Changeset 0.5.0 (#35)

2026-02-19 15:03:20 +00:00
parent a93a6b9635
commit 30d0c11219
65 changed files with 5345 additions and 8070 deletions
--- a/.gitea/workflows/ci.yaml
+++ b/.gitea/workflows/ci.yaml
@@ -1,16 +1,30 @@
 # .gitea/workflows/ci.yaml
 # ============================================
-# Chat Switchboard - CI/CD Pipeline
+# Chat Switchboard - CI/CD Pipeline (v0.5.4)
 # ============================================
 # Cluster deployments use SEPARATE FE + BE images.
-# Unified image is for Docker Hub only (Docker Desktop use).
+# Unified image is for Docker Hub only (docker-compose use).
+#
+# Pipeline:
+#   1. Go test (all PRs and pushes)
+#   2. Build + Deploy (depends on test passing)
 #
 # Deployment mapping:
-#   PR              → FE + BE :dev    → wipe/migrate → dev.DOMAIN
-#   Push to main    → FE + BE :test   → migrate      → test.DOMAIN
-#   Push to main    → FE-only :lite   → (no DB)      → lite.DOMAIN
-#   Tag v*          → FE + BE :latest → migrate      → www.DOMAIN
-#                   → Unified :latest → Docker Hub only (not deployed)
+#   PR              → FE + BE :dev    → dev.DOMAIN   (DB wipe + fresh schema)
+#   Push to main    → FE + BE :test   → test.DOMAIN  (migrate only)
+#   Tag v*          → FE + BE :latest → www.DOMAIN   (migrate only)
+#                   → Unified :latest → Docker Hub    (not deployed)
+#
+# Database lifecycle:
+#   CI: bootstrap (admin creds) → create DB + role + extensions
+#   CI: dev wipe (app creds)    → drop tables for fresh install test
+#   BE: auto-migrate on startup → schema_migrations tracking
+#
+# Shared PG safety:
+#   - Bootstrap uses IF NOT EXISTS (idempotent)
+#   - Dev wipe REFUSES to run on databases not ending in _dev
+#   - Admin creds never reach the backend pods
+#   - Each env has its own DB name (chat_switchboard_{dev,test,})
 #
 # Required Gitea Variables:
 #   REGISTRY, NAMESPACE, DOMAIN, POSTGRES_HOST
@@ -46,7 +60,6 @@ env:
  POSTGRES_PORT: ${{ vars.POSTGRES_PORT || '5432' }}
  FE_IMAGE: ${{ vars.REGISTRY }}/xcaliber/switchboard-fe
  BE_IMAGE: ${{ vars.REGISTRY }}/xcaliber/switchboard-be
-  UNIFIED_IMAGE: ${{ vars.REGISTRY }}/xcaliber/switchboard
  DOCKERHUB_IMAGE: ${{ vars.DOCKERHUB_IMAGE || 'gobha/chat-switchboard' }}

 jobs:
@@ -157,7 +170,10 @@ jobs:
            echo "env_label=test (main)"                     >> "$GITHUB_OUTPUT"
          fi

-      # ── Database Bootstrap ───────────────────────
+      # ── Database Bootstrap (admin creds) ───────
+      # Creates the database and app role if they don't exist.
+      # Idempotent — safe to run every build. Uses admin creds
+      # that the backend pods never see.
      - name: Bootstrap database
        env:
          PGHOST: ${{ env.POSTGRES_HOST }}
@@ -171,18 +187,114 @@ jobs:
          chmod +x scripts/db-bootstrap.sh
          scripts/db-bootstrap.sh

-      # ── Database Migration ───────────────────────
-      - name: Run migrations
+      # ── Dev: Upgrade Test (dev only) ───────────
+      # The dev DB still has the PREVIOUS PR's schema.
+      # Apply THIS PR's migrations to test the upgrade path.
+      # If migration fails → CI fails before build.
+      - name: "Dev: test migration upgrade"
+        if: steps.env.outputs.DB_WIPE == 'true'
        env:
          PGHOST: ${{ env.POSTGRES_HOST }}
          PGPORT: ${{ env.POSTGRES_PORT }}
          PGUSER: ${{ secrets.POSTGRES_USER }}
          PGPASSWORD: ${{ secrets.POSTGRES_PASSWORD }}
-          DB_NAME: ${{ steps.env.outputs.DB_NAME }}
-          DB_WIPE: ${{ steps.env.outputs.DB_WIPE }}
+          PGDATABASE: ${{ steps.env.outputs.DB_NAME }}
        run: |
-          chmod +x scripts/db-migrate.sh
-          scripts/db-migrate.sh
+          echo "━━━ Upgrade test: applying migrations to existing dev DB ━━━"
+
+          # Ensure tracking table exists (first run or after manual wipe)
+          psql -v ON_ERROR_STOP=1 <<'SQL'
+          CREATE TABLE IF NOT EXISTS schema_migrations (
+              version    VARCHAR(255) PRIMARY KEY,
+              applied_at TIMESTAMPTZ DEFAULT NOW()
+          );
+          SQL
+
+          APPLIED=0
+          SKIPPED=0
+          FAILED=0
+
+          for migration in server/database/migrations/*.sql; do
+            [ -f "${migration}" ] || continue
+            VERSION=$(basename "${migration}")
+
+            ALREADY=$(psql -tAc "SELECT 1 FROM schema_migrations WHERE version = '${VERSION}';" || echo "0")
+            if [[ "${ALREADY}" == "1" ]]; then
+              echo "  skip: ${VERSION}"
+              SKIPPED=$((SKIPPED + 1))
+              continue
+            fi
+
+            echo "  apply: ${VERSION}..."
+            if psql -v ON_ERROR_STOP=1 -f "${migration}"; then
+              psql -c "INSERT INTO schema_migrations (version) VALUES ('${VERSION}');"
+              echo "  ✓ ${VERSION}"
+              APPLIED=$((APPLIED + 1))
+            else
+              echo "  ✗ ${VERSION} FAILED"
+              FAILED=$((FAILED + 1))
+            fi
+          done
+
+          echo ""
+          echo "━━━ Upgrade test: ${APPLIED} applied, ${SKIPPED} skipped, ${FAILED} failed ━━━"
+          if [[ ${FAILED} -gt 0 ]]; then
+            echo "❌ Migration upgrade test failed — fix before merge"
+            exit 1
+          fi
+
+      # ── Dev: Validate Schema ───────────────────
+      # After upgrade, verify all expected tables/columns exist.
+      # Catches migrations that apply cleanly but produce wrong schema.
+      - name: "Dev: validate schema"
+        if: steps.env.outputs.DB_WIPE == 'true'
+        env:
+          PGHOST: ${{ env.POSTGRES_HOST }}
+          PGPORT: ${{ env.POSTGRES_PORT }}
+          PGUSER: ${{ secrets.POSTGRES_USER }}
+          PGPASSWORD: ${{ secrets.POSTGRES_PASSWORD }}
+          PGDATABASE: ${{ steps.env.outputs.DB_NAME }}
+        run: |
+          chmod +x scripts/db-validate.sh
+          scripts/db-validate.sh
+
+      # ── Dev Wipe (fresh install test) ──────────
+      # Upgrade test passed. Now wipe so the deploy tests a
+      # full fresh-install migration via the backend binary.
+      # SAFETY: hard-refuses on any database not ending in _dev.
+      - name: "Dev: wipe for fresh install test"
+        if: steps.env.outputs.DB_WIPE == 'true'
+        env:
+          PGHOST: ${{ env.POSTGRES_HOST }}
+          PGPORT: ${{ env.POSTGRES_PORT }}
+          PGUSER: ${{ secrets.POSTGRES_USER }}
+          PGPASSWORD: ${{ secrets.POSTGRES_PASSWORD }}
+          PGDATABASE: ${{ steps.env.outputs.DB_NAME }}
+        run: |
+          DB="${{ steps.env.outputs.DB_NAME }}"
+
+          # ── SAFETY: only wipe databases ending in _dev ──
+          if [[ "${DB}" != *_dev ]]; then
+            echo "❌ REFUSING to wipe '${DB}' — only *_dev databases can be wiped"
+            exit 1
+          fi
+
+          echo "⚠ Dev wipe: dropping all tables in ${DB}"
+          psql <<'SQL'
+          DO $$ DECLARE
+              r RECORD;
+          BEGIN
+              FOR r IN (SELECT tablename FROM pg_tables WHERE schemaname = 'public') LOOP
+                  EXECUTE 'DROP TABLE IF EXISTS public.' || quote_ident(r.tablename) || ' CASCADE';
+              END LOOP;
+          END $$;
+          SQL
+          echo "✓ Dev wipe complete — backend will rebuild schema on startup"
+
+      # Dev exercises BOTH paths:
+      #   1. Upgrade test above:     existing schema → new migrations (psql)
+      #   2. Fresh install on deploy: empty DB → all migrations (backend binary)
+      # Test/Prod: backend applies only pending migrations on startup.

      # ── Build Backend Image ──────────────────────
      - name: Build backend image
@@ -191,7 +303,7 @@ jobs:
            -t ${BE_IMAGE}:${{ steps.env.outputs.IMAGE_TAG }} \
            ./server

-      # ── Build Frontend Image (cluster mode) ──────
+      # ── Build Frontend Image ─────────────────────
      - name: Build frontend image
        run: |
          docker build -f Dockerfile.frontend \
@@ -215,27 +327,22 @@ jobs:
            docker push ${FE_IMAGE}:${{ steps.env.outputs.EXTRA_TAG }}
          fi

-      # ── Build + Push Unified (release only) ──────
-      - name: Build unified image
+      # ── Build + Push Unified to Docker Hub (release only) ──
+      - name: Build and push unified image
        if: steps.env.outputs.is_release == 'true'
        run: |
-          docker build -f Dockerfile.unified \
-            -t ${UNIFIED_IMAGE}:latest \
-            -t ${UNIFIED_IMAGE}:${{ steps.env.outputs.EXTRA_TAG }} \
+          docker build -f Dockerfile \
+            -t ${DOCKERHUB_IMAGE}:latest \
+            -t ${DOCKERHUB_IMAGE}:${{ steps.env.outputs.EXTRA_TAG }} \
            .
-          docker push ${UNIFIED_IMAGE}:latest
-          docker push ${UNIFIED_IMAGE}:${{ steps.env.outputs.EXTRA_TAG }}
-
-      - name: Push unified to Docker Hub
-        if: steps.env.outputs.is_release == 'true' && secrets.DOCKERHUB_TOKEN != ''
-        continue-on-error: true
-        run: |
-          echo "${{ secrets.DOCKERHUB_TOKEN }}" | \
-            docker login -u "${{ secrets.DOCKERHUB_USERNAME }}" --password-stdin
-          docker tag ${UNIFIED_IMAGE}:latest ${DOCKERHUB_IMAGE}:latest
-          docker tag ${UNIFIED_IMAGE}:${{ steps.env.outputs.EXTRA_TAG }} ${DOCKERHUB_IMAGE}:${{ steps.env.outputs.EXTRA_TAG }}
-          docker push ${DOCKERHUB_IMAGE}:latest
-          docker push ${DOCKERHUB_IMAGE}:${{ steps.env.outputs.EXTRA_TAG }}
+          if [[ -n "${{ secrets.DOCKERHUB_TOKEN }}" ]]; then
+            echo "${{ secrets.DOCKERHUB_TOKEN }}" | \
+              docker login -u "${{ secrets.DOCKERHUB_USERNAME }}" --password-stdin
+            docker push ${DOCKERHUB_IMAGE}:latest
+            docker push ${DOCKERHUB_IMAGE}:${{ steps.env.outputs.EXTRA_TAG }}
+          else
+            echo "⚠ Docker Hub credentials not configured, skipping push"
+          fi

      # ── Deploy to Kubernetes ─────────────────────
      - name: Ensure namespace
@@ -278,10 +385,10 @@ jobs:
          kubectl apply -f /tmp/ingress.yaml

      - name: Rollout and verify
+        env:
+          SUFFIX: ${{ steps.env.outputs.DEPLOY_SUFFIX }}
+          ENV: ${{ steps.env.outputs.ENVIRONMENT }}
        run: |
-          SUFFIX="${{ steps.env.outputs.DEPLOY_SUFFIX }}"
-          ENV="${{ steps.env.outputs.ENVIRONMENT }}"
-
          kubectl rollout restart deployment/switchboard-be${SUFFIX} -n ${NAMESPACE}
          kubectl rollout restart deployment/switchboard-fe${SUFFIX} -n ${NAMESPACE}

@@ -303,6 +410,25 @@ jobs:
            exit 1
          fi

+      # ── Post-deploy: verify schema ─────────────
+      - name: Verify schema migration
+        run: |
+          sleep 5
+          HOST="${{ steps.env.outputs.DEPLOY_HOST }}"
+          HEALTH=$(curl -sf "https://${HOST}/api/v1/health" 2>/dev/null || echo "unreachable")
+          echo "Health: ${HEALTH}"
+
+          # Extract fields (|| true prevents set -e from killing on no-match)
+          SCHEMA=$(echo "${HEALTH}" | grep -o '"schema_version":"[^"]*"' | cut -d'"' -f4 || true)
+          DB_OK=$(echo "${HEALTH}" | grep -o '"database":true' || true)
+
+          if [[ -n "${SCHEMA}" ]] && [[ -n "${DB_OK}" ]]; then
+            echo "✅ Schema: ${SCHEMA} | DB: connected"
+          else
+            echo "⚠ Could not verify schema (backend may still be starting)"
+            echo "  Response: ${HEALTH:0:200}"
+          fi
+
      - name: Summary
        run: |
          HOST="${{ steps.env.outputs.DEPLOY_HOST }}"
@@ -314,46 +440,7 @@ jobs:
          echo "| **Backend** | \`${BE_IMAGE}:${{ steps.env.outputs.IMAGE_TAG }}\` |" >> $GITHUB_STEP_SUMMARY
          echo "| **Frontend** | \`${FE_IMAGE}:${{ steps.env.outputs.IMAGE_TAG }}\` |" >> $GITHUB_STEP_SUMMARY
          echo "| **Database** | \`${{ steps.env.outputs.DB_NAME }}\` |" >> $GITHUB_STEP_SUMMARY
+          echo "| **DB Wipe** | ${{ steps.env.outputs.DB_WIPE }} |" >> $GITHUB_STEP_SUMMARY
          if [[ "${{ steps.env.outputs.is_release }}" == "true" ]]; then
            echo "| **Docker Hub** | \`${DOCKERHUB_IMAGE}:${{ steps.env.outputs.EXTRA_TAG }}\` (unified) |" >> $GITHUB_STEP_SUMMARY
          fi
-
-  # ── Stage 3: FE-Only Lite Deploy (main only) ─
-  deploy-lite:
-    runs-on: ubuntu-latest
-    if: gitea.ref == 'refs/heads/main'
-    steps:
-      - name: Checkout
-        uses: actions/checkout@v4
-
-      - name: Add /tools to PATH
-        run: echo "/tools" >> $GITHUB_PATH
-
-      - name: Install tools
-        run: |
-          apt-get update -qq && apt-get install -y -qq gettext-base > /dev/null
-
-      - name: Build lite image
-        run: |
-          docker build -f Dockerfile.frontend \
-            -t ${FE_IMAGE}:lite \
-            .
-
-      - name: Push lite image
-        run: docker push ${FE_IMAGE}:lite
-
-      - name: Deploy lite
-        env:
-          ENVIRONMENT: lite
-          IMAGE_TAG: lite
-          DEPLOY_HOST: "lite.${{ vars.DOMAIN || 'gobha.ai' }}"
-        run: |
-          export NAMESPACE="${NAMESPACE}"
-          export FE_IMAGE="${FE_IMAGE}"
-          export CERT_ISSUER="${CERT_ISSUER}"
-          envsubst < k8s/lite.yaml > /tmp/lite.yaml
-          kubectl create namespace ${NAMESPACE} 2>/dev/null || true
-          kubectl apply -f /tmp/lite.yaml
-          kubectl rollout restart deployment/switchboard-lite -n ${NAMESPACE}
-          kubectl rollout status deployment/switchboard-lite -n ${NAMESPACE} --timeout=120s
-          echo "✅ Lite deployed to https://${DEPLOY_HOST}"