feat: make infra provisioning idempotent + aeries-daeya public discovery feed

- Make postgres and redis provisioning idempotent: return success when already
  provisioned with credentials stored, allowing cookbook trees to safely include
  explicit add-db/add-redis steps alongside auto-provisioned project creation
- Update tests to reflect new idempotent behavior
- Consolidate docs CI into single multi-stage Docker build (remove separate
  build-docs step; Dockerfile.nginx now builds Slate then serves with nginx)
- Delete redundant skeleton docs/Dockerfile (replaced by multi-stage nginx image)
- Add watch verb to woodpecker-deployer RBAC (required by kubectl rollout status)
- Aeries Daeya cookbook: add public discovery feed (/) + character profiles (/c/:handle),
  characters.published/handle/tagline fields, dark pink design system, /studio/* routes,
  verify-public-discovery + verify-otp-endpoint smoke test steps
- Fix Input.tsx: remove non-existent --border-hover CSS variable hover effect

cookbooks/scripts/aeries-daeya-test.sh

@@ -47,29 +47,32 @@ run_flow() {
     echo "Running tree: aeries-daeya for project: $PROJECT_NAME"
     echo ""
     echo "  Phase 1 — Infrastructure (DB + Redis + service + worker + app)"
-    echo "  Phase 2 — Avatar & Look Data Model (characters, looks, albums, posts, mutations)"
+    echo "  Phase 2 — Avatar & Look Data Model (characters + public discovery endpoints)"
     echo "  Phase 3 — AI Generation Pipeline (portrait gen, outfit styling, mutation explorer)"
-    echo "  Phase 4 — Studio UI (creation wizard, mutation explorer, look panel, albums)"
+    echo "  Phase 4 — Studio UI (public feed + character profiles + studio + dark pink theme)"
     echo ""
 
     "$SCRIPT_DIR/tree-runner.sh" run aeries-daeya \
         --project-name "$PROJECT_NAME" \
-        ${AUTO_TEARDOWN:+--auto-teardown}
+        $([[ "$AUTO_TEARDOWN" == "true" ]] && echo "--auto-teardown")
 
     DOMAIN=$(api_call GET "/project/$PROJECT_NAME" | jq -r '.data.domain // empty')
 
     if [[ -n "$DOMAIN" ]]; then
         print_success "Aeries Daeya is live at https://$DOMAIN"
         echo ""
-        echo "  Studio:       https://$DOMAIN"
+        echo "  Discovery:    https://$DOMAIN                         (public — no login)"
+        echo "  Studio:       https://$DOMAIN/studio                  (requires login)"
         echo "  API health:   https://$DOMAIN/api/daeya-api/health"
-        echo "  Characters:   https://$DOMAIN/api/daeya-api/characters  (requires auth)"
+        echo "  Public feed:  https://$DOMAIN/api/daeya-api/characters/public"
         echo ""
         echo "  Flow:"
-        echo "    1. Open https://$DOMAIN → login with OTP"
+        echo "    1. Open https://$DOMAIN → browse published characters (no login needed)"
-        echo "    2. Create Character → 4-step wizard (describe, shape, soul, generate)"
+        echo "    2. Click 'Create Your Character' → OTP login → /studio"
-        echo "    3. Open Character → use Mutation Explorer to adjust skin tone, background, lighting"
+        echo "    3. Create Character → 4-step wizard (describe, shape, soul, generate)"
-        echo "    4. Add Look → upload a photo of an outfit to style on your character"
+        echo "    4. In studio: toggle 'Public' → character appears on discovery feed"
+        echo "    5. Open Character → Mutation Explorer (skin tone, background, lighting, style)"
+        echo "    6. Add Look → upload a photo of an outfit to style on your character"
     fi
 }
 
@@ -114,6 +117,18 @@ diagnose() {
         diagnose_site_failure "$domain" "$PROJECT_NAME"
     fi
 
+    print_diagnostic_header "Notify / Email Checks"
+    echo ""
+    echo "  If OTP email never arrives:"
+    print_fix "Check NOTIFY_API_KEY is injected into the daeya-api pod"
+    print_cmd "kubectl exec -n projects deploy/$PROJECT_NAME-daeya-api -- env | grep NOTIFY"
+    echo ""
+    print_fix "If NOTIFY_API_KEY is empty, notify provisioning failed during project creation"
+    print_cmd "curl -s -H 'X-API-Key: \$RDEV_API_KEY' \$RDEV_API_URL/projects/$PROJECT_NAME/notify/status | jq"
+    echo ""
+    print_fix "Test OTP endpoint directly (200/202/429 = notify working; 404/500 = broken)"
+    print_cmd "curl -s -o /dev/null -w '%{http_code}' -X POST https://$domain/api/daeya-api/auth/otp/send -H 'Content-Type: application/json' -d '{\"email\":\"test@example.com\"}'"
+
     print_diagnostic_header "AI Generation Checks"
     echo ""
     echo "  If characters are stuck in 'pending' status:"
@@ -127,6 +142,16 @@ diagnose() {
     echo "  If look generation fails with photo upload:"
     print_fix "GEMINI_API_KEY may not be injected or vision endpoint unreachable"
     print_cmd "kubectl exec -n projects deploy/$PROJECT_NAME-media-worker -- env | grep GEMINI"
+
+    print_diagnostic_header "Public Discovery Checks"
+    echo ""
+    echo "  If /characters/public returns 404:"
+    print_fix "implement-avatar-model build may not have created public endpoints"
+    print_cmd "curl -s https://$domain/api/daeya-api/characters/public | jq"
+    echo ""
+    echo "  If published characters don't show on the discovery feed:"
+    print_fix "Character may not be toggled to published=true in the studio"
+    print_cmd "curl -s -H 'Authorization: Bearer <token>' https://$domain/api/daeya-api/characters | jq '.[].published'"
 }
 
 teardown() {

deployments/k8s/base/woodpecker-deployer-rbac.yaml

@@ -22,15 +22,19 @@ metadata:
     app.kubernetes.io/part-of: rdev
 rules:
   # Deploy steps: set image, patch replicas, verify rollout
-  # - get/list: read deployment and replicaset state
+  # - get/list/watch: read deployment and replicaset state (watch required by kubectl rollout status)
   # - patch: kubectl set image, kubectl patch (replicas)
   - apiGroups: ["apps"]
     resources: ["deployments"]
-    verbs: ["get", "list", "patch"]
+    verbs: ["get", "list", "patch", "watch"]
-  # rollout status needs to watch replicasets
+  # rollout status watches replicasets to track new/old replica counts
   - apiGroups: ["apps"]
     resources: ["replicasets"]
-    verbs: ["get", "list"]
+    verbs: ["get", "list", "watch"]
+  # rollout status watches pods to detect readiness and crash loops
+  - apiGroups: [""]
+    resources: ["pods"]
+    verbs: ["get", "list", "watch"]
 ---
 apiVersion: rbac.authorization.k8s.io/v1
 kind: RoleBinding

internal/adapter/templates/templates/skeleton/.woodpecker.yml.tmpl

@@ -143,29 +143,11 @@ steps:
       branch: main
       event: push
 
-  # Build Slate static documentation (skipped if no docs infrastructure)
+  # Build and push docs-nginx image (multi-stage: builds Slate + serves with nginx)
-  build-docs:
+  # Depends on generate-docs which produces markdown includes from OpenAPI specs
-    image: ruby:3.2-slim
+  # failure: ignore allows pipeline to continue if docs build fails
-    depends_on: [generate-docs]
-    failure: ignore
-    commands:
-      - |
-        if [ ! -d "docs" ] || [ ! -f "docs/Gemfile" ]; then
-          echo "==> No docs/ directory or Gemfile found, skipping Slate build"
-          exit 0
-        fi
-      - apt-get update && apt-get install -y build-essential nodejs
-      - cd docs && bundle install --jobs 4
-      - cd docs && bundle exec middleman build --clean
-      - echo "==> Docs built to docs/build/"
-    when:
-      branch: main
-      event: push
-
-  # Build and push docs-nginx image (skipped if no docs build output)
-  # failure: ignore allows pipeline to continue if docs weren't built
   build-docs-image:
-    depends_on: [build-docs]
+    depends_on: [generate-docs]
     image: woodpeckerci/plugin-kaniko
     failure: ignore
     settings:
@@ -194,16 +176,8 @@ steps:
         REPO="{{PROJECT_NAME}}-docs"
         REGISTRY="registry.threesix.ai"
 
-        # Check if docs were built (same check as deploy-docs)
-        if [ ! -d "docs/build" ]; then
-          echo "==> No docs build output, skipping verification"
-          exit 0
-        fi
-
         echo "==> Verifying image $REGISTRY/$REPO:$TAG exists in registry"
 
-        # Query registry v2 API to check if manifest exists
-        # Returns 200 if image exists, 404 if not
         HTTP_CODE=$(curl -s -o /dev/null -w "%{http_code}" \
           --insecure \
           "https://$REGISTRY/v2/$REPO/manifests/$TAG" \
@@ -211,13 +185,10 @@ steps:
 
         if [ "$HTTP_CODE" = "200" ]; then
           echo "==> Image verified: $REGISTRY/$REPO:$TAG"
-          # Create marker file for deploy-docs to check
-          touch /tmp/image-verified
           exit 0
         elif [ "$HTTP_CODE" = "404" ]; then
           echo "==> WARNING: Image $REGISTRY/$REPO:$TAG not found in registry"
-          echo "    This may indicate the build step failed or is still pushing"
+          echo "    Build step may have failed. Deploy will be skipped."
-          echo "    Deploy step will be skipped to prevent ImagePullBackOff"
           exit 1
         else
           echo "==> WARNING: Registry check returned HTTP $HTTP_CODE"

internal/adapter/templates/templates/skeleton/docs/Dockerfile

@@ -1,27 +0,0 @@
-# Slate documentation builder
-# Used by CI to generate static HTML from OpenAPI specs
-
-FROM ruby:3.2-slim
-
-# Install build dependencies
-RUN apt-get update && apt-get install -y \
-    build-essential \
-    nodejs \
-    npm \
-    git \
-    && rm -rf /var/lib/apt/lists/*
-
-# Install widdershins globally for OpenAPI to Slate markdown conversion
-RUN npm install -g widdershins
-
-WORKDIR /docs
-
-# Copy Gemfile first for layer caching
-COPY Gemfile Gemfile.lock* ./
-RUN bundle install
-
-# Copy the rest of the docs source
-COPY . .
-
-# Build static site
-CMD ["bundle", "exec", "middleman", "build", "--clean"]

internal/adapter/templates/templates/skeleton/docs/Dockerfile.nginx

@@ -1,13 +1,35 @@
 # Production nginx image for serving Slate documentation
-# Built by CI after Slate generates static HTML
+# Multi-stage build: generates static HTML then serves with nginx
+# No dependency on external build steps — self-contained
 
+# Stage 1: Build Slate static site
+FROM ruby:3.2-slim AS builder
+
+RUN apt-get update && apt-get install -y \
+    build-essential \
+    nodejs \
+    && rm -rf /var/lib/apt/lists/*
+
+WORKDIR /docs
+
+# Copy Gemfile first for layer caching
+COPY Gemfile Gemfile.lock* ./
+RUN bundle install --jobs 4
+
+# Copy the rest of the docs source
+COPY . .
+
+# Build static site — produces build/ directory
+RUN bundle exec middleman build --clean
+
+# Stage 2: Serve with nginx
 FROM nginx:alpine
 
 # Remove default nginx content
 RUN rm -rf /usr/share/nginx/html/*
 
-# Copy built static files from Slate
+# Copy built static files from Slate builder
-COPY build/ /usr/share/nginx/html/
+COPY --from=builder /docs/build/ /usr/share/nginx/html/
 
 # Custom nginx config for SPA-style routing
 RUN cat > /etc/nginx/conf.d/default.conf << 'EOF'

internal/adapter/templates/templates/skeleton/packages/ui/src/components/Input.tsx

@@ -14,7 +14,7 @@ const Input = React.forwardRef<HTMLInputElement, InputProps>(
           'flex h-9 w-full rounded-md border bg-transparent px-3 py-1 text-sm shadow-sm transition-colors file:border-0 file:bg-transparent file:text-sm file:font-medium placeholder:text-[var(--text-muted)] focus-visible:outline-none focus-visible:ring-2 focus-visible:ring-[var(--accent)] disabled:cursor-not-allowed disabled:opacity-50',
           error
             ? 'border-[var(--error)] focus-visible:ring-[var(--error)]'
-            : 'border-[var(--border)] hover:border-[var(--border-hover)]',
+            : 'border-[var(--border)]',
           className
         )}
         ref={ref}

internal/service/component_infra.go

@@ -24,6 +24,7 @@ func (s *ComponentService) addInfraComponent(ctx context.Context, projectID stri
 }
 
 // provisionPostgres provisions a PostgreSQL/CockroachDB database for the project.
+// Idempotent: returns existing component if already provisioned with credentials stored.
 func (s *ComponentService) provisionPostgres(ctx context.Context, projectID, name string) (*domain.Component, error) {
 	if s.dbProvisioner == nil {
 		return nil, fmt.Errorf("database provisioner not configured")
@@ -35,7 +36,41 @@ func (s *ComponentService) provisionPostgres(ctx context.Context, projectID, nam
 		return nil, fmt.Errorf("failed to check existing database: %w", err)
 	}
 	if existing != nil {
-		return nil, fmt.Errorf("%w: postgres already provisioned for project %s", domain.ErrDuplicateComponent, projectID)
+		// Already provisioned — return success if credentials are stored (idempotent).
+		// This allows cookbook trees to have explicit add-db steps even though
+		// project creation auto-provisions the database.
+		if s.credentialStore != nil {
+			storedURL, storeErr := s.credentialStore.Get(ctx, projectID+":DATABASE_URL")
+			if storeErr == nil && storedURL != "" {
+				log := logging.FromContext(ctx).WithService("component")
+				log.Info("postgres already provisioned, returning existing (idempotent)",
+					logging.FieldProjectID, projectID)
+				return &domain.Component{
+					Type:         domain.ComponentTypePostgres,
+					Name:         name,
+					Path:         "infra/postgres",
+					Port:         existing.Port,
+					Template:     "postgres",
+					Dependencies: []string{},
+				}, nil
+			}
+			// Credentials missing — fall through to re-provision
+			log := logging.FromContext(ctx).WithService("component")
+			log.Warn("database exists but DATABASE_URL not in credential store, re-provisioning",
+				logging.FieldProjectID, projectID)
+		} else {
+			log := logging.FromContext(ctx).WithService("component")
+			log.Info("postgres already provisioned, returning existing (idempotent)",
+				logging.FieldProjectID, projectID)
+			return &domain.Component{
+				Type:         domain.ComponentTypePostgres,
+				Name:         name,
+				Path:         "infra/postgres",
+				Port:         existing.Port,
+				Template:     "postgres",
+				Dependencies: []string{},
+			}, nil
+		}
 	}
 
 	// Provision the database
@@ -88,19 +123,41 @@ func (s *ComponentService) provisionRedis(ctx context.Context, projectID, name s
 		return nil, fmt.Errorf("failed to check existing cache: %w", err)
 	}
 	if existing != nil {
-		// Redis user exists — check if credentials are stored. If they are, it's a true duplicate.
+		// Redis user exists — check if credentials are stored.
-		// If not (credentials were lost), fall through to re-provision (CreateProjectCache resets the password).
+		// If they are, return success (idempotent) so cookbook trees can safely
+		// have add-redis steps even though project creation auto-provisions Redis.
+		// If not (credentials were lost), fall through to re-provision.
 		if s.credentialStore != nil {
 			storedURL, storeErr := s.credentialStore.Get(ctx, projectID+":REDIS_URL")
 			if storeErr == nil && storedURL != "" {
-				return nil, fmt.Errorf("%w: redis already provisioned for project %s", domain.ErrDuplicateComponent, projectID)
+				log := logging.FromContext(ctx).WithService("component")
+				log.Info("redis already provisioned, returning existing (idempotent)",
+					logging.FieldProjectID, projectID)
+				return &domain.Component{
+					Type:         domain.ComponentTypeRedis,
+					Name:         name,
+					Path:         "infra/redis",
+					Port:         existing.Port,
+					Template:     "redis",
+					Dependencies: []string{},
+				}, nil
 			}
 			// Credentials missing from store — re-provision to recover
 			log := logging.FromContext(ctx).WithService("component")
 			log.Warn("redis user exists but REDIS_URL not in credential store, re-provisioning",
 				logging.FieldProjectID, projectID)
 		} else {
-			return nil, fmt.Errorf("%w: redis already provisioned for project %s", domain.ErrDuplicateComponent, projectID)
+			log := logging.FromContext(ctx).WithService("component")
+			log.Info("redis already provisioned, returning existing (idempotent)",
+				logging.FieldProjectID, projectID)
+			return &domain.Component{
+				Type:         domain.ComponentTypeRedis,
+				Name:         name,
+				Path:         "infra/redis",
+				Port:         existing.Port,
+				Template:     "redis",
+				Dependencies: []string{},
+			}, nil
 		}
 	}
 

internal/service/component_test.go

@@ -283,15 +283,18 @@ func TestProvisionPostgres(t *testing.T) {
 			wantErr:       false,
 		},
 		{
-			name:          "postgres already exists",
+			name:          "postgres already exists with credentials (idempotent)",
 			projectID:     "test-project",
 			componentName: "main-db",
 			dbProvisioner: &mockDatabaseProvisioner{
-				existingDB: &domain.DatabaseCredentials{ProjectID: "test-project"},
+				existingDB: &domain.DatabaseCredentials{ProjectID: "test-project", Port: 26257},
 			},
-			credStore:       newMockCredentialStore(),
+			credStore: func() *mockCredentialStore {
-			wantErr:         true,
+				cs := newMockCredentialStore()
-			wantErrContains: "already provisioned",
+				cs.stored["test-project:DATABASE_URL"] = "postgresql://proj-test-project:pass@localhost:26257/project_test_project"
+				return cs
+			}(),
+			wantErr: false,
 		},
 		{
 			name:          "provisioning fails",
@@ -369,20 +372,19 @@ func TestProvisionRedis(t *testing.T) {
 			wantErr:          false,
 		},
 		{
-			name:          "redis already exists",
+			name:          "redis already exists with credentials (idempotent)",
 			projectID:     "test-project",
 			componentName: "cache",
 			cacheProvisioner: &mockCacheProvisioner{
-				existingCache: &domain.CacheCredentials{ProjectID: "test-project"},
+				existingCache: &domain.CacheCredentials{ProjectID: "test-project", Port: 6379},
 			},
 			credStore: func() *mockCredentialStore {
-				// Simulate credentials already stored — this is a true duplicate
+				// Simulate credentials already stored — idempotent return
 				cs := newMockCredentialStore()
 				cs.stored["test-project:REDIS_URL"] = "redis://proj-test-project:pass@localhost:6379"
 				return cs
 			}(),
-			wantErr:         true,
+			wantErr: false,
-			wantErrContains: "already provisioned",
 		},
 		{
 			name:          "provisioning fails",