

          apt-get update
          DEBIAN_FRONTEND=noninteractive apt-get install -y redis-server

      - name: Start Redis
        run: |
          redis-server --daemonize yes --port 6379
          sleep 2
          redis-cli ping || exit 1
          echo "✅ Redis started"

      - name: Start agent coordinator
        run: |
          cd /var/lib/aitbc-workspaces/production-tests/repo
          export PYTHONPATH="apps/agent-coordinator/src:$PYTHONPATH"

          echo "Ensuring default port 9001 is available..."
          pkill -f "uvicorn app.main:app" 2>/dev/null || true
          if command -v lsof >/dev/null 2>&1; then
            pids=$(lsof -tiTCP:9001 -sTCP:LISTEN || true)
            if [ -n "$pids" ]; then
              echo "⚠️ Port 9001 already in use by PID(s): $pids; terminating"
              kill $pids 2>/dev/null || true
              sleep 2
            fi
            if lsof -tiTCP:9001 -sTCP:LISTEN >/dev/null 2>&1; then
              echo "❌ Port 9001 is still in use; aborting startup"
              lsof -iTCP:9001 -sTCP:LISTEN || true
              exit 1
            fi
          fi

          # Start agent coordinator in background
          nohup env PYTHONUNBUFFERED=1 venv/bin/uvicorn app.main:app \
            --host 0.0.0.0 \
            --port 9001 \
            --log-level info \
            > /tmp/agent-coordinator.log 2>&1 &

          echo $! > /tmp/agent-coordinator.pid
          sleep 2
          if ! kill -0 "$(cat /tmp/agent-coordinator.pid)" 2>/dev/null; then
            echo "❌ Agent coordinator exited during startup"
            cat /tmp/agent-coordinator.log
            exit 1
          fi
          echo "✅ Agent coordinator started (PID: $(cat /tmp/agent-coordinator.pid))"

      - name: Wait for agent coordinator ready
        run: |
          echo "Waiting for agent coordinator on port 9001..."
          for i in $(seq 1 30); do
            code=$(curl -so /dev/null -w '%{http_code}' "http://localhost:9001/health" 2>/dev/null) || code=0
            if [ "$code" -ge 200 ] && [ "$code" -lt 600 ]; then
              echo "✅ Agent coordinator ready (HTTP $code)"
              exit 0
            fi
            sleep 2
          done
          echo "❌ Agent coordinator not ready"
          cat /tmp/agent-coordinator.log
          exit 1

      - name: Run production tests
        run: |
          cd /var/lib/aitbc-workspaces/production-tests/repo
          export PYTHONPATH="apps/agent-coordinator/src:$PYTHONPATH"

          # Test both chains
          export CHAINS="ait-mainnet,ait-testnet"

          venv/bin/python -m pytest -c /dev/null --rootdir "$PWD" --import-mode=importlib tests/production/ \
            -v \
            --tb=short \
            --timeout=30 \
            -k "not test_error_handling"

          echo "✅ Production tests completed"

      - name: Agent coordinator logs
        if: always()
        run: |
          if [ -f /tmp/agent-coordinator.log ]; then
            echo "=== Agent Coordinator Logs ==="
            cat /tmp/agent-coordinator.log
          fi

      - name: Cleanup
        if: always()
        run: |
          if [ -f /tmp/agent-coordinator.pid ]; then
            kill $(cat /tmp/agent-coordinator.pid) 2>/dev/null || true
            rm -f /tmp/agent-coordinator.pid
          fi
          pkill -f "uvicorn app.main:app" 2>/dev/null || true
          redis-cli shutdown 2>/dev/null || true
          rm -rf /var/lib/aitbc-workspaces/production-tests