feat: add GCS-based persistent media storage, AI generation pipeline, and composable skeleton packages

Adds complete media storage pipeline with GCS presigned uploads, AI image/video/text generation
via queue-based workers, realtime SSE event streaming, and comprehensive skeleton packages
(storage, mediagen, textgen, generation, realtime, persona, routing, ai-client). Includes
security fixes for media delete authorization, nil pointer guards in handlers, video persistence
via download-then-upload, consistent signed URLs, and Image→ImageIcon rename to avoid DOM collision.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

.gitignore

@@ -43,3 +43,7 @@ tmp/
 /rdev-api
 /claudebox-sidecar
 /sdlc
+/render-skeleton
+
+# Rendered example monorepo (regenerated from templates)
+examples/full-monorepo/

CLAUDE.md

@@ -37,6 +37,7 @@ When discussing code: "add to **platform**" = edit rdev; "add to **skeleton**" =
 | **Cookbook tree system (commands)** | [services/cookbook-trees.md](.claude/guides/services/cookbook-trees.md) |
 | **Slackpath reference architectures** | [services/cookbook-trees.md](.claude/guides/services/cookbook-trees.md#slackpath-trees-reference-architectures) |
 | **Write cookbook trees** | [cookbook-trees/SKILL.md](.claude/skills/cookbook-trees/SKILL.md) |
+| **Build/maintain skeleton packages** | [skeleton-craftsman/SKILL.md](.claude/skills/skeleton-craftsman/SKILL.md) |
 | **Build orchestration** | [services/build-orchestration.md](.claude/guides/services/build-orchestration.md) |
 | **Build event streaming** | [services/build-streaming.md](.claude/guides/services/build-streaming.md) |
 | **Resource provisioning plan** | [services/resource-provisioning-plan.md](.claude/guides/services/resource-provisioning-plan.md) |
@@ -55,10 +56,13 @@ When discussing code: "add to **platform**" = edit rdev; "add to **skeleton**" =
 | **Woodpecker CI v3 pipelines** | [ops/woodpecker-v3.md](.claude/guides/ops/woodpecker-v3.md) |
 | **Traefik v3 ingress & middleware** | [ops/traefik-v3.md](.claude/guides/ops/traefik-v3.md) |
 | **Zot container registry** | [ops/zot-registry.md](.claude/guides/ops/zot-registry.md) |
+| **cert-manager / TLS certificates** | [ops/cert-manager.md](.claude/guides/ops/cert-manager.md) |
 | **Structured logging** | `internal/logging/` - field constants, context propagation, redaction |
 
 ## Critical Rules
 
+- **Frustration = systemic fix:** When the user says they're tired of repeating something, stop what you're doing and find or create a systemic fix in `.claude/**/*` or `CLAUDE.md` — don't just apologize and do the same thing again.
+- **AI credentials are provisioned:** rdev injects `LAOZHANG_API_KEY` and `GEMINI_API_KEY` as env vars into every deployed component (`component_deploy.go:fetchProjectCredentials`). Skeleton code reads them with `os.Getenv()`. Never treat AI packages as needing external setup.
 - **Root cause fixes:** When diagnosing failures in generated projects, NEVER patch the project directly. Find the systemic root cause in: (1) **platform** - rdev handlers/services that create resources, (2) **skeleton** - templates that ship in generated projects, or (3) **cookbook** - test scripts with wrong assumptions. Fix the source, not the symptom. Every project-specific fix is technical debt that will recur.
 - **LLM vs rdev:** LLMs generate code; rdev executes deterministic operations (git, lint, deploy). Never rely on LLMs for runbook tasks.
 - **Pod git ops:** Git operations run inside pods via `PodGitOperations` (kubectl exec), never locally.

cmd/rdev-api/config.go

@@ -34,6 +34,11 @@ type Config struct {
 	// Internal API token for service-to-service callbacks
 	InternalToken string
 
+	// Citadel logging integration
+	CitadelURL              string // e.g., "https://citadel-staging.orchard9.ai"
+	CitadelAPIKey           string // API key for Citadel (starts with ck_live_ or ck_dev_)
+	CitadelPlatformTenantID string // Tenant ID for the rdev-platform environment
+
 	// Infrastructure adapters (threesix.ai) - fallback values if not in credential store
 	GiteaURL                string
 	GiteaToken              string
@@ -107,6 +112,11 @@ func loadConfig() Config {
 		// Internal API token for service-to-service callbacks (e.g., SDLC callbacks)
 		InternalToken: os.Getenv("INTERNAL_TOKEN"),
 
+		// Citadel logging integration
+		CitadelURL:              os.Getenv("CITADEL_URL"),                // e.g., "https://citadel-staging.orchard9.ai"
+		CitadelAPIKey:           os.Getenv("CITADEL_API_KEY"),            // API key for Citadel
+		CitadelPlatformTenantID: os.Getenv("CITADEL_PLATFORM_TENANT_ID"), // rdev-platform tenant ID
+
 		// Infrastructure adapters (fallback if not in credential store)
 		GiteaURL:                envutil.GetEnv("GITEA_URL", "https://git.threesix.ai"),
 		GiteaToken:              os.Getenv("GITEA_TOKEN"),

cmd/rdev-api/main.go

@@ -7,6 +7,7 @@ import (
 	"strings"
 	"time"
 
+	citadeladapter "github.com/orchard9/rdev/internal/adapter/citadel"
 	"github.com/orchard9/rdev/internal/adapter/cloudflare"
 	"github.com/orchard9/rdev/internal/adapter/cockroach"
 	"github.com/orchard9/rdev/internal/adapter/codeagent"
@@ -96,6 +97,16 @@ func main() {
 	// Load infrastructure config from credential store (falls back to env vars)
 	infraCfg := loadInfraConfig(context.Background(), credentialStore, cfg, logger)
 
+	// Initialize Citadel client (optional - for log environment provisioning and audit shipping)
+	var citadelClient *citadeladapter.Client
+	if cfg.CitadelURL != "" && cfg.CitadelAPIKey != "" {
+		citadelClient = citadeladapter.NewClient(citadeladapter.Config{
+			URL:    cfg.CitadelURL,
+			APIKey: cfg.CitadelAPIKey,
+		}, logger)
+		logger.Info("citadel client initialized", "url", cfg.CitadelURL)
+	}
+
 	// Create adapters (dependency injection)
 	namespace := envutil.GetEnv("K8S_NAMESPACE", "rdev")
 
@@ -113,7 +124,16 @@ func main() {
 		}
 	}
 
-	auditLogger := postgres.NewAuditLogger(database.DB)
+	var auditLogger port.AuditLogger
+	var auditShipper *citadeladapter.AuditShipper
+	pgAuditLogger := postgres.NewAuditLogger(database.DB)
+	if citadelClient != nil && cfg.CitadelPlatformTenantID != "" {
+		auditShipper = citadeladapter.NewAuditShipper(pgAuditLogger, citadelClient, cfg.CitadelPlatformTenantID, logger)
+		auditLogger = auditShipper
+		logger.Info("audit logger wrapped with citadel shipper", "tenant_id", cfg.CitadelPlatformTenantID)
+	} else {
+		auditLogger = pgAuditLogger
+	}
 	rateLimiter := postgres.NewRateLimiter(database.DB)
 	stopRateLimitCleanup := rateLimiter.StartCleanupWorker(context.Background(), 5*time.Minute)
 	commandQueue := postgres.NewCommandQueueRepository(database.DB)
@@ -459,6 +479,9 @@ func main() {
 	if registryClient != nil {
 		projectInfraService = projectInfraService.WithRegistryProvider(registryClient)
 	}
+	if citadelClient != nil {
+		projectInfraService = projectInfraService.WithCitadelClient(citadelClient)
+	}
 
 	// Create domain service adapter for infrastructure handler
 	domainServiceAdapter := handlers.NewDomainServiceAdapter(projectInfraService)
@@ -761,6 +784,9 @@ func main() {
 		}
 		queueProcessor.Stop()
 		webhookDispatcher.Stop()
+		if auditShipper != nil {
+			auditShipper.Close()
+		}
 		projectRepo.StopWatching()
 		stopRateLimitCleanup()
 		closeProvisioner(dbProvisioner, "database", logger)

cmd/render-skeleton/main.go

@@ -0,0 +1,241 @@
+// Package main provides a CLI tool for rendering skeleton templates locally.
+//
+// This is used for testing templates without needing Gitea/rdev infrastructure.
+// The tool renders the monorepo skeleton plus all component types to a local directory.
+//
+// Usage:
+//
+//	go run ./cmd/render-skeleton -output ./examples/full-monorepo
+package main
+
+import (
+	"flag"
+	"fmt"
+	"maps"
+	"os"
+	"path/filepath"
+	"strings"
+
+	"github.com/orchard9/rdev/internal/adapter/templates"
+)
+
+// Component defines a component to render.
+type Component struct {
+	Type    string
+	Name    string
+	Port    string
+	DestDir string // "services", "workers", "apps", or "cli"
+}
+
+// defaultComponents are the components to render in a full example project.
+var defaultComponents = []Component{
+	{Type: "service", Name: "example-api", Port: "8001", DestDir: "services"},
+	{Type: "worker", Name: "example-worker", Port: "", DestDir: "workers"},
+	{Type: "app-astro", Name: "example-astro", Port: "4321", DestDir: "apps"},
+	{Type: "app-react", Name: "example-react", Port: "5173", DestDir: "apps"},
+	{Type: "app-nextjs", Name: "example-nextjs", Port: "3000", DestDir: "apps"},
+	{Type: "cli", Name: "example-cli", Port: "", DestDir: "cli"},
+}
+
+func main() {
+	outputDir := flag.String("output", "", "Output directory for rendered skeleton")
+	projectName := flag.String("project", "test-project", "Project name")
+	goModule := flag.String("module", "git.threesix.ai/threesix/test-project", "Go module path")
+	domain := flag.String("domain", "test.threesix.ai", "Domain for the project")
+	flag.Parse()
+
+	if *outputDir == "" {
+		fmt.Fprintln(os.Stderr, "Usage: render-skeleton -output <dir> [-project <name>] [-module <path>] [-domain <domain>]")
+		fmt.Fprintln(os.Stderr, "")
+		fmt.Fprintln(os.Stderr, "Example:")
+		fmt.Fprintln(os.Stderr, "  go run ./cmd/render-skeleton -output ./examples/full-monorepo")
+		os.Exit(1)
+	}
+
+	// Base variables for the project
+	vars := map[string]string{
+		"PROJECT_NAME": *projectName,
+		"GO_MODULE":    *goModule,
+		"DOMAIN":       *domain,
+		"DESCRIPTION":  "Test project for skeleton verification",
+		"GIT_URL":      "https://" + strings.TrimPrefix(*goModule, "git."),
+	}
+
+	fmt.Printf("Rendering skeleton to %s\n", *outputDir)
+	fmt.Printf("  PROJECT_NAME: %s\n", vars["PROJECT_NAME"])
+	fmt.Printf("  GO_MODULE:    %s\n", vars["GO_MODULE"])
+	fmt.Printf("  DOMAIN:       %s\n", vars["DOMAIN"])
+	fmt.Println()
+
+	// Clean output directory if it exists (but preserve .git if present)
+	if err := cleanOutputDir(*outputDir); err != nil {
+		fmt.Fprintf(os.Stderr, "Error cleaning output directory: %v\n", err)
+		os.Exit(1)
+	}
+
+	// Render skeleton
+	fmt.Println("Rendering skeleton...")
+	if err := templates.RenderSkeletonToDir(*outputDir, vars); err != nil {
+		fmt.Fprintf(os.Stderr, "Error rendering skeleton: %v\n", err)
+		os.Exit(1)
+	}
+
+	// Render all components
+	for _, c := range defaultComponents {
+		fmt.Printf("Rendering component: %s (%s)\n", c.Name, c.Type)
+
+		// Component-specific variables
+		componentVars := copyVars(vars)
+		componentVars["COMPONENT_NAME"] = c.Name
+		componentVars["PORT"] = c.Port
+
+		// For frontend apps, inject the primary service name/port for API proxy
+		if strings.HasPrefix(c.Type, "app-") {
+			if svc := findFirstService(defaultComponents); svc != nil {
+				componentVars["SERVICE_NAME"] = svc.Name
+				componentVars["SERVICE_PORT"] = svc.Port
+			}
+		}
+
+		// Determine destination path
+		destPath := filepath.Join(c.DestDir, c.Name)
+
+		if err := templates.RenderComponentToDir(*outputDir, c.Type, destPath, componentVars); err != nil {
+			fmt.Fprintf(os.Stderr, "Error rendering component %s: %v\n", c.Name, err)
+			os.Exit(1)
+		}
+	}
+
+	// Update monorepo files
+	fmt.Println("\nUpdating monorepo files...")
+	if err := updateGoWork(*outputDir, defaultComponents); err != nil {
+		fmt.Fprintf(os.Stderr, "Error updating go.work: %v\n", err)
+		os.Exit(1)
+	}
+	if err := updateProcfile(*outputDir, defaultComponents); err != nil {
+		fmt.Fprintf(os.Stderr, "Error updating Procfile: %v\n", err)
+		os.Exit(1)
+	}
+
+	// Remove .gitkeep files from directories that now have content
+	if err := removeGitkeeps(*outputDir); err != nil {
+		fmt.Fprintf(os.Stderr, "Error removing .gitkeep files: %v\n", err)
+		os.Exit(1)
+	}
+
+	fmt.Println("\nDone!")
+	fmt.Printf("Rendered %d components to %s\n", len(defaultComponents), *outputDir)
+	fmt.Println("\nNext steps:")
+	fmt.Println("  cd " + *outputDir)
+	fmt.Println("  go work sync")
+	fmt.Println("  go build ./...")
+	fmt.Println("  pnpm install")
+	fmt.Println("  pnpm -r typecheck")
+}
+
+// cleanOutputDir removes existing content but preserves .git directory.
+func cleanOutputDir(outputDir string) error {
+	entries, err := os.ReadDir(outputDir)
+	if os.IsNotExist(err) {
+		return os.MkdirAll(outputDir, 0755)
+	}
+	if err != nil {
+		return err
+	}
+
+	for _, entry := range entries {
+		if entry.Name() == ".git" {
+			continue // preserve .git
+		}
+		path := filepath.Join(outputDir, entry.Name())
+		if err := os.RemoveAll(path); err != nil {
+			return fmt.Errorf("failed to remove %s: %w", path, err)
+		}
+	}
+	return nil
+}
+
+// copyVars creates a copy of a variables map.
+func copyVars(vars map[string]string) map[string]string {
+	return maps.Clone(vars)
+}
+
+// updateGoWork adds component modules to go.work.
+func updateGoWork(outputDir string, components []Component) error {
+	var sb strings.Builder
+	sb.WriteString("go 1.25\n\n")
+	sb.WriteString("use ./pkg\n")
+
+	for _, c := range components {
+		// Only Go components get added to go.work
+		switch c.Type {
+		case "service", "worker", "cli":
+			fmt.Fprintf(&sb, "use ./%s/%s\n", c.DestDir, c.Name)
+		}
+	}
+
+	return os.WriteFile(filepath.Join(outputDir, "go.work"), []byte(sb.String()), 0644)
+}
+
+// updateProcfile adds component processes to Procfile.
+func updateProcfile(outputDir string, components []Component) error {
+	var lines []string
+	lines = append(lines, "# Local development processes")
+	lines = append(lines, "")
+
+	for _, c := range components {
+		var cmd string
+		switch c.Type {
+		case "service":
+			cmd = fmt.Sprintf("%s: cd %s/%s && make dev", c.Name, c.DestDir, c.Name)
+		case "worker":
+			cmd = fmt.Sprintf("%s: cd %s/%s && make dev", c.Name, c.DestDir, c.Name)
+		case "cli":
+			// CLIs don't run as processes
+			continue
+		case "app-astro", "app-react":
+			cmd = fmt.Sprintf("%s: cd %s/%s && pnpm dev", c.Name, c.DestDir, c.Name)
+		case "app-nextjs":
+			cmd = fmt.Sprintf("%s: cd %s/%s && pnpm dev", c.Name, c.DestDir, c.Name)
+		default:
+			continue
+		}
+		lines = append(lines, cmd)
+	}
+	lines = append(lines, "")
+
+	content := strings.Join(lines, "\n")
+	return os.WriteFile(filepath.Join(outputDir, "Procfile"), []byte(content), 0644)
+}
+
+// findFirstService returns the first service component, or nil if none.
+func findFirstService(components []Component) *Component {
+	for i := range components {
+		if components[i].Type == "service" {
+			return &components[i]
+		}
+	}
+	return nil
+}
+
+// removeGitkeeps removes .gitkeep files from directories that have other content.
+func removeGitkeeps(outputDir string) error {
+	keepDirs := []string{"services", "workers", "apps", "cli", "packages"}
+
+	for _, dir := range keepDirs {
+		gitkeepPath := filepath.Join(outputDir, dir, ".gitkeep")
+		dirPath := filepath.Join(outputDir, dir)
+
+		entries, err := os.ReadDir(dirPath)
+		if err != nil {
+			continue // directory might not exist
+		}
+
+		// If there's more than just .gitkeep, remove it
+		if len(entries) > 1 {
+			_ = os.Remove(gitkeepPath) // Ignore error - file may not exist
+		}
+	}
+
+	return nil
+}

cookbooks/scripts/common.sh

@@ -95,7 +95,7 @@ api_call() {
 # Returns: 0 on success, 1 on failure, 2 on timeout
 wait_for_build() {
     local task_id="$1"
-    local max_attempts="${2:-60}"  # 5 minutes default (5s * 60)
+    local max_attempts="${2:-120}"  # 10 minutes default (5s * 120)
     local poll_interval="${3:-5}"
     local attempt=0
 
@@ -155,7 +155,7 @@ wait_for_build() {
 # the pipeline has already failed.
 wait_for_pipeline() {
     local project_id="$1"
-    local max_attempts="${2:-60}"  # 5 minutes default
+    local max_attempts="${2:-120}"  # 10 minutes default
     local poll_interval="${3:-5}"
     local attempt=0
     local tracked_pipeline=""  # Track specific pipeline once found

cookbooks/scripts/tree-runner.sh

@@ -120,7 +120,7 @@ execute_wait_pipeline_step() {
 
     local project_id max_attempts poll_interval
     project_id=$(echo "$step_json" | jq -r '.project_id')
-    max_attempts=$(echo "$step_json" | jq -r '.max_attempts // 60')
+    max_attempts=$(echo "$step_json" | jq -r '.max_attempts // 120')
     poll_interval=$(echo "$step_json" | jq -r '.poll_interval // 5')
 
     wait_for_pipeline "$project_id" "$max_attempts" "$poll_interval"

cookbooks/trees/aeries-1-genesis.yaml

@@ -77,7 +77,7 @@ steps:
     depends_on: [spec-feature]
     action: wait_build
     build_id: "{{ .outputs.spec-feature.build_id }}"
-    max_attempts: 60
+    max_attempts: 120
     poll_interval: 5
 
   implement-backend:

cookbooks/trees/composable-app.yaml

@@ -62,7 +62,7 @@ steps:
     depends_on: [verify-components]
     action: wait_pipeline
     project_id: "{{ .outputs.create-project.project_id }}"
-    max_attempts: 60
+    max_attempts: 120
     poll_interval: 5
     on_error: continue
 

cookbooks/trees/evolving-app.yaml

@@ -39,7 +39,7 @@ steps:
     depends_on: [add-service]
     action: wait_pipeline
     project_id: "{{ .outputs.create-project.project_id }}"
-    max_attempts: 60
+    max_attempts: 120
 
   # --- Phase 2: Evolve (Add Feature) ---
   create-feature:
@@ -71,7 +71,7 @@ steps:
     depends_on: [generate-spec]
     action: wait_build
     build_id: "{{ .outputs.generate-spec.build_id }}"
-    max_attempts: 60
+    max_attempts: 120
     poll_interval: 5
 
   check-artifact:

cookbooks/trees/foundary-refine.yaml

@@ -136,7 +136,7 @@ steps:
     depends_on: [wait-deploy-2]
     action: wait_site
     domain: "{{ .vars.domain }}"
-    max_attempts: 60
+    max_attempts: 120
 
   verify-complete:
     description: "Print success summary"

cookbooks/trees/foundary.yaml

@@ -86,7 +86,7 @@ steps:
     depends_on: [wait-components]
     action: wait_site
     domain: "{{ .outputs.create-project.domain }}"
-    max_attempts: 60
+    max_attempts: 120
     on_error: continue
 
   # ============================================================

cookbooks/trees/full-stack-feature.yaml

@@ -39,7 +39,7 @@ steps:
     depends_on: [add-service]
     action: wait_pipeline
     project_id: "{{ .outputs.create-project.project_id }}"
-    max_attempts: 60
+    max_attempts: 120
 
   # --- Phase 2: SDLC Process (Spec & Design) ---
   create-feature:
@@ -71,7 +71,7 @@ steps:
     depends_on: [generate-spec]
     action: wait_build
     build_id: "{{ .outputs.generate-spec.build_id }}"
-    max_attempts: 60
+    max_attempts: 120
     poll_interval: 5
 
   approve-spec:
@@ -101,7 +101,7 @@ steps:
     depends_on: [generate-design]
     action: wait_build
     build_id: "{{ .outputs.generate-design.build_id }}"
-    max_attempts: 60
+    max_attempts: 120
     poll_interval: 5
 
   approve-design:
@@ -152,7 +152,7 @@ steps:
     depends_on: [wait-implementation]
     action: wait_pipeline
     project_id: "{{ .outputs.create-project.project_id }}"
-    max_attempts: 60
+    max_attempts: 120
 
   # --- Phase 4: Verification ---
   verify-crud:

cookbooks/trees/genkit-test.yaml

@@ -0,0 +1,208 @@
+name: genkit-test
+description: "Validate AI generation packages with a chat UI and API backend"
+version: 1
+
+vars:
+  project_name: ""
+  service_name: "ai-gateway"
+  app_name: "chat-ui"
+  feature_slug: "ai-chat"
+
+steps:
+  # --- Infrastructure ---
+  create-project:
+    action: api
+    method: POST
+    endpoint: /project
+    body:
+      name: "{{ .vars.project_name }}"
+      description: "Genkit validation: AI chat with text and image generation"
+    outputs:
+      - project_id: .data.name
+      - domain: .data.domain
+
+  add-service:
+    description: Add AI gateway service
+    depends_on: [create-project]
+    action: api
+    method: POST
+    endpoint: "/projects/{{ .outputs.create-project.project_id }}/components"
+    body:
+      type: service
+      name: "{{ .vars.service_name }}"
+
+  add-ui:
+    description: Add React chat UI
+    depends_on: [create-project]
+    action: api
+    method: POST
+    endpoint: "/projects/{{ .outputs.create-project.project_id }}/components"
+    body:
+      type: app-react
+      name: "{{ .vars.app_name }}"
+
+  wait-init:
+    depends_on: [add-service, add-ui]
+    action: wait_pipeline
+    project_id: "{{ .outputs.create-project.project_id }}"
+    max_attempts: 120
+    poll_interval: 5
+
+  # --- SDLC: Build AI Endpoints ---
+  create-feature:
+    depends_on: [wait-init]
+    action: api
+    method: POST
+    endpoint: "/projects/{{ .outputs.create-project.project_id }}/sdlc/features"
+    body:
+      slug: "{{ .vars.feature_slug }}"
+      title: "AI Generation Endpoints"
+
+  implement-ai:
+    description: "Agent implements text and image generation endpoints"
+    depends_on: [create-feature]
+    action: api
+    method: POST
+    endpoint: "/projects/{{ .outputs.create-project.project_id }}/builds"
+    body:
+      prompt: |
+        /implement-feature {{ .vars.feature_slug }} --requirements '
+        Use pkg/textgen and pkg/mediagen to create AI generation endpoints.
+
+        Environment variables available: LAOZHANG_API_KEY, GEMINI_API_KEY
+
+        Endpoints to implement:
+
+        1. GET /health/providers - Check AI provider connectivity
+           - Initialize textgen.Manager with ProductionConfig
+           - Return provider names and health status
+
+        2. POST /chat - Text generation
+           Request: {"message": "string", "system_prompt": "string (optional)"}
+           - Use textgen.Manager.GenerateText()
+           - Return: {"response": "string", "provider": "string"}
+
+        3. POST /generate-image - Image generation
+           Request: {"prompt": "string"}
+           - Use mediagen.Manager.GenerateImage()
+           - Return: {"image_base64": "string", "provider": "string"}
+
+        Use ProductionConfig for both managers (LaoZhang primary, Gemini fallback).
+        Handle errors gracefully with proper HTTP status codes.
+        '
+      auto_commit: true
+      auto_push: true
+      git_clone_url: "https://git.threesix.ai/jordan/{{ .outputs.create-project.project_id }}.git"
+    outputs:
+      - build_id: .data.task_id
+
+  wait-build:
+    description: Wait for agent code generation
+    depends_on: [implement-ai]
+    action: wait_build
+    build_id: "{{ .outputs.implement-ai.build_id }}"
+    max_attempts: 120
+    poll_interval: 5
+
+  wait-deploy:
+    depends_on: [wait-build]
+    action: wait_pipeline
+    project_id: "{{ .outputs.create-project.project_id }}"
+    max_attempts: 120
+    poll_interval: 5
+
+  # --- Verification ---
+  verify-service-health:
+    description: "Verify the AI service is running"
+    depends_on: [wait-deploy]
+    action: shell
+    command: |
+      DOMAIN="{{ .outputs.create-project.domain }}"
+      SERVICE_NAME="{{ .vars.service_name }}"
+
+      HEALTH=$(curl -s "https://$DOMAIN/api/$SERVICE_NAME/health" | jq -r '.data.status // .status // empty')
+      if [ "$HEALTH" == "healthy" ] || [ "$HEALTH" == "ok" ]; then
+        echo "Service healthy"
+        exit 0
+      else
+        echo "Service not healthy: $HEALTH"
+        curl -s "https://$DOMAIN/api/$SERVICE_NAME/health" | jq .
+        exit 1
+      fi
+
+  verify-provider-health:
+    description: "Verify AI providers are accessible"
+    depends_on: [verify-service-health]
+    on_error: continue
+    action: shell
+    command: |
+      DOMAIN="{{ .outputs.create-project.domain }}"
+      SERVICE_NAME="{{ .vars.service_name }}"
+
+      echo "Checking provider health..."
+      RESP=$(curl -s "https://$DOMAIN/api/$SERVICE_NAME/health/providers")
+      echo "$RESP" | jq .
+
+      # Check if we got any provider info (success even if some providers are down)
+      if echo "$RESP" | jq -e '.data // .providers // .' > /dev/null 2>&1; then
+        echo "Provider health endpoint working"
+        exit 0
+      else
+        echo "Provider health check failed"
+        exit 1
+      fi
+
+  verify-text-generation:
+    description: "Test text generation endpoint"
+    depends_on: [verify-provider-health]
+    on_error: continue
+    action: shell
+    command: |
+      DOMAIN="{{ .outputs.create-project.domain }}"
+      SERVICE_NAME="{{ .vars.service_name }}"
+
+      echo "Testing text generation..."
+      RESP=$(curl -s -X POST "https://$DOMAIN/api/$SERVICE_NAME/chat" \
+        -H "Content-Type: application/json" \
+        -d '{"message": "What is 2+2? Reply with just the number."}')
+      echo "$RESP" | jq .
+
+      # Check if we got a response
+      RESPONSE=$(echo "$RESP" | jq -r '.response // .data.response // .text // empty')
+      if [ -n "$RESPONSE" ]; then
+        echo "Text generation working: $RESPONSE"
+        exit 0
+      else
+        echo "Text generation failed"
+        exit 1
+      fi
+
+  verify-image-generation:
+    description: "Test image generation endpoint"
+    depends_on: [verify-text-generation]
+    on_error: continue
+    action: shell
+    command: |
+      DOMAIN="{{ .outputs.create-project.domain }}"
+      SERVICE_NAME="{{ .vars.service_name }}"
+
+      echo "Testing image generation..."
+      RESP=$(curl -s -X POST "https://$DOMAIN/api/$SERVICE_NAME/generate-image" \
+        -H "Content-Type: application/json" \
+        -d '{"prompt": "a simple red circle on white background"}')
+
+      # Check if we got base64 image data
+      IMAGE=$(echo "$RESP" | jq -r '.image_base64 // .data.image_base64 // .image // empty')
+      if [ -n "$IMAGE" ] && [ ${#IMAGE} -gt 100 ]; then
+        echo "Image generation working (got ${#IMAGE} chars of base64)"
+        exit 0
+      else
+        echo "Image generation failed or returned empty"
+        echo "$RESP" | jq .
+        exit 1
+      fi
+
+teardown:
+  - action: api
+    method: DELETE
+    endpoint: "/project/{{ .outputs.create-project.project_id }}"

cookbooks/trees/landing-page.yaml

@@ -38,7 +38,7 @@ steps:
     depends_on: [add-component]
     action: wait_pipeline
     project_id: "{{ .outputs.create-project.project_id }}"
-    max_attempts: 60
+    max_attempts: 120
     poll_interval: 5
     on_error: continue
 

deployments/k8s/base/citadel-agent/configmap.yaml

@@ -0,0 +1,27 @@
+apiVersion: v1
+kind: ConfigMap
+metadata:
+  name: citadel-agent-config
+  namespace: observability
+  labels:
+    app.kubernetes.io/name: citadel-agent
+    app.kubernetes.io/part-of: citadel
+data:
+  # Agent ships logs via HTTP to partner-hosted Citadel
+  CITADEL_HTTP: "true"
+  CITADEL_HTTP_URL: "https://citadel-staging.orchard9.ai"
+
+  # Log routing: agent reads these k8s labels to determine target environment
+  # citadel.io/environment → target Citadel tenant ID
+  # citadel.io/service     → service name tag
+  CITADEL_ROUTE_LABEL_ENVIRONMENT: "citadel.io/environment"
+  CITADEL_ROUTE_LABEL_SERVICE: "citadel.io/service"
+
+  # Namespaces to collect logs from
+  # rdev namespace = platform components (rdev-api, rdev-worker, claudebox)
+  # projects namespace = generated project pods
+  CITADEL_NAMESPACES: "rdev,projects,observability"
+
+  # Default environment for pods without citadel.io/environment label
+  # Platform pods (rdev-api, rdev-worker) route here
+  CITADEL_DEFAULT_ENVIRONMENT: "rdev-platform"

deployments/k8s/base/citadel-agent/daemonset.yaml

@@ -0,0 +1,92 @@
+apiVersion: apps/v1
+kind: DaemonSet
+metadata:
+  name: citadel-agent
+  namespace: observability
+  labels:
+    app.kubernetes.io/name: citadel-agent
+    app.kubernetes.io/part-of: citadel
+spec:
+  selector:
+    matchLabels:
+      app.kubernetes.io/name: citadel-agent
+  updateStrategy:
+    type: RollingUpdate
+    rollingUpdate:
+      maxUnavailable: 1
+  template:
+    metadata:
+      labels:
+        app.kubernetes.io/name: citadel-agent
+        app.kubernetes.io/part-of: citadel
+        # Label this pod so it routes its OWN logs to rdev-platform
+        citadel.io/environment: rdev-platform
+        citadel.io/service: citadel-agent
+    spec:
+      serviceAccountName: citadel-agent
+      tolerations:
+        # Run on all nodes including control plane
+        - operator: Exists
+      containers:
+        - name: agent
+          image: gcr.io/orchard9/citadel-agent:latest
+          resources:
+            requests:
+              cpu: 50m
+              memory: 64Mi
+            limits:
+              cpu: 200m
+              memory: 256Mi
+          envFrom:
+            - configMapRef:
+                name: citadel-agent-config
+          env:
+            - name: CITADEL_API_KEY
+              valueFrom:
+                secretKeyRef:
+                  name: citadel-agent
+                  key: api-key
+            - name: CITADEL_TENANT_ID
+              valueFrom:
+                secretKeyRef:
+                  name: citadel-agent
+                  key: tenant-id
+            - name: NODE_NAME
+              valueFrom:
+                fieldRef:
+                  fieldPath: spec.nodeName
+          volumeMounts:
+            # Container log files on the node
+            - name: varlog
+              mountPath: /var/log
+              readOnly: true
+            # Container runtime data (for resolving container IDs to pod metadata)
+            - name: containers
+              mountPath: /var/lib/docker/containers
+              readOnly: true
+            # Persistent state (checkpoint offsets survive agent restarts)
+            - name: agent-state
+              mountPath: /var/lib/citadel-agent
+          livenessProbe:
+            httpGet:
+              path: /healthz
+              port: 9090
+            initialDelaySeconds: 10
+            periodSeconds: 30
+          readinessProbe:
+            httpGet:
+              path: /readyz
+              port: 9090
+            initialDelaySeconds: 5
+            periodSeconds: 10
+      volumes:
+        - name: varlog
+          hostPath:
+            path: /var/log
+        - name: containers
+          hostPath:
+            path: /var/lib/docker/containers
+        - name: agent-state
+          hostPath:
+            path: /var/lib/citadel-agent
+            type: DirectoryOrCreate

deployments/k8s/base/citadel-agent/kustomization.yaml

@@ -0,0 +1,13 @@
+apiVersion: kustomize.config.k8s.io/v1beta1
+kind: Kustomization
+
+namespace: observability
+
+resources:
+  - namespace.yaml
+  - serviceaccount.yaml
+  - daemonset.yaml
+  - configmap.yaml
+  # NOTE: secret.yaml contains real keys and is gitignored.
+  # Copy from secret.yaml.example and fill in real values before deploying.
+  - secret.yaml

deployments/k8s/base/citadel-agent/namespace.yaml

@@ -0,0 +1,6 @@
+apiVersion: v1
+kind: Namespace
+metadata:
+  name: observability
+  labels:
+    app.kubernetes.io/part-of: citadel

deployments/k8s/base/citadel-agent/secret.yaml.example

@@ -0,0 +1,11 @@
+apiVersion: v1
+kind: Secret
+metadata:
+  name: citadel-agent
+  namespace: observability
+type: Opaque
+stringData:
+  # API key for Citadel (get from: citadel api-key create --name "rdev-agent" --environment live)
+  api-key: "ck_live_REPLACE_ME"
+  # Tenant ID for rdev-platform environment (get from: citadel env create rdev-platform)
+  tenant-id: "REPLACE_ME"

deployments/k8s/base/citadel-agent/serviceaccount.yaml

@@ -0,0 +1,37 @@
+apiVersion: v1
+kind: ServiceAccount
+metadata:
+  name: citadel-agent
+  namespace: observability
+  labels:
+    app.kubernetes.io/name: citadel-agent
+    app.kubernetes.io/part-of: citadel
+---
+# ClusterRole to read pod metadata for log enrichment
+apiVersion: rbac.authorization.k8s.io/v1
+kind: ClusterRole
+metadata:
+  name: citadel-agent
+  labels:
+    app.kubernetes.io/name: citadel-agent
+    app.kubernetes.io/part-of: citadel
+rules:
+  - apiGroups: [""]
+    resources: ["pods", "namespaces"]
+    verbs: ["get", "list", "watch"]
+---
+apiVersion: rbac.authorization.k8s.io/v1
+kind: ClusterRoleBinding
+metadata:
+  name: citadel-agent
+  labels:
+    app.kubernetes.io/name: citadel-agent
+    app.kubernetes.io/part-of: citadel
+roleRef:
+  apiGroup: rbac.authorization.k8s.io
+  kind: ClusterRole
+  name: citadel-agent
+subjects:
+  - kind: ServiceAccount
+    name: citadel-agent
+    namespace: observability

deployments/k8s/base/claudebox.yaml

@@ -19,6 +19,9 @@ spec:
         app.kubernetes.io/name: claudebox
         app.kubernetes.io/part-of: rdev
         rdev.orchard9.ai/role: worker
+        # Citadel agent routes these logs to the rdev-platform environment
+        citadel.io/environment: rdev-platform
+        citadel.io/service: claudebox
     spec:
       containers:
         - name: claudebox

deployments/k8s/base/kustomization.yaml

@@ -36,3 +36,6 @@ resources:
   # Wildcard TLS for session preview URLs
   - preview-cert.yaml
 
+  # Citadel log agent (ships container logs to partner-hosted Citadel)
+  - citadel-agent/
+

deployments/k8s/base/rdev-api.yaml

@@ -20,6 +20,9 @@ spec:
         app: rdev-api
         app.kubernetes.io/name: rdev-api
         app.kubernetes.io/part-of: rdev
+        # Citadel agent routes these logs to the rdev-platform environment
+        citadel.io/environment: rdev-platform
+        citadel.io/service: rdev-api
     spec:
       serviceAccountName: rdev-api
       containers:

docs/citadel-integration.md

@@ -0,0 +1,186 @@
+# Citadel Integration Plan
+
+rdev integrates with a **partner-hosted Citadel** instance at `citadel-staging.orchard9.ai` for centralized log aggregation, querying, and alerting across both the platform and generated projects.
+
+## Environment Architecture
+
+```
+Organization: orchard9 (on citadel-staging.orchard9.ai)
+│
+├── Environment: "rdev-platform"
+│   ├── rdev-api logs (stdout via agent)
+│   ├── rdev-worker logs (stdout via agent)
+│   ├── claudebox logs (stdout via agent)
+│   └── audit events (shipped from AuditLogger)
+│
+├── Environment: "<project-slug>"       (auto-created per project)
+│   └── Project pod logs (stdout via agent, routed by k8s labels)
+│
+├── Tenant Group: "platform"            → rdev-platform
+└── Tenant Group: "projects"            → all project-* environments
+```
+
+Each project gets its own Citadel environment, matching rdev's isolation model (each project gets its own DB, Redis, DNS, registry namespace).
+
+## Integration Points
+
+### 1. Agent DaemonSet (Log Collection)
+
+**What:** A `citadel-agent` DaemonSet on every k3s node collects container stdout/stderr and ships to Citadel via HTTPS.
+
+**Why:** rdev already outputs structured JSON slog to stdout. Zero code changes needed.
+
+**Routing:** Agent reads k8s labels to route logs to the correct Citadel environment:
+- `citadel.io/environment` label → determines target environment
+- `citadel.io/service` label → tags the service name in Citadel
+
+**Manifests:** `deployments/k8s/base/citadel-agent/`
+
+### 2. Citadel Client Adapter (API Integration)
+
+**What:** Go HTTP client for Citadel's API, following rdev's hexagonal architecture (port interface + adapter).
+
+**Why:** Needed for auto-provisioning environments and shipping audit logs.
+
+**Files:**
+- `internal/port/citadel.go` — Port interface
+- `internal/adapter/citadel/client.go` — HTTP client implementation
+- `internal/adapter/citadel/audit_shipper.go` — Audit log shipping
+
+### 3. Project Provisioning Step
+
+**What:** When `ProjectInfraService.CreateProject()` runs, a new step creates a Citadel environment for the project.
+
+**Why:** Each project needs its own isolated log environment. Auto-provisioning eliminates manual setup.
+
+**Where in the flow:**
+```
+CreateProject()
+  1. Generate slug
+  2. Create project in DB
+  3. Create git repo
+  4. Create DNS
+  5. Activate CI
+  6. Seed template
+  7. Provision DB + cache
+  8. *** Create Citadel environment ***  ← NEW
+  9. Create initial deployment
+  10. Trigger CI build
+```
+
+**Rollback:** On project deletion, the Citadel environment is deleted too.
+
+### 4. Skeleton Template Labels
+
+**What:** Add `citadel.io/*` labels to the skeleton's k8s deployment templates.
+
+**Why:** The agent uses these labels to route project logs to the correct Citadel environment.
+
+**Template vars:** `{{CITADEL_TENANT_ID}}` injected during provisioning.
+
+### 5. Audit Log Shipping
+
+**What:** Hook into rdev's existing `AuditLogger` to also ship audit events to Citadel.
+
+**Why:** Unified search across application logs and security events.
+
+**Pattern:** Wrap the existing PostgreSQL `AuditLogger` with a multi-writer that sends to both PostgreSQL and Citadel.
+
+## Configuration
+
+### Environment Variables (rdev-api)
+
+| Variable | Description | Example |
+|----------|-------------|---------|
+| `CITADEL_URL` | Partner Citadel instance URL | `https://citadel-staging.orchard9.ai` |
+| `CITADEL_API_KEY` | API key for environment management | `ck_live_...` |
+| `CITADEL_TENANT_ID` | Platform environment tenant ID | `uuid` |
+| `CITADEL_ENABLED` | Enable/disable Citadel integration | `true` |
+
+### Secrets (k8s)
+
+```yaml
+# deployments/k8s/base/citadel-agent/secret.yaml
+apiVersion: v1
+kind: Secret
+metadata:
+  name: citadel-agent
+  namespace: observability
+type: Opaque
+stringData:
+  api-key: "ck_live_..."
+  tenant-id: "..."
+```
+
+## Queries (What You Get)
+
+### Platform Operations
+
+```bash
+# All errors across platform
+citadel query "level:error" --tenant rdev-platform --last 1h
+
+# Track a project lifecycle
+citadel query "project_id:my-project" --tenant rdev-platform --last 24h
+
+# Build failures
+citadel query "component:build level:error" --tenant rdev-platform
+
+# Saga failures
+citadel query "component:saga result:failure" --tenant rdev-platform
+
+# Slow API requests
+citadel query "duration_ms:>5000 component:http" --tenant rdev-platform
+
+# Audit trail
+citadel query "action:project.create" --tenant rdev-platform --last 7d
+```
+
+### Generated Projects
+
+```bash
+# Query a specific project's logs
+citadel query "level:error" --tenant <project-slug> --last 1h
+
+# All project errors at once
+citadel query "level:error" --group projects --last 1h
+
+# Correlate platform + project
+citadel query "project_id:my-project" --last 24h  # org-wide
+```
+
+## Implementation Order
+
+| Phase | Task | Effort |
+|-------|------|--------|
+| 1 | Write Citadel client adapter (`internal/adapter/citadel/`) | 2h |
+| 1 | Write port interface (`internal/port/citadel.go`) | 30m |
+| 2 | Deploy agent DaemonSet (`deployments/k8s/base/citadel-agent/`) | 1h |
+| 2 | Add citadel labels to existing rdev-api/worker manifests | 15m |
+| 3 | Add Citadel env creation to `ProjectInfraService.CreateProject()` | 2h |
+| 3 | Add Citadel env deletion to `ProjectInfraService.DeleteProject()` | 30m |
+| 3 | Add migration for `citadel_tenant_id` column on projects table | 15m |
+| 4 | Add `citadel.io/*` labels to skeleton k8s templates | 30m |
+| 4 | Add `CITADEL_TENANT_ID` to template variables | 15m |
+| 5 | Implement audit log shipper | 2h |
+| 5 | Wire into main.go startup | 30m |
+
+## Files Created/Modified
+
+### New Files
+- `internal/port/citadel.go`
+- `internal/adapter/citadel/client.go`
+- `internal/adapter/citadel/audit_shipper.go`
+- `deployments/k8s/base/citadel-agent/kustomization.yaml`
+- `deployments/k8s/base/citadel-agent/daemonset.yaml`
+- `deployments/k8s/base/citadel-agent/serviceaccount.yaml`
+- `deployments/k8s/base/citadel-agent/configmap.yaml`
+- `internal/db/migrations/024_citadel_tenant_id.sql`
+
+### Modified Files
+- `deployments/k8s/base/kustomization.yaml` — include citadel-agent
+- `deployments/k8s/base/rdev-api.yaml` — add citadel labels + env vars
+- `internal/service/project_infra.go` — add CitadelClient field
+- `internal/service/project_infra_crud.go` — add provisioning + cleanup steps
+- `internal/adapter/templates/templates/skeleton/` — k8s template labels
+- `cmd/rdev-api/main.go` — wire Citadel adapter

examples/README.md

@@ -0,0 +1,45 @@
+# Examples
+
+This directory contains example projects that are rendered from rdev's skeleton templates.
+
+## full-monorepo
+
+A fully rendered monorepo skeleton with all component types:
+
+| Component | Type | Purpose |
+|-----------|------|---------|
+| `services/example-api` | service | Go REST API |
+| `workers/example-worker` | worker | Go background worker |
+| `apps/example-astro` | app-astro | Astro landing page |
+| `apps/example-react` | app-react | React SPA |
+| `apps/example-nextjs` | app-nextjs | Next.js dashboard |
+| `cli/example-cli` | cli | Go CLI tool |
+
+### Purpose
+
+1. **Template testing**: Ensures templates render correctly and compile
+2. **IDE debugging**: Step through generated code with full syntax highlighting
+3. **Documentation**: Shows what rendered projects look like
+
+### Regenerating
+
+To regenerate after template changes:
+
+```bash
+./scripts/verify-skeleton.sh --update
+```
+
+### Verification
+
+To verify templates are in sync:
+
+```bash
+./scripts/verify-skeleton.sh       # Full verification (Go + TypeScript)
+./scripts/verify-skeleton.sh --quick  # Skip TypeScript
+```
+
+## How This Works
+
+1. `cmd/render-skeleton/main.go` - CLI tool that renders templates using the actual `templates.Provider`
+2. `internal/adapter/templates/provider.go` - Has `RenderSkeletonToDir` and `RenderComponentToDir` functions
+3. `scripts/verify-skeleton.sh` - CI script that checks sync and attempts builds

go.mod

@@ -3,6 +3,7 @@ module github.com/orchard9/rdev
 go 1.25.0
 
 require (
+	cloud.google.com/go/storage v1.59.2
 	code.gitea.io/sdk/gitea v0.22.1
 	github.com/bdpiprava/scalar-go v0.13.0
 	github.com/go-chi/chi/v5 v5.1.0
@@ -18,6 +19,7 @@ require (
 	go.opentelemetry.io/otel/sdk v1.39.0
 	go.opentelemetry.io/otel/trace v1.39.0
 	go.woodpecker-ci.org/woodpecker/v3 v3.13.0
+	google.golang.org/api v0.265.0
 	gopkg.in/yaml.v3 v3.0.1
 	k8s.io/api v0.35.0
 	k8s.io/apimachinery v0.35.0
@@ -32,7 +34,6 @@ require (
 	cloud.google.com/go/compute/metadata v0.9.0 // indirect
 	cloud.google.com/go/iam v1.5.3 // indirect
 	cloud.google.com/go/monitoring v1.24.3 // indirect
-	cloud.google.com/go/storage v1.59.2 // indirect
 	github.com/42wim/httpsig v1.2.3 // indirect
 	github.com/GoogleCloudPlatform/opentelemetry-operations-go/detectors/gcp v1.30.0 // indirect
 	github.com/GoogleCloudPlatform/opentelemetry-operations-go/exporter/metric v0.54.0 // indirect
@@ -95,7 +96,6 @@ require (
 	golang.org/x/term v0.39.0 // indirect
 	golang.org/x/text v0.33.0 // indirect
 	golang.org/x/time v0.14.0 // indirect
-	google.golang.org/api v0.265.0 // indirect
 	google.golang.org/genproto v0.0.0-20251202230838-ff82c1b0f217 // indirect
 	google.golang.org/genproto/googleapis/api v0.0.0-20251202230838-ff82c1b0f217 // indirect
 	google.golang.org/genproto/googleapis/rpc v0.0.0-20260128011058-8636f8732409 // indirect

go.sum

@@ -10,10 +10,16 @@ cloud.google.com/go/compute/metadata v0.9.0 h1:pDUj4QMoPejqq20dK0Pg2N4yG9zIkYGdB
 cloud.google.com/go/compute/metadata v0.9.0/go.mod h1:E0bWwX5wTnLPedCKqk3pJmVgCBSM6qQI1yTBdEb3C10=
 cloud.google.com/go/iam v1.5.3 h1:+vMINPiDF2ognBJ97ABAYYwRgsaqxPbQDlMnbHMjolc=
 cloud.google.com/go/iam v1.5.3/go.mod h1:MR3v9oLkZCTlaqljW6Eb2d3HGDGK5/bDv93jhfISFvU=
+cloud.google.com/go/logging v1.13.1 h1:O7LvmO0kGLaHY/gq8cV7T0dyp6zJhYAOtZPX4TF3QtY=
+cloud.google.com/go/logging v1.13.1/go.mod h1:XAQkfkMBxQRjQek96WLPNze7vsOmay9H5PqfsNYDqvw=
+cloud.google.com/go/longrunning v0.7.0 h1:FV0+SYF1RIj59gyoWDRi45GiYUMM3K1qO51qoboQT1E=
+cloud.google.com/go/longrunning v0.7.0/go.mod h1:ySn2yXmjbK9Ba0zsQqunhDkYi0+9rlXIwnoAf+h+TPY=
 cloud.google.com/go/monitoring v1.24.3 h1:dde+gMNc0UhPZD1Azu6at2e79bfdztVDS5lvhOdsgaE=
 cloud.google.com/go/monitoring v1.24.3/go.mod h1:nYP6W0tm3N9H/bOw8am7t62YTzZY+zUeQ+Bi6+2eonI=
 cloud.google.com/go/storage v1.59.2 h1:gmOAuG1opU8YvycMNpP+DvHfT9BfzzK5Cy+arP+Nocw=
 cloud.google.com/go/storage v1.59.2/go.mod h1:cMWbtM+anpC74gn6qjLh+exqYcfmB9Hqe5z6adx+CLI=
+cloud.google.com/go/trace v1.11.7 h1:kDNDX8JkaAG3R2nq1lIdkb7FCSi1rCmsEtKVsty7p+U=
+cloud.google.com/go/trace v1.11.7/go.mod h1:TNn9d5V3fQVf6s4SCveVMIBS2LJUqo73GACmq/Tky0s=
 code.gitea.io/sdk/gitea v0.22.1 h1:7K05KjRORyTcTYULQ/AwvlVS6pawLcWyXZcTr7gHFyA=
 code.gitea.io/sdk/gitea v0.22.1/go.mod h1:yyF5+GhljqvA30sRDreoyHILruNiy4ASufugzYg0VHM=
 github.com/42wim/httpsig v1.2.3 h1:xb0YyWhkYj57SPtfSttIobJUPJZB9as1nsfo7KWVcEs=
@@ -22,6 +28,8 @@ github.com/GoogleCloudPlatform/opentelemetry-operations-go/detectors/gcp v1.30.0
 github.com/GoogleCloudPlatform/opentelemetry-operations-go/detectors/gcp v1.30.0/go.mod h1:P4WPRUkOhJC13W//jWpyfJNDAIpvRbAUIYLX/4jtlE0=
 github.com/GoogleCloudPlatform/opentelemetry-operations-go/exporter/metric v0.54.0 h1:lhhYARPUu3LmHysQ/igznQphfzynnqI3D75oUyw1HXk=
 github.com/GoogleCloudPlatform/opentelemetry-operations-go/exporter/metric v0.54.0/go.mod h1:l9rva3ApbBpEJxSNYnwT9N4CDLrWgtq3u8736C5hyJw=
+github.com/GoogleCloudPlatform/opentelemetry-operations-go/internal/cloudmock v0.54.0 h1:xfK3bbi6F2RDtaZFtUdKO3osOBIhNb+xTs8lFW6yx9o=
+github.com/GoogleCloudPlatform/opentelemetry-operations-go/internal/cloudmock v0.54.0/go.mod h1:vB2GH9GAYYJTO3mEn8oYwzEdhlayZIdQz6zdzgUIRvA=
 github.com/GoogleCloudPlatform/opentelemetry-operations-go/internal/resourcemapping v0.54.0 h1:s0WlVbf9qpvkh1c/uDAPElam0WrL7fHRIidgZJ7UqZI=
 github.com/GoogleCloudPlatform/opentelemetry-operations-go/internal/resourcemapping v0.54.0/go.mod h1:Mf6O40IAyB9zR/1J8nGDDPirZQQPbYJni8Yisy7NTMc=
 github.com/Masterminds/semver/v3 v3.4.0 h1:Zog+i5UMtVoCU8oKka5P7i9q9HgrJeGzI9SA1Xbatp0=
@@ -52,8 +60,11 @@ github.com/dgryski/go-rendezvous v0.0.0-20200823014737-9f7001d12a5f/go.mod h1:cu
 github.com/emicklei/go-restful/v3 v3.12.2 h1:DhwDP0vY3k8ZzE0RunuJy8GhNpPL6zqLkDf9B/a0/xU=
 github.com/emicklei/go-restful/v3 v3.12.2/go.mod h1:6n3XBCmQQb25CM2LCACGz8ukIrRry+4bhvbpWn3mrbc=
 github.com/envoyproxy/go-control-plane v0.13.5-0.20251024222203-75eaa193e329 h1:K+fnvUM0VZ7ZFJf0n4L/BRlnsb9pL/GuDG6FqaH+PwM=
+github.com/envoyproxy/go-control-plane v0.13.5-0.20251024222203-75eaa193e329/go.mod h1:Alz8LEClvR7xKsrq3qzoc4N0guvVNSS8KmSChGYr9hs=
 github.com/envoyproxy/go-control-plane/envoy v1.35.0 h1:ixjkELDE+ru6idPxcHLj8LBVc2bFP7iBytj353BoHUo=
 github.com/envoyproxy/go-control-plane/envoy v1.35.0/go.mod h1:09qwbGVuSWWAyN5t/b3iyVfz5+z8QWGrzkoqm/8SbEs=
+github.com/envoyproxy/go-control-plane/ratelimit v0.1.0 h1:/G9QYbddjL25KvtKTv3an9lx6VBE2cnb8wp1vEGNYGI=
+github.com/envoyproxy/go-control-plane/ratelimit v0.1.0/go.mod h1:Wk+tMFAFbCXaJPzVVHnPgRKdUdwW/KdbRt94AzgRee4=
 github.com/envoyproxy/protoc-gen-validate v1.2.1 h1:DEo3O99U8j4hBFwbJfrz9VtgcDfUKS7KJ7spH3d86P8=
 github.com/envoyproxy/protoc-gen-validate v1.2.1/go.mod h1:d/C80l/jxXLdfEIhX1W2TmLfsJ31lvEjwamM4DxlWXU=
 github.com/felixge/httpsnoop v1.0.4 h1:NFTV2Zj1bL4mc9sqWACXbQFVBBg2W3GPvqp8/ESS2Wg=
@@ -86,6 +97,8 @@ github.com/google/gnostic-models v0.7.0/go.mod h1:whL5G0m6dmc5cPxKc5bdKdEN3UjI7O
 github.com/google/go-cmp v0.7.0 h1:wk8382ETsv4JYUZwIsn6YpYiWiBsYLSJiTsyBybVuN8=
 github.com/google/go-cmp v0.7.0/go.mod h1:pXiqmnSA92OHEEa9HXL2W4E7lf9JzCmGVUdgjX3N/iU=
 github.com/google/gofuzz v1.0.0/go.mod h1:dBl0BpW6vV/+mYPU4Po3pmUjxk6FQPldtuIdl/M65Eg=
+github.com/google/martian/v3 v3.3.3 h1:DIhPTQrbPkgs2yJYdXU/eNACCG5DVQjySNRNlflZ9Fc=
+github.com/google/martian/v3 v3.3.3/go.mod h1:iEPrYcgCF7jA9OtScMFQyAlZZ4YXTKEtJ1E6RWzmBA0=
 github.com/google/pprof v0.0.0-20250403155104-27863c87afa6 h1:BHT72Gu3keYf3ZEu2J0b1vyeLSOYI8bm5wbJM/8yDe8=
 github.com/google/pprof v0.0.0-20250403155104-27863c87afa6/go.mod h1:boTsfXsheKC2y+lKOCMpSfarhxDeIzfZG1jqGcPl3cA=
 github.com/google/s2a-go v0.1.9 h1:LGD7gtMgezd8a/Xak7mEWL0PjoTQFvpRudN895yqKW0=
@@ -179,6 +192,8 @@ go.opentelemetry.io/otel/exporters/otlp/otlptrace v1.39.0 h1:f0cb2XPmrqn4XMy9PNl
 go.opentelemetry.io/otel/exporters/otlp/otlptrace v1.39.0/go.mod h1:vnakAaFckOMiMtOIhFI2MNH4FYrZzXCYxmb1LlhoGz8=
 go.opentelemetry.io/otel/exporters/otlp/otlptrace/otlptracegrpc v1.39.0 h1:in9O8ESIOlwJAEGTkkf34DesGRAc/Pn8qJ7k3r/42LM=
 go.opentelemetry.io/otel/exporters/otlp/otlptrace/otlptracegrpc v1.39.0/go.mod h1:Rp0EXBm5tfnv0WL+ARyO/PHBEaEAT8UUHQ6AGJcSq6c=
+go.opentelemetry.io/otel/exporters/stdout/stdoutmetric v1.38.0 h1:wm/Q0GAAykXv83wzcKzGGqAnnfLFyFe7RslekZuv+VI=
+go.opentelemetry.io/otel/exporters/stdout/stdoutmetric v1.38.0/go.mod h1:ra3Pa40+oKjvYh+ZD3EdxFZZB0xdMfuileHAm4nNN7w=
 go.opentelemetry.io/otel/metric v1.39.0 h1:d1UzonvEZriVfpNKEVmHXbdf909uGTOQjA0HF0Ls5Q0=
 go.opentelemetry.io/otel/metric v1.39.0/go.mod h1:jrZSWL33sD7bBxg1xjrqyDjnuzTUB0x1nBERXd7Ftcs=
 go.opentelemetry.io/otel/sdk v1.39.0 h1:nMLYcjVsvdui1B/4FRkwjzoRVsMK8uL/cj0OyhKzt18=
@@ -237,8 +252,6 @@ google.golang.org/genproto v0.0.0-20251202230838-ff82c1b0f217 h1:GvESR9BIyHUahIb
 google.golang.org/genproto v0.0.0-20251202230838-ff82c1b0f217/go.mod h1:yJ2HH4EHEDTd3JiLmhds6NkJ17ITVYOdV3m3VKOnws0=
 google.golang.org/genproto/googleapis/api v0.0.0-20251202230838-ff82c1b0f217 h1:fCvbg86sFXwdrl5LgVcTEvNC+2txB5mgROGmRL5mrls=
 google.golang.org/genproto/googleapis/api v0.0.0-20251202230838-ff82c1b0f217/go.mod h1:+rXWjjaukWZun3mLfjmVnQi18E1AsFbDN9QdJ5YXLto=
-google.golang.org/genproto/googleapis/rpc v0.0.0-20251202230838-ff82c1b0f217 h1:gRkg/vSppuSQoDjxyiGfN4Upv/h/DQmIR10ZU8dh4Ww=
-google.golang.org/genproto/googleapis/rpc v0.0.0-20251202230838-ff82c1b0f217/go.mod h1:7i2o+ce6H/6BluujYR+kqX3GKH+dChPTQU19wjRPiGk=
 google.golang.org/genproto/googleapis/rpc v0.0.0-20260128011058-8636f8732409 h1:H86B94AW+VfJWDqFeEbBPhEtHzJwJfTbgE2lZa54ZAQ=
 google.golang.org/genproto/googleapis/rpc v0.0.0-20260128011058-8636f8732409/go.mod h1:j9x/tPzZkyxcgEFkiKEEGxfvyumM01BEtsW8xzOahRQ=
 google.golang.org/grpc v1.78.0 h1:K1XZG/yGDJnzMdd/uZHAkVqJE+xIDOcmdSFZkBUicNc=

internal/adapter/citadel/audit_shipper.go

@@ -0,0 +1,161 @@
+package citadel
+
+import (
+	"context"
+	"log/slog"
+	"sync"
+	"time"
+
+	"github.com/orchard9/rdev/internal/domain"
+	"github.com/orchard9/rdev/internal/port"
+)
+
+// AuditShipper wraps an existing AuditLogger and also ships audit events to Citadel.
+// It buffers events and flushes them in batches for efficiency.
+type AuditShipper struct {
+	inner    port.AuditLogger
+	client   port.CitadelClient
+	tenantID string // rdev-platform tenant ID
+	logger   *slog.Logger
+
+	mu     sync.Mutex
+	buffer []map[string]any
+	done   chan struct{}
+}
+
+// NewAuditShipper wraps an existing AuditLogger with Citadel shipping.
+// tenantID is the Citadel tenant ID for the rdev-platform environment.
+func NewAuditShipper(inner port.AuditLogger, client port.CitadelClient, tenantID string, logger *slog.Logger) *AuditShipper {
+	if logger == nil {
+		logger = slog.Default()
+	}
+	s := &AuditShipper{
+		inner:    inner,
+		client:   client,
+		tenantID: tenantID,
+		logger:   logger.With("component", "audit_shipper"),
+		buffer:   make([]map[string]any, 0, 64),
+		done:     make(chan struct{}),
+	}
+	go s.flushLoop()
+	return s
+}
+
+// LogCommandStart records the start of a command and ships to Citadel.
+func (s *AuditShipper) LogCommandStart(ctx context.Context, entry *domain.AuditLogEntry) error {
+	// Always write to primary store first
+	if err := s.inner.LogCommandStart(ctx, entry); err != nil {
+		return err
+	}
+
+	// Buffer for Citadel (best-effort, don't block)
+	s.enqueue(map[string]any{
+		"message":      "audit: command started",
+		"level":        "info",
+		"service":      "rdev-platform",
+		"event_type":   "audit",
+		"audit_action": "command_start",
+		"command_id":   entry.CommandID,
+		"command_type": string(entry.CommandType),
+		"project_id":   entry.ProjectID,
+		"api_key_id":   entry.APIKeyID,
+		"client_ip":    entry.ClientIP,
+		"timestamp":    entry.StartedAt.Format(time.RFC3339Nano),
+	})
+
+	return nil
+}
+
+// LogCommandEnd records the completion of a command and ships to Citadel.
+func (s *AuditShipper) LogCommandEnd(ctx context.Context, commandID string, result *domain.AuditResult) error {
+	// Always write to primary store first
+	if err := s.inner.LogCommandEnd(ctx, commandID, result); err != nil {
+		return err
+	}
+
+	// Buffer for Citadel (best-effort)
+	s.enqueue(map[string]any{
+		"message":           "audit: command completed",
+		"level":             auditStatusToLevel(result.Status),
+		"service":           "rdev-platform",
+		"event_type":        "audit",
+		"audit_action":      "command_end",
+		"command_id":        commandID,
+		"status":            string(result.Status),
+		"exit_code":         result.ExitCode,
+		"duration_ms":       result.DurationMs,
+		"error_message":     result.ErrorMessage,
+		"output_size_bytes": result.OutputSizeBytes,
+		"timestamp":         time.Now().Format(time.RFC3339Nano),
+	})
+
+	return nil
+}
+
+// List delegates to the inner AuditLogger.
+func (s *AuditShipper) List(ctx context.Context, filters domain.AuditFilters) ([]domain.AuditLogEntry, error) {
+	return s.inner.List(ctx, filters)
+}
+
+// Get delegates to the inner AuditLogger.
+func (s *AuditShipper) Get(ctx context.Context, commandID string) (*domain.AuditLogEntry, error) {
+	return s.inner.Get(ctx, commandID)
+}
+
+// Close flushes remaining events and stops the background goroutine.
+func (s *AuditShipper) Close() {
+	close(s.done)
+	s.flush()
+}
+
+func (s *AuditShipper) enqueue(event map[string]any) {
+	s.mu.Lock()
+	defer s.mu.Unlock()
+	s.buffer = append(s.buffer, event)
+}
+
+func (s *AuditShipper) flushLoop() {
+	ticker := time.NewTicker(5 * time.Second)
+	defer ticker.Stop()
+
+	for {
+		select {
+		case <-ticker.C:
+			s.flush()
+		case <-s.done:
+			return
+		}
+	}
+}
+
+func (s *AuditShipper) flush() {
+	s.mu.Lock()
+	if len(s.buffer) == 0 {
+		s.mu.Unlock()
+		return
+	}
+	events := s.buffer
+	s.buffer = make([]map[string]any, 0, 64)
+	s.mu.Unlock()
+
+	ctx, cancel := context.WithTimeout(context.Background(), 10*time.Second)
+	defer cancel()
+
+	if err := s.client.IngestBatch(ctx, s.tenantID, events); err != nil {
+		s.logger.Warn("failed to ship audit events to citadel",
+			"error", err,
+			"event_count", len(events),
+		)
+	}
+}
+
+func auditStatusToLevel(status domain.AuditStatus) string {
+	switch status {
+	case domain.AuditStatusError:
+		return "error"
+	case domain.AuditStatusCancelled:
+		return "warn"
+	default:
+		return "info"
+	}
+}

internal/adapter/citadel/client.go

@@ -0,0 +1,244 @@
+// Package citadel provides an HTTP client adapter for the Citadel logging platform.
+//
+// This adapter communicates with a partner-hosted Citadel instance
+// (e.g., citadel-staging.orchard9.ai) to manage environments and ship logs.
+package citadel
+
+import (
+	"bytes"
+	"context"
+	"encoding/json"
+	"fmt"
+	"io"
+	"log/slog"
+	"net/http"
+	"time"
+
+	"github.com/orchard9/rdev/internal/port"
+)
+
+// Client implements port.CitadelClient via HTTP.
+type Client struct {
+	baseURL    string
+	apiKey     string
+	httpClient *http.Client
+	logger     *slog.Logger
+}
+
+// Ensure Client implements port.CitadelClient.
+var _ port.CitadelClient = (*Client)(nil)
+
+// Config holds configuration for the Citadel client.
+type Config struct {
+	// URL is the base URL of the Citadel instance (e.g., "https://citadel-staging.orchard9.ai").
+	URL string
+	// APIKey is the API key for authentication (starts with "ck_live_" or "ck_dev_").
+	APIKey string
+}
+
+// NewClient creates a new Citadel HTTP client.
+func NewClient(cfg Config, logger *slog.Logger) *Client {
+	if logger == nil {
+		logger = slog.Default()
+	}
+	return &Client{
+		baseURL: cfg.URL,
+		apiKey:  cfg.APIKey,
+		httpClient: &http.Client{
+			Timeout: 30 * time.Second,
+		},
+		logger: logger.With("component", "citadel_client"),
+	}
+}
+
+// CreateEnvironment creates a new Citadel environment.
+func (c *Client) CreateEnvironment(ctx context.Context, name string) (*port.CitadelEnvironment, error) {
+	body := map[string]string{"name": name}
+	data, err := json.Marshal(body)
+	if err != nil {
+		return nil, fmt.Errorf("marshal request: %w", err)
+	}
+
+	req, err := http.NewRequestWithContext(ctx, http.MethodPost, c.baseURL+"/api/v1/environments", bytes.NewReader(data))
+	if err != nil {
+		return nil, fmt.Errorf("create request: %w", err)
+	}
+	c.setHeaders(req)
+
+	resp, err := c.httpClient.Do(req)
+	if err != nil {
+		return nil, fmt.Errorf("create environment request failed: %w", err)
+	}
+	defer func() { _ = resp.Body.Close() }()
+
+	if resp.StatusCode == http.StatusConflict {
+		// Environment already exists — fetch and return it
+		c.logger.Info("citadel environment already exists, fetching", "name", name)
+		return c.GetEnvironment(ctx, name)
+	}
+
+	if resp.StatusCode < 200 || resp.StatusCode >= 300 {
+		return nil, c.readError(resp)
+	}
+
+	var result struct {
+		TenantID string `json:"tenant_id"`
+		Name     string `json:"name"`
+	}
+	if err := json.NewDecoder(resp.Body).Decode(&result); err != nil {
+		return nil, fmt.Errorf("decode response: %w", err)
+	}
+
+	c.logger.Info("citadel environment created", "name", name, "tenant_id", result.TenantID)
+
+	return &port.CitadelEnvironment{
+		TenantID: result.TenantID,
+		Name:     result.Name,
+	}, nil
+}
+
+// DeleteEnvironment removes a Citadel environment.
+func (c *Client) DeleteEnvironment(ctx context.Context, tenantID string) error {
+	req, err := http.NewRequestWithContext(ctx, http.MethodDelete, c.baseURL+"/api/v1/environments/"+tenantID, nil)
+	if err != nil {
+		return fmt.Errorf("create request: %w", err)
+	}
+	c.setHeaders(req)
+
+	resp, err := c.httpClient.Do(req)
+	if err != nil {
+		return fmt.Errorf("delete environment request failed: %w", err)
+	}
+	defer func() { _ = resp.Body.Close() }()
+
+	// 404 is fine — environment may already be gone
+	if resp.StatusCode == http.StatusNotFound {
+		return nil
+	}
+
+	if resp.StatusCode < 200 || resp.StatusCode >= 300 {
+		return c.readError(resp)
+	}
+
+	c.logger.Info("citadel environment deleted", "tenant_id", tenantID)
+	return nil
+}
+
+// GetEnvironment returns an environment by name.
+func (c *Client) GetEnvironment(ctx context.Context, name string) (*port.CitadelEnvironment, error) {
+	req, err := http.NewRequestWithContext(ctx, http.MethodGet, c.baseURL+"/api/v1/environments?name="+name, nil)
+	if err != nil {
+		return nil, fmt.Errorf("create request: %w", err)
+	}
+	c.setHeaders(req)
+
+	resp, err := c.httpClient.Do(req)
+	if err != nil {
+		return nil, fmt.Errorf("get environment request failed: %w", err)
+	}
+	defer func() { _ = resp.Body.Close() }()
+
+	if resp.StatusCode == http.StatusNotFound {
+		return nil, nil
+	}
+
+	if resp.StatusCode < 200 || resp.StatusCode >= 300 {
+		return nil, c.readError(resp)
+	}
+
+	var result struct {
+		TenantID string `json:"tenant_id"`
+		Name     string `json:"name"`
+	}
+	if err := json.NewDecoder(resp.Body).Decode(&result); err != nil {
+		return nil, fmt.Errorf("decode response: %w", err)
+	}
+
+	return &port.CitadelEnvironment{
+		TenantID: result.TenantID,
+		Name:     result.Name,
+	}, nil
+}
+
+// IngestEvent sends a single log event to Citadel.
+func (c *Client) IngestEvent(ctx context.Context, tenantID string, event map[string]any) error {
+	data, err := json.Marshal(event)
+	if err != nil {
+		return fmt.Errorf("marshal event: %w", err)
+	}
+
+	req, err := http.NewRequestWithContext(ctx, http.MethodPost, c.baseURL+"/api/v1/ingest/event", bytes.NewReader(data))
+	if err != nil {
+		return fmt.Errorf("create request: %w", err)
+	}
+	c.setHeaders(req)
+	req.Header.Set("X-Tenant-ID", tenantID)
+
+	resp, err := c.httpClient.Do(req)
+	if err != nil {
+		return fmt.Errorf("ingest event request failed: %w", err)
+	}
+	defer func() { _ = resp.Body.Close() }()
+
+	if resp.StatusCode < 200 || resp.StatusCode >= 300 {
+		return c.readError(resp)
+	}
+
+	return nil
+}
+
+// IngestBatch sends a batch of log events to Citadel.
+func (c *Client) IngestBatch(ctx context.Context, tenantID string, events []map[string]any) error {
+	data, err := json.Marshal(events)
+	if err != nil {
+		return fmt.Errorf("marshal events: %w", err)
+	}
+
+	req, err := http.NewRequestWithContext(ctx, http.MethodPost, c.baseURL+"/api/v1/ingest", bytes.NewReader(data))
+	if err != nil {
+		return fmt.Errorf("create request: %w", err)
+	}
+	c.setHeaders(req)
+	req.Header.Set("X-Tenant-ID", tenantID)
+
+	resp, err := c.httpClient.Do(req)
+	if err != nil {
+		return fmt.Errorf("ingest batch request failed: %w", err)
+	}
+	defer func() { _ = resp.Body.Close() }()
+
+	if resp.StatusCode < 200 || resp.StatusCode >= 300 {
+		return c.readError(resp)
+	}
+
+	return nil
+}
+
+// Healthy returns true if the Citadel instance is reachable.
+func (c *Client) Healthy(ctx context.Context) bool {
+	ctx, cancel := context.WithTimeout(ctx, 5*time.Second)
+	defer cancel()
+
+	req, err := http.NewRequestWithContext(ctx, http.MethodGet, c.baseURL+"/health", nil)
+	if err != nil {
+		return false
+	}
+
+	resp, err := c.httpClient.Do(req)
+	if err != nil {
+		return false
+	}
+	defer func() { _ = resp.Body.Close() }()
+
+	return resp.StatusCode == http.StatusOK
+}
+
+func (c *Client) setHeaders(req *http.Request) {
+	req.Header.Set("Content-Type", "application/json")
+	req.Header.Set("Authorization", "Bearer "+c.apiKey)
+}
+
+func (c *Client) readError(resp *http.Response) error {
+	body, _ := io.ReadAll(io.LimitReader(resp.Body, 1024))
+	return fmt.Errorf("citadel API error (HTTP %d): %s", resp.StatusCode, string(body))
+}

internal/adapter/deployer/resources.go

@@ -156,6 +156,10 @@ func (d *Deployer) buildDeployment(spec domain.DeploySpec, ns string, replicas i
 	if spec.ComponentPath != "" {
 		labels["component"] = sanitizeLabelValue(spec.ComponentPath)
 	}
+	// Apply extra labels (e.g., citadel.io/environment for log routing)
+	for k, v := range spec.ExtraLabels {
+		labels[k] = v
+	}
 
 	return &appsv1.Deployment{
 		ObjectMeta: metav1.ObjectMeta{

internal/adapter/templates/components.go

@@ -0,0 +1,341 @@
+package templates
+
+import (
+	"context"
+	"fmt"
+	"io/fs"
+	"os"
+	"path/filepath"
+	"strings"
+
+	"github.com/orchard9/rdev/internal/domain"
+	"github.com/orchard9/rdev/internal/port"
+)
+
+// availableComponentTemplates lists all supported component templates.
+var availableComponentTemplates = []port.ComponentTemplateInfo{
+	{
+		Type:        "service",
+		Description: "Go API service using pkg/ shared packages",
+		Stack:       "go",
+		DefaultPort: 8080,
+		DestDir:     "services",
+	},
+	{
+		Type:        "worker",
+		Description: "Go background worker for async job processing",
+		Stack:       "go",
+		DefaultPort: 0, // Workers don't expose ports
+		DestDir:     "workers",
+	},
+	{
+		Type:        "app-astro",
+		Description: "Astro landing page with Tailwind CSS",
+		Stack:       "astro",
+		DefaultPort: 4321,
+		DestDir:     "apps",
+	},
+	{
+		Type:        "app-react",
+		Description: "React SPA with Vite, TypeScript, and Tailwind",
+		Stack:       "react",
+		DefaultPort: 5173,
+		DestDir:     "apps",
+	},
+	{
+		Type:        "app-nextjs",
+		Description: "Next.js 14 dashboard with App Router and design system",
+		Stack:       "nextjs",
+		DefaultPort: 3000,
+		DestDir:     "apps",
+	},
+	{
+		Type:        "cli",
+		Description: "Go CLI tool using Cobra",
+		Stack:       "go",
+		DefaultPort: 0, // CLIs don't expose ports
+		DestDir:     "cli",
+	},
+}
+
+// GetComponentTemplate returns info about a specific component template.
+func (p *Provider) GetComponentTemplate(ctx context.Context, componentType string) (*port.ComponentTemplateInfo, error) {
+	// Check for context cancellation
+	select {
+	case <-ctx.Done():
+		return nil, ctx.Err()
+	default:
+	}
+
+	for _, t := range availableComponentTemplates {
+		if t.Type == componentType {
+			result := t
+			files, err := listComponentTemplateFiles(componentType)
+			if err == nil {
+				result.Files = files
+			}
+			return &result, nil
+		}
+	}
+	return nil, fmt.Errorf("%w: component type %s", domain.ErrTemplateNotFound, componentType)
+}
+
+// ListComponentTemplates returns available component templates.
+// If componentType is empty, returns all templates; otherwise filters by type.
+func (p *Provider) ListComponentTemplates(ctx context.Context, componentType string) ([]port.ComponentTemplateInfo, error) {
+	// Check for context cancellation
+	select {
+	case <-ctx.Done():
+		return nil, ctx.Err()
+	default:
+	}
+
+	var result []port.ComponentTemplateInfo
+	for _, t := range availableComponentTemplates {
+		if componentType != "" && t.Type != componentType {
+			continue
+		}
+		info := t
+		files, err := listComponentTemplateFiles(t.Type)
+		if err == nil {
+			info.Files = files
+		}
+		result = append(result, info)
+	}
+	return result, nil
+}
+
+// GetComponentFiles returns the files for a component template with variables interpolated.
+func (p *Provider) GetComponentFiles(ctx context.Context, componentType string, destPath string, vars map[string]string) ([]port.ComponentFile, error) {
+	// Check for context cancellation
+	select {
+	case <-ctx.Done():
+		return nil, ctx.Err()
+	default:
+	}
+
+	// Validate component type exists
+	found := false
+	for _, t := range availableComponentTemplates {
+		if t.Type == componentType {
+			found = true
+			break
+		}
+	}
+	if !found {
+		return nil, fmt.Errorf("%w: component type %s", domain.ErrTemplateNotFound, componentType)
+	}
+
+	templateDir := "templates/components/" + componentType
+	var files []port.ComponentFile
+
+	err := fs.WalkDir(templatesFS, templateDir, func(path string, d fs.DirEntry, err error) error {
+		if err != nil {
+			return err
+		}
+		if d.IsDir() {
+			return nil
+		}
+
+		// Read file content
+		content, err := templatesFS.ReadFile(path)
+		if err != nil {
+			return fmt.Errorf("failed to read component template file %s: %w", path, err)
+		}
+
+		// Interpolate variables
+		interpolated := interpolateVars(string(content), vars)
+
+		// Calculate relative path from component template root
+		relPath, err := filepath.Rel(templateDir, path)
+		if err != nil {
+			return fmt.Errorf("failed to get relative path: %w", err)
+		}
+
+		// Strip .tmpl extension
+		relPath = strings.TrimSuffix(relPath, ".tmpl")
+
+		// Prepend destination path
+		fullPath := filepath.Join(destPath, relPath)
+
+		files = append(files, port.ComponentFile{
+			Path:    fullPath,
+			Content: interpolated,
+		})
+
+		return nil
+	})
+
+	if err != nil {
+		return nil, fmt.Errorf("failed to collect component template files: %w", err)
+	}
+
+	if len(files) == 0 {
+		return nil, fmt.Errorf("component template %s contains no files", componentType)
+	}
+
+	p.logger.Debug("prepared component files",
+		"component_type", componentType,
+		"dest_path", destPath,
+		"file_count", len(files),
+	)
+
+	return files, nil
+}
+
+// listComponentTemplateFiles returns the list of files in a component template.
+func listComponentTemplateFiles(componentType string) ([]string, error) {
+	templateDir := "templates/components/" + componentType
+	var files []string
+
+	err := fs.WalkDir(templatesFS, templateDir, func(path string, d fs.DirEntry, err error) error {
+		if err != nil {
+			return err
+		}
+		if d.IsDir() {
+			return nil
+		}
+		relPath, err := filepath.Rel(templateDir, path)
+		if err != nil {
+			return err
+		}
+		// Strip .tmpl extension for display
+		relPath = strings.TrimSuffix(relPath, ".tmpl")
+		files = append(files, relPath)
+		return nil
+	})
+
+	return files, err
+}
+
+// GetComponentWoodpeckerStep returns the .woodpecker.step.yml content for a component.
+// This is the CI step that should be inserted into the main .woodpecker.yml file.
+func (p *Provider) GetComponentWoodpeckerStep(ctx context.Context, componentType string, vars map[string]string) (string, error) {
+	// Check for context cancellation
+	select {
+	case <-ctx.Done():
+		return "", ctx.Err()
+	default:
+	}
+
+	stepPath := "templates/components/" + componentType + "/.woodpecker.step.yml.tmpl"
+	content, err := templatesFS.ReadFile(stepPath)
+	if err != nil {
+		return "", fmt.Errorf("failed to read woodpecker step template: %w", err)
+	}
+
+	return interpolateVars(string(content), vars), nil
+}
+
+// RenderSkeletonToDir renders the monorepo skeleton template to a local directory.
+// This is used for testing templates locally without needing Gitea.
+func RenderSkeletonToDir(outputDir string, vars map[string]string) error {
+	templateDir := "templates/skeleton"
+
+	return fs.WalkDir(templatesFS, templateDir, func(path string, d fs.DirEntry, err error) error {
+		if err != nil {
+			return err
+		}
+		if d.IsDir() {
+			return nil
+		}
+
+		// Read file content
+		content, err := templatesFS.ReadFile(path)
+		if err != nil {
+			return fmt.Errorf("failed to read template file %s: %w", path, err)
+		}
+
+		// Interpolate variables
+		interpolated := interpolateVars(string(content), vars)
+
+		// Calculate relative path from template root
+		relPath, err := filepath.Rel(templateDir, path)
+		if err != nil {
+			return fmt.Errorf("failed to get relative path: %w", err)
+		}
+
+		// Strip .tmpl extension
+		relPath = strings.TrimSuffix(relPath, ".tmpl")
+
+		// Create output path
+		outPath := filepath.Join(outputDir, relPath)
+
+		// Create parent directories
+		if err := os.MkdirAll(filepath.Dir(outPath), 0755); err != nil {
+			return fmt.Errorf("failed to create directory for %s: %w", outPath, err)
+		}
+
+		// Write file
+		if err := os.WriteFile(outPath, []byte(interpolated), 0644); err != nil {
+			return fmt.Errorf("failed to write file %s: %w", outPath, err)
+		}
+
+		return nil
+	})
+}
+
+// RenderComponentToDir renders a component template to a local directory.
+// The destPath is relative to outputDir (e.g., "services/my-api").
+func RenderComponentToDir(outputDir, componentType, destPath string, vars map[string]string) error {
+	// Validate component type exists
+	found := false
+	for _, t := range availableComponentTemplates {
+		if t.Type == componentType {
+			found = true
+			break
+		}
+	}
+	if !found {
+		return fmt.Errorf("unknown component type: %s", componentType)
+	}
+
+	templateDir := "templates/components/" + componentType
+
+	return fs.WalkDir(templatesFS, templateDir, func(path string, d fs.DirEntry, err error) error {
+		if err != nil {
+			return err
+		}
+		if d.IsDir() {
+			return nil
+		}
+
+		// Skip .woodpecker.step.yml.tmpl - it's for CI insertion, not file creation
+		if strings.HasSuffix(path, ".woodpecker.step.yml.tmpl") {
+			return nil
+		}
+
+		// Read file content
+		content, err := templatesFS.ReadFile(path)
+		if err != nil {
+			return fmt.Errorf("failed to read component template file %s: %w", path, err)
+		}
+
+		// Interpolate variables
+		interpolated := interpolateVars(string(content), vars)
+
+		// Calculate relative path from component template root
+		relPath, err := filepath.Rel(templateDir, path)
+		if err != nil {
+			return fmt.Errorf("failed to get relative path: %w", err)
+		}
+
+		// Strip .tmpl extension
+		relPath = strings.TrimSuffix(relPath, ".tmpl")
+
+		// Create output path under destPath
+		outPath := filepath.Join(outputDir, destPath, relPath)
+
+		// Create parent directories
+		if err := os.MkdirAll(filepath.Dir(outPath), 0755); err != nil {
+			return fmt.Errorf("failed to create directory for %s: %w", outPath, err)
+		}
+
+		// Write file
+		if err := os.WriteFile(outPath, []byte(interpolated), 0644); err != nil {
+			return fmt.Errorf("failed to write file %s: %w", outPath, err)
+		}
+
+		return nil
+	})
+}

internal/adapter/templates/provider.go

@@ -43,52 +43,6 @@ var skeletonTemplate = port.TemplateInfo{
 	Stack:       "monorepo",
 }
 
-// availableComponentTemplates lists all supported component templates.
-var availableComponentTemplates = []port.ComponentTemplateInfo{
-	{
-		Type:        "service",
-		Description: "Go API service using pkg/ shared packages",
-		Stack:       "go",
-		DefaultPort: 8080,
-		DestDir:     "services",
-	},
-	{
-		Type:        "worker",
-		Description: "Go background worker for async job processing",
-		Stack:       "go",
-		DefaultPort: 0, // Workers don't expose ports
-		DestDir:     "workers",
-	},
-	{
-		Type:        "app-astro",
-		Description: "Astro landing page with Tailwind CSS",
-		Stack:       "astro",
-		DefaultPort: 4321,
-		DestDir:     "apps",
-	},
-	{
-		Type:        "app-react",
-		Description: "React SPA with Vite, TypeScript, and Tailwind",
-		Stack:       "react",
-		DefaultPort: 5173,
-		DestDir:     "apps",
-	},
-	{
-		Type:        "app-nextjs",
-		Description: "Next.js 14 dashboard with App Router and design system",
-		Stack:       "nextjs",
-		DefaultPort: 3000,
-		DestDir:     "apps",
-	},
-	{
-		Type:        "cli",
-		Description: "Go CLI tool using Cobra",
-		Stack:       "go",
-		DefaultPort: 0, // CLIs don't expose ports
-		DestDir:     "cli",
-	},
-}
-
 // templateNameRegex validates template names (alphanumeric, dash only).
 var templateNameRegex = regexp.MustCompile(`^[a-z][a-z0-9-]*$`)
 
@@ -323,172 +277,3 @@ func (p *Provider) GetSkeleton(ctx context.Context) (*port.TemplateInfo, error)
 	}
 	return &result, nil
 }
-
-// GetComponentTemplate returns info about a specific component template.
-func (p *Provider) GetComponentTemplate(ctx context.Context, componentType string) (*port.ComponentTemplateInfo, error) {
-	// Check for context cancellation
-	select {
-	case <-ctx.Done():
-		return nil, ctx.Err()
-	default:
-	}
-
-	for _, t := range availableComponentTemplates {
-		if t.Type == componentType {
-			result := t
-			files, err := listComponentTemplateFiles(componentType)
-			if err == nil {
-				result.Files = files
-			}
-			return &result, nil
-		}
-	}
-	return nil, fmt.Errorf("%w: component type %s", domain.ErrTemplateNotFound, componentType)
-}
-
-// ListComponentTemplates returns available component templates.
-// If componentType is empty, returns all templates; otherwise filters by type.
-func (p *Provider) ListComponentTemplates(ctx context.Context, componentType string) ([]port.ComponentTemplateInfo, error) {
-	// Check for context cancellation
-	select {
-	case <-ctx.Done():
-		return nil, ctx.Err()
-	default:
-	}
-
-	var result []port.ComponentTemplateInfo
-	for _, t := range availableComponentTemplates {
-		if componentType != "" && t.Type != componentType {
-			continue
-		}
-		info := t
-		files, err := listComponentTemplateFiles(t.Type)
-		if err == nil {
-			info.Files = files
-		}
-		result = append(result, info)
-	}
-	return result, nil
-}
-
-// GetComponentFiles returns the files for a component template with variables interpolated.
-func (p *Provider) GetComponentFiles(ctx context.Context, componentType string, destPath string, vars map[string]string) ([]port.ComponentFile, error) {
-	// Check for context cancellation
-	select {
-	case <-ctx.Done():
-		return nil, ctx.Err()
-	default:
-	}
-
-	// Validate component type exists
-	found := false
-	for _, t := range availableComponentTemplates {
-		if t.Type == componentType {
-			found = true
-			break
-		}
-	}
-	if !found {
-		return nil, fmt.Errorf("%w: component type %s", domain.ErrTemplateNotFound, componentType)
-	}
-
-	templateDir := "templates/components/" + componentType
-	var files []port.ComponentFile
-
-	err := fs.WalkDir(templatesFS, templateDir, func(path string, d fs.DirEntry, err error) error {
-		if err != nil {
-			return err
-		}
-		if d.IsDir() {
-			return nil
-		}
-
-		// Read file content
-		content, err := templatesFS.ReadFile(path)
-		if err != nil {
-			return fmt.Errorf("failed to read component template file %s: %w", path, err)
-		}
-
-		// Interpolate variables
-		interpolated := interpolateVars(string(content), vars)
-
-		// Calculate relative path from component template root
-		relPath, err := filepath.Rel(templateDir, path)
-		if err != nil {
-			return fmt.Errorf("failed to get relative path: %w", err)
-		}
-
-		// Strip .tmpl extension
-		relPath = strings.TrimSuffix(relPath, ".tmpl")
-
-		// Prepend destination path
-		fullPath := filepath.Join(destPath, relPath)
-
-		files = append(files, port.ComponentFile{
-			Path:    fullPath,
-			Content: interpolated,
-		})
-
-		return nil
-	})
-
-	if err != nil {
-		return nil, fmt.Errorf("failed to collect component template files: %w", err)
-	}
-
-	if len(files) == 0 {
-		return nil, fmt.Errorf("component template %s contains no files", componentType)
-	}
-
-	p.logger.Debug("prepared component files",
-		"component_type", componentType,
-		"dest_path", destPath,
-		"file_count", len(files),
-	)
-
-	return files, nil
-}
-
-// listComponentTemplateFiles returns the list of files in a component template.
-func listComponentTemplateFiles(componentType string) ([]string, error) {
-	templateDir := "templates/components/" + componentType
-	var files []string
-
-	err := fs.WalkDir(templatesFS, templateDir, func(path string, d fs.DirEntry, err error) error {
-		if err != nil {
-			return err
-		}
-		if d.IsDir() {
-			return nil
-		}
-		relPath, err := filepath.Rel(templateDir, path)
-		if err != nil {
-			return err
-		}
-		// Strip .tmpl extension for display
-		relPath = strings.TrimSuffix(relPath, ".tmpl")
-		files = append(files, relPath)
-		return nil
-	})
-
-	return files, err
-}
-
-// GetComponentWoodpeckerStep returns the .woodpecker.step.yml content for a component.
-// This is the CI step that should be inserted into the main .woodpecker.yml file.
-func (p *Provider) GetComponentWoodpeckerStep(ctx context.Context, componentType string, vars map[string]string) (string, error) {
-	// Check for context cancellation
-	select {
-	case <-ctx.Done():
-		return "", ctx.Err()
-	default:
-	}
-
-	stepPath := "templates/components/" + componentType + "/.woodpecker.step.yml.tmpl"
-	content, err := templatesFS.ReadFile(stepPath)
-	if err != nil {
-		return "", fmt.Errorf("failed to read woodpecker step template: %w", err)
-	}
-
-	return interpolateVars(string(content), vars), nil
-}

internal/adapter/templates/templates/components/app-nextjs/next.config.mjs.tmpl

@@ -1,6 +1,5 @@
-import type { NextConfig } from 'next';
+/** @type {import('next').NextConfig} */
-
+const nextConfig = {
-const nextConfig: NextConfig = {
   // Enable React strict mode for better development experience
   reactStrictMode: true,
 

internal/adapter/templates/templates/components/app-react/package.json.tmpl

@@ -11,9 +11,13 @@
     "format": "prettier --write src/"
   },
   "dependencies": {
-    "@{{PROJECT_NAME}}/logger": "workspace:*",
+    "@{{PROJECT_NAME}}/ai-client": "workspace:*",
-    "@{{PROJECT_NAME}}/ui": "workspace:*",
+    "@{{PROJECT_NAME}}/api-client": "workspace:*",
+    "@{{PROJECT_NAME}}/auth": "workspace:*",
     "@{{PROJECT_NAME}}/layout": "workspace:*",
+    "@{{PROJECT_NAME}}/logger": "workspace:*",
+    "@{{PROJECT_NAME}}/realtime": "workspace:*",
+    "@{{PROJECT_NAME}}/ui": "workspace:*",
     "react": "^18.3.1",
     "react-dom": "^18.3.1",
     "react-router-dom": "^6.23.1"

internal/adapter/templates/templates/components/app-react/postcss.config.cjs

@@ -1,4 +1,4 @@
-export default {
+module.exports = {
   plugins: {
     tailwindcss: {},
     autoprefixer: {},

internal/adapter/templates/templates/components/app-react/src/App.tsx.tmpl

@@ -1,4 +1,5 @@
 import { Routes, Route, useLocation, useNavigate } from 'react-router-dom';
+import { AuthProvider, useAuth, ProtectedRoute } from '@{{PROJECT_NAME}}/auth';
 import { DashboardShell, Sidebar, Header, type NavItem } from '@{{PROJECT_NAME}}/layout';
 import {
   Button,
@@ -9,13 +10,24 @@ import {
   CardContent,
   Badge,
   Home,
+  ImageIcon,
   Users,
   Settings,
   BarChart3,
+  MessageSquare,
+  Sparkles,
+  Loader2,
 } from '@{{PROJECT_NAME}}/ui';
+import { LoginPage } from './pages/LoginPage';
+import { ChatPage } from './pages/ChatPage';
+import { GeneratePage } from './pages/GeneratePage';
+import { MediaPage } from './pages/MediaPage';
 
 const navItems: NavItem[] = [
   { label: 'Dashboard', href: '/', icon: Home },
+  { label: 'Chat', href: '/chat', icon: MessageSquare },
+  { label: 'Generate', href: '/generate', icon: Sparkles },
+  { label: 'Media', href: '/media', icon: ImageIcon },
   { label: 'Analytics', href: '/analytics', icon: BarChart3 },
   { label: 'Users', href: '/users', icon: Users, badge: '12' },
   { label: 'Settings', href: '/settings', icon: Settings },
@@ -23,6 +35,9 @@ const navItems: NavItem[] = [
 
 const pageTitles: Record<string, string> = {
   '/': 'Dashboard',
+  '/chat': 'Chat',
+  '/generate': 'Generate',
+  '/media': 'Media',
   '/analytics': 'Analytics',
   '/users': 'Users',
   '/settings': 'Settings',
@@ -195,6 +210,14 @@ function AnalyticsPage() {
 }
 
 function SettingsPage() {
+  const { logout } = useAuth();
+  const navigate = useNavigate();
+
+  const handleLogout = async () => {
+    await logout();
+    navigate('/login');
+  };
+
   return (
     <div className="space-y-6">
       <Card>
@@ -218,6 +241,22 @@ function SettingsPage() {
         </CardContent>
       </Card>
 
+      <Card>
+        <CardHeader>
+          <CardTitle>Account</CardTitle>
+          <CardDescription>Manage your account settings.</CardDescription>
+        </CardHeader>
+        <CardContent>
+          <div className="flex items-center justify-between">
+            <div>
+              <p className="text-sm font-medium text-[var(--text-primary)]">Sign Out</p>
+              <p className="text-sm text-[var(--text-muted)]">Sign out of your account on this device.</p>
+            </div>
+            <Button variant="outline" onClick={handleLogout}>Sign Out</Button>
+          </div>
+        </CardContent>
+      </Card>
+
       <Card>
         <CardHeader>
           <CardTitle>Danger Zone</CardTitle>
@@ -237,9 +276,21 @@ function SettingsPage() {
   );
 }
 
-function App() {
+function LoadingScreen() {
+  return (
+    <div className="min-h-screen flex items-center justify-center bg-[var(--surface-100)]">
+      <div className="flex flex-col items-center gap-4">
+        <Loader2 className="h-8 w-8 animate-spin text-[var(--accent)]" />
+        <p className="text-sm text-[var(--text-muted)]">Loading...</p>
+      </div>
+    </div>
+  );
+}
+
+function AppLayout() {
   const location = useLocation();
   const navigate = useNavigate();
+  const { user } = useAuth();
 
   const itemsWithActive = navItems.map((item) => ({
     ...item,
@@ -259,7 +310,7 @@ function App() {
           onNavigate={(href) => navigate(href)}
           footer={
             <div className="text-sm text-[var(--text-muted)]">
-              v0.0.1
+              {user?.email || 'v0.0.1'}
             </div>
           }
         />
@@ -274,6 +325,9 @@ function App() {
     >
       <Routes>
         <Route path="/" element={<DashboardPage />} />
+        <Route path="/chat" element={<ChatPage />} />
+        <Route path="/generate" element={<GeneratePage />} />
+        <Route path="/media" element={<MediaPage />} />
         <Route path="/users" element={<UsersPage />} />
         <Route path="/analytics" element={<AnalyticsPage />} />
         <Route path="/settings" element={<SettingsPage />} />
@@ -282,4 +336,41 @@ function App() {
   );
 }
 
+function AppRoutes() {
+  const location = useLocation();
+  const navigate = useNavigate();
+
+  return (
+    <Routes>
+      <Route path="/login" element={<LoginPage />} />
+      <Route
+        path="/*"
+        element={
+          <ProtectedRoute
+            redirectTo="/login"
+            onRedirect={(path) => {
+              // Navigate to login, storing current location for redirect after login
+              navigate(path, { state: { from: location.pathname }, replace: true });
+            }}
+            fallback={<LoadingScreen />}
+          >
+            <AppLayout />
+          </ProtectedRoute>
+        }
+      />
+    </Routes>
+  );
+}
+
+function App() {
+  // Determine API base URL from environment or current origin
+  const apiBaseUrl = import.meta.env.VITE_API_URL || '';
+
+  return (
+    <AuthProvider loginUrl={`${apiBaseUrl}/api/{{SERVICE_NAME}}/auth/login`}>
+      <AppRoutes />
+    </AuthProvider>
+  );
+}
+
 export default App;

internal/adapter/templates/templates/components/app-react/src/main.tsx.tmpl

@@ -7,7 +7,12 @@ import './lib/logger';
 
 ReactDOM.createRoot(document.getElementById('root')!).render(
   <React.StrictMode>
-    <BrowserRouter>
+    <BrowserRouter
+      future={{
+        v7_startTransition: true,
+        v7_relativeSplatPath: true,
+      }}
+    >
       <App />
     </BrowserRouter>
   </React.StrictMode>

internal/adapter/templates/templates/components/app-react/src/pages/ChatPage.tsx.tmpl

@@ -0,0 +1,190 @@
+import { useRef, useEffect, useState, useCallback, useMemo } from 'react';
+import { useAuth } from '@{{PROJECT_NAME}}/auth';
+import { useChat } from '@{{PROJECT_NAME}}/realtime';
+import {
+  Card,
+  CardHeader,
+  CardTitle,
+  CardDescription,
+  CardContent,
+  ChatBubble,
+  ChatInput,
+  Badge,
+  ProviderBadge,
+} from '@{{PROJECT_NAME}}/ui';
+
+interface TimelineMessage {
+  id: string;
+  content: string;
+  role: 'user' | 'assistant' | 'system';
+  timestamp: Date;
+  provider?: string;
+  isStreaming?: boolean;
+}
+
+export function ChatPage() {
+  const { user, getToken } = useAuth();
+  const messagesEndRef = useRef<HTMLDivElement>(null);
+
+  // API base URL from environment
+  const apiBaseUrl = import.meta.env.VITE_API_URL || '';
+
+  const authHeaders = useMemo(() => {
+    const token = getToken();
+    return token ? { Authorization: `Bearer ${token}` } : undefined;
+  }, [getToken]);
+
+  const {
+    messages,
+    aiMessages,
+    streamingMessages,
+    sendMessage,
+    connectionState,
+    onlineUsers,
+  } = useChat({
+    endpoint: `${apiBaseUrl}/api/{{SERVICE_NAME}}/chat/messages`,
+    sseEndpoint: `${apiBaseUrl}/api/{{SERVICE_NAME}}/events`,
+    channel: 'channel:general',
+    userId: user?.id || 'anonymous',
+    userName: user?.name || user?.email || 'Anonymous',
+    headers: authHeaders,
+  });
+
+  // Track send errors for user feedback
+  const [sendError, setSendError] = useState<string | null>(null);
+
+  // Merge user messages + AI messages into a single sorted timeline
+  const timeline = useMemo<TimelineMessage[]>(() => {
+    const combined: TimelineMessage[] = [];
+
+    for (const msg of messages) {
+      combined.push({
+        id: msg.id,
+        content: msg.content,
+        role: msg.userId === user?.id ? 'user' : 'assistant',
+        timestamp: new Date(msg.timestamp),
+      });
+    }
+
+    for (const msg of aiMessages) {
+      combined.push({
+        id: msg.id,
+        content: msg.content,
+        role: 'assistant',
+        timestamp: new Date(msg.timestamp),
+        provider: msg.provider,
+      });
+    }
+
+    // Add in-progress streaming messages
+    for (const [, stream] of streamingMessages) {
+      combined.push({
+        id: stream.streamId,
+        content: stream.content,
+        role: 'assistant',
+        timestamp: new Date(stream.timestamp),
+        isStreaming: true,
+      });
+    }
+
+    combined.sort((a, b) => a.timestamp.getTime() - b.timestamp.getTime());
+    return combined;
+  }, [messages, aiMessages, streamingMessages, user?.id]);
+
+  // Handle sending a message (wraps async sendMessage for ChatInput)
+  const handleSendMessage = useCallback((content: string) => {
+    sendMessage(content).catch(() => {
+      setSendError('Failed to send message. Please try again.');
+      setTimeout(() => setSendError(null), 3000);
+    });
+  }, [sendMessage]);
+
+  // Auto-scroll to bottom when new messages arrive
+  useEffect(() => {
+    messagesEndRef.current?.scrollIntoView({ behavior: 'smooth' });
+  }, [timeline]);
+
+  const connectionBadge = () => {
+    switch (connectionState) {
+      case 'connected':
+        return <Badge variant="success">Connected</Badge>;
+      case 'connecting':
+        return <Badge variant="warning">Connecting...</Badge>;
+      case 'disconnected':
+        return <Badge variant="error">Disconnected</Badge>;
+      case 'error':
+        return <Badge variant="error">Error</Badge>;
+      default:
+        return null;
+    }
+  };
+
+  return (
+    <div className="h-[calc(100vh-8rem)] flex flex-col">
+      <Card className="flex-1 flex flex-col">
+        <CardHeader className="flex-shrink-0 flex flex-row items-center justify-between">
+          <div>
+            <CardTitle>AI Chat</CardTitle>
+            <CardDescription>
+              Chat with AI in real-time
+            </CardDescription>
+          </div>
+          <div className="flex items-center gap-3">
+            <span className="text-sm text-[var(--text-muted)]">
+              {onlineUsers.length} online
+            </span>
+            {connectionBadge()}
+          </div>
+        </CardHeader>
+
+        <CardContent className="flex-1 flex flex-col min-h-0">
+          {/* Messages area */}
+          <div className="flex-1 overflow-y-auto space-y-4 pr-2 mb-4">
+            {timeline.length === 0 ? (
+              <div className="h-full flex items-center justify-center">
+                <p className="text-[var(--text-muted)] text-sm">
+                  No messages yet. Start the conversation!
+                </p>
+              </div>
+            ) : (
+              timeline.map((msg) => (
+                <div key={msg.id}>
+                  <ChatBubble
+                    role={msg.role}
+                    content={msg.content}
+                    timestamp={msg.timestamp}
+                    isStreaming={msg.isStreaming}
+                  />
+                  {msg.provider && (
+                    <div className={msg.role === 'user' ? 'text-right' : 'text-left'}>
+                      <ProviderBadge provider={msg.provider} className="mt-1" />
+                    </div>
+                  )}
+                </div>
+              ))
+            )}
+            <div ref={messagesEndRef} />
+          </div>
+
+          {/* Input area */}
+          <div className="flex-shrink-0 space-y-2">
+            {sendError && (
+              <div className="px-3 py-2 text-sm text-[var(--error)] bg-[var(--error)]/10 rounded-lg">
+                {sendError}
+              </div>
+            )}
+            <ChatInput
+              onSubmit={handleSendMessage}
+              disabled={connectionState !== 'connected'}
+              placeholder={
+                connectionState === 'connected'
+                  ? 'Type a message... (Cmd+Enter to send)'
+                  : 'Connecting...'
+              }
+            />
+          </div>
+        </CardContent>
+      </Card>
+    </div>
+  );
+}

internal/adapter/templates/templates/components/app-react/src/pages/GeneratePage.tsx.tmpl

@@ -0,0 +1,248 @@
+import { useState, useMemo } from 'react';
+import { useNavigate } from 'react-router-dom';
+import { useAuth } from '@{{PROJECT_NAME}}/auth';
+import { useMediaGeneration } from '@{{PROJECT_NAME}}/realtime';
+import {
+  Card,
+  CardHeader,
+  CardTitle,
+  CardDescription,
+  CardContent,
+  Button,
+  FormField,
+  Select,
+  SelectTrigger,
+  SelectValue,
+  SelectContent,
+  SelectItem,
+  ImageGrid,
+  VideoGrid,
+  GenerationProgress,
+  ProviderBadge,
+  Loader2,
+} from '@{{PROJECT_NAME}}/ui';
+
+type GenerateMode = 'image' | 'video';
+
+interface ImageResult {
+  images: Array<{ data: string; isUrl: boolean; seed?: number }>;
+  provider: string;
+  latencyMs: number;
+}
+
+interface VideoResult {
+  videos: Array<{ data: string; isUrl: boolean; mimeType: string }>;
+  provider: string;
+  latencyMs: number;
+}
+
+export function GeneratePage() {
+  const { user, getToken } = useAuth();
+  const navigate = useNavigate();
+  const [mode, setMode] = useState<GenerateMode>('image');
+  const [prompt, setPrompt] = useState('');
+  const [aspectRatio, setAspectRatio] = useState('1:1');
+  const [count, setCount] = useState(1);
+  const [duration, setDuration] = useState('5s');
+
+  const apiPrefix = import.meta.env.VITE_API_URL || '';
+
+  const authHeaders = useMemo(() => {
+    const token = getToken();
+    return token ? { Authorization: `Bearer ${token}` } : undefined;
+  }, [getToken]);
+
+  const imageGen = useMediaGeneration<ImageResult>({
+    endpoint: `${apiPrefix}/api/{{SERVICE_NAME}}/generate/image`,
+    sseEndpoint: `${apiPrefix}/api/{{SERVICE_NAME}}/events`,
+    userId: user?.id || 'anonymous',
+    headers: authHeaders,
+  });
+
+  const videoGen = useMediaGeneration<VideoResult>({
+    endpoint: `${apiPrefix}/api/{{SERVICE_NAME}}/generate/video`,
+    sseEndpoint: `${apiPrefix}/api/{{SERVICE_NAME}}/events`,
+    userId: user?.id || 'anonymous',
+    headers: authHeaders,
+  });
+
+  const gen = mode === 'image' ? imageGen : videoGen;
+  const isGenerating = gen.status === 'pending' || gen.status === 'generating';
+
+  const handleGenerate = async () => {
+    if (!prompt.trim()) return;
+    gen.reset();
+    const request = mode === 'image'
+      ? { prompt, count, aspectRatio }
+      : { prompt, aspectRatio, duration };
+    await gen.generate(request);
+  };
+
+  return (
+    <div className="space-y-6">
+      <Card>
+        <CardHeader>
+          <CardTitle>AI Generation</CardTitle>
+          <CardDescription>
+            Generate images and videos using AI (Gemini / LaoZhang)
+          </CardDescription>
+        </CardHeader>
+        <CardContent className="space-y-4">
+          {/* Mode toggle */}
+          <div className="flex gap-2">
+            <Button
+              variant={mode === 'image' ? 'default' : 'outline'}
+              onClick={() => setMode('image')}
+              disabled={isGenerating}
+            >
+              Images
+            </Button>
+            <Button
+              variant={mode === 'video' ? 'default' : 'outline'}
+              onClick={() => {
+                setMode('video');
+                if (aspectRatio === '1:1') setAspectRatio('16:9');
+              }}
+              disabled={isGenerating}
+            >
+              Video
+            </Button>
+          </div>
+
+          <FormField
+            label="Prompt"
+            name="prompt"
+            value={prompt}
+            onChange={(e) => setPrompt(e.target.value)}
+            placeholder={
+              mode === 'image'
+                ? 'A serene mountain landscape at sunset...'
+                : 'A cat playing piano in a jazz club...'
+            }
+          />
+
+          <div className="grid grid-cols-2 gap-4">
+            <div className="space-y-2">
+              <label className="text-sm font-medium text-[var(--text-primary)]">Aspect Ratio</label>
+              <Select value={aspectRatio} onValueChange={(v) => setAspectRatio(v)}>
+                <SelectTrigger>
+                  <SelectValue />
+                </SelectTrigger>
+                <SelectContent>
+                  {mode === 'image' && <SelectItem value="1:1">Square (1:1)</SelectItem>}
+                  <SelectItem value="16:9">Landscape (16:9)</SelectItem>
+                  <SelectItem value="9:16">Portrait (9:16)</SelectItem>
+                </SelectContent>
+              </Select>
+            </div>
+
+            {mode === 'image' ? (
+              <div className="space-y-2">
+                <label className="text-sm font-medium text-[var(--text-primary)]">Count</label>
+                <Select value={String(count)} onValueChange={(v) => setCount(Number(v))}>
+                  <SelectTrigger>
+                    <SelectValue />
+                  </SelectTrigger>
+                  <SelectContent>
+                    <SelectItem value="1">1 image</SelectItem>
+                    <SelectItem value="2">2 images</SelectItem>
+                    <SelectItem value="4">4 images</SelectItem>
+                  </SelectContent>
+                </Select>
+              </div>
+            ) : (
+              <div className="space-y-2">
+                <label className="text-sm font-medium text-[var(--text-primary)]">Duration</label>
+                <Select value={duration} onValueChange={(v) => setDuration(v)}>
+                  <SelectTrigger>
+                    <SelectValue />
+                  </SelectTrigger>
+                  <SelectContent>
+                    <SelectItem value="5s">5 seconds</SelectItem>
+                    <SelectItem value="10s">10 seconds</SelectItem>
+                  </SelectContent>
+                </Select>
+              </div>
+            )}
+          </div>
+
+          <Button onClick={handleGenerate} disabled={isGenerating || !prompt.trim()}>
+            {isGenerating && <Loader2 className="mr-2 h-4 w-4 animate-spin" />}
+            {isGenerating ? 'Generating...' : `Generate ${mode === 'image' ? 'Images' : 'Video'}`}
+          </Button>
+        </CardContent>
+      </Card>
+
+      {isGenerating && (
+        <Card>
+          <CardContent className="py-4 space-y-3">
+            <div className="flex items-center justify-between text-sm">
+              <span className="text-[var(--text-muted)]">{gen.message || 'Starting...'}</span>
+              <span className="text-[var(--text-muted)]">{gen.progress}%</span>
+            </div>
+            <GenerationProgress percent={gen.progress} />
+            {gen.sseState !== 'connected' && (
+              <p className="text-xs text-[var(--warning)]">
+                SSE {gen.sseState} — events may be delayed
+              </p>
+            )}
+          </CardContent>
+        </Card>
+      )}
+
+      {gen.status === 'failed' && gen.error && (
+        <Card className="border-[var(--error)]">
+          <CardContent className="py-4 text-[var(--error)]">
+            {gen.error}
+          </CardContent>
+        </Card>
+      )}
+
+      {gen.status === 'complete' && imageGen.result && mode === 'image' && (
+        <Card>
+          <CardHeader className="flex flex-row items-center justify-between">
+            <CardTitle>Results</CardTitle>
+            <div className="flex items-center gap-2">
+              {imageGen.result.provider && <ProviderBadge provider={imageGen.result.provider} />}
+              <Button variant="outline" size="sm" onClick={() => navigate('/media')}>
+                View in Library
+              </Button>
+            </div>
+          </CardHeader>
+          <CardContent>
+            <ImageGrid
+              images={imageGen.result.images.map((img) => ({
+                src: img.isUrl ? img.data : `data:image/png;base64,${img.data}`,
+                alt: prompt,
+              }))}
+              columns={imageGen.result.images.length > 1 ? 2 : 1}
+            />
+          </CardContent>
+        </Card>
+      )}
+
+      {gen.status === 'complete' && videoGen.result && mode === 'video' && (
+        <Card>
+          <CardHeader className="flex flex-row items-center justify-between">
+            <CardTitle>Results</CardTitle>
+            <div className="flex items-center gap-2">
+              {videoGen.result.provider && <ProviderBadge provider={videoGen.result.provider} />}
+              <Button variant="outline" size="sm" onClick={() => navigate('/media')}>
+                View in Library
+              </Button>
+            </div>
+          </CardHeader>
+          <CardContent>
+            <VideoGrid
+              videos={videoGen.result.videos.map((vid) => ({
+                src: vid.isUrl ? vid.data : `data:${vid.mimeType};base64,${vid.data}`,
+                mimeType: vid.mimeType,
+                alt: prompt,
+              }))}
+            />
+          </CardContent>
+        </Card>
+      )}
+    </div>
+  );
+}

internal/adapter/templates/templates/components/app-react/src/pages/LoginPage.tsx.tmpl

@@ -0,0 +1,109 @@
+import { useState } from 'react';
+import { useNavigate, useLocation } from 'react-router-dom';
+import { useAuth } from '@{{PROJECT_NAME}}/auth';
+import {
+  Button,
+  Card,
+  CardHeader,
+  CardTitle,
+  CardDescription,
+  CardContent,
+  CardFooter,
+  FormField,
+  useFormErrors,
+  Alert,
+  AlertDescription,
+  Loader2,
+} from '@{{PROJECT_NAME}}/ui';
+import { isApiClientError } from '@{{PROJECT_NAME}}/api-client';
+
+export function LoginPage() {
+  const navigate = useNavigate();
+  const location = useLocation();
+  const { login, isLoading } = useAuth();
+  const { setErrors, clearErrors, getError } = useFormErrors();
+  const [generalError, setGeneralError] = useState<string | null>(null);
+
+  // Get the redirect path from location state, default to dashboard
+  const from = (location.state as { from?: string })?.from || '/';
+
+  const handleSubmit = async (e: React.FormEvent<HTMLFormElement>) => {
+    e.preventDefault();
+    clearErrors();
+    setGeneralError(null);
+
+    const formData = new FormData(e.currentTarget);
+    const email = formData.get('email') as string;
+    const password = formData.get('password') as string;
+
+    try {
+      await login({ email, password });
+      navigate(from, { replace: true });
+    } catch (error) {
+      if (isApiClientError(error)) {
+        if (error.isValidationError()) {
+          setErrors(error.getFieldErrors());
+        } else {
+          setGeneralError(error.message);
+        }
+      } else {
+        setGeneralError('An unexpected error occurred. Please try again.');
+      }
+    }
+  };
+
+  return (
+    <div className="min-h-screen flex items-center justify-center bg-[var(--surface-100)] p-4">
+      <Card className="w-full max-w-md">
+        <CardHeader className="text-center">
+          <CardTitle className="text-2xl">Welcome back</CardTitle>
+          <CardDescription>Sign in to your {{PROJECT_NAME}} account</CardDescription>
+        </CardHeader>
+
+        <form onSubmit={handleSubmit}>
+          <CardContent className="space-y-4">
+            {generalError && (
+              <Alert variant="destructive">
+                <AlertDescription>{generalError}</AlertDescription>
+              </Alert>
+            )}
+
+            <FormField
+              label="Email"
+              name="email"
+              type="email"
+              placeholder="you@example.com"
+              error={getError('email')}
+              required
+              autoComplete="email"
+              autoFocus
+            />
+
+            <FormField
+              label="Password"
+              name="password"
+              type="password"
+              placeholder="Enter your password"
+              error={getError('password')}
+              required
+              autoComplete="current-password"
+            />
+          </CardContent>
+
+          <CardFooter className="flex flex-col gap-4">
+            <Button type="submit" className="w-full" disabled={isLoading}>
+              {isLoading && <Loader2 className="mr-2 h-4 w-4 animate-spin" />}
+              Sign in
+            </Button>
+
+            <p className="text-sm text-center text-[var(--text-muted)]">
+              Demo accounts: test@example.com / password123
+              <br />
+              or admin@example.com / admin123
+            </p>
+          </CardFooter>
+        </form>
+      </Card>
+    </div>
+  );
+}

internal/adapter/templates/templates/components/app-react/src/pages/MediaPage.tsx.tmpl

@@ -0,0 +1,129 @@
+import { useState, useEffect, useMemo, useCallback } from 'react';
+import { useAuth } from '@{{PROJECT_NAME}}/auth';
+import { useMediaUpload } from '@{{PROJECT_NAME}}/realtime';
+import {
+  Card,
+  CardHeader,
+  CardTitle,
+  CardDescription,
+  CardContent,
+  MediaUploader,
+  MediaLibrary,
+  type MediaItem,
+  Badge,
+} from '@{{PROJECT_NAME}}/ui';
+
+export function MediaPage() {
+  const { getToken } = useAuth();
+  const [items, setItems] = useState<MediaItem[]>([]);
+  const [isLoading, setIsLoading] = useState(true);
+  const [deleteError, setDeleteError] = useState<string | null>(null);
+
+  const apiPrefix = import.meta.env.VITE_API_URL || '';
+
+  const authHeaders = useMemo(() => {
+    const token = getToken();
+    return token ? { Authorization: `Bearer ${token}` } : undefined;
+  }, [getToken]);
+
+  const mediaUpload = useMediaUpload({
+    apiPrefix,
+    serviceName: '{{SERVICE_NAME}}',
+    headers: authHeaders,
+  });
+
+  const [fetchError, setFetchError] = useState<string | null>(null);
+
+  const fetchMedia = useCallback(async () => {
+    setFetchError(null);
+    try {
+      const res = await fetch(`${apiPrefix}/api/{{SERVICE_NAME}}/media`, {
+        headers: { ...authHeaders },
+      });
+      if (!res.ok) {
+        setFetchError(`Failed to load media (${res.status})`);
+        return;
+      }
+      const data = await res.json();
+      setItems(data.items || []);
+    } catch (err) {
+      setFetchError(err instanceof Error ? err.message : 'Failed to load media');
+    } finally {
+      setIsLoading(false);
+    }
+  }, [apiPrefix, authHeaders]);
+
+  useEffect(() => {
+    fetchMedia();
+  }, [fetchMedia]);
+
+  const handleUploadComplete = useCallback(() => {
+    fetchMedia();
+  }, [fetchMedia]);
+
+  const handleDelete = useCallback(async (path: string) => {
+    setDeleteError(null);
+    try {
+      const res = await fetch(`${apiPrefix}/api/{{SERVICE_NAME}}/media/${path}`, {
+        method: 'DELETE',
+        headers: { ...authHeaders },
+      });
+      if (!res.ok) throw new Error(`Delete failed: ${res.status}`);
+      setItems((prev) => prev.filter((item) => item.path !== path));
+    } catch (err) {
+      setDeleteError(err instanceof Error ? err.message : 'Delete failed');
+    }
+  }, [apiPrefix, authHeaders]);
+
+  return (
+    <div className="space-y-6">
+      <Card>
+        <CardHeader>
+          <CardTitle>Upload Media</CardTitle>
+          <CardDescription>
+            Upload images and videos to your media library
+          </CardDescription>
+        </CardHeader>
+        <CardContent>
+          <MediaUploader
+            upload={mediaUpload.upload}
+            isUploading={mediaUpload.isUploading}
+            progress={mediaUpload.progress}
+            onUploadComplete={handleUploadComplete}
+            onError={(err) => console.error('Upload error:', err)}
+          />
+          {mediaUpload.error && (
+            <p className="mt-2 text-sm text-[var(--error)]">{mediaUpload.error}</p>
+          )}
+        </CardContent>
+      </Card>
+
+      <Card>
+        <CardHeader className="flex flex-row items-center justify-between">
+          <div>
+            <CardTitle>Media Library</CardTitle>
+            <CardDescription>
+              Your uploaded and generated media files
+            </CardDescription>
+          </div>
+          {items.length > 0 && (
+            <Badge variant="outline">{items.length} files</Badge>
+          )}
+        </CardHeader>
+        <CardContent>
+          {(deleteError || fetchError) && (
+            <p className="mb-4 text-sm text-[var(--error)]">{deleteError || fetchError}</p>
+          )}
+          {isLoading ? (
+            <div className="text-center py-8 text-[var(--text-muted)]">Loading...</div>
+          ) : (
+            <MediaLibrary
+              items={items}
+              onDelete={handleDelete}
+            />
+          )}
+        </CardContent>
+      </Card>
+    </div>
+  );
+}

internal/adapter/templates/templates/components/app-react/tailwind.config.ts

@@ -1,5 +1,6 @@
-/** @type {import('tailwindcss').Config} */
+import type { Config } from 'tailwindcss';
-export default {
+
+const config: Config = {
   content: [
     './index.html',
     './src/**/*.{js,ts,jsx,tsx}',
@@ -12,3 +13,5 @@ export default {
   },
   plugins: [],
 };
+
+export default config;

internal/adapter/templates/templates/components/app-react/vite.config.ts.tmpl

@@ -6,6 +6,27 @@ export default defineConfig({
   plugins: [react()],
   server: {
     port: {{PORT}},
+    proxy: {
+      // SSE events endpoint — must disable buffering for streaming
+      '/api/{{SERVICE_NAME}}/events': {
+        target: 'http://localhost:{{SERVICE_PORT}}',
+        changeOrigin: true,
+        // Disable response buffering so SSE events stream immediately
+        configure: (proxy) => {
+          proxy.on('proxyRes', (proxyRes) => {
+            // Prevent Vite from buffering SSE responses
+            if (proxyRes.headers['content-type']?.includes('text/event-stream')) {
+              proxyRes.headers['cache-control'] = 'no-cache';
+              proxyRes.headers['x-accel-buffering'] = 'no';
+            }
+          });
+        },
+      },
+      '/api': {
+        target: 'http://localhost:{{SERVICE_PORT}}',
+        changeOrigin: true,
+      },
+    },
   },
   preview: {
     port: {{PORT}},

internal/adapter/templates/templates/components/service/Makefile.tmpl

@@ -1,4 +1,4 @@
-.PHONY: build run test lint fmt docker-build clean
+.PHONY: build run dev test lint fmt docker-build clean
 
 SERVICE := {{COMPONENT_NAME}}
 BINARY := bin/$(SERVICE)
@@ -12,6 +12,10 @@ build:
 run:
 	go run ./cmd/server
 
+# Run the service in development mode (alias for run)
+dev:
+	go run ./cmd/server
+
 # Run tests
 test:
 	go test -v ./...

internal/adapter/templates/templates/components/service/cmd/server/main.go.tmpl

@@ -2,14 +2,30 @@
 package main
 
 import (
+	"context"
 	"flag"
 	"fmt"
 	"os"
+	"time"
+
+	"github.com/redis/go-redis/v9"
 
 	"{{GO_MODULE}}/pkg/app"
+	"{{GO_MODULE}}/pkg/database"
+	"{{GO_MODULE}}/pkg/gemini"
+	"{{GO_MODULE}}/pkg/laozhang"
 	"{{GO_MODULE}}/pkg/logging"
+	"{{GO_MODULE}}/pkg/mediagen"
+	mediagenAdapters "{{GO_MODULE}}/pkg/mediagen/adapters"
+	"{{GO_MODULE}}/pkg/generation"
+	"{{GO_MODULE}}/pkg/queue"
+	"{{GO_MODULE}}/pkg/realtime"
+	"{{GO_MODULE}}/pkg/storage"
+	"{{GO_MODULE}}/pkg/textgen"
+	textgenAdapters "{{GO_MODULE}}/pkg/textgen/adapters"
 	"{{GO_MODULE}}/services/{{COMPONENT_NAME}}/internal/adapter/memory"
 	"{{GO_MODULE}}/services/{{COMPONENT_NAME}}/internal/api"
+	"{{GO_MODULE}}/services/{{COMPONENT_NAME}}/internal/config"
 	"{{GO_MODULE}}/services/{{COMPONENT_NAME}}/internal/service"
 )
 
@@ -30,21 +46,247 @@ func main() {
 		os.Exit(0)
 	}
 
+	// Load config
+	cfg := config.Load()
+
 	// Create logger
 	logger := logging.Default()
 
+	ctx := context.Background()
+
+	// Create SSE hub for async event delivery (generation progress, chat, etc.)
+	sseHub := realtime.NewSSEHub(logger.Logger)
+
+	// Initialize storage backend (before queue, since standalone queue handlers use it).
+	// GCS_BUCKET set = production (GCS). Otherwise = dev (in-memory).
+	port := fmt.Sprintf("%d", {{PORT}})
+	var mediaStore storage.Store
+	if bucket := os.Getenv("GCS_BUCKET"); bucket != "" {
+		gcsStore, err := storage.NewGCSStore(bucket, os.Getenv("GCS_SERVICE_ACCOUNT_JSON"), logger.Logger)
+		if err != nil {
+			logger.Error("failed to create GCS store", "error", err)
+			os.Exit(1)
+		}
+		defer func() { _ = gcsStore.Close() }()
+		mediaStore = gcsStore
+		logger.Info("storage initialized (GCS)", "bucket", bucket)
+	} else {
+		memStore := storage.NewMemoryStore("http://localhost:" + port + "/storage")
+		mediaStore = memStore
+		logger.Info("storage initialized (in-memory dev mode)")
+	}
+
+	// Select queue backend based on DATABASE_URL availability.
+	// With DATABASE_URL: DB queue + separate worker process (production)
+	// Without DATABASE_URL: in-memory queue + in-process handlers (development)
+	var jobQueue queue.Producer
+	if cfg.Database.URL != "" {
+		jobQueue = setupDBQueue(ctx, cfg, sseHub, logger)
+	} else {
+		logger.Info("DATABASE_URL not set — running in standalone mode (in-memory queue + in-process AI)")
+		jobQueue = setupStandaloneQueue(ctx, mediaStore, sseHub, logger)
+	}
+
 	// Create adapters (repositories)
 	exampleRepo := memory.NewExampleRepository()
+	userRepo := memory.NewUserRepository()
 
 	// Create services (business logic)
 	exampleService := service.NewExampleService(exampleRepo, logger)
+	authService := service.NewAuthService(userRepo, cfg.JWTSecret, logger)
 
 	// Create application
 	application := app.New("{{COMPONENT_NAME}}", app.WithDefaultPort({{PORT}}))
 
+	// Mount in-memory storage HTTP handler for dev mode
+	if memStore, ok := mediaStore.(*storage.MemoryStore); ok {
+		application.Router().Handle("/storage/*", memStore)
+	}
+
 	// Register routes with dependency injection
-	api.RegisterRoutes(application, exampleService)
+	api.RegisterRoutes(application, &api.Dependencies{
+		ExampleService: exampleService,
+		AuthService:    authService,
+		Queue:          jobQueue,
+		SSEHub:         sseHub,
+		Store:          mediaStore,
+	})
 
 	// Start server
 	application.Run()
 }
+
+// setupDBQueue initializes the production queue backend with database + optional Redis.
+func setupDBQueue(ctx context.Context, cfg *config.Config, sseHub *realtime.SSEHub, logger *logging.Logger) queue.Producer {
+	pool, err := database.Connect(ctx, cfg.Database.URL, database.Options{
+		MaxOpenConns:    cfg.Database.MaxOpenConns,
+		MaxIdleConns:    cfg.Database.MaxIdleConns,
+		ConnMaxLifetime: cfg.Database.ConnMaxLifetime,
+	})
+	if err != nil {
+		logger.Error("failed to connect to database", "error", err)
+		os.Exit(1)
+	}
+	// Note: pool is not deferred here since it's needed for the lifetime of the process.
+	// The OS reclaims resources on exit.
+	logger.Info("connected to database")
+
+	if err := queue.RunMigrations(ctx, pool); err != nil {
+		logger.Error("failed to run queue migrations", "error", err)
+		os.Exit(1)
+	}
+	logger.Info("queue migrations complete")
+
+	jobQueue := queue.NewQueue(pool.DB, logger)
+
+	// Start Redis SSE subscriber if configured.
+	if cfg.RedisURL != "" {
+		opts, err := redis.ParseURL(cfg.RedisURL)
+		if err != nil {
+			logger.Error("failed to parse REDIS_URL", "error", err)
+			os.Exit(1)
+		}
+		redisClient := redis.NewClient(opts)
+		if err := redisClient.Ping(ctx).Err(); err != nil {
+			logger.Error("failed to connect to Redis", "error", err)
+			os.Exit(1)
+		}
+		logger.Info("connected to Redis")
+
+		go func() {
+			if err := realtime.RunSSESubscriber(ctx, redisClient, sseHub, logger.Logger); err != nil {
+				logger.Error("SSE Redis subscriber stopped", "error", err)
+			}
+		}()
+	} else {
+		logger.Warn("REDIS_URL not set — SSE events from worker will not be delivered")
+	}
+
+	return jobQueue
+}
+
+// setupStandaloneQueue initializes an in-memory queue with in-process AI handlers.
+// This mode requires no database or Redis — everything runs in a single process.
+func setupStandaloneQueue(ctx context.Context, store storage.Store, sseHub *realtime.SSEHub, logger *logging.Logger) queue.Producer {
+	memQueue := queue.NewMemoryQueue(logger.Logger)
+
+	// LocalPublisher delivers events directly to the SSE hub (no Redis needed).
+	pub := realtime.NewLocalPublisher(sseHub)
+
+	// Initialize AI providers
+	mediagenManager := initMediagen(ctx, logger)
+	textgenManager := initTextgen(ctx, logger)
+
+	// Register job handlers (same handlers the worker uses).
+	if mediagenManager != nil {
+		memQueue.RegisterHandler("generate_image", generation.ImageHandler(mediagenManager, store, pub, logger))
+		memQueue.RegisterHandler("generate_video", generation.VideoHandler(mediagenManager, store, pub, logger))
+	}
+	if textgenManager != nil {
+		memQueue.RegisterHandler("generate_text", generation.TextHandler(textgenManager, pub, logger))
+		memQueue.RegisterHandler("ai_chat_response", generation.ChatResponseHandler(textgenManager, pub, logger))
+	}
+
+	return memQueue
+}
+
+// initMediagen creates a mediagen manager from available AI provider credentials.
+func initMediagen(ctx context.Context, logger *logging.Logger) *mediagen.Manager {
+	var laozhangMediaProvider *mediagenAdapters.LaoZhangProvider
+	var geminiMediaProvider *mediagenAdapters.GeminiProvider
+
+	if apiKey := os.Getenv("LAOZHANG_API_KEY"); apiKey != "" {
+		client, err := laozhang.NewClient(laozhang.Config{
+			APIKey:       apiKey,
+			VideoTimeout: 5 * time.Minute,
+			Logger:       logger.Logger,
+		})
+		if err != nil {
+			logger.Warn("failed to create LaoZhang client", "error", err)
+		} else {
+			laozhangMediaProvider = mediagenAdapters.NewLaoZhangProvider(client)
+			logger.Info("LaoZhang media provider initialized")
+		}
+	}
+
+	if apiKey := os.Getenv("GEMINI_API_KEY"); apiKey != "" {
+		client, err := gemini.NewClient(ctx, gemini.Config{
+			APIKey: apiKey,
+			Logger: logger.Logger,
+		})
+		if err != nil {
+			logger.Warn("failed to create Gemini client", "error", err)
+		} else {
+			geminiMediaProvider = mediagenAdapters.NewGeminiProvider(client)
+			logger.Info("Gemini media provider initialized")
+		}
+	}
+
+	if laozhangMediaProvider == nil && geminiMediaProvider == nil {
+		logger.Warn("no media generation providers available (set LAOZHANG_API_KEY or GEMINI_API_KEY)")
+		return nil
+	}
+
+	mgCfg := mediagen.ProductionConfig(mediagen.ProviderSet{
+		LaoZhang: laozhangMediaProvider,
+		Gemini:   geminiMediaProvider,
+	}, mediagen.WithLogger(logger.Logger))
+	if laozhangMediaProvider != nil {
+		mgCfg.VideoProviders = append(mgCfg.VideoProviders, laozhangMediaProvider)
+	}
+	if geminiMediaProvider != nil {
+		mgCfg.VideoProviders = append(mgCfg.VideoProviders, geminiMediaProvider)
+	}
+
+	mgr, err := mediagen.NewManager(mgCfg)
+	if err != nil {
+		logger.Warn("failed to create mediagen manager", "error", err)
+		return nil
+	}
+	logger.Info("mediagen manager initialized (image + video)")
+	return mgr
+}
+
+// initTextgen creates a textgen manager from available AI provider credentials.
+func initTextgen(ctx context.Context, logger *logging.Logger) *textgen.Manager {
+	var textProviders []textgen.TextGenerator
+
+	if apiKey := os.Getenv("LAOZHANG_API_KEY"); apiKey != "" {
+		client, err := laozhang.NewClient(laozhang.Config{
+			APIKey: apiKey,
+			Logger: logger.Logger,
+		})
+		if err != nil {
+			logger.Warn("failed to create LaoZhang text client", "error", err)
+		} else {
+			textProviders = append(textProviders, textgenAdapters.NewLaoZhangTextProvider(client, ""))
+		}
+	}
+
+	if apiKey := os.Getenv("GEMINI_API_KEY"); apiKey != "" {
+		provider, err := textgenAdapters.NewGeminiTextProvider(ctx, textgenAdapters.GeminiTextConfig{
+			APIKey: apiKey,
+		})
+		if err != nil {
+			logger.Warn("failed to create Gemini text provider", "error", err)
+		} else {
+			textProviders = append(textProviders, provider)
+		}
+	}
+
+	if len(textProviders) == 0 {
+		logger.Warn("no text generation providers available")
+		return nil
+	}
+
+	tgCfg := textgen.ProductionConfig(textgen.ProviderSet{}, textgen.WithLogger(logger.Logger))
+	tgCfg.Providers = textProviders
+
+	mgr, err := textgen.NewManager(tgCfg)
+	if err != nil {
+		logger.Warn("failed to create textgen manager", "error", err)
+		return nil
+	}
+	logger.Info("textgen manager initialized")
+	return mgr
+}

internal/adapter/templates/templates/components/service/internal/adapter/memory/user.go.tmpl

@@ -0,0 +1,92 @@
+package memory
+
+import (
+	"context"
+	"sync"
+
+	"{{GO_MODULE}}/pkg/auth"
+)
+
+// userEntry stores a user with their password for demo purposes.
+type userEntry struct {
+	user     *auth.User
+	password string
+}
+
+// UserRepository is an in-memory user store for demo/testing purposes.
+// Pre-populated with demo users.
+type UserRepository struct {
+	mu    sync.RWMutex
+	users map[string]*userEntry // keyed by email
+}
+
+// NewUserRepository creates a new in-memory user repository with demo users.
+func NewUserRepository() *UserRepository {
+	repo := &UserRepository{
+		users: make(map[string]*userEntry),
+	}
+
+	// Add demo users
+	repo.users["test@example.com"] = &userEntry{
+		user: &auth.User{
+			ID:    "usr_test_001",
+			Email: "test@example.com",
+			Roles: []string{"user"},
+			Metadata: map[string]any{
+				"name": "Test User",
+			},
+		},
+		password: "password123",
+	}
+
+	repo.users["admin@example.com"] = &userEntry{
+		user: &auth.User{
+			ID:    "usr_admin_001",
+			Email: "admin@example.com",
+			Roles: []string{"admin", "user"},
+			Metadata: map[string]any{
+				"name": "Admin User",
+			},
+		},
+		password: "admin123",
+	}
+
+	return repo
+}
+
+// FindByEmail returns a user by email address.
+func (r *UserRepository) FindByEmail(ctx context.Context, email string) (*auth.User, error) {
+	r.mu.RLock()
+	defer r.mu.RUnlock()
+
+	entry, ok := r.users[email]
+	if !ok {
+		return nil, nil
+	}
+	return entry.user, nil
+}
+
+// FindByID returns a user by ID.
+func (r *UserRepository) FindByID(ctx context.Context, id string) (*auth.User, error) {
+	r.mu.RLock()
+	defer r.mu.RUnlock()
+
+	for _, entry := range r.users {
+		if entry.user.ID == id {
+			return entry.user, nil
+		}
+	}
+	return nil, nil
+}
+
+// ValidatePassword checks if the password matches for a user.
+func (r *UserRepository) ValidatePassword(ctx context.Context, user *auth.User, password string) bool {
+	r.mu.RLock()
+	defer r.mu.RUnlock()
+
+	entry, ok := r.users[user.Email]
+	if !ok {
+		return false
+	}
+	return entry.password == password
+}

internal/adapter/templates/templates/components/service/internal/api/handlers/auth.go.tmpl

@@ -0,0 +1,127 @@
+package handlers
+
+import (
+	"errors"
+	"net/http"
+
+	"{{GO_MODULE}}/pkg/app"
+	"{{GO_MODULE}}/pkg/auth"
+	"{{GO_MODULE}}/pkg/httperror"
+	"{{GO_MODULE}}/pkg/httpresponse"
+	"{{GO_MODULE}}/pkg/logging"
+	"{{GO_MODULE}}/services/{{COMPONENT_NAME}}/internal/service"
+)
+
+// Auth handles authentication HTTP requests.
+type Auth struct {
+	svc    *service.AuthService
+	logger *logging.Logger
+}
+
+// NewAuth creates a new Auth handler with injected dependencies.
+func NewAuth(svc *service.AuthService, logger *logging.Logger) *Auth {
+	return &Auth{
+		svc:    svc,
+		logger: logger.WithComponent("AuthHandler"),
+	}
+}
+
+// LoginRequest is the request body for login.
+type LoginRequest struct {
+	Email    string `json:"email" validate:"required,email"`
+	Password string `json:"password" validate:"required,min=1"`
+}
+
+// LoginResponse is the response for successful login.
+type LoginResponse struct {
+	Token string       `json:"token"`
+	User  UserResponse `json:"user"`
+}
+
+// UserResponse is the user data returned in auth responses.
+type UserResponse struct {
+	ID    string   `json:"id"`
+	Email string   `json:"email"`
+	Name  string   `json:"name,omitempty"`
+	Roles []string `json:"roles,omitempty"`
+}
+
+// toUserResponse converts an auth.User to UserResponse.
+func toUserResponse(u *auth.User) UserResponse {
+	name := ""
+	if u.Metadata != nil {
+		if n, ok := u.Metadata["name"].(string); ok {
+			name = n
+		}
+	}
+	return UserResponse{
+		ID:    u.ID,
+		Email: u.Email,
+		Name:  name,
+		Roles: u.Roles,
+	}
+}
+
+// Login authenticates a user and returns a JWT token.
+//
+// POST /api/{service}/auth/login
+func (h *Auth) Login(w http.ResponseWriter, r *http.Request) error {
+	var req LoginRequest
+	if err := app.BindAndValidate(r, &req); err != nil {
+		return err
+	}
+
+	output, err := h.svc.Login(r.Context(), service.LoginInput{
+		Email:    req.Email,
+		Password: req.Password,
+	})
+	if err != nil {
+		if errors.Is(err, service.ErrInvalidCredentials) {
+			return httperror.Unauthorized("invalid email or password")
+		}
+		return err
+	}
+
+	httpresponse.OK(w, r, LoginResponse{
+		Token: output.Token,
+		User:  toUserResponse(output.User),
+	})
+	return nil
+}
+
+// Me returns the current authenticated user.
+//
+// GET /api/{service}/auth/me
+func (h *Auth) Me(w http.ResponseWriter, r *http.Request) error {
+	user, err := auth.GetUserOrError(r.Context())
+	if err != nil {
+		return httperror.Unauthorized("not authenticated")
+	}
+
+	// Optionally refresh user data from repository
+	freshUser, err := h.svc.GetCurrentUser(r.Context(), user.ID)
+	if err != nil {
+		if errors.Is(err, service.ErrUserNotFound) {
+			return httperror.Unauthorized("user not found")
+		}
+		return err
+	}
+
+	httpresponse.OK(w, r, toUserResponse(freshUser))
+	return nil
+}
+
+// Logout handles user logout.
+// This is a stateless operation since we use JWTs.
+//
+// POST /api/{service}/auth/logout
+func (h *Auth) Logout(w http.ResponseWriter, r *http.Request) error {
+	// With JWT-based auth, logout is handled client-side by discarding the token.
+	// This endpoint exists for API completeness and could be extended to:
+	// - Add the token to a blacklist
+	// - Clear server-side sessions if using hybrid auth
+	// - Log the logout event
+
+	httpresponse.NoContent(w)
+	return nil
+}

internal/adapter/templates/templates/components/service/internal/api/handlers/chat.go.tmpl

@@ -0,0 +1,94 @@
+package handlers
+
+import (
+	"net/http"
+	"time"
+
+	"github.com/google/uuid"
+
+	"{{GO_MODULE}}/pkg/app"
+	"{{GO_MODULE}}/pkg/auth"
+	"{{GO_MODULE}}/pkg/httpresponse"
+	"{{GO_MODULE}}/pkg/logging"
+	"{{GO_MODULE}}/pkg/queue"
+	"{{GO_MODULE}}/pkg/realtime"
+)
+
+// Chat handles HTTP requests for chat messaging with AI responses.
+// User messages are broadcast immediately via SSE.
+// AI responses are enqueued and processed by the worker with streaming chunks.
+type Chat struct {
+	queue  queue.Producer
+	sseHub *realtime.SSEHub
+	logger *logging.Logger
+}
+
+// NewChat creates a new Chat handler.
+func NewChat(q queue.Producer, hub *realtime.SSEHub, logger *logging.Logger) *Chat {
+	return &Chat{
+		queue:  q,
+		sseHub: hub,
+		logger: logger.WithComponent("ChatHandler"),
+	}
+}
+
+// SendMessageRequest is the request body for sending a chat message.
+type SendMessageRequest struct {
+	Content string `json:"content" validate:"required,min=1,max=5000"`
+}
+
+// SendMessage broadcasts a chat message to a channel via SSE
+// and enqueues an AI response job for the worker.
+func (h *Chat) SendMessage(w http.ResponseWriter, r *http.Request) error {
+	var req SendMessageRequest
+	if err := app.BindAndValidate(r, &req); err != nil {
+		return err
+	}
+
+	// Get user info
+	userID := "anonymous"
+	userName := "Anonymous"
+	if user := auth.GetUser(r.Context()); user != nil {
+		userID = user.ID
+		if name, ok := user.Metadata["name"].(string); ok && name != "" {
+			userName = name
+		} else if user.Email != "" {
+			userName = user.Email
+		}
+	}
+
+	msgID := uuid.New().String()
+	now := time.Now().UTC()
+
+	// Broadcast user message to channel:general immediately (synchronous — users
+	// see their own messages instantly without waiting for the queue)
+	h.sseHub.SendToChannel("channel:general", &realtime.SSEEvent{
+		Type:      "chat",
+		Timestamp: now,
+		JobID:     msgID,
+		Message:   req.Content,
+		Result: map[string]any{
+			"id":        msgID,
+			"content":   req.Content,
+			"userId":    userID,
+			"userName":  userName,
+			"timestamp": now.Format(time.RFC3339),
+		},
+	})
+
+	// Enqueue AI response job — worker streams chunks via Redis → SSE
+	if _, err := h.queue.Enqueue(r.Context(), "ai_chat_response", map[string]any{
+		"content": req.Content,
+		"userID":  userID,
+		"channel": "channel:general",
+	}); err != nil {
+		h.logger.Error("failed to enqueue AI chat response", "error", err)
+		// Don't fail the request — user message was already delivered
+	}
+
+	httpresponse.OK(w, r, map[string]string{
+		"id":     msgID,
+		"status": "sent",
+	})
+	return nil
+}

internal/adapter/templates/templates/components/service/internal/api/handlers/generate.go.tmpl

@@ -0,0 +1,188 @@
+package handlers
+
+import (
+	"net/http"
+
+	"{{GO_MODULE}}/pkg/app"
+	"{{GO_MODULE}}/pkg/auth"
+	"{{GO_MODULE}}/pkg/httperror"
+	"{{GO_MODULE}}/pkg/httpresponse"
+	"{{GO_MODULE}}/pkg/logging"
+	"{{GO_MODULE}}/pkg/queue"
+	"{{GO_MODULE}}/pkg/realtime"
+)
+
+// Generate handles HTTP requests for AI generation endpoints.
+// All generation is async: validate request, enqueue job, return 202 with job ID.
+// The worker processes jobs and sends results via Redis → SSE.
+type Generate struct {
+	queue  queue.Producer
+	sseHub *realtime.SSEHub
+	logger *logging.Logger
+}
+
+// NewGenerate creates a new Generate handler with injected dependencies.
+func NewGenerate(q queue.Producer, hub *realtime.SSEHub, logger *logging.Logger) *Generate {
+	return &Generate{
+		queue:  q,
+		sseHub: hub,
+		logger: logger.WithComponent("GenerateHandler"),
+	}
+}
+
+// ---------------------------------------------------------------------------
+// Image generation (async - returns job ID, results via SSE)
+// ---------------------------------------------------------------------------
+
+// GenerateImageRequest is the request body for image generation.
+type GenerateImageRequest struct {
+	Prompt      string `json:"prompt" validate:"required,min=1,max=2000"`
+	Count       int    `json:"count"`
+	AspectRatio string `json:"aspectRatio"`
+}
+
+// GenerateAccepted is the immediate HTTP response with the job ID.
+type GenerateAccepted struct {
+	JobID string `json:"jobId"`
+}
+
+// GenerateImage queues an image generation job.
+// Returns immediately with job ID. Results come via SSE events:
+//   - generation_started: Job accepted
+//   - generation_progress: Progress updates
+//   - generation_complete: Images available
+//   - generation_failed: Error occurred
+//
+// Client should subscribe to SSE channel `user:<userId>` before calling.
+func (h *Generate) GenerateImage(w http.ResponseWriter, r *http.Request) error {
+	var req GenerateImageRequest
+	if err := app.BindAndValidate(r, &req); err != nil {
+		return err
+	}
+
+	// Set defaults
+	if req.Count == 0 {
+		req.Count = 1
+	}
+	if req.Count > 4 {
+		req.Count = 4
+	}
+
+	user := auth.GetUser(r.Context())
+	if user == nil {
+		return httperror.Unauthorized("authentication required")
+	}
+
+	jobID, err := h.queue.Enqueue(r.Context(), "generate_image", map[string]any{
+		"prompt":      req.Prompt,
+		"count":       req.Count,
+		"aspectRatio": req.AspectRatio,
+		"userID":      user.ID,
+	})
+	if err != nil {
+		h.logger.Error("failed to enqueue image job", "error", err)
+		return httperror.Internal("failed to queue image generation")
+	}
+
+	h.logger.Info("image generation queued", "jobId", jobID, "userID", user.ID)
+
+	httpresponse.Accepted(w, r, GenerateAccepted{JobID: jobID})
+	return nil
+}
+
+// ---------------------------------------------------------------------------
+// Video generation (async - takes 2-5 minutes)
+// ---------------------------------------------------------------------------
+
+// GenerateVideoRequest is the request body for video generation.
+type GenerateVideoRequest struct {
+	Prompt      string `json:"prompt" validate:"required,min=1,max=2000"`
+	AspectRatio string `json:"aspectRatio"`
+	Duration    string `json:"duration"`
+}
+
+// GenerateVideo queues a video generation job.
+// Returns immediately with job ID. Results come via SSE events.
+func (h *Generate) GenerateVideo(w http.ResponseWriter, r *http.Request) error {
+	var req GenerateVideoRequest
+	if err := app.BindAndValidate(r, &req); err != nil {
+		return err
+	}
+
+	// Validate video aspect ratio (Veo only supports 16:9 and 9:16)
+	if req.AspectRatio != "" && req.AspectRatio != "16:9" && req.AspectRatio != "9:16" {
+		return httperror.BadRequest("video only supports 16:9 and 9:16 aspect ratios")
+	}
+
+	user := auth.GetUser(r.Context())
+	if user == nil {
+		return httperror.Unauthorized("authentication required")
+	}
+
+	jobID, err := h.queue.Enqueue(r.Context(), "generate_video", map[string]any{
+		"prompt":      req.Prompt,
+		"aspectRatio": req.AspectRatio,
+		"duration":    req.Duration,
+		"userID":      user.ID,
+	})
+	if err != nil {
+		h.logger.Error("failed to enqueue video job", "error", err)
+		return httperror.Internal("failed to queue video generation")
+	}
+
+	h.logger.Info("video generation queued", "jobId", jobID, "userID", user.ID)
+
+	httpresponse.Accepted(w, r, GenerateAccepted{JobID: jobID})
+	return nil
+}
+
+// ---------------------------------------------------------------------------
+// Text generation (async - returns job ID, results via SSE with streaming chunks)
+// ---------------------------------------------------------------------------
+
+// GenerateTextRequest is the request body for text generation.
+type GenerateTextRequest struct {
+	Prompt       string `json:"prompt" validate:"required,min=1,max=5000"`
+	SystemPrompt string `json:"systemPrompt"`
+	MaxTokens    int    `json:"maxTokens"`
+}
+
+// GenerateText queues a text generation job.
+// Returns immediately with job ID. Chunks come via SSE as ai_chat_chunk events.
+func (h *Generate) GenerateText(w http.ResponseWriter, r *http.Request) error {
+	var req GenerateTextRequest
+	if err := app.BindAndValidate(r, &req); err != nil {
+		return err
+	}
+
+	user := auth.GetUser(r.Context())
+	if user == nil {
+		return httperror.Unauthorized("authentication required")
+	}
+
+	jobID, err := h.queue.Enqueue(r.Context(), "generate_text", map[string]any{
+		"prompt":       req.Prompt,
+		"systemPrompt": req.SystemPrompt,
+		"maxTokens":    req.MaxTokens,
+		"userID":       user.ID,
+	})
+	if err != nil {
+		h.logger.Error("failed to enqueue text job", "error", err)
+		return httperror.Internal("failed to queue text generation")
+	}
+
+	h.logger.Info("text generation queued", "jobId", jobID, "userID", user.ID)
+
+	httpresponse.Accepted(w, r, GenerateAccepted{JobID: jobID})
+	return nil
+}
+
+// ---------------------------------------------------------------------------
+// SSE Events endpoint
+// ---------------------------------------------------------------------------
+
+// Events returns the SSE handler for event subscriptions.
+// Mount at /api/events.
+func (h *Generate) Events() http.Handler {
+	return realtime.NewSSEHandler(h.sseHub, h.logger.Logger)
+}

internal/adapter/templates/templates/components/service/internal/api/handlers/media.go.tmpl

@@ -0,0 +1,161 @@
+package handlers
+
+import (
+	"fmt"
+	"net/http"
+	"strings"
+
+	"{{GO_MODULE}}/pkg/app"
+	"{{GO_MODULE}}/pkg/auth"
+	"{{GO_MODULE}}/pkg/httperror"
+	"{{GO_MODULE}}/pkg/httpresponse"
+	"{{GO_MODULE}}/pkg/logging"
+	"{{GO_MODULE}}/pkg/storage"
+
+	"github.com/go-chi/chi/v5"
+	"github.com/google/uuid"
+)
+
+// Media handles media upload and library operations.
+type Media struct {
+	store  storage.Store
+	logger *logging.Logger
+}
+
+// NewMedia creates a new media handler.
+func NewMedia(store storage.Store, logger *logging.Logger) *Media {
+	return &Media{store: store, logger: logger.WithComponent("MediaHandler")}
+}
+
+// Routes returns the media subrouter.
+func (h *Media) Routes() http.Handler {
+	r := chi.NewRouter()
+	r.Post("/upload/init", app.Wrap(h.InitUpload))
+	r.Post("/upload/complete", app.Wrap(h.CompleteUpload))
+	r.Get("/", app.Wrap(h.List))
+	r.Delete("/*", app.Wrap(h.Delete))
+	return r
+}
+
+// initUploadRequest is the request body for POST /media/upload/init.
+type initUploadRequest struct {
+	Filename    string `json:"filename" validate:"required"`
+	ContentType string `json:"contentType" validate:"required"`
+}
+
+// InitUpload returns a presigned URL for direct client-to-storage upload.
+func (h *Media) InitUpload(w http.ResponseWriter, r *http.Request) error {
+	var req initUploadRequest
+	if err := app.BindAndValidate(r, &req); err != nil {
+		return err
+	}
+
+	user := auth.GetUser(r.Context())
+	userID := "anonymous"
+	if user != nil {
+		userID = user.ID
+	}
+
+	// Build object path: media/{userID}/{uuid}/{filename}
+	objectPath := fmt.Sprintf("media/%s/%s/%s", userID, uuid.New().String(), req.Filename)
+
+	presigned, err := h.store.UploadPresigned(r.Context(), objectPath, req.ContentType)
+	if err != nil {
+		h.logger.Error("failed to create presigned upload", "error", err)
+		return httperror.Internal("failed to create upload URL")
+	}
+
+	httpresponse.OK(w, r, map[string]any{
+		"uploadURL":  presigned.URL,
+		"objectPath": objectPath,
+		"headers":    presigned.Headers,
+		"method":     presigned.Method,
+		"expires":    presigned.Expires,
+	})
+	return nil
+}
+
+// completeUploadRequest is the request body for POST /media/upload/complete.
+type completeUploadRequest struct {
+	ObjectPath string `json:"objectPath" validate:"required"`
+}
+
+// CompleteUpload confirms an upload is done and returns the final URL.
+func (h *Media) CompleteUpload(w http.ResponseWriter, r *http.Request) error {
+	var req completeUploadRequest
+	if err := app.BindAndValidate(r, &req); err != nil {
+		return err
+	}
+
+	url, err := h.store.GetURL(r.Context(), req.ObjectPath)
+	if err != nil {
+		h.logger.Error("failed to get object URL", "error", err, "path", req.ObjectPath)
+		return httperror.Internal("failed to confirm upload")
+	}
+
+	httpresponse.OK(w, r, map[string]any{
+		"url":  url,
+		"path": req.ObjectPath,
+	})
+	return nil
+}
+
+// List returns the user's media objects.
+func (h *Media) List(w http.ResponseWriter, r *http.Request) error {
+	user := auth.GetUser(r.Context())
+	userID := "anonymous"
+	if user != nil {
+		userID = user.ID
+	}
+
+	prefix := fmt.Sprintf("media/%s/", userID)
+
+	// Allow filtering by sub-prefix (e.g., ?prefix=images)
+	if subPrefix := r.URL.Query().Get("prefix"); subPrefix != "" {
+		prefix = fmt.Sprintf("media/%s/%s", userID, subPrefix)
+	}
+
+	objects, err := h.store.List(r.Context(), prefix)
+	if err != nil {
+		h.logger.Error("failed to list media", "error", err)
+		return httperror.Internal("failed to list media")
+	}
+
+	if objects == nil {
+		objects = []storage.MediaObject{}
+	}
+
+	httpresponse.OK(w, r, map[string]any{
+		"items": objects,
+		"count": len(objects),
+	})
+	return nil
+}
+
+// Delete removes a media object.
+// Users can only delete objects under their own media/{userID}/ prefix.
+func (h *Media) Delete(w http.ResponseWriter, r *http.Request) error {
+	// Extract path from URL (everything after /media/)
+	path := strings.TrimPrefix(r.URL.Path, "/")
+	if path == "" {
+		return httperror.BadRequest("path is required")
+	}
+
+	// Verify the path belongs to the authenticated user
+	user := auth.GetUser(r.Context())
+	if user == nil {
+		return httperror.Unauthorized("authentication required")
+	}
+	expectedPrefix := fmt.Sprintf("media/%s/", user.ID)
+	if !strings.HasPrefix(path, expectedPrefix) {
+		return httperror.Forbidden("cannot delete another user's media")
+	}
+
+	if err := h.store.Delete(r.Context(), path); err != nil {
+		h.logger.Error("failed to delete media", "error", err, "path", path)
+		return httperror.Internal("failed to delete media")
+	}
+
+	httpresponse.OK(w, r, map[string]any{"deleted": path})
+	return nil
+}

internal/adapter/templates/templates/components/service/internal/api/routes.go.tmpl

@@ -4,6 +4,9 @@ package api
 import (
 	"{{GO_MODULE}}/pkg/app"
 	"{{GO_MODULE}}/pkg/auth"
+	"{{GO_MODULE}}/pkg/queue"
+	"{{GO_MODULE}}/pkg/realtime"
+	"{{GO_MODULE}}/pkg/storage"
 	"{{GO_MODULE}}/services/{{COMPONENT_NAME}}/internal/api/handlers"
 	"{{GO_MODULE}}/services/{{COMPONENT_NAME}}/internal/config"
 	"{{GO_MODULE}}/services/{{COMPONENT_NAME}}/internal/service"
@@ -14,23 +17,52 @@ import (
 // This allows the monorepo to expose multiple services under a single domain:
 //   - https://domain/api/{{COMPONENT_NAME}}/health
 //   - https://domain/api/{{COMPONENT_NAME}}/examples
-func RegisterRoutes(application *app.App, exampleService *service.ExampleService) {
+//   - https://domain/api/{{COMPONENT_NAME}}/events?channel=user:123 (SSE)
+func RegisterRoutes(application *app.App, deps *Dependencies) {
 	logger := application.Logger()
 	cfg := config.Load()
 
 	// Initialize handlers with injected services
 	healthHandler := handlers.NewHealth(logger)
-	exampleHandler := handlers.NewExample(exampleService, logger)
+	exampleHandler := handlers.NewExample(deps.ExampleService, logger)
+	authHandler := handlers.NewAuth(deps.AuthService, logger)
+	generateHandler := handlers.NewGenerate(deps.Queue, deps.SSEHub, logger)
+	chatHandler := handlers.NewChat(deps.Queue, deps.SSEHub, logger)
+	mediaHandler := handlers.NewMedia(deps.Store, logger)
 
 	// Build and mount OpenAPI spec
 	spec := NewServiceSpec()
 	application.EnableDocs(spec)
 
+	// JWT validator for protected routes
+	jwtValidator := auth.NewJWTValidator(auth.JWTConfig{
+		Secret: []byte(cfg.JWTSecret),
+		Issuer: "{{PROJECT_NAME}}",
+	})
+
 	// Register API routes under /api/{service-name} to match ingress path routing.
 	// The ingress routes /api/{{COMPONENT_NAME}}/* to this service.
 	application.Route("/api/{{COMPONENT_NAME}}", func(r app.Router) {
 		r.Get("/health", healthHandler.Check)
 
+		// ----- Auth routes -----
+		// Public auth routes
+		r.Post("/auth/login", app.Wrap(authHandler.Login))
+		r.Post("/auth/logout", app.Wrap(authHandler.Logout))
+
+		// Protected auth routes
+		r.Group(func(r app.Router) {
+			r.Use(auth.Middleware(auth.MiddlewareConfig{
+				Validator: jwtValidator,
+			}))
+			r.Get("/auth/me", app.Wrap(authHandler.Me))
+		})
+
+		// ----- SSE Events -----
+		// Server-Sent Events for async job updates (generation progress, etc.)
+		r.Mount("/events", generateHandler.Events())
+
+		// ----- Example routes -----
 		// Public routes (no auth required)
 		r.Get("/examples", app.Wrap(exampleHandler.List))
 		r.Get("/examples/{id}", app.Wrap(exampleHandler.Get))
@@ -39,10 +71,7 @@ func RegisterRoutes(application *app.App, exampleService *service.ExampleService
 		r.Group(func(r app.Router) {
 			if cfg.AuthEnabled {
 				r.Use(auth.Middleware(auth.MiddlewareConfig{
-					Validator: auth.NewJWTValidator(auth.JWTConfig{
+					Validator: jwtValidator,
-						Secret: []byte(cfg.JWTSecret),
-						Issuer: "{{PROJECT_NAME}}",
-					}),
 				}))
 			}
 
@@ -50,5 +79,34 @@ func RegisterRoutes(application *app.App, exampleService *service.ExampleService
 			r.Put("/examples/{id}", app.Wrap(exampleHandler.Update))
 			r.Delete("/examples/{id}", app.Wrap(exampleHandler.Delete))
 		})
+
+		// ----- Chat + Generate + Media routes (auth required) -----
+		// Auth is required because SSE events are delivered to user:<userId> channels.
+		// Without a real user identity, events go to user:anonymous and never reach the client.
+		r.Group(func(r app.Router) {
+			r.Use(auth.Middleware(auth.MiddlewareConfig{
+				Validator: jwtValidator,
+			}))
+
+			// Chat messaging
+			r.Post("/chat/messages", app.Wrap(chatHandler.SendMessage))
+
+			// Media generation (all queue-based, returns 202)
+			r.Post("/generate/image", app.Wrap(generateHandler.GenerateImage))
+			r.Post("/generate/video", app.Wrap(generateHandler.GenerateVideo))
+			r.Post("/generate/text", app.Wrap(generateHandler.GenerateText))
+
+			// Media library (upload, list, delete)
+			r.Mount("/media", mediaHandler.Routes())
+		})
 	})
 }
+
+// Dependencies holds all service dependencies for route registration.
+type Dependencies struct {
+	ExampleService *service.ExampleService
+	AuthService    *service.AuthService
+	Queue          queue.Producer
+	SSEHub         *realtime.SSEHub
+	Store          storage.Store
+}

internal/adapter/templates/templates/components/service/internal/config/config.go.tmpl

@@ -18,6 +18,9 @@ type Config struct {
 	// Auth
 	AuthEnabled bool
 	JWTSecret   string
+
+	// Redis for cross-process SSE event delivery
+	RedisURL string
 }
 
 // Load reads configuration from environment variables.
@@ -30,5 +33,6 @@ func Load() *Config {
 
 		AuthEnabled: strings.EqualFold(os.Getenv("AUTH_ENABLED"), "true"),
 		JWTSecret:   os.Getenv("JWT_SECRET"),
+		RedisURL:    os.Getenv("REDIS_URL"),
 	}
 }

internal/adapter/templates/templates/components/service/internal/port/user.go.tmpl

@@ -0,0 +1,23 @@
+package port
+
+import (
+	"context"
+
+	"{{GO_MODULE}}/pkg/auth"
+)
+
+// UserRepository defines the interface for user lookup operations.
+// Used by AuthService for authentication.
+type UserRepository interface {
+	// FindByEmail returns a user by email address.
+	// Returns nil if not found (no error).
+	FindByEmail(ctx context.Context, email string) (*auth.User, error)
+
+	// FindByID returns a user by ID.
+	// Returns nil if not found (no error).
+	FindByID(ctx context.Context, id string) (*auth.User, error)
+
+	// ValidatePassword checks if the password matches for a user.
+	// Returns true if valid, false otherwise.
+	ValidatePassword(ctx context.Context, user *auth.User, password string) bool
+}

internal/adapter/templates/templates/components/service/internal/service/auth.go.tmpl

@@ -0,0 +1,97 @@
+package service
+
+import (
+	"context"
+	"errors"
+	"time"
+
+	"{{GO_MODULE}}/pkg/auth"
+	"{{GO_MODULE}}/pkg/logging"
+	"{{GO_MODULE}}/services/{{COMPONENT_NAME}}/internal/port"
+)
+
+// Auth errors.
+var (
+	ErrInvalidCredentials = errors.New("invalid email or password")
+	ErrUserNotFound       = errors.New("user not found")
+)
+
+// AuthService handles authentication logic.
+type AuthService struct {
+	userRepo  port.UserRepository
+	jwtSecret []byte
+	issuer    string
+	logger    *logging.Logger
+}
+
+// NewAuthService creates a new auth service.
+func NewAuthService(userRepo port.UserRepository, jwtSecret string, logger *logging.Logger) *AuthService {
+	return &AuthService{
+		userRepo:  userRepo,
+		jwtSecret: []byte(jwtSecret),
+		issuer:    "{{PROJECT_NAME}}",
+		logger:    logger.WithService("AuthService"),
+	}
+}
+
+// LoginInput contains the data needed to log in.
+type LoginInput struct {
+	Email    string
+	Password string
+}
+
+// LoginOutput contains the login result.
+type LoginOutput struct {
+	Token string
+	User  *auth.User
+}
+
+// Login authenticates a user and returns a JWT token.
+func (s *AuthService) Login(ctx context.Context, input LoginInput) (*LoginOutput, error) {
+	// Find user by email
+	user, err := s.userRepo.FindByEmail(ctx, input.Email)
+	if err != nil {
+		return nil, err
+	}
+	if user == nil {
+		s.logger.Warn("login attempt for unknown email", "email", input.Email)
+		return nil, ErrInvalidCredentials
+	}
+
+	// Validate password
+	if !s.userRepo.ValidatePassword(ctx, user, input.Password) {
+		s.logger.Warn("invalid password attempt", "email", input.Email)
+		return nil, ErrInvalidCredentials
+	}
+
+	// Generate JWT token
+	token, err := auth.GenerateTokenWithIssuer(
+		s.jwtSecret,
+		user,
+		24*time.Hour, // 24 hour expiration
+		s.issuer,
+		s.issuer, // audience = issuer for simplicity
+	)
+	if err != nil {
+		return nil, err
+	}
+
+	s.logger.Info("user logged in", "user_id", user.ID, "email", user.Email)
+
+	return &LoginOutput{
+		Token: token,
+		User:  user,
+	}, nil
+}
+
+// GetCurrentUser returns the user for the given ID.
+func (s *AuthService) GetCurrentUser(ctx context.Context, userID string) (*auth.User, error) {
+	user, err := s.userRepo.FindByID(ctx, userID)
+	if err != nil {
+		return nil, err
+	}
+	if user == nil {
+		return nil, ErrUserNotFound
+	}
+	return user, nil
+}

internal/adapter/templates/templates/components/worker/Makefile.tmpl

@@ -1,4 +1,4 @@
-.PHONY: build run test lint fmt docker-build clean
+.PHONY: build run dev test lint fmt docker-build clean
 
 WORKER := {{COMPONENT_NAME}}
 BINARY := bin/$(WORKER)
@@ -12,6 +12,10 @@ build:
 run:
 	go run ./cmd/worker
 
+# Run the worker in development mode (alias for run)
+dev:
+	go run ./cmd/worker
+
 # Run tests
 test:
 	go test -v ./...

internal/adapter/templates/templates/components/worker/cmd/worker/main.go.tmpl

@@ -3,22 +3,28 @@ package main
 
 import (
 	"context"
-	"embed"
 	"os"
 	"os/signal"
 	"syscall"
 	"time"
 
+	"github.com/redis/go-redis/v9"
+
 	"{{GO_MODULE}}/pkg/database"
+	"{{GO_MODULE}}/pkg/gemini"
+	"{{GO_MODULE}}/pkg/laozhang"
 	"{{GO_MODULE}}/pkg/logging"
+	"{{GO_MODULE}}/pkg/mediagen"
+	mediagenAdapters "{{GO_MODULE}}/pkg/mediagen/adapters"
 	"{{GO_MODULE}}/pkg/queue"
+	"{{GO_MODULE}}/pkg/realtime"
+	"{{GO_MODULE}}/pkg/storage"
+	"{{GO_MODULE}}/pkg/textgen"
+	textgenAdapters "{{GO_MODULE}}/pkg/textgen/adapters"
 	"{{GO_MODULE}}/workers/{{COMPONENT_NAME}}/internal/config"
 	"{{GO_MODULE}}/workers/{{COMPONENT_NAME}}/internal/handlers"
 )
 
-//go:embed migrations/*.sql
-var migrationsFS embed.FS
-
 func main() {
 	// Initialize logger first (with defaults) so we can log config errors
 	logger := logging.New(logging.Config{
@@ -60,12 +66,125 @@ func main() {
 	defer pool.Close()
 	logger.Info("connected to database", "url", pool.URL)
 
-	// Run migrations
+	// Run queue migrations (idempotent — safe for both service and worker)
-	database.MustRunMigrations(ctx, pool, migrationsFS, "migrations")
+	if err := queue.RunMigrations(ctx, pool); err != nil {
-	logger.Info("migrations complete")
+		logger.Error("failed to run queue migrations", "error", err)
+		os.Exit(1)
+	}
+	logger.Info("queue migrations complete")
 
 	// Initialize queue
-	jobQueue := queue.NewPostgresQueue(pool.DB, logger)
+	jobQueue := queue.NewQueue(pool.DB, logger)
+
+	// Initialize Redis for SSE event publishing
+	if cfg.RedisURL == "" {
+		logger.Error("REDIS_URL is required for worker to publish SSE events")
+		os.Exit(1)
+	}
+	redisOpts, err := redis.ParseURL(cfg.RedisURL)
+	if err != nil {
+		logger.Error("failed to parse REDIS_URL", "error", err)
+		os.Exit(1)
+	}
+	redisClient := redis.NewClient(redisOpts)
+	if err := redisClient.Ping(ctx).Err(); err != nil {
+		logger.Error("failed to connect to Redis", "error", err)
+		os.Exit(1)
+	}
+	logger.Info("connected to Redis")
+
+	ssePub := realtime.NewSSEPublisher(redisClient, logger.Logger)
+
+	// Initialize AI providers
+	// LaoZhang client (primary provider — pay-per-use, OpenAI-compatible)
+	var laozhangClient *laozhang.Client
+	if apiKey := os.Getenv("LAOZHANG_API_KEY"); apiKey != "" {
+		laozhangClient, err = laozhang.NewClient(laozhang.Config{
+			APIKey:       apiKey,
+			VideoTimeout: 5 * time.Minute,
+			Logger:       logger.Logger,
+		})
+		if err != nil {
+			logger.Warn("failed to create LaoZhang client", "error", err)
+		} else {
+			logger.Info("LaoZhang client initialized")
+		}
+	}
+
+	// Gemini client for media generation
+	var geminiClient *gemini.Client
+	if apiKey := os.Getenv("GEMINI_API_KEY"); apiKey != "" {
+		geminiClient, err = gemini.NewClient(ctx, gemini.Config{
+			APIKey: apiKey,
+			Logger: logger.Logger,
+		})
+		if err != nil {
+			logger.Warn("failed to create Gemini client", "error", err)
+		} else {
+			logger.Info("Gemini client initialized")
+		}
+	}
+
+	// Create mediagen manager (image + video)
+	var mediagenManager *mediagen.Manager
+	{
+		var laozhangMediaProvider *mediagenAdapters.LaoZhangProvider
+		var geminiMediaProvider *mediagenAdapters.GeminiProvider
+		if laozhangClient != nil {
+			laozhangMediaProvider = mediagenAdapters.NewLaoZhangProvider(laozhangClient)
+		}
+		if geminiClient != nil {
+			geminiMediaProvider = mediagenAdapters.NewGeminiProvider(geminiClient)
+		}
+
+		if geminiMediaProvider != nil || laozhangMediaProvider != nil {
+			mgCfg := mediagen.ProductionConfig(mediagen.ProviderSet{
+				LaoZhang: laozhangMediaProvider,
+				Gemini:   geminiMediaProvider,
+			}, mediagen.WithLogger(logger.Logger))
+			if laozhangMediaProvider != nil {
+				mgCfg.VideoProviders = append(mgCfg.VideoProviders, laozhangMediaProvider)
+			}
+			if geminiMediaProvider != nil {
+				mgCfg.VideoProviders = append(mgCfg.VideoProviders, geminiMediaProvider)
+			}
+			mediagenManager, err = mediagen.NewManager(mgCfg)
+			if err != nil {
+				logger.Warn("failed to create mediagen manager", "error", err)
+			} else {
+				logger.Info("mediagen manager initialized (image + video)")
+			}
+		}
+	}
+
+	// Create textgen manager (text + streaming)
+	var textgenManager *textgen.Manager
+	{
+		var textProviders []textgen.TextGenerator
+		if laozhangClient != nil {
+			textProviders = append(textProviders, textgenAdapters.NewLaoZhangTextProvider(laozhangClient, ""))
+		}
+		if apiKey := os.Getenv("GEMINI_API_KEY"); apiKey != "" {
+			geminiTextProvider, err := textgenAdapters.NewGeminiTextProvider(ctx, textgenAdapters.GeminiTextConfig{
+				APIKey: apiKey,
+			})
+			if err != nil {
+				logger.Warn("failed to create gemini text provider", "error", err)
+			} else {
+				textProviders = append(textProviders, geminiTextProvider)
+			}
+		}
+		if len(textProviders) > 0 {
+			textgenCfg := textgen.ProductionConfig(textgen.ProviderSet{}, textgen.WithLogger(logger.Logger))
+			textgenCfg.Providers = textProviders
+			textgenManager, err = textgen.NewManager(textgenCfg)
+			if err != nil {
+				logger.Warn("failed to create textgen manager", "error", err)
+			} else {
+				logger.Info("textgen manager initialized")
+			}
+		}
+	}
 
 	// Initialize and start handler
 	handler := handlers.New(logger, jobQueue, handlers.Config{
@@ -74,10 +193,29 @@ func main() {
 		JobTimeout:      cfg.Worker.JobTimeout,
 	})
 
+	// Initialize storage backend for persisting generated media.
+	// GCS_BUCKET is injected by the platform; if absent, store is nil (media not persisted).
+	var mediaStore storage.Store
+	if bucket := os.Getenv("GCS_BUCKET"); bucket != "" {
+		gcsStore, err := storage.NewGCSStore(bucket, os.Getenv("GCS_SERVICE_ACCOUNT_JSON"), logger.Logger)
+		if err != nil {
+			logger.Warn("failed to create GCS store, generated media will not be persisted", "error", err)
+		} else {
+			defer func() { _ = gcsStore.Close() }()
+			mediaStore = gcsStore
+			logger.Info("storage initialized (GCS)", "bucket", bucket)
+		}
+	}
+
 	// Register job handlers
-	// TODO: Register your job handlers here
+	if mediagenManager != nil {
-	// handler.RegisterHandler("send_email", emailHandler)
+		handler.RegisterHandler("generate_image", handlers.ImageHandler(mediagenManager, mediaStore, ssePub, logger))
-	// handler.RegisterHandler("process_image", imageHandler)
+		handler.RegisterHandler("generate_video", handlers.VideoHandler(mediagenManager, mediaStore, ssePub, logger))
+	}
+	if textgenManager != nil {
+		handler.RegisterHandler("generate_text", handlers.TextHandler(textgenManager, ssePub, logger))
+		handler.RegisterHandler("ai_chat_response", handlers.ChatResponseHandler(textgenManager, ssePub, logger))
+	}
 
 	// Setup signal handling
 	sigCh := make(chan os.Signal, 1)
@@ -98,7 +236,6 @@ func main() {
 	cancel()
 
 	// Give in-flight jobs time to complete (grace period)
-	// This allows handlers to notice context cancellation and finish cleanly.
 	const shutdownGracePeriod = 5 * time.Second
 	time.Sleep(shutdownGracePeriod)
 
@@ -106,7 +243,7 @@ func main() {
 }
 
 // runStaleJobRecovery periodically requeues jobs that have been running too long.
-func runStaleJobRecovery(ctx context.Context, q *queue.PostgresQueue, timeout time.Duration, logger *logging.Logger) {
+func runStaleJobRecovery(ctx context.Context, q *queue.DBQueue, timeout time.Duration, logger *logging.Logger) {
 	const staleCheckInterval = time.Minute
 	ticker := time.NewTicker(staleCheckInterval)
 	defer ticker.Stop()

internal/adapter/templates/templates/components/worker/internal/config/config.go.tmpl

@@ -2,6 +2,7 @@
 package config
 
 import (
+	"os"
 	"time"
 
 	"github.com/spf13/viper"
@@ -15,6 +16,9 @@ type Config struct {
 	Database config.DatabaseConfig
 	Logging  config.LoggingConfig
 	Worker   WorkerConfig
+
+	// Redis for publishing SSE events to the service
+	RedisURL string
 }
 
 // WorkerConfig holds worker-specific settings.
@@ -62,5 +66,6 @@ func Load() (*Config, error) {
 			StaleJobTimeout: viper.GetDuration("WORKER_STALE_JOB_TIMEOUT"),
 			JobTimeout:      viper.GetDuration("WORKER_JOB_TIMEOUT"),
 		},
+		RedisURL: os.Getenv("REDIS_URL"),
 	}, nil
 }

internal/adapter/templates/templates/components/worker/internal/handlers/generate.go.tmpl

@@ -0,0 +1,33 @@
+// Package handlers re-exports generation job handlers from the shared package.
+// The worker registers these handlers to process queue jobs.
+package handlers
+
+import (
+	"{{GO_MODULE}}/pkg/generation"
+	"{{GO_MODULE}}/pkg/logging"
+	"{{GO_MODULE}}/pkg/mediagen"
+	"{{GO_MODULE}}/pkg/queue"
+	"{{GO_MODULE}}/pkg/realtime"
+	"{{GO_MODULE}}/pkg/storage"
+	"{{GO_MODULE}}/pkg/textgen"
+)
+
+// ImageHandler returns a queue.Handler that processes image generation jobs.
+func ImageHandler(mg *mediagen.Manager, store storage.Store, pub realtime.EventPublisher, logger *logging.Logger) queue.Handler {
+	return generation.ImageHandler(mg, store, pub, logger)
+}
+
+// VideoHandler returns a queue.Handler that processes video generation jobs.
+func VideoHandler(mg *mediagen.Manager, store storage.Store, pub realtime.EventPublisher, logger *logging.Logger) queue.Handler {
+	return generation.VideoHandler(mg, store, pub, logger)
+}
+
+// TextHandler returns a queue.Handler that processes text generation jobs with streaming.
+func TextHandler(tg *textgen.Manager, pub realtime.EventPublisher, logger *logging.Logger) queue.Handler {
+	return generation.TextHandler(tg, pub, logger)
+}
+
+// ChatResponseHandler returns a queue.Handler that generates AI chat responses.
+func ChatResponseHandler(tg *textgen.Manager, pub realtime.EventPublisher, logger *logging.Logger) queue.Handler {
+	return generation.ChatResponseHandler(tg, pub, logger)
+}

internal/adapter/templates/templates/skeleton/.claude/agents/realtime-specialist.md

@@ -1,40 +1,42 @@
 ---
 name: realtime-specialist
-description: WebSocket and real-time communication patterns for {{PROJECT_NAME}} - connection management, room-based broadcasting, Redis pub/sub scaling
+description: SSE and real-time communication patterns for {{PROJECT_NAME}} - HTTP2 POST for input, SSE for output, Redis pub/sub for scaling
 color: cyan
 ---
 
 # Realtime Specialist
 
-You design and implement real-time communication features for {{PROJECT_NAME}} using pkg/realtime. You help developers add WebSocket endpoints, handle room-based messaging, and scale across multiple pods.
+You design and implement real-time communication features for {{PROJECT_NAME}} using HTTP2 + SSE. You help developers add event streams, handle channel-based messaging, and scale across multiple pods.
+
+## Critical Rules
+
+- **NO WEBSOCKETS. EVER.** All real-time communication uses HTTP2 + SSE.
+- **User → Server:** HTTP2 POST/PUT/DELETE. Standard REST endpoints.
+- **Server → User:** SSE only. One-way event stream.
+- **Event flow:** `server → redis → redis listeners → SSE hub → user`
 
 ## When to Use
 
-- Adding WebSocket endpoints to a service
+- Adding SSE endpoints to a service
-- Implementing chat or notification features
+- Implementing chat, notifications, or progress features
-- Broadcasting messages to connected clients
+- Broadcasting events to connected clients
 - Scaling real-time features across multiple pods
 - Handling client reconnection and presence
 
 ## Architecture Overview
 
 ```
-                    ┌─────────────────────────────────────┐
+┌─────────────┐   HTTP2 POST    ┌─────────────┐   publish    ┌─────────────┐
-                    │            Redis Pub/Sub            │
+│   Browser   │ ───────────────▶│   API       │ ────────────▶│   Redis     │
-                    └─────────────┬───────────┬───────────┘
+│             │                 │   Handler   │              │   Pub/Sub   │
-                                  │           │
+│             │                 └─────────────┘              └──────┬──────┘
-          ┌───────────────────────┼───────────┼───────────────────────┐
+│             │                                                     │
-          │                       │           │                       │
+│             │                                              subscribe
-    ┌─────▼─────┐           ┌─────▼─────┐           ┌─────▼─────┐
+│             │                                                     │
-    │   Pod A   │           │   Pod B   │           │   Pod C   │
+│             │                                              ┌──────▼──────┐
-    │           │           │           │           │           │
+│             │   SSE stream    ┌─────────────┐   notify     │   Redis     │
-    │ ┌───────┐ │           │ ┌───────┐ │           │ ┌───────┐ │
+│             │ ◀───────────────│   SSE Hub   │ ◀────────────│   Listener  │
-    │ │  Hub  │ │           │ │  Hub  │ │           │ │  Hub  │ │
+└─────────────┘                 └─────────────┘              └─────────────┘
-    │ └───┬───┘ │           │ └───┬───┘ │           │ └───┬───┘ │
-    │     │     │           │     │     │           │     │     │
-    │ ┌───▼───┐ │           │ ┌───▼───┐ │           │ ┌───▼───┐ │
-    │ │Clients│ │           │ │Clients│ │           │ │Clients│ │
-    └─────────┘             └─────────┘             └─────────┘
 ```
 
 ## Quick Start
@@ -45,15 +47,14 @@ You design and implement real-time communication features for {{PROJECT_NAME}} u
 func main() {
     logger := logging.NewDevelopment()
 
-    // Create hub
+    // Create SSE hub
-    hub := realtime.NewHub(logger)
+    sseHub := realtime.NewSSEHub(logger)
-    go hub.Run(ctx)
 
-    // Create handler (no Redis needed for single pod)
+    // Create handler
-    wsHandler := realtime.NewHandler(hub, logger, realtime.HandlerConfig{})
+    sseHandler := realtime.NewSSEHandler(sseHub, logger)
 
     // Mount on router
-    r.Mount("/ws", wsHandler.Routes())
+    r.Mount("/api/events", sseHandler.Routes())
 }
 ```
 
@@ -63,160 +64,212 @@ func main() {
 func main() {
     logger := logging.NewProduction()
 
-    // Create hub
+    // Create SSE hub
-    hub := realtime.NewHub(logger)
+    sseHub := realtime.NewSSEHub(logger)
-    go hub.Run(ctx)
 
     // Create Redis broadcaster for cross-pod messaging
     redisClient := redis.NewClient(&redis.Options{Addr: os.Getenv("REDIS_URL")})
-    broadcaster := realtime.NewRedisBroadcaster(redisClient, hub, logger)
+    broadcaster := realtime.NewRedisBroadcaster(redisClient, sseHub, logger)
     go broadcaster.Run(ctx)
 
     // Create handler with broadcaster
-    wsHandler := realtime.NewHandler(hub, logger, realtime.HandlerConfig{
+    sseHandler := realtime.NewSSEHandler(sseHub, logger)
-        Broadcaster: broadcaster,
-    })
 
-    r.Mount("/ws", wsHandler.Routes())
+    r.Mount("/api/events", sseHandler.Routes())
 }
 ```
 
-## Message Protocol
+## Channel Types
 
-Messages use JSON format:
+| Pattern | Use For | Example |
+|---------|---------|---------|
+| `user:<id>` | Private events for one user | `user:u_abc123` |
+| `channel:<id>` | Shared events for a room/topic | `channel:general` |
+
+## Event Structure
+
+Every event MUST follow this structure:
 
 ```json
 {
-  "id": "uuid",
   "type": "chat",
-  "room": "general",
+  "timestamp": "2024-01-15T10:30:00Z",
-  "from": "client-id",
+  "userId": "u_abc123",
-  "data": { "text": "Hello world" },
+  "content": "Hello world"
-  "timestamp": "2024-01-15T10:30:00Z"
 }
 ```
 
-### Message Types
-
-| Type | Description |
-|------|-------------|
-| `chat` | User-generated chat message |
-| `presence` | User online/offline/away status |
-| `notification` | System notification to user |
-| `system` | Broadcast from server |
-| `error` | Error response to client |
-| `ping` / `pong` | Application-level keepalive |
-
 ## Patterns
 
-### Room-Based Chat
+### Chat Room
 
-```go
+**Client sends message (HTTP POST):**
-wsHandler := realtime.NewHandler(hub, logger, realtime.HandlerConfig{
-    OnConnect: func(conn realtime.Connection) {
-        // Notify room of new member
-        msg, _ := realtime.SystemMessage("presence", realtime.PresenceData{
-            Status: realtime.PresenceOnline,
-            UserID: conn.UserID(),
-        })
-        hub.Broadcast(msg)
-    },
-    OnDisconnect: func(conn realtime.Connection) {
-        msg, _ := realtime.SystemMessage("presence", realtime.PresenceData{
-            Status: realtime.PresenceOffline,
-            UserID: conn.UserID(),
-        })
-        hub.Broadcast(msg)
-    },
-})
 
-// Connect: ws://host/ws/room-name
+```typescript
+// POST /api/chat/messages
+await fetch('/api/chat/messages', {
+  method: 'POST',
+  headers: { 'Content-Type': 'application/json' },
+  body: JSON.stringify({
+    channel: 'general',
+    content: 'Hello world',
+  }),
+});
 ```
 
-### Message Filtering
+**Server handles POST, publishes to Redis:**
 
 ```go
-wsHandler := realtime.NewHandler(hub, logger, realtime.HandlerConfig{
+func (h *ChatHandler) PostMessage(w http.ResponseWriter, r *http.Request) error {
-    OnMessage: func(conn realtime.Connection, msg *realtime.Message) *realtime.Message {
-        // Filter profanity
-        if containsProfanity(msg.Data) {
-            return nil // Suppress message
-        }
-
-        // Add server metadata
-        msg.From = conn.UserID() // Use user ID instead of connection ID
-
-        return msg
-    },
-})
-```
-
-### Authenticated Connections
-
-```go
-wsHandler := realtime.NewHandler(hub, logger, realtime.HandlerConfig{
-    AuthRequired: true, // Requires valid JWT
-})
-
-// Client connects with token:
-// ws://host/ws?token=<jwt>
-// OR
-// ws://host/ws with Authorization header
-```
-
-### Sending from HTTP Handlers
-
-```go
-// Broadcast to a room from REST endpoint
-func (h *ChatHandler) PostMessage(w http.ResponseWriter, r *http.Request) {
     var req struct {
-        Room string `json:"room"`
+        Channel string `json:"channel"`
-        Text string `json:"text"`
+        Content string `json:"content"`
+    }
+    if err := app.Bind(r, &req); err != nil {
+        return err
     }
-    // ... decode request ...
 
-    msg := &realtime.Message{
+    user := auth.GetUser(r.Context())
-        Type: realtime.MessageTypeChat,
+
-        Room: req.Room,
+    // Publish to Redis (reaches all pods)
-        Data: json.RawMessage(`{"text":"` + req.Text + `"}`),
+    h.broadcaster.Publish(r.Context(), &realtime.Event{
+        Type:      "chat",
+        Channel:   req.Channel,
+        UserID:    user.ID,
+        UserName:  user.Name,
+        Content:   req.Content,
         Timestamp: time.Now().UTC(),
+    })
+
+    return httpresponse.NoContent(w, r)
+}
+```
+
+**Client receives via SSE:**
+
+```typescript
+useEventChannel(`channel:general`, {
+  onEvent: (event) => {
+    if (event.type === 'chat') {
+      addMessage(event);
+    }
+  }
+});
+```
+
+### Async Job Progress
+
+**Client initiates job (HTTP POST):**
+
+```typescript
+const { jobId } = await api.post('/generate/video', {
+  prompt: 'A cat playing piano',
+  aspectRatio: '16:9',
+});
+```
+
+**Client listens for progress (SSE):**
+
+```typescript
+useEventChannel(`user:${userId}`, {
+  onEvent: (event) => {
+    if (event.jobId !== jobId) return;
+    switch (event.type) {
+      case 'generation_progress':
+        setProgress(event.progress);
+        break;
+      case 'generation_complete':
+        setResult(event.result);
+        break;
+    }
+  }
+});
+```
+
+**Worker sends progress events:**
+
+```go
+func (w *Worker) ProcessJob(ctx context.Context, job *domain.Job) error {
+    // Send progress
+    w.hub.SendToUser(job.UserID, &realtime.Event{
+        Type:     "generation_progress",
+        JobID:    job.ID,
+        Progress: 50,
+        Message:  "Processing...",
+    })
+
+    // ... do work ...
+
+    // Send complete
+    w.hub.SendToUser(job.UserID, &realtime.Event{
+        Type:   "generation_complete",
+        JobID:  job.ID,
+        Result: result,
+    })
+    return nil
+}
+```
+
+### Presence
+
+**Client connects, server broadcasts presence:**
+
+```go
+// In SSE connection handler
+func (h *SSEHandler) onConnect(userID string, channel string) {
+    h.broadcaster.Publish(ctx, &realtime.Event{
+        Type:    "presence",
+        Channel: channel,
+        UserID:  userID,
+        Status:  "online",
+    })
 }
 
-    // Publish via broadcaster (reaches all pods)
+func (h *SSEHandler) onDisconnect(userID string, channel string) {
-    if h.broadcaster != nil {
+    h.broadcaster.Publish(ctx, &realtime.Event{
-        h.broadcaster.Publish(r.Context(), msg)
+        Type:    "presence",
-    } else {
+        Channel: channel,
-        h.hub.Broadcast(msg)
+        UserID:  userID,
-    }
+        Status:  "offline",
+    })
 }
 ```
 
 ## Client Reconnection
 
-Clients should implement reconnection with exponential backoff:
+SSE clients should implement reconnection with exponential backoff:
 
-```javascript
+```typescript
-class RealtimeClient {
+function useEventChannel(channel: string, config: Config) {
-    connect() {
+  const [retries, setRetries] = useState(0);
-        this.ws = new WebSocket(`${this.url}?last_id=${this.lastMessageId}`);
+
-        this.ws.onclose = () => this.scheduleReconnect();
+  const connect = useCallback(() => {
-        this.ws.onmessage = (e) => {
+    const eventSource = new EventSource(`/api/events?channel=${channel}`);
-            const msg = JSON.parse(e.data);
+
-            this.lastMessageId = msg.id;
+    eventSource.onopen = () => setRetries(0);
-            this.onMessage(msg);
+
+    eventSource.onerror = () => {
+      eventSource.close();
+      const delay = Math.min(1000 * Math.pow(2, retries), 30000);
+      setTimeout(connect, delay);
+      setRetries(r => r + 1);
     };
-    }
 
-    scheduleReconnect() {
+    eventSource.onmessage = (e) => {
-        const delay = Math.min(1000 * Math.pow(2, this.retries), 30000);
+      config.onEvent(JSON.parse(e.data));
-        setTimeout(() => this.connect(), delay);
+    };
-        this.retries++;
+  }, [channel, retries]);
-    }
 }
 ```
 
 ## Scaling Considerations
 
+### Redis Channel Strategy
+
+- One channel per room: `realtime:channel:{channelId}`
+- One channel per user: `realtime:user:{userId}`
+- Pattern subscription: `realtime:*`
+
 ### Connection Limits
 
 Set reasonable limits per pod:
@@ -224,26 +277,20 @@ Set reasonable limits per pod:
 ```go
 const maxConnectionsPerPod = 10000
 
-func (h *Handler) HandleWebSocket(w http.ResponseWriter, r *http.Request) {
+func (h *SSEHandler) HandleSSE(w http.ResponseWriter, r *http.Request) {
     if h.hub.ConnectionCount() >= maxConnectionsPerPod {
         http.Error(w, "server at capacity", http.StatusServiceUnavailable)
         return
     }
-    // ... continue upgrade ...
+    // ... continue ...
 }
 ```
 
-### Redis Channel Strategy
-
-- One channel per room: `realtime:room:{roomId}`
-- Global channel for broadcasts: `realtime:global`
-- Pattern subscription: `realtime:room:*`
-
 ### Memory Considerations
 
-Each connection uses ~10KB for buffers. Plan accordingly:
+Each SSE connection uses ~5KB for buffers. Plan accordingly:
-- 10,000 connections ≈ 100MB
+- 10,000 connections ≈ 50MB
-- 100,000 connections ≈ 1GB
+- 100,000 connections ≈ 500MB
 
 ## Monitoring
 
@@ -251,50 +298,26 @@ Track these metrics:
 
 | Metric | Description |
 |--------|-------------|
-| `realtime_connections_total` | Total active connections |
+| `sse_connections_total` | Total active SSE connections |
-| `realtime_rooms_total` | Number of active rooms |
+| `sse_channels_total` | Number of active channels |
-| `realtime_messages_sent` | Messages sent per second |
+| `sse_events_sent` | Events sent per second |
-| `realtime_messages_received` | Messages received per second |
+| `redis_publish_errors` | Failed Redis publishes |
-| `realtime_redis_publish_errors` | Failed Redis publishes |
-
-## Error Handling
-
-### Client Errors
-
-```go
-OnMessage: func(conn realtime.Connection, msg *realtime.Message) *realtime.Message {
-    if err := validate(msg); err != nil {
-        errMsg, _ := realtime.SystemMessage(realtime.MessageTypeError, map[string]string{
-            "error": err.Error(),
-        })
-        conn.Send(errMsg)
-        return nil // Don't broadcast invalid message
-    }
-    return msg
-}
-```
-
-### Redis Failures
-
-RedisBroadcaster degrades gracefully:
-- If publish fails, message still broadcasts locally
-- Subscriber reconnects automatically on disconnect
-- Log warnings for monitoring
 
 ## Do
 
-1. ALWAYS use room-based broadcasting for multi-tenant apps
+1. USE HTTP POST for all client→server messages
-2. SET connection limits per pod
+2. USE SSE for all server→client events
-3. IMPLEMENT client reconnection with backoff
+3. USE Redis pub/sub for multi-pod deployments
-4. USE Redis for multi-pod deployments
+4. SET connection limits per pod
-5. AUTHENTICATE WebSocket connections in production
+5. IMPLEMENT client reconnection with backoff
-6. MONITOR connection count and message rates
+6. AUTHENTICATE SSE connections in production
+7. DOCUMENT all channels in `docs/channels.md`
 
 ## Do Not
 
-1. STORE large payloads in messages (send IDs, fetch data separately)
+1. USE WebSocket for anything — SSE only
-2. BROADCAST without rate limiting
+2. STORE large payloads in events (send IDs, fetch data separately)
-3. RELY on message ordering (out-of-order is possible)
+3. BROADCAST without rate limiting
-4. SKIP ping/pong (connections will time out)
+4. SIMULATE progress with fake timers
-5. USE synchronous operations in message handlers (blocks hub)
+5. SKIP ping/pong (connections will time out)
 6. TRUST client-provided user IDs (extract from auth token)

internal/adapter/templates/templates/skeleton/.claude/guides/events.md

@@ -0,0 +1,164 @@
+# Event Channels
+
+## Critical Rules
+
+- **NO WEBSOCKETS. EVER.** All real-time communication uses HTTP2 + SSE.
+- **User → Server:** HTTP2 POST/PUT/DELETE. Standard REST endpoints.
+- **Server → User:** SSE only. One-way event stream.
+- **Event flow:** `POST → Service (enqueue) → Queue → Worker (generate) → Redis pub/sub → Service SSE subscriber → SSE Hub → User`
+- **This applies to EVERYTHING:** Chat, notifications, progress updates, generation — all SSE.
+- **Channel format is non-negotiable:** `user:<id>` or `channel:<id>`. No exceptions.
+- **All events are JSON** with `type` as the first field.
+- **Document every channel** in this file before using it.
+- **Service is thin:** Validates, enqueues, returns 202. No AI work in the service.
+- **Worker does all AI work:** Initializes providers, processes jobs, publishes events via Redis.
+
+## Architecture
+
+```
+┌─────────────┐   POST /generate/*    ┌─────────────┐   enqueue    ┌─────────────┐
+│   Browser   │ ──────────────────────▶│   Service   │ ────────────▶│   CRDB      │
+│             │   { jobId } (202)      │   (thin)    │              │   Queue     │
+│             │ ◀──────────────────────│             │              └──────┬──────┘
+│             │                        │             │                     │
+│             │                        │             │              dequeue│
+│             │                        │             │                     ▼
+│             │                        │             │              ┌──────────────┐
+│             │                        │             │              │   Worker     │
+│             │                        │             │              │   (AI work)  │
+│             │                        │             │              └──────┬───────┘
+│             │                        │             │                     │
+│             │                        │             │              publish│(SSE events)
+│             │                        │             │                     ▼
+│             │                        │             │              ┌──────────────┐
+│             │   SSE stream           │  SSE Hub    │  subscribe   │   Redis      │
+│             │ ◀──────────────────────│  ◀──────────│──────────────│   Pub/Sub    │
+└─────────────┘                        └─────────────┘              └──────────────┘
+```
+
+**Generation flow:**
+1. User sends `POST /api/generate/image` with prompt
+2. Service validates, enqueues job in CRDB, returns `202 {jobId}`
+3. Worker dequeues job, calls AI provider (LaoZhang/Gemini)
+4. Worker publishes progress/result SSE events to Redis
+5. Service's SSE subscriber receives events from Redis
+6. SSE Hub delivers to the user's connected SSE stream
+
+**Chat flow:**
+1. User sends `POST /api/chat/messages`
+2. Service broadcasts user message to `channel:general` via SSE Hub (immediate)
+3. Service enqueues `ai_chat_response` job
+4. Worker dequeues, streams AI response tokens via Redis pub/sub
+5. Service's SSE subscriber delivers `ai_chat_chunk` events to channel
+
+## Channel Types
+
+| Pattern | Use For | Example |
+|---------|---------|---------|
+| `user:<id>` | Private events for one user | `user:u_abc123` |
+| `channel:<id>` | Shared events for a room/topic | `channel:general` |
+
+## Event Structure
+
+Every event MUST follow this structure:
+
+```typescript
+interface Event {
+  type: string;        // REQUIRED: event type identifier
+  timestamp: string;   // REQUIRED: ISO 8601
+  jobId?: string;      // Job correlation ID
+  progress?: number;   // 0-100 percentage
+  message?: string;    // Human-readable status
+  result?: any;        // Payload (type-specific)
+  error?: string;      // Error message
+}
+```
+
+## Standard Event Types
+
+### User Channel Events
+
+#### Media Generation Events
+
+| Event | Trigger | Payload |
+|-------|---------|---------|
+| `generation_started` | Worker picks up job | `{ jobId, message }` |
+| `generation_progress` | Progress update | `{ jobId, progress, message }` |
+| `generation_complete` | Generation done | `{ jobId, progress: 100, result }` |
+| `generation_failed` | Error occurred | `{ jobId, error }` |
+
+#### Text Generation Events (streaming)
+
+| Event | Trigger | Payload |
+|-------|---------|---------|
+| `ai_chat_chunk` | Token generated | `{ jobId, result: { streamId, text, done, provider? } }` |
+
+#### Media Upload Events
+
+| Event | Trigger | Payload |
+|-------|---------|---------|
+| `upload_started` | Upload job begins | `{ jobId }` |
+| `upload_progress` | Chunk uploaded | `{ jobId, progress, bytesUploaded }` |
+| `upload_complete` | Processing done | `{ jobId, result: { original, optimized, thumbnail } }` |
+| `upload_failed` | Error occurred | `{ jobId, error }` |
+
+### Room Channel Events
+
+| Event | Trigger | Payload |
+|-------|---------|---------|
+| `chat` | User sends message | `{ result: { id, content, userId, userName, timestamp } }` |
+| `ai_chat_chunk` | AI streaming chunk | `{ result: { streamId, text, done, provider? } }` |
+| `ai_chat` | AI response complete | `{ result: { id, content, provider, timestamp } }` |
+| `presence` | User joins/leaves | `{ status, userId, userName }` |
+| `typing` | User typing indicator | `{ userId, isTyping }` |
+
+## Implementation Pattern
+
+### Backend (Go) — Service Handler
+
+```go
+// Enqueue generation job (service is thin — no AI work)
+jobID, err := h.queue.Enqueue(r.Context(), "generate_image", map[string]any{
+    "prompt": req.Prompt,
+    "userID": userID,
+})
+httpresponse.Accepted(w, r, GenerateAccepted{JobID: jobID})
+```
+
+### Backend (Go) — Worker Job Handler
+
+```go
+// Worker publishes events via Redis SSE publisher
+pub.SendToUser(userID, &realtime.SSEEvent{
+    Type:     realtime.EventGenerationComplete,
+    JobID:    jobID,
+    Progress: 100,
+    Result:   result,
+})
+```
+
+### Frontend (TypeScript)
+
+```typescript
+// Subscribe to user channel for generation events
+const { status, progress, result } = useMediaGeneration<ImageResult>({
+  endpoint: '/api/generate/image',
+  userId: currentUser.id,
+});
+
+// Subscribe to room channel for chat events
+const { messages, aiMessages, streamingMessages } = useChat({
+  endpoint: '/api/chat/messages',
+  channel: 'channel:general',
+  userId: currentUser.id,
+});
+```
+
+## Active Channels
+
+<!-- Document all channels used in this project below -->
+
+| Channel | Events | Purpose |
+|---------|--------|---------|
+| `user:<userId>` | `generation_*`, `ai_chat_chunk` | Async job results, text streaming |
+| `channel:<room>` | `chat`, `ai_chat_chunk`, `presence` | Real-time chat |

internal/adapter/templates/templates/skeleton/.claude/guides/media.md

@@ -0,0 +1,200 @@
+# Media Pipeline
+
+## Critical Rules
+
+- **ALL media operations are async jobs.** Upload, process, generate - everything goes through the job queue.
+- **NEVER wait synchronously.** POST returns a job ID immediately. Results come via SSE.
+- **NEVER simulate progress.** Real progress comes from real events. Fake progress is a lie.
+- **Storage is opaque.** Backend returns URLs. Frontend never constructs storage paths.
+- **GCS in production, MemoryStore in dev.** When `GCS_BUCKET` env var is set, storage uses GCS. Otherwise, an in-memory store serves files at `/storage/`.
+
+## Architecture Overview
+
+```
+┌─────────────┐   POST /generate/*   ┌──────────────┐   enqueue   ┌──────────────┐
+│   Frontend  │ ─────────────────────▶│   Service    │ ───────────▶│   CRDB       │
+│             │   { jobId } (202)     │   (thin)     │             │   Queue      │
+│             │ ◀─────────────────────│              │             └──────┬───────┘
+│             │                       │              │                    │
+│             │                       │              │             dequeue│
+│             │                       │              │                    ▼
+│             │                       │              │             ┌──────────────┐
+│             │                       │              │             │   Worker     │
+│             │                       │              │             │   (AI work)  │
+│             │                       │              │             └──────┬───────┘
+│             │                       │              │                    │
+│             │   SSE stream          │   SSE Hub    │  Redis sub  │  persist to
+│             │ ◀─────────────────────│   ◀──────────│─────────────│◀── storage
+└─────────────┘                       └──────────────┘             │    (GCS)
+                                                                   └─────┘
+```
+
+## Storage
+
+### Backend: `pkg/storage/`
+
+The storage package provides a `Store` interface with two implementations:
+
+| Implementation | When | Env Vars |
+|---|---|---|
+| `GCSStore` | `GCS_BUCKET` is set (production, deployed) | `GCS_BUCKET`, `GCS_SERVICE_ACCOUNT_JSON` |
+| `MemoryStore` | No `GCS_BUCKET` (local dev, standalone) | None |
+
+```go
+type Store interface {
+    Upload(ctx context.Context, path string, data []byte, contentType string) (string, error)
+    UploadPresigned(ctx context.Context, path string, contentType string) (*PresignedUpload, error)
+    GetURL(ctx context.Context, path string) (string, error)
+    Delete(ctx context.Context, path string) error
+    List(ctx context.Context, prefix string) ([]MediaObject, error)
+}
+```
+
+### Initialization (service main.go)
+
+Storage is initialized early in `main()` — before the queue, since standalone queue handlers need it:
+
+```go
+var mediaStore storage.Store
+if bucket := os.Getenv("GCS_BUCKET"); bucket != "" {
+    mediaStore, _ = storage.NewGCSStore(bucket, os.Getenv("GCS_SERVICE_ACCOUNT_JSON"), logger)
+} else {
+    memStore := storage.NewMemoryStore("http://localhost:" + port + "/storage")
+    mediaStore = memStore
+    // Mount memStore.ServeHTTP at /storage/* for dev mode
+}
+```
+
+### Object Path Convention
+
+All media is stored under `media/{userID}/`:
+- Generated images: `media/{userID}/images/{jobID}_{index}.png`
+- Generated videos: `media/{userID}/videos/{jobID}_{index}.mp4`
+- Uploads: `media/{userID}/{uuid}/{filename}`
+
+### Generation Auto-Persist
+
+Image and video generation handlers accept a `storage.Store`. When non-nil, generated results are automatically persisted and SSE events contain permanent URLs instead of temporary provider URLs or base64.
+
+```go
+generation.ImageHandler(mediagenManager, store, pub, logger)
+generation.VideoHandler(mediagenManager, store, pub, logger)
+```
+
+## Upload Flow (Presigned URL)
+
+```
+Frontend                    Backend                     Storage (GCS/Memory)
+   │                           │                              │
+   │ POST /media/upload/init   │                              │
+   │ {filename, contentType}   │                              │
+   │──────────────────────────▶│                              │
+   │                           │  UploadPresigned()           │
+   │                           │─────────────────────────────▶│
+   │  {uploadURL, objectPath}  │                              │
+   │◀──────────────────────────│                              │
+   │                           │                              │
+   │ PUT uploadURL (file body) │                              │
+   │─────────────────────────────────────────────────────────▶│
+   │          200 OK           │                              │
+   │◀─────────────────────────────────────────────────────────│
+   │                           │                              │
+   │ POST /media/upload/complete│                             │
+   │ {objectPath}              │                              │
+   │──────────────────────────▶│  GetURL()                    │
+   │                           │─────────────────────────────▶│
+   │  {url, path}              │                              │
+   │◀──────────────────────────│                              │
+```
+
+### Frontend Hook
+
+```typescript
+import { useMediaUpload } from '@project/realtime';
+
+const { upload, isUploading, progress, error, reset } = useMediaUpload({
+  apiPrefix: '',
+  serviceName: 'example-api',
+  headers: { Authorization: `Bearer ${token}` },
+});
+
+// Upload a file
+const result = await upload(file); // { url, path }
+```
+
+## Media Library
+
+### Endpoints
+
+| Method | Path | Description |
+|--------|------|-------------|
+| `POST` | `/media/upload/init` | Get presigned upload URL |
+| `POST` | `/media/upload/complete` | Confirm upload, get final URL |
+| `GET` | `/media` | List user's media (optional `?prefix=images`) |
+| `DELETE` | `/media/{path...}` | Delete a media object |
+
+All endpoints require authentication.
+
+### Frontend Components
+
+```tsx
+import { MediaUploader, MediaLibrary } from '@project/ui';
+
+// Upload drop zone
+<MediaUploader
+  upload={mediaUpload.upload}
+  isUploading={mediaUpload.isUploading}
+  progress={mediaUpload.progress}
+  onUploadComplete={() => refetchMedia()}
+/>
+
+// Media grid with preview
+<MediaLibrary
+  items={items}
+  onDelete={(path) => deleteMedia(path)}
+/>
+```
+
+## Event Reference
+
+| Event | When | Payload |
+|-------|------|---------|
+| `generation_started` | Generation begins | `{ jobId }` |
+| `generation_progress` | Progress update | `{ jobId, progress, message }` |
+| `generation_complete` | Generation done (URLs are persistent) | `{ jobId, result }` |
+| `generation_failed` | Error occurred | `{ jobId, error }` |
+| `upload_started` | Upload job begins | `{ jobId }` |
+| `upload_progress` | Chunk uploaded | `{ jobId, progress }` |
+| `upload_complete` | Upload done | `{ jobId, result: { url, path } }` |
+| `upload_failed` | Error occurred | `{ jobId, error }` |
+
+## Common Mistakes
+
+### DON'T: Construct storage URLs
+
+```typescript
+// WRONG
+const url = `/storage/uploads/${userId}/${fileId}.jpg`;
+```
+
+### DO: Use URLs from backend
+
+```typescript
+// CORRECT
+const url = media.url; // Backend provides complete URL
+```
+
+### DON'T: Proxy uploads through the backend
+
+```typescript
+// WRONG - wastes backend bandwidth
+await fetch('/api/upload', { body: file });
+```
+
+### DO: Upload directly to storage via presigned URL
+
+```typescript
+// CORRECT - frontend uploads directly to GCS
+const { uploadURL } = await initUpload(file);
+await fetch(uploadURL, { method: 'PUT', body: file });
+```

internal/adapter/templates/templates/skeleton/.gitignore

@@ -17,8 +17,9 @@ coverage.html
 # Dependency directories
 vendor/
 
-# Go workspace file (local only)
+# Go workspace files (local only)
 go.work.sum
+*.go.sum
 
 # IDE
 .idea/
@@ -34,6 +35,7 @@ go.work.sum
 # Node
 node_modules/
 .npm/
+pnpm-lock.yaml
 
 # Shared packages
 packages/*/node_modules/

internal/adapter/templates/templates/skeleton/CLAUDE.md.tmpl

@@ -10,6 +10,8 @@
 | **Build a feature** | [feature-development.md](.claude/guides/feature-development.md) |
 | **Backend API patterns** | [backend/api-patterns.md](.claude/guides/backend/api-patterns.md) |
 | **Frontend design system** | [frontend/design-system.md](.claude/guides/frontend/design-system.md) |
+| **Event channels** | [events.md](.claude/guides/events.md) |
+| **Media pipeline** | [media.md](.claude/guides/media.md) |
 | **Deploy** | [ops/deploying.md](.claude/guides/ops/deploying.md) |
 
 ## Quick Reference
@@ -36,6 +38,13 @@
 - **OpenAPI first:** Document endpoints in `spec.go` using `openapi.*` helpers. Mount with `application.EnableDocs(spec)`.
 - **CSS variables:** All UI components use CSS custom properties (`var(--background)`, `var(--accent)`, etc.). Never hardcode colors.
 - **Monorepo imports:** Go packages from `{{GO_MODULE}}/pkg/*`, TypeScript from `@{{PROJECT_NAME}}/*`.
+- **NO WEBSOCKETS. EVER.** All real-time communication uses HTTP2 + SSE. User → server is HTTP2 POST. Server → user is SSE. This includes chat, notifications, progress, everything.
+- **Event flow:** `POST → Service (enqueue) → Queue → Worker (generate) → Redis pub/sub → Service SSE subscriber → User`. Service is thin, worker does AI work.
+- **Channel naming:** `user:<id>` = events for a specific user. `channel:<id>` = events for a topic/room/resource. Document all channels in `./docs/channels.md`.
+- **Media uploads:** POST returns job ID immediately. Progress and result come via SSE events. Never wait synchronously.
+- **Media generation:** Same pattern - POST queues job, returns ID, results via SSE. Video takes 2-5 min; never block HTTP. Text generation streams `ai_chat_chunk` events token-by-token.
+- **Media storage:** Backend returns complete URLs. Never construct storage paths in frontend. Variants (thumbnail, optimized) auto-generated.
+- **No fake progress:** Never simulate progress with timers. Real progress comes from real events.
 
 ## Architecture
 

internal/adapter/templates/templates/skeleton/packages/ai-client/package.json.tmpl

@@ -0,0 +1,22 @@
+{
+  "name": "@{{PROJECT_NAME}}/ai-client",
+  "version": "0.0.1",
+  "private": true,
+  "type": "module",
+  "main": "src/index.ts",
+  "types": "src/index.ts",
+  "exports": {
+    ".": {
+      "types": "./src/index.ts",
+      "import": "./src/index.ts"
+    }
+  },
+  "scripts": {
+    "typecheck": "tsc --noEmit",
+    "build": "tsc",
+    "lint": "eslint . --ext ts,tsx --report-unused-disable-directives --max-warnings 0"
+  },
+  "devDependencies": {
+    "typescript": "^5.5.3"
+  }
+}

internal/adapter/templates/templates/skeleton/packages/ai-client/src/index.ts

@@ -0,0 +1,44 @@
+// Types
+export type {
+  TokenUsage,
+  AIResponseMeta,
+  AIError,
+  ErrorCode,
+} from './types';
+
+export { createAIError, ErrorCodes } from './types';
+
+// Text generation
+export type {
+  MessageRole,
+  TextGenMessage,
+  TextGenRequest,
+  TextGenResponse,
+  TextGenChunk,
+  TextGenOptions,
+} from './textgen';
+
+export { streamText, generateText } from './textgen';
+
+// Image generation
+export type {
+  ImageSize,
+  AspectRatio,
+  ImageGenRequest,
+  GeneratedImage,
+  ImageGenResponse,
+  GenerationProgress,
+  MediaGenOptions,
+} from './mediagen';
+
+export { generateImage, generateImageWithProgress } from './mediagen';
+
+// Video generation
+export type {
+  VideoDuration,
+  VideoGenRequest,
+  GeneratedVideo,
+  VideoGenResponse,
+} from './mediagen';
+
+export { generateVideo, generateVideoWithProgress } from './mediagen';

internal/adapter/templates/templates/skeleton/packages/ai-client/src/mediagen.ts

@@ -0,0 +1,384 @@
+/**
+ * Media generation types and utilities.
+ *
+ * IMPORTANT: For generation, use @{{PROJECT_NAME}}/realtime's useMediaGeneration hook.
+ * That hook implements the correct async pattern:
+ * 1. POST returns job ID immediately
+ * 2. Progress and results come via SSE events
+ * 3. Never blocks waiting for generation
+ *
+ * @example
+ * ```tsx
+ * import { useMediaGeneration } from '@{{PROJECT_NAME}}/realtime';
+ *
+ * const { status, progress, result, generate } = useMediaGeneration({
+ *   endpoint: '/api/generate/video',
+ *   userId: user.id,
+ * });
+ *
+ * await generate({ prompt: 'A cat playing piano' });
+ * // Results arrive via SSE - no blocking
+ * ```
+ *
+ * The synchronous functions in this file are DEPRECATED.
+ * Only use them for simple image generation where blocking is acceptable.
+ */
+
+import { AIResponseMeta, createAIError, ErrorCodes } from './types';
+
+/** Image size presets. */
+export type ImageSize = '256x256' | '512x512' | '1024x1024' | '1024x1792' | '1792x1024';
+
+/** Aspect ratio presets. */
+export type AspectRatio = '1:1' | '16:9' | '9:16' | '4:3' | '3:4';
+
+/** Video duration presets. */
+export type VideoDuration = '5s' | '10s';
+
+// ---------------------------------------------------------------------------
+// Image types
+// ---------------------------------------------------------------------------
+
+/** Request for image generation. */
+export interface ImageGenRequest {
+  prompt: string;
+  model?: string;
+  size?: ImageSize;
+  aspectRatio?: AspectRatio;
+  count?: number;
+  negativePrompt?: string;
+  style?: string;
+}
+
+/** A single generated image. */
+export interface GeneratedImage {
+  /** Base64-encoded image data or URL. */
+  data: string;
+  /** Whether data is a URL (true) or base64 (false). */
+  isUrl: boolean;
+  /** Provider that generated this image. */
+  provider: string;
+  /** Generation seed if available. */
+  seed?: number;
+  /** Revised prompt if the provider modified it. */
+  revisedPrompt?: string;
+}
+
+/** Response from image generation. */
+export interface ImageGenResponse extends AIResponseMeta {
+  images: GeneratedImage[];
+}
+
+// ---------------------------------------------------------------------------
+// Video types
+// ---------------------------------------------------------------------------
+
+/** Request for video generation. */
+export interface VideoGenRequest {
+  prompt: string;
+  model?: string;
+  aspectRatio?: AspectRatio;
+  duration?: VideoDuration;
+}
+
+/** A single generated video. */
+export interface GeneratedVideo {
+  /** Base64-encoded video data or URL. */
+  data: string;
+  /** Whether data is a URL (true) or base64 (false). */
+  isUrl: boolean;
+  /** MIME type (e.g. "video/mp4"). */
+  mimeType: string;
+  /** Provider that generated this video. */
+  provider: string;
+  /** Generation seed if available. */
+  seed?: number;
+}
+
+/** Response from video generation. */
+export interface VideoGenResponse extends AIResponseMeta {
+  videos: GeneratedVideo[];
+}
+
+// ---------------------------------------------------------------------------
+// Shared types
+// ---------------------------------------------------------------------------
+
+/** Progress update during generation. */
+export interface GenerationProgress {
+  percent: number;
+  stage: string;
+}
+
+/** Options for API calls. */
+export interface MediaGenOptions {
+  timeout?: number;
+  headers?: Record<string, string>;
+}
+
+// ---------------------------------------------------------------------------
+// Internal helpers
+// ---------------------------------------------------------------------------
+
+/** Parse image array from backend response. */
+function parseImages(data: Record<string, unknown>): GeneratedImage[] {
+  const images = data.images as Record<string, unknown>[] | undefined;
+  if (!images) return [];
+  return images.map((img) => ({
+    data: (img.data as string) || (img.url as string) || '',
+    isUrl: !!img.isUrl || !!img.url,
+    provider: (img.provider as string) || (data.provider as string) || '',
+    seed: img.seed as number | undefined,
+    revisedPrompt: img.revised_prompt as string | undefined,
+  }));
+}
+
+/** Parse video array from backend response. */
+function parseVideos(data: Record<string, unknown>): GeneratedVideo[] {
+  const videos = data.videos as Record<string, unknown>[] | undefined;
+  if (!videos) return [];
+  return videos.map((v) => ({
+    data: (v.data as string) || (v.url as string) || '',
+    isUrl: !!v.isUrl || !!v.url,
+    mimeType: (v.mimeType as string) || 'video/mp4',
+    provider: (v.provider as string) || (data.provider as string) || '',
+    seed: v.seed as number | undefined,
+  }));
+}
+
+/** Handle HTTP errors from generation endpoints. */
+function handleHttpError(status: number, errorData: Record<string, unknown>): never {
+  throw createAIError(
+    (errorData.message as string) || `HTTP ${status}`,
+    status === 429 ? ErrorCodes.RATE_LIMITED : ErrorCodes.PROVIDER_ERROR,
+    undefined,
+    status === 429 || status >= 500,
+  );
+}
+
+/**
+ * Simulate progress ticks during a synchronous fetch.
+ * Returns a cleanup function.
+ */
+function simulateProgress(
+  onProgress: (p: GenerationProgress) => void,
+  stage: string,
+  maxPercent: number,
+  intervalMs: number,
+): () => void {
+  let current = 10;
+  const id = setInterval(() => {
+    if (current < maxPercent) {
+      current += Math.random() * 8 + 2;
+      current = Math.min(current, maxPercent);
+      onProgress({ percent: Math.round(current), stage });
+    }
+  }, intervalMs);
+  return () => clearInterval(id);
+}
+
+// ---------------------------------------------------------------------------
+// Image generation
+// ---------------------------------------------------------------------------
+
+/**
+ * Generate images from a text prompt.
+ *
+ * @param endpoint - The image generation API endpoint URL
+ * @param request - The generation request
+ * @param options - Additional options
+ * @returns Promise resolving to generated images
+ *
+ * @example
+ * ```ts
+ * const response = await generateImage('/api/my-api/generate/image', {
+ *   prompt: 'A sunset over mountains',
+ *   count: 2,
+ * });
+ * ```
+ */
+export async function generateImage(
+  endpoint: string,
+  request: ImageGenRequest,
+  options: MediaGenOptions = {},
+): Promise<ImageGenResponse> {
+  const startTime = Date.now();
+
+  const controller = new AbortController();
+  const timeoutId = options.timeout
+    ? setTimeout(() => controller.abort(), options.timeout)
+    : undefined;
+
+  try {
+    const response = await fetch(endpoint, {
+      method: 'POST',
+      headers: { 'Content-Type': 'application/json', ...options.headers },
+      body: JSON.stringify(request),
+      signal: controller.signal,
+    });
+
+    if (!response.ok) {
+      const errorData = await response.json().catch(() => ({}));
+      handleHttpError(response.status, errorData);
+    }
+
+    const data = await response.json();
+
+    return {
+      images: parseImages(data),
+      provider: (data.provider as string) || '',
+      model: data.model as string | undefined,
+      latencyMs: Date.now() - startTime,
+    };
+  } catch (error) {
+    if (error instanceof Error && error.name === 'AbortError') {
+      throw createAIError('Request timed out', ErrorCodes.TIMEOUT, undefined, true);
+    }
+    throw error;
+  } finally {
+    if (timeoutId) clearTimeout(timeoutId);
+  }
+}
+
+/**
+ * Generate images with progress feedback.
+ *
+ * The backend generates images synchronously. Progress is simulated
+ * to give the user visual feedback while waiting.
+ *
+ * @param endpoint - The image generation API endpoint URL
+ * @param request - The generation request
+ * @param onProgress - Callback for progress updates
+ * @param options - Additional options
+ * @returns Promise resolving to generated images
+ *
+ * @example
+ * ```ts
+ * const response = await generateImageWithProgress(
+ *   '/api/my-api/generate/image',
+ *   { prompt: 'A detailed fantasy landscape', count: 4 },
+ *   (progress) => setProgress(progress.percent),
+ * );
+ * ```
+ */
+export async function generateImageWithProgress(
+  endpoint: string,
+  request: ImageGenRequest,
+  onProgress: (progress: GenerationProgress) => void,
+  options: MediaGenOptions = {},
+): Promise<ImageGenResponse> {
+  onProgress({ percent: 0, stage: 'Starting generation...' });
+
+  const stopSimulation = simulateProgress(onProgress, 'Generating images...', 85, 800);
+
+  try {
+    const result = await generateImage(endpoint, request, options);
+    onProgress({ percent: 100, stage: 'Complete' });
+    return result;
+  } catch (error) {
+    onProgress({ percent: 0, stage: 'Failed' });
+    throw error;
+  } finally {
+    stopSimulation();
+  }
+}
+
+// ---------------------------------------------------------------------------
+// Video generation
+// ---------------------------------------------------------------------------
+
+/**
+ * @deprecated Use useMediaGeneration from @{{PROJECT_NAME}}/realtime instead.
+ * Video generation takes 2-5 minutes. Blocking HTTP requests is wrong.
+ *
+ * Correct pattern:
+ * ```tsx
+ * import { useMediaGeneration } from '@{{PROJECT_NAME}}/realtime';
+ * const { generate, progress, result } = useMediaGeneration({
+ *   endpoint: '/api/generate/video',
+ *   userId: user.id,
+ * });
+ * await generate({ prompt: '...' }); // Returns immediately
+ * // Result arrives via SSE
+ * ```
+ */
+export async function generateVideo(
+  endpoint: string,
+  request: VideoGenRequest,
+  options: MediaGenOptions = {},
+): Promise<VideoGenResponse> {
+  const startTime = Date.now();
+
+  const controller = new AbortController();
+  // Video generation can take minutes; default timeout is 5 min
+  const timeout = options.timeout || 300_000;
+  const timeoutId = setTimeout(() => controller.abort(), timeout);
+
+  try {
+    const response = await fetch(endpoint, {
+      method: 'POST',
+      headers: { 'Content-Type': 'application/json', ...options.headers },
+      body: JSON.stringify(request),
+      signal: controller.signal,
+    });
+
+    if (!response.ok) {
+      const errorData = await response.json().catch(() => ({}));
+      handleHttpError(response.status, errorData);
+    }
+
+    const data = await response.json();
+
+    return {
+      videos: parseVideos(data),
+      provider: (data.provider as string) || '',
+      model: data.model as string | undefined,
+      latencyMs: Date.now() - startTime,
+    };
+  } catch (error) {
+    if (error instanceof Error && error.name === 'AbortError') {
+      throw createAIError('Video generation timed out', ErrorCodes.TIMEOUT, undefined, true);
+    }
+    throw error;
+  } finally {
+    clearTimeout(timeoutId);
+  }
+}
+
+/**
+ * @deprecated Use useMediaGeneration from @{{PROJECT_NAME}}/realtime instead.
+ * This function uses FAKE simulated progress. Real progress comes from SSE events.
+ *
+ * Correct pattern:
+ * ```tsx
+ * import { useMediaGeneration } from '@{{PROJECT_NAME}}/realtime';
+ * const { generate, progress, result } = useMediaGeneration({
+ *   endpoint: '/api/generate/video',
+ *   userId: user.id,
+ * });
+ * await generate({ prompt: '...' });
+ * // progress updates automatically from real SSE events
+ * ```
+ */
+export async function generateVideoWithProgress(
+  endpoint: string,
+  request: VideoGenRequest,
+  onProgress: (progress: GenerationProgress) => void,
+  options: MediaGenOptions = {},
+): Promise<VideoGenResponse> {
+  onProgress({ percent: 0, stage: 'Starting video generation...' });
+
+  // Video takes longer; simulate progress more slowly
+  const stopSimulation = simulateProgress(onProgress, 'Rendering video...', 80, 2000);
+
+  try {
+    const result = await generateVideo(endpoint, request, options);
+    onProgress({ percent: 100, stage: 'Complete' });
+    return result;
+  } catch (error) {
+    onProgress({ percent: 0, stage: 'Failed' });
+    throw error;
+  } finally {
+    stopSimulation();
+  }
+}

internal/adapter/templates/templates/skeleton/packages/ai-client/src/textgen.ts

@@ -0,0 +1,226 @@
+/**
+ * Text generation client for calling textgen API endpoints.
+ */
+
+import { TokenUsage, AIResponseMeta, createAIError, ErrorCodes } from './types';
+
+/** Role for chat messages. */
+export type MessageRole = 'system' | 'user' | 'assistant';
+
+/** A single message in a conversation. */
+export interface TextGenMessage {
+  role: MessageRole;
+  content: string;
+}
+
+/** Request for text generation. */
+export interface TextGenRequest {
+  messages: TextGenMessage[];
+  model?: string;
+  maxTokens?: number;
+  temperature?: number;
+  stream?: boolean;
+}
+
+/** Response from text generation. */
+export interface TextGenResponse extends AIResponseMeta {
+  text: string;
+  usage?: TokenUsage;
+  finishReason?: 'stop' | 'length' | 'content_filter';
+}
+
+/** Chunk received during streaming. */
+export interface TextGenChunk {
+  text: string;
+  done: boolean;
+}
+
+/** Options for API calls. */
+export interface TextGenOptions {
+  timeout?: number;
+  headers?: Record<string, string>;
+}
+
+/**
+ * Stream text generation with Server-Sent Events.
+ *
+ * @param endpoint - The textgen API endpoint URL
+ * @param request - The generation request
+ * @param onChunk - Callback for each text chunk received
+ * @param onDone - Callback when generation completes
+ * @param onError - Callback for errors
+ * @param options - Additional options
+ * @returns Abort function to cancel the stream
+ *
+ * @example
+ * ```ts
+ * const abort = streamText(
+ *   '/api/textgen',
+ *   { messages: [{ role: 'user', content: 'Hello!' }] },
+ *   (chunk) => console.log(chunk),
+ *   (response) => console.log('Done:', response),
+ *   (error) => console.error(error)
+ * );
+ *
+ * // Cancel if needed
+ * abort();
+ * ```
+ */
+export function streamText(
+  endpoint: string,
+  request: TextGenRequest,
+  onChunk: (chunk: string) => void,
+  onDone: (response: TextGenResponse) => void,
+  onError: (error: Error) => void,
+  options: TextGenOptions = {}
+): () => void {
+  const controller = new AbortController();
+  const startTime = Date.now();
+
+  const streamRequest = { ...request, stream: true };
+
+  fetch(endpoint, {
+    method: 'POST',
+    headers: {
+      'Content-Type': 'application/json',
+      ...options.headers,
+    },
+    body: JSON.stringify(streamRequest),
+    signal: controller.signal,
+  })
+    .then(async (response) => {
+      if (!response.ok) {
+        const errorData = await response.json().catch(() => ({}));
+        throw createAIError(
+          errorData.message || `HTTP ${response.status}`,
+          response.status === 429 ? ErrorCodes.RATE_LIMITED : ErrorCodes.PROVIDER_ERROR,
+          undefined,
+          response.status === 429 || response.status >= 500
+        );
+      }
+
+      const reader = response.body?.getReader();
+      if (!reader) {
+        throw createAIError('No response body', ErrorCodes.NETWORK_ERROR);
+      }
+
+      const decoder = new TextDecoder();
+      let fullText = '';
+      let provider = '';
+      let usage: TokenUsage | undefined;
+
+      while (true) {
+        const { done, value } = await reader.read();
+        if (done) break;
+
+        const chunk = decoder.decode(value, { stream: true });
+        const lines = chunk.split('\n');
+
+        for (const line of lines) {
+          if (line.startsWith('data: ')) {
+            const data = line.slice(6);
+            if (data === '[DONE]') continue;
+
+            try {
+              const parsed = JSON.parse(data);
+              if (parsed.text) {
+                fullText += parsed.text;
+                onChunk(parsed.text);
+              }
+              if (parsed.provider) provider = parsed.provider;
+              if (parsed.usage) usage = parsed.usage;
+            } catch {
+              // Skip invalid JSON lines
+            }
+          }
+        }
+      }
+
+      onDone({
+        text: fullText,
+        provider,
+        usage,
+        latencyMs: Date.now() - startTime,
+      });
+    })
+    .catch((error) => {
+      if (error.name === 'AbortError') return;
+      onError(error);
+    });
+
+  return () => controller.abort();
+}
+
+/**
+ * Generate text (non-streaming).
+ *
+ * @param endpoint - The textgen API endpoint URL
+ * @param request - The generation request
+ * @param options - Additional options
+ * @returns Promise resolving to the generation response
+ *
+ * @example
+ * ```ts
+ * const response = await generateText('/api/textgen', {
+ *   messages: [
+ *     { role: 'system', content: 'You are a helpful assistant.' },
+ *     { role: 'user', content: 'What is 2 + 2?' }
+ *   ],
+ *   temperature: 0.7,
+ * });
+ * console.log(response.text);
+ * ```
+ */
+export async function generateText(
+  endpoint: string,
+  request: TextGenRequest,
+  options: TextGenOptions = {}
+): Promise<TextGenResponse> {
+  const startTime = Date.now();
+  const nonStreamRequest = { ...request, stream: false };
+
+  const controller = new AbortController();
+  const timeoutId = options.timeout
+    ? setTimeout(() => controller.abort(), options.timeout)
+    : undefined;
+
+  try {
+    const response = await fetch(endpoint, {
+      method: 'POST',
+      headers: {
+        'Content-Type': 'application/json',
+        ...options.headers,
+      },
+      body: JSON.stringify(nonStreamRequest),
+      signal: controller.signal,
+    });
+
+    if (!response.ok) {
+      const errorData = await response.json().catch(() => ({}));
+      throw createAIError(
+        errorData.message || `HTTP ${response.status}`,
+        response.status === 429 ? ErrorCodes.RATE_LIMITED : ErrorCodes.PROVIDER_ERROR,
+        undefined,
+        response.status === 429 || response.status >= 500
+      );
+    }
+
+    const data = await response.json();
+
+    return {
+      text: data.text || '',
+      provider: data.provider || '',
+      model: data.model,
+      usage: data.usage,
+      finishReason: data.finish_reason,
+      latencyMs: Date.now() - startTime,
+    };
+  } catch (error) {
+    if (error instanceof Error && error.name === 'AbortError') {
+      throw createAIError('Request timed out', ErrorCodes.TIMEOUT, undefined, true);
+    }
+    throw error;
+  } finally {
+    if (timeoutId) clearTimeout(timeoutId);
+  }
+}

internal/adapter/templates/templates/skeleton/packages/ai-client/src/types.ts

@@ -0,0 +1,51 @@
+/**
+ * Shared types for AI client operations.
+ */
+
+/** Token usage information from AI providers. */
+export interface TokenUsage {
+  promptTokens: number;
+  completionTokens: number;
+  totalTokens?: number;
+}
+
+/** Base response metadata from any AI operation. */
+export interface AIResponseMeta {
+  provider: string;
+  model?: string;
+  latencyMs: number;
+}
+
+/** Error from AI operations. */
+export interface AIError extends Error {
+  code: string;
+  provider?: string;
+  retryable: boolean;
+}
+
+/** Create a typed AI error. */
+export function createAIError(
+  message: string,
+  code: string,
+  provider?: string,
+  retryable = false
+): AIError {
+  const error = new Error(message) as AIError;
+  error.code = code;
+  error.provider = provider;
+  error.retryable = retryable;
+  return error;
+}
+
+/** Common error codes. */
+export const ErrorCodes = {
+  NETWORK_ERROR: 'NETWORK_ERROR',
+  TIMEOUT: 'TIMEOUT',
+  RATE_LIMITED: 'RATE_LIMITED',
+  INVALID_REQUEST: 'INVALID_REQUEST',
+  PROVIDER_ERROR: 'PROVIDER_ERROR',
+  QUOTA_EXCEEDED: 'QUOTA_EXCEEDED',
+  CONTENT_FILTERED: 'CONTENT_FILTERED',
+} as const;
+
+export type ErrorCode = (typeof ErrorCodes)[keyof typeof ErrorCodes];

internal/adapter/templates/templates/skeleton/packages/ai-client/tsconfig.json

@@ -0,0 +1,19 @@
+{
+  "compilerOptions": {
+    "target": "ES2022",
+    "lib": ["ES2022", "DOM"],
+    "module": "ESNext",
+    "moduleResolution": "bundler",
+    "strict": true,
+    "esModuleInterop": true,
+    "skipLibCheck": true,
+    "forceConsistentCasingInFileNames": true,
+    "declaration": true,
+    "declarationMap": true,
+    "sourceMap": true,
+    "outDir": "dist",
+    "rootDir": "src"
+  },
+  "include": ["src/**/*"],
+  "exclude": ["node_modules", "dist"]
+}

internal/adapter/templates/templates/skeleton/packages/api-client/package.json.tmpl

@@ -5,6 +5,12 @@
   "type": "module",
   "main": "src/index.ts",
   "types": "src/index.ts",
+  "exports": {
+    ".": {
+      "types": "./src/index.ts",
+      "import": "./src/index.ts"
+    }
+  },
   "scripts": {
     "generate": "../scripts/generate-client.sh",
     "typecheck": "tsc --noEmit",

internal/adapter/templates/templates/skeleton/packages/api-client/src/client.ts

@@ -1,3 +1,6 @@
+import type { ApiError } from './types';
+import { ApiClientError } from './errors';
+
 /**
  * API Client Configuration
  */
@@ -6,7 +9,85 @@ export interface ClientConfig {
   apiKey?: string;
   bearerToken?: string;
   headers?: Record<string, string>;
-  onError?: (error: Error) => void;
+  /** Called on any error (after parsing) */
+  onError?: (error: ApiClientError) => void;
+  /** Called specifically on auth errors (401) - useful for triggering logout */
+  onAuthError?: (error: ApiClientError) => void;
+}
+
+/**
+ * Parse error response body into ApiError structure.
+ */
+async function parseErrorResponse(response: Response): Promise<ApiError> {
+  try {
+    const body = await response.json();
+    // Handle wrapped response format { error: { ... } }
+    if (body.error && typeof body.error === 'object') {
+      return {
+        status: response.status,
+        code: body.error.code || 'UNKNOWN_ERROR',
+        message: body.error.message || response.statusText,
+        details: body.error.details,
+      };
+    }
+    // Handle direct error format { code, message, details }
+    if (body.code && body.message) {
+      return {
+        status: response.status,
+        code: body.code,
+        message: body.message,
+        details: body.details,
+      };
+    }
+    // Handle simple message format { message: "..." }
+    if (body.message) {
+      return {
+        status: response.status,
+        code: codeFromStatus(response.status),
+        message: body.message,
+        details: body.details,
+      };
+    }
+    // Fallback
+    return {
+      status: response.status,
+      code: codeFromStatus(response.status),
+      message: response.statusText || 'An error occurred',
+    };
+  } catch {
+    // Failed to parse JSON
+    return {
+      status: response.status,
+      code: codeFromStatus(response.status),
+      message: response.statusText || 'An error occurred',
+    };
+  }
+}
+
+/**
+ * Map HTTP status to error code.
+ */
+function codeFromStatus(status: number): string {
+  switch (status) {
+    case 400:
+      return 'BAD_REQUEST';
+    case 401:
+      return 'UNAUTHORIZED';
+    case 403:
+      return 'FORBIDDEN';
+    case 404:
+      return 'NOT_FOUND';
+    case 409:
+      return 'CONFLICT';
+    case 422:
+      return 'UNPROCESSABLE_ENTITY';
+    case 429:
+      return 'TOO_MANY_REQUESTS';
+    case 503:
+      return 'SERVICE_UNAVAILABLE';
+    default:
+      return status >= 500 ? 'INTERNAL_ERROR' : 'UNKNOWN_ERROR';
+  }
 }
 
 /**
@@ -20,9 +101,20 @@ export interface ClientConfig {
  *
  * const users = await client.get('/users');
  * const newUser = await client.post('/users', { name: 'John' });
+ *
+ * @example
+ * // Handling validation errors
+ * try {
+ *   await client.post('/users', { email: 'invalid' });
+ * } catch (error) {
+ *   if (isApiClientError(error) && error.isValidationError()) {
+ *     const fieldErrors = error.getFieldErrors();
+ *     console.log(fieldErrors.email); // "email must be a valid email address"
+ *   }
+ * }
  */
 export function createClient(config: ClientConfig) {
-  const { baseUrl, apiKey, bearerToken, headers = {}, onError } = config;
+  const { baseUrl, apiKey, bearerToken, headers = {}, onError, onAuthError } = config;
 
   async function request<T>(
     method: string,
@@ -65,10 +157,19 @@ export function createClient(config: ClientConfig) {
     });
 
     if (!response.ok) {
-      const error = new Error(`API error: ${response.status}`);
+      const apiError = await parseErrorResponse(response);
+      const error = new ApiClientError(apiError, response);
+
+      // Call auth error handler for 401s
+      if (error.isAuthError() && onAuthError) {
+        onAuthError(error);
+      }
+
+      // Call general error handler
       if (onError) {
         onError(error);
       }
+
       throw error;
     }
 
@@ -77,19 +178,22 @@ export function createClient(config: ClientConfig) {
       return undefined as T;
     }
 
-    return response.json();
+    const json = await response.json();
+
+    // Handle wrapped response format { data: { ... } }
+    if (json.data !== undefined) {
+      return json.data as T;
+    }
+
+    return json as T;
   }
 
   return {
     get: <T>(path: string, params?: Record<string, string | number | boolean | undefined>) =>
       request<T>('GET', path, { params }),
-    post: <T>(path: string, body?: unknown) =>
+    post: <T>(path: string, body?: unknown) => request<T>('POST', path, { body }),
-      request<T>('POST', path, { body }),
+    put: <T>(path: string, body?: unknown) => request<T>('PUT', path, { body }),
-    put: <T>(path: string, body?: unknown) =>
+    patch: <T>(path: string, body?: unknown) => request<T>('PATCH', path, { body }),
-      request<T>('PUT', path, { body }),
+    delete: <T>(path: string) => request<T>('DELETE', path),
-    patch: <T>(path: string, body?: unknown) =>
-      request<T>('PATCH', path, { body }),
-    delete: <T>(path: string) =>
-      request<T>('DELETE', path),
   };
 }

internal/adapter/templates/templates/skeleton/packages/api-client/src/errors.ts

@@ -0,0 +1,114 @@
+import type { ApiError, ValidationDetail, ErrorCodeType } from './types';
+
+/**
+ * API client error with typed error information.
+ * Provides convenient methods for accessing validation errors.
+ */
+export class ApiClientError extends Error {
+  /** HTTP status code */
+  readonly status: number;
+  /** Machine-readable error code */
+  readonly code: string;
+  /** Optional validation details or other error details */
+  readonly details?: ValidationDetail[] | Record<string, unknown>;
+  /** The original response */
+  readonly response?: Response;
+
+  constructor(apiError: ApiError, response?: Response) {
+    super(apiError.message);
+    this.name = 'ApiClientError';
+    this.status = apiError.status;
+    this.code = apiError.code;
+    this.details = apiError.details;
+    this.response = response;
+
+    // Maintains proper stack trace in V8 environments
+    if ('captureStackTrace' in Error && typeof Error.captureStackTrace === 'function') {
+      Error.captureStackTrace(this, ApiClientError);
+    }
+  }
+
+  /**
+   * Check if this is a validation error.
+   */
+  isValidationError(): boolean {
+    return this.code === 'VALIDATION_ERROR' || this.code === 'BAD_REQUEST';
+  }
+
+  /**
+   * Check if this is an authentication error.
+   */
+  isAuthError(): boolean {
+    return this.status === 401 || this.code === 'UNAUTHORIZED';
+  }
+
+  /**
+   * Check if this is a permission error.
+   */
+  isForbiddenError(): boolean {
+    return this.status === 403 || this.code === 'FORBIDDEN';
+  }
+
+  /**
+   * Check if this is a not found error.
+   */
+  isNotFoundError(): boolean {
+    return this.status === 404 || this.code === 'NOT_FOUND';
+  }
+
+  /**
+   * Check if the error has a specific code.
+   */
+  hasCode(code: ErrorCodeType | string): boolean {
+    return this.code === code;
+  }
+
+  /**
+   * Get validation errors as an array.
+   * Returns empty array if no validation details are present.
+   */
+  getValidationErrors(): ValidationDetail[] {
+    if (!this.details) return [];
+    if (Array.isArray(this.details)) return this.details;
+    return [];
+  }
+
+  /**
+   * Get validation errors mapped by field name.
+   * Useful for displaying errors next to form fields.
+   *
+   * @example
+   * const errors = error.getFieldErrors();
+   * // { email: 'email is required', password: 'password must be at least 8 characters' }
+   */
+  getFieldErrors(): Record<string, string> {
+    const errors: Record<string, string> = {};
+    for (const detail of this.getValidationErrors()) {
+      errors[detail.field] = detail.message;
+    }
+    return errors;
+  }
+
+  /**
+   * Get the error message for a specific field.
+   * Returns undefined if no error for that field.
+   */
+  getFieldError(field: string): string | undefined {
+    const errors = this.getValidationErrors();
+    return errors.find((e) => e.field === field)?.message;
+  }
+
+  /**
+   * Check if a specific field has an error.
+   */
+  hasFieldError(field: string): boolean {
+    return this.getFieldError(field) !== undefined;
+  }
+}
+
+/**
+ * Type guard for ApiClientError.
+ */
+export function isApiClientError(error: unknown): error is ApiClientError {
+  return error instanceof ApiClientError;
+}

internal/adapter/templates/templates/skeleton/packages/api-client/src/index.ts

@@ -1,3 +1,5 @@
 export * from './client';
+export * from './types';
+export * from './errors';
 // Note: schema.d.ts is generated by running `pnpm generate`
 // export type { paths, components, operations } from './schema';

internal/adapter/templates/templates/skeleton/packages/api-client/src/types.ts

@@ -0,0 +1,67 @@
+/**
+ * Validation error detail from the API.
+ * Matches the Go httpvalidation.ValidationDetail struct.
+ */
+export interface ValidationDetail {
+  /** The field name that failed validation */
+  field: string;
+  /** Human-readable error message */
+  message: string;
+}
+
+/**
+ * Standard API error response structure.
+ * Matches the Go httperror.HTTPError struct.
+ */
+export interface ApiError {
+  /** HTTP status code */
+  status: number;
+  /** Machine-readable error code (e.g., "VALIDATION_ERROR", "NOT_FOUND") */
+  code: string;
+  /** Human-readable error message */
+  message: string;
+  /** Optional additional details (validation errors, etc.) */
+  details?: ValidationDetail[] | Record<string, unknown>;
+}
+
+/**
+ * Standard API response wrapper.
+ */
+export interface ApiResponse<T> {
+  data?: T;
+  error?: ApiError;
+}
+
+/**
+ * Common error codes returned by the API.
+ */
+export const ErrorCode = {
+  BAD_REQUEST: 'BAD_REQUEST',
+  VALIDATION_ERROR: 'VALIDATION_ERROR',
+  UNAUTHORIZED: 'UNAUTHORIZED',
+  FORBIDDEN: 'FORBIDDEN',
+  NOT_FOUND: 'NOT_FOUND',
+  CONFLICT: 'CONFLICT',
+  UNPROCESSABLE_ENTITY: 'UNPROCESSABLE_ENTITY',
+  TOO_MANY_REQUESTS: 'TOO_MANY_REQUESTS',
+  INTERNAL_ERROR: 'INTERNAL_ERROR',
+  SERVICE_UNAVAILABLE: 'SERVICE_UNAVAILABLE',
+} as const;
+
+export type ErrorCodeType = (typeof ErrorCode)[keyof typeof ErrorCode];
+
+/**
+ * Auth-related types for login/user responses.
+ */
+export interface AuthUser {
+  id: string;
+  email: string;
+  name?: string;
+  roles?: string[];
+  scopes?: string[];
+}
+
+export interface LoginResponse {
+  token: string;
+  user: AuthUser;
+}

internal/adapter/templates/templates/skeleton/packages/auth/package.json.tmpl

@@ -5,6 +5,12 @@
   "type": "module",
   "main": "src/index.ts",
   "types": "src/index.ts",
+  "exports": {
+    ".": {
+      "types": "./src/index.ts",
+      "import": "./src/index.ts"
+    }
+  },
   "scripts": {
     "typecheck": "tsc --noEmit",
     "build": "tsc",

internal/adapter/templates/templates/skeleton/packages/auth/src/AuthProvider.tsx

@@ -34,8 +34,6 @@ export interface AuthProviderProps {
   loginUrl?: string;
   /** API endpoint for logout */
   logoutUrl?: string;
-  /** API endpoint for fetching current user */
-  userUrl?: string;
   /** Custom login handler */
   onLogin?: (credentials: LoginCredentials) => Promise<{ token: string; user: User }>;
   /** Custom logout handler */
@@ -68,7 +66,6 @@ export function AuthProvider({
   children,
   loginUrl = '/api/auth/login',
   logoutUrl = '/api/auth/logout',
-  userUrl = '/api/auth/me',
   onLogin,
   onLogout,
   storage = 'localStorage',
@@ -140,8 +137,9 @@ export function AuthProvider({
           });
 
           if (!response.ok) {
-            const error = await response.json().catch(() => ({}));
+            const errBody = await response.json().catch(() => ({}));
-            throw new Error(error.message || 'Login failed');
+            const errMsg = errBody.error?.message || errBody.message || 'Login failed';
+            throw new Error(errMsg);
           }
 
           const data = await response.json();

internal/adapter/templates/templates/skeleton/packages/layout/package.json.tmpl

@@ -5,6 +5,12 @@
   "type": "module",
   "main": "src/index.ts",
   "types": "src/index.ts",
+  "exports": {
+    ".": {
+      "types": "./src/index.ts",
+      "import": "./src/index.ts"
+    }
+  },
   "scripts": {
     "typecheck": "tsc --noEmit",
     "build": "tsc",

internal/adapter/templates/templates/skeleton/packages/logger/package.json.tmpl

@@ -5,6 +5,12 @@
   "type": "module",
   "main": "src/index.ts",
   "types": "src/index.ts",
+  "exports": {
+    ".": {
+      "types": "./src/index.ts",
+      "import": "./src/index.ts"
+    }
+  },
   "scripts": {
     "typecheck": "tsc --noEmit",
     "build": "tsc"

internal/adapter/templates/templates/skeleton/packages/realtime/package.json.tmpl

@@ -0,0 +1,28 @@
+{
+  "name": "@{{PROJECT_NAME}}/realtime",
+  "version": "0.0.1",
+  "private": true,
+  "type": "module",
+  "main": "src/index.ts",
+  "types": "src/index.ts",
+  "exports": {
+    ".": {
+      "types": "./src/index.ts",
+      "import": "./src/index.ts"
+    }
+  },
+  "scripts": {
+    "typecheck": "tsc --noEmit",
+    "build": "tsc",
+    "lint": "eslint . --ext ts,tsx --report-unused-disable-directives --max-warnings 0"
+  },
+  "peerDependencies": {
+    "react": "^18.0.0",
+    "react-dom": "^18.0.0"
+  },
+  "devDependencies": {
+    "@types/react": "^18.3.3",
+    "@types/react-dom": "^18.3.0",
+    "typescript": "^5.5.3"
+  }
+}

internal/adapter/templates/templates/skeleton/packages/realtime/src/index.ts

@@ -0,0 +1,5 @@
+export * from './types';
+export { useEventChannel, type ChannelEvent, type SSEState, type UseEventChannelConfig } from './useEventChannel';
+export { useMediaGeneration, type GenerationStatus, type UseMediaGenerationConfig, type UseMediaGenerationResult } from './useMediaGeneration';
+export { useChat, type UseChatConfig, type UseChatResult, type ChatMessage } from './useChat';
+export { useMediaUpload, type UploadProgress, type UploadResult, type UseMediaUploadConfig, type UseMediaUploadResult } from './useMediaUpload';

internal/adapter/templates/templates/skeleton/packages/realtime/src/types.ts

@@ -0,0 +1,79 @@
+/**
+ * SSE event types matching Go realtime constants.
+ */
+export type EventType =
+  | 'chat'
+  | 'ai_chat'
+  | 'ai_chat_chunk'
+  | 'presence'
+  | 'notification'
+  | 'system'
+  | 'error'
+  | 'generation_started'
+  | 'generation_progress'
+  | 'generation_complete'
+  | 'generation_failed'
+  | 'upload_started'
+  | 'upload_progress'
+  | 'upload_complete'
+  | 'upload_failed';
+
+/**
+ * Chat message data payload.
+ */
+export interface ChatData {
+  content: string;
+  userId: string;
+  userName?: string;
+}
+
+/**
+ * AI response data payload.
+ */
+export interface AIResponseData {
+  content: string;
+  provider: string;
+}
+
+/**
+ * AI chunk data payload (for streaming).
+ */
+export interface AIChunkData {
+  streamId: string;
+  text: string;
+  done: boolean;
+  provider?: string;
+}
+
+/**
+ * Presence message data payload.
+ */
+export interface PresenceData {
+  userId: string;
+  userName?: string;
+  status: PresenceStatus;
+}
+
+/**
+ * Presence status values.
+ */
+export type PresenceStatus = 'online' | 'offline' | 'away';
+
+/**
+ * Generation result payload.
+ */
+export interface GenerationResult {
+  url: string;
+  provider: string;
+  latencyMs: number;
+}
+
+/**
+ * Upload result payload.
+ */
+export interface UploadResult {
+  id: string;
+  original: string;
+  optimized: string;
+  thumbnail: string;
+}

internal/adapter/templates/templates/skeleton/packages/realtime/src/useChat.ts

@@ -0,0 +1,296 @@
+'use client';
+
+import { useState, useCallback, useRef } from 'react';
+import { useEventChannel, type ChannelEvent } from './useEventChannel';
+
+/**
+ * Chat message from a user.
+ */
+export interface ChatMessage {
+  id: string;
+  content: string;
+  userId: string;
+  userName?: string;
+  timestamp: string;
+}
+
+/**
+ * AI response message.
+ */
+export interface AIMessage {
+  id: string;
+  content: string;
+  provider: string;
+  timestamp: string;
+}
+
+/**
+ * In-progress AI streaming message.
+ */
+export interface AIStreamingMessage {
+  streamId: string;
+  content: string;
+  isStreaming: boolean;
+  timestamp: string;
+}
+
+/**
+ * Online user in the room.
+ */
+export interface OnlineUser {
+  id: string;
+  name?: string;
+  status: 'online' | 'away' | 'offline';
+}
+
+/**
+ * Configuration for the chat hook.
+ */
+export interface UseChatConfig {
+  /** API endpoint for sending messages (e.g., '/api/chat/messages') */
+  endpoint: string;
+  /** SSE endpoint for events (default: '/api/events') */
+  sseEndpoint?: string;
+  /** Channel to subscribe to (e.g., 'channel:room123') */
+  channel: string;
+  /** Current user ID */
+  userId: string;
+  /** Current user display name */
+  userName?: string;
+  /** Additional headers for POST requests */
+  headers?: Record<string, string>;
+}
+
+/**
+ * Result from the chat hook.
+ */
+export interface UseChatResult {
+  /** List of chat messages */
+  messages: ChatMessage[];
+  /** List of AI responses */
+  aiMessages: AIMessage[];
+  /** Map of in-progress AI streams */
+  streamingMessages: Map<string, AIStreamingMessage>;
+  /** List of online users */
+  onlineUsers: OnlineUser[];
+  /** SSE connection state */
+  connectionState: 'connecting' | 'connected' | 'disconnected' | 'error';
+  /** Send a chat message */
+  sendMessage: (content: string) => Promise<void>;
+  /** Clear all messages */
+  clearMessages: () => void;
+}
+
+/** Maximum message IDs to track for deduplication */
+const MAX_SEEN_MESSAGES = 500;
+
+/**
+ * Chat hook using HTTP2 + SSE pattern.
+ *
+ * - Send messages: HTTP POST to endpoint
+ * - Receive messages: SSE subscription to channel
+ *
+ * @example
+ * ```tsx
+ * const {
+ *   messages,
+ *   aiMessages,
+ *   sendMessage,
+ *   connectionState,
+ * } = useChat({
+ *   endpoint: '/api/chat/messages',
+ *   channel: 'channel:room123',
+ *   userId: currentUser.id,
+ *   userName: currentUser.name,
+ * });
+ *
+ * // Send a message (HTTP POST)
+ * await sendMessage('Hello, world!');
+ *
+ * // Messages arrive via SSE automatically
+ * return (
+ *   <div>
+ *     {messages.map(msg => (
+ *       <ChatBubble key={msg.id} {...msg} />
+ *     ))}
+ *   </div>
+ * );
+ * ```
+ */
+export function useChat(config: UseChatConfig): UseChatResult {
+  const {
+    endpoint,
+    sseEndpoint = '/api/events',
+    channel,
+    userId,
+    userName,
+    headers,
+  } = config;
+
+  const [messages, setMessages] = useState<ChatMessage[]>([]);
+  const [aiMessages, setAIMessages] = useState<AIMessage[]>([]);
+  const [streamingMessages, setStreamingMessages] = useState<Map<string, AIStreamingMessage>>(new Map());
+  const [onlineUsers, setOnlineUsers] = useState<OnlineUser[]>([]);
+
+  const seenMessageIds = useRef(new Set<string>());
+
+  // Handle incoming SSE events
+  const handleEvent = useCallback((event: ChannelEvent) => {
+    const eventId = event.id as string | undefined;
+
+    switch (event.type) {
+      case 'chat': {
+        // Deduplicate
+        if (eventId && seenMessageIds.current.has(eventId)) return;
+        if (eventId) {
+          seenMessageIds.current.add(eventId);
+          if (seenMessageIds.current.size > MAX_SEEN_MESSAGES) {
+            const arr = Array.from(seenMessageIds.current);
+            seenMessageIds.current = new Set(arr.slice(-MAX_SEEN_MESSAGES));
+          }
+        }
+
+        // Chat data may be at top level or nested in event.result (SSEEvent envelope)
+        const payload = (event.result as Record<string, unknown>) || event;
+        const msg: ChatMessage = {
+          id: (payload.id as string) || eventId || crypto.randomUUID(),
+          content: (payload.content as string) || (event.message as string) || '',
+          userId: (payload.userId as string) || 'unknown',
+          userName: payload.userName as string | undefined,
+          timestamp: (payload.timestamp as string) || event.timestamp || new Date().toISOString(),
+        };
+        setMessages((prev) => [...prev, msg]);
+        break;
+      }
+
+      case 'ai_chat_chunk': {
+        // Chunk data may be at top level or nested in event.result (SSEEvent envelope)
+        const chunkPayload = (event.result as Record<string, unknown>) || event;
+        const streamId = (chunkPayload.streamId as string) || (event.streamId as string);
+        const text = (chunkPayload.text as string) || (event.text as string) || '';
+        const done = (chunkPayload.done as boolean) || (event.done as boolean) || false;
+
+        setStreamingMessages((prev) => {
+          const updated = new Map(prev);
+          const existing = updated.get(streamId);
+
+          if (done) {
+            // Move completed stream to aiMessages
+            const finalContent = (existing?.content || '') + text;
+            if (finalContent) {
+              const provider = (chunkPayload.provider as string) || '';
+              setAIMessages((prevAI) => [...prevAI, {
+                id: streamId || crypto.randomUUID(),
+                content: finalContent,
+                provider,
+                timestamp: existing?.timestamp || new Date().toISOString(),
+              }]);
+            }
+            updated.delete(streamId);
+          } else {
+            updated.set(streamId, {
+              streamId,
+              content: (existing?.content || '') + text,
+              isStreaming: true,
+              timestamp: existing?.timestamp || new Date().toISOString(),
+            });
+          }
+          return updated;
+        });
+        break;
+      }
+
+      case 'ai_chat': {
+        // AI chat data may be at top level or nested in event.result (SSEEvent envelope)
+        const aiPayload = (event.result as Record<string, unknown>) || event;
+        const aiId = (aiPayload.id as string) || event.jobId || eventId || crypto.randomUUID();
+
+        // Deduplicate
+        if (seenMessageIds.current.has(aiId)) return;
+        seenMessageIds.current.add(aiId);
+        if (seenMessageIds.current.size > MAX_SEEN_MESSAGES) {
+          const arr = Array.from(seenMessageIds.current);
+          seenMessageIds.current = new Set(arr.slice(-MAX_SEEN_MESSAGES));
+        }
+
+        const aiMsg: AIMessage = {
+          id: aiId,
+          content: (aiPayload.content as string) || (event.message as string) || '',
+          provider: (aiPayload.provider as string) || '',
+          timestamp: (aiPayload.timestamp as string) || event.timestamp || new Date().toISOString(),
+        };
+        setAIMessages((prev) => [...prev, aiMsg]);
+        break;
+      }
+
+      case 'presence': {
+        const presenceUserId = event.userId as string;
+        const status = event.status as 'online' | 'away' | 'offline';
+        const name = event.userName as string | undefined;
+
+        setOnlineUsers((prev) => {
+          if (status === 'offline') {
+            return prev.filter((u) => u.id !== presenceUserId);
+          }
+          const existing = prev.find((u) => u.id === presenceUserId);
+          if (existing) {
+            return prev.map((u) =>
+              u.id === presenceUserId ? { ...u, status, name } : u
+            );
+          }
+          return [...prev, { id: presenceUserId, name, status }];
+        });
+        break;
+      }
+    }
+  }, []);
+
+  // Subscribe to SSE channel
+  const { state: sseState } = useEventChannel({
+    endpoint: sseEndpoint,
+    channel,
+    onEvent: handleEvent,
+  });
+
+  // Send message via HTTP POST
+  const sendMessage = useCallback(
+    async (content: string) => {
+      const response = await fetch(endpoint, {
+        method: 'POST',
+        headers: {
+          'Content-Type': 'application/json',
+          ...headers,
+        },
+        body: JSON.stringify({
+          content,
+          userId,
+          userName,
+          channel,
+        }),
+      });
+
+      if (!response.ok) {
+        const error = await response.json().catch(() => ({}));
+        throw new Error((error as { message?: string }).message || `HTTP ${response.status}`);
+      }
+    },
+    [endpoint, headers, userId, userName, channel]
+  );
+
+  const clearMessages = useCallback(() => {
+    setMessages([]);
+    setAIMessages([]);
+    setStreamingMessages(new Map());
+    seenMessageIds.current.clear();
+  }, []);
+
+  return {
+    messages,
+    aiMessages,
+    streamingMessages,
+    onlineUsers,
+    connectionState: sseState,
+    sendMessage,
+    clearMessages,
+  };
+}

internal/adapter/templates/templates/skeleton/packages/realtime/src/useEventChannel.ts

@@ -0,0 +1,164 @@
+'use client';
+
+import { useEffect, useRef, useCallback, useState } from 'react';
+
+/**
+ * Event from SSE channel.
+ */
+export interface ChannelEvent {
+  type: string;
+  timestamp: string;
+  jobId?: string;
+  progress?: number;
+  message?: string;
+  result?: unknown;
+  error?: string;
+  [key: string]: unknown;
+}
+
+/**
+ * Connection state for SSE.
+ */
+export type SSEState = 'connecting' | 'connected' | 'disconnected' | 'error';
+
+/**
+ * Configuration for useEventChannel hook.
+ */
+export interface UseEventChannelConfig {
+  /** SSE endpoint URL (e.g., '/api/events') */
+  endpoint: string;
+  /** Channel to subscribe to (e.g., 'user:123') */
+  channel: string;
+  /** Handler for incoming events */
+  onEvent?: (event: ChannelEvent) => void;
+  /** Handler for connection state changes */
+  onStateChange?: (state: SSEState) => void;
+  /** Auto-reconnect on disconnect (default: true) */
+  autoReconnect?: boolean;
+  /** Reconnect delay in ms (default: 3000) */
+  reconnectDelay?: number;
+}
+
+/**
+ * Hook for subscribing to SSE event channels.
+ *
+ * Connects to an SSE endpoint and receives events for a specific channel.
+ * Automatically reconnects on disconnect.
+ *
+ * @example
+ * ```tsx
+ * const { state } = useEventChannel({
+ *   endpoint: '/api/events',
+ *   channel: `user:${userId}`,
+ *   onEvent: (event) => {
+ *     switch (event.type) {
+ *       case 'generation_complete':
+ *         setResult(event.result);
+ *         break;
+ *       case 'generation_progress':
+ *         setProgress(event.progress);
+ *         break;
+ *     }
+ *   },
+ * });
+ * ```
+ */
+export function useEventChannel(config: UseEventChannelConfig) {
+  const {
+    endpoint,
+    channel,
+    onEvent,
+    onStateChange,
+    autoReconnect = true,
+    reconnectDelay = 3000,
+  } = config;
+
+  const [state, setState] = useState<SSEState>('disconnected');
+  const eventSourceRef = useRef<EventSource | null>(null);
+  const reconnectTimeoutRef = useRef<ReturnType<typeof setTimeout>>();
+
+  // Refs for callbacks to avoid stale closures
+  const onEventRef = useRef(onEvent);
+  const onStateChangeRef = useRef(onStateChange);
+
+  useEffect(() => {
+    onEventRef.current = onEvent;
+    onStateChangeRef.current = onStateChange;
+  });
+
+  const updateState = useCallback((newState: SSEState) => {
+    setState(newState);
+    onStateChangeRef.current?.(newState);
+  }, []);
+
+  const connect = useCallback(() => {
+    // Close existing connection
+    if (eventSourceRef.current) {
+      eventSourceRef.current.close();
+    }
+
+    // Build URL with channel parameter
+    const url = new URL(endpoint, window.location.origin);
+    url.searchParams.set('channel', channel);
+
+    updateState('connecting');
+
+    const eventSource = new EventSource(url.toString());
+    eventSourceRef.current = eventSource;
+
+    eventSource.onopen = () => {
+      updateState('connected');
+    };
+
+    eventSource.onmessage = (event) => {
+      try {
+        const data = JSON.parse(event.data) as ChannelEvent;
+        onEventRef.current?.(data);
+      } catch {
+        console.error('[SSE] Failed to parse event:', event.data);
+      }
+    };
+
+    eventSource.onerror = () => {
+      updateState('error');
+      eventSource.close();
+      eventSourceRef.current = null;
+
+      // Reconnect if enabled
+      if (autoReconnect) {
+        reconnectTimeoutRef.current = setTimeout(() => {
+          connect();
+        }, reconnectDelay);
+      }
+    };
+  }, [endpoint, channel, autoReconnect, reconnectDelay, updateState]);
+
+  const disconnect = useCallback(() => {
+    if (reconnectTimeoutRef.current) {
+      clearTimeout(reconnectTimeoutRef.current);
+    }
+    if (eventSourceRef.current) {
+      eventSourceRef.current.close();
+      eventSourceRef.current = null;
+    }
+    updateState('disconnected');
+  }, [updateState]);
+
+  // Connect on mount, disconnect on unmount
+  useEffect(() => {
+    connect();
+
+    return () => {
+      disconnect();
+    };
+  }, [connect, disconnect]);
+
+  return {
+    /** Current connection state */
+    state,
+    /** Manually reconnect */
+    reconnect: connect,
+    /** Manually disconnect */
+    disconnect,
+  };
+}

internal/adapter/templates/templates/skeleton/packages/realtime/src/useMediaGeneration.ts

@@ -0,0 +1,241 @@
+'use client';
+
+import { useState, useCallback, useRef } from 'react';
+import { useEventChannel, type ChannelEvent, type SSEState } from './useEventChannel';
+
+/**
+ * Generation status.
+ */
+export type GenerationStatus = 'idle' | 'pending' | 'generating' | 'complete' | 'failed';
+
+/**
+ * Configuration for media generation hook.
+ */
+export interface UseMediaGenerationConfig {
+  /** API endpoint for starting the generation job */
+  endpoint: string;
+  /** SSE endpoint for events (default: '/api/events') */
+  sseEndpoint?: string;
+  /** User ID for subscribing to user channel */
+  userId: string;
+  /** Additional headers for the POST request */
+  headers?: Record<string, string>;
+}
+
+/**
+ * Result from the media generation hook.
+ */
+export interface UseMediaGenerationResult<T> {
+  /** Current generation status */
+  status: GenerationStatus;
+  /** SSE connection state */
+  sseState: SSEState;
+  /** Progress percentage (0-100) */
+  progress: number;
+  /** Current stage message */
+  message: string;
+  /** Generated result (available when status is 'complete') */
+  result: T | null;
+  /** Error message (available when status is 'failed') */
+  error: string | null;
+  /** Job ID of the current generation */
+  jobId: string | null;
+  /** Start a new generation */
+  generate: (request: Record<string, unknown>) => Promise<void>;
+  /** Reset state for a new generation */
+  reset: () => void;
+}
+
+/**
+ * Hook for async media generation with real-time progress via SSE events.
+ *
+ * This hook implements the correct async pattern:
+ * 1. POST to endpoint returns job ID immediately
+ * 2. Progress and results come via SSE events
+ * 3. Never blocks waiting for generation
+ *
+ * @example
+ * ```tsx
+ * const {
+ *   status,
+ *   progress,
+ *   message,
+ *   result,
+ *   error,
+ *   generate,
+ * } = useMediaGeneration<VideoResult>({
+ *   endpoint: '/api/generate/video',
+ *   userId: currentUser.id,
+ * });
+ *
+ * // Start generation
+ * await generate({ prompt: 'A cat playing piano', aspectRatio: '16:9' });
+ *
+ * // UI automatically updates as SSE events arrive
+ * return (
+ *   <div>
+ *     <p>Status: {status}</p>
+ *     <p>Progress: {progress}%</p>
+ *     <p>{message}</p>
+ *     {result && <video src={result.url} />}
+ *     {error && <p className="error">{error}</p>}
+ *   </div>
+ * );
+ * ```
+ */
+export function useMediaGeneration<T = unknown>(
+  config: UseMediaGenerationConfig
+): UseMediaGenerationResult<T> {
+  const {
+    endpoint,
+    sseEndpoint = '/api/events',
+    userId,
+    headers,
+  } = config;
+
+  const [status, setStatus] = useState<GenerationStatus>('idle');
+  const [progress, setProgress] = useState(0);
+  const [message, setMessage] = useState('');
+  const [result, setResult] = useState<T | null>(null);
+  const [error, setError] = useState<string | null>(null);
+  const [jobId, setJobId] = useState<string | null>(null);
+
+  const jobIdRef = useRef<string | null>(null);
+
+  // Handle incoming SSE events
+  const handleEvent = useCallback((event: ChannelEvent) => {
+    // Only handle events for our job
+    if (!jobIdRef.current || event.jobId !== jobIdRef.current) return;
+
+    switch (event.type) {
+      case 'generation_started':
+        setStatus('generating');
+        setProgress(0);
+        setMessage('Starting generation...');
+        break;
+
+      case 'generation_progress':
+        setProgress(event.progress ?? 0);
+        setMessage(event.message ?? 'Generating...');
+        break;
+
+      case 'generation_complete':
+        setStatus('complete');
+        setProgress(100);
+        setMessage('Complete');
+        setResult(event.result as T);
+        break;
+
+      case 'generation_failed':
+        setStatus('failed');
+        setError(event.error ?? 'Generation failed');
+        break;
+
+      case 'upload_started':
+        setStatus('generating');
+        setProgress(0);
+        setMessage('Starting upload...');
+        break;
+
+      case 'upload_progress':
+        setProgress(event.progress ?? 0);
+        setMessage('Uploading...');
+        break;
+
+      case 'upload_complete':
+        setStatus('complete');
+        setProgress(100);
+        setMessage('Complete');
+        setResult(event.result as T);
+        break;
+
+      case 'upload_failed':
+        setStatus('failed');
+        setError(event.error ?? 'Upload failed');
+        break;
+    }
+  }, []);
+
+  // Subscribe to SSE events for user channel
+  const { state: sseState } = useEventChannel({
+    endpoint: sseEndpoint,
+    channel: `user:${userId}`,
+    onEvent: handleEvent,
+  });
+
+  const generate = useCallback(
+    async (request: Record<string, unknown>) => {
+      // Reset state
+      setStatus('pending');
+      setProgress(0);
+      setMessage('Submitting job...');
+      setResult(null);
+      setError(null);
+
+      try {
+        const response = await fetch(endpoint, {
+          method: 'POST',
+          headers: {
+            'Content-Type': 'application/json',
+            ...headers,
+          },
+          body: JSON.stringify(request),
+        });
+
+        if (!response.ok) {
+          const errorData = await response.json().catch(() => ({}));
+          throw new Error(
+            (errorData as { message?: string }).message || `HTTP ${response.status}`
+          );
+        }
+
+        // Handle both envelope-wrapped ({ data: { jobId } }) and direct ({ jobId }) responses
+        const json = (await response.json()) as
+          | { data: { jobId: string }; meta?: unknown }
+          | { jobId: string };
+        const resolvedJobId =
+          'data' in json && json.data?.jobId
+            ? json.data.jobId
+            : (json as { jobId: string }).jobId;
+
+        if (!resolvedJobId) {
+          throw new Error('No job ID returned from server');
+        }
+
+        // Store job ID and transition to generating immediately.
+        // The 202 response confirms the job was accepted — don't wait for
+        // the generation_started SSE event to leave the 'pending' state.
+        jobIdRef.current = resolvedJobId;
+        setJobId(resolvedJobId);
+        setStatus('generating');
+        setMessage('Generation in progress...');
+      } catch (err) {
+        setStatus('failed');
+        setError(err instanceof Error ? err.message : 'Failed to start generation');
+      }
+    },
+    [endpoint, headers]
+  );
+
+  const reset = useCallback(() => {
+    jobIdRef.current = null;
+    setJobId(null);
+    setStatus('idle');
+    setProgress(0);
+    setMessage('');
+    setResult(null);
+    setError(null);
+  }, []);
+
+  return {
+    status,
+    sseState,
+    progress,
+    message,
+    result,
+    error,
+    jobId,
+    generate,
+    reset,
+  };
+}

internal/adapter/templates/templates/skeleton/packages/realtime/src/useMediaUpload.ts

@@ -0,0 +1,141 @@
+import { useState, useCallback } from 'react';
+
+export interface UploadProgress {
+  loaded: number;
+  total: number;
+  percent: number;
+}
+
+export interface UploadResult {
+  url: string;
+  path: string;
+}
+
+export interface UseMediaUploadConfig {
+  /** Base URL for the API (e.g., "" or "http://localhost:8001") */
+  apiPrefix?: string;
+  /** Service name for API path (e.g., "example-api") */
+  serviceName: string;
+  /** Auth headers to include in API calls */
+  headers?: Record<string, string>;
+  /** Called on progress updates during the upload to GCS */
+  onProgress?: (progress: UploadProgress) => void;
+}
+
+export interface UseMediaUploadResult {
+  /** Upload a file. Returns the final URL on success. */
+  upload: (file: File) => Promise<UploadResult>;
+  /** Whether an upload is in progress */
+  isUploading: boolean;
+  /** Current upload progress (0-100) */
+  progress: number;
+  /** Error message if upload failed */
+  error: string | null;
+  /** Reset state */
+  reset: () => void;
+}
+
+export function useMediaUpload(config: UseMediaUploadConfig): UseMediaUploadResult {
+  const [isUploading, setIsUploading] = useState(false);
+  const [progress, setProgress] = useState(0);
+  const [error, setError] = useState<string | null>(null);
+
+  const reset = useCallback(() => {
+    setIsUploading(false);
+    setProgress(0);
+    setError(null);
+  }, []);
+
+  const upload = useCallback(async (file: File): Promise<UploadResult> => {
+    setIsUploading(true);
+    setProgress(0);
+    setError(null);
+
+    try {
+      // 1. Get presigned upload URL from backend
+      const initRes = await fetch(
+        `${config.apiPrefix || ''}/api/${config.serviceName}/media/upload/init`,
+        {
+          method: 'POST',
+          headers: {
+            'Content-Type': 'application/json',
+            ...config.headers,
+          },
+          body: JSON.stringify({
+            filename: file.name,
+            contentType: file.type,
+          }),
+        }
+      );
+
+      if (!initRes.ok) {
+        throw new Error(`Upload init failed: ${initRes.status}`);
+      }
+
+      const { uploadURL, objectPath, headers: uploadHeaders } = await initRes.json();
+
+      // 2. Upload directly to storage using XHR (for progress tracking)
+      await new Promise<void>((resolve, reject) => {
+        const xhr = new XMLHttpRequest();
+        xhr.open('PUT', uploadURL);
+        xhr.timeout = 5 * 60 * 1000; // 5 minutes for large uploads
+
+        // Set headers from presigned response
+        if (uploadHeaders) {
+          Object.entries(uploadHeaders).forEach(([key, value]) => {
+            xhr.setRequestHeader(key, value as string);
+          });
+        }
+
+        xhr.upload.onprogress = (e) => {
+          if (e.lengthComputable) {
+            const pct = Math.round((e.loaded / e.total) * 100);
+            setProgress(pct);
+            config.onProgress?.({ loaded: e.loaded, total: e.total, percent: pct });
+          }
+        };
+
+        xhr.onload = () => {
+          if (xhr.status >= 200 && xhr.status < 300) {
+            resolve();
+          } else {
+            reject(new Error(`Upload failed: ${xhr.status}`));
+          }
+        };
+
+        xhr.onerror = () => reject(new Error('Upload failed: network error'));
+        xhr.ontimeout = () => reject(new Error('Upload timed out'));
+        xhr.send(file);
+      });
+
+      // 3. Confirm upload with backend
+      const completeRes = await fetch(
+        `${config.apiPrefix || ''}/api/${config.serviceName}/media/upload/complete`,
+        {
+          method: 'POST',
+          headers: {
+            'Content-Type': 'application/json',
+            ...config.headers,
+          },
+          body: JSON.stringify({ objectPath }),
+        }
+      );
+
+      if (!completeRes.ok) {
+        throw new Error(`Upload complete failed: ${completeRes.status}`);
+      }
+
+      const result = await completeRes.json();
+      setProgress(100);
+      setIsUploading(false);
+      return { url: result.url, path: result.path };
+    } catch (err) {
+      const msg = err instanceof Error ? err.message : 'Upload failed';
+      setError(msg);
+      setIsUploading(false);
+      throw err;
+    }
+  }, [config]);
+
+  return { upload, isUploading, progress, error, reset };
+}

internal/adapter/templates/templates/skeleton/packages/realtime/tsconfig.json

@@ -0,0 +1,20 @@
+{
+  "compilerOptions": {
+    "target": "ES2020",
+    "useDefineForClassFields": true,
+    "lib": ["ES2020", "DOM", "DOM.Iterable"],
+    "module": "ESNext",
+    "skipLibCheck": true,
+    "moduleResolution": "bundler",
+    "allowImportingTsExtensions": true,
+    "resolveJsonModule": true,
+    "isolatedModules": true,
+    "noEmit": true,
+    "jsx": "react-jsx",
+    "strict": true,
+    "noUnusedLocals": true,
+    "noUnusedParameters": true,
+    "noFallthroughCasesInSwitch": true
+  },
+  "include": ["src"]
+}

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

@@ -0,0 +1,75 @@
+import * as React from 'react';
+import { cva, type VariantProps } from 'class-variance-authority';
+import { cn } from '../utils/cn';
+
+const chatBubbleVariants = cva(
+  'relative max-w-[85%] rounded-2xl px-4 py-3 text-sm leading-relaxed',
+  {
+    variants: {
+      role: {
+        user: 'ml-auto bg-[var(--accent)] text-[var(--accent-foreground)] rounded-br-sm',
+        assistant: 'mr-auto bg-[var(--surface-200)] text-[var(--text-primary)] rounded-bl-sm',
+        system: 'mx-auto bg-[var(--surface-100)] text-[var(--text-secondary)] text-xs italic max-w-[90%]',
+      },
+    },
+    defaultVariants: {
+      role: 'assistant',
+    },
+  }
+);
+
+export interface ChatBubbleProps
+  extends Omit<React.HTMLAttributes<HTMLDivElement>, 'role'>,
+    VariantProps<typeof chatBubbleVariants> {
+  content: string;
+  timestamp?: Date;
+  avatar?: string;
+  isStreaming?: boolean;
+}
+
+const ChatBubble = React.forwardRef<HTMLDivElement, ChatBubbleProps>(
+  ({ className, role, content, timestamp, avatar, isStreaming, ...props }, ref) => {
+    const isUser = role === 'user';
+
+    return (
+      <div
+        ref={ref}
+        className={cn(
+          'flex items-end gap-2',
+          isUser ? 'flex-row-reverse' : 'flex-row',
+          className
+        )}
+        {...props}
+      >
+        {avatar && role !== 'system' && (
+          <div className="flex-shrink-0 w-8 h-8 rounded-full bg-[var(--surface-300)] overflow-hidden">
+            <img src={avatar} alt="" className="w-full h-full object-cover" />
+          </div>
+        )}
+
+        <div className="flex flex-col gap-1">
+          <div className={cn(chatBubbleVariants({ role }))}>
+            <span className="whitespace-pre-wrap">{content}</span>
+            {isStreaming && (
+              <span className="inline-block w-2 h-4 ml-1 bg-current animate-pulse" />
+            )}
+          </div>
+
+          {timestamp && (
+            <span
+              className={cn(
+                'text-xs text-[var(--text-muted)]',
+                isUser ? 'text-right' : 'text-left'
+              )}
+            >
+              {timestamp.toLocaleTimeString([], { hour: '2-digit', minute: '2-digit' })}
+            </span>
+          )}
+        </div>
+      </div>
+    );
+  }
+);
+ChatBubble.displayName = 'ChatBubble';
+
+export { ChatBubble, chatBubbleVariants };

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

@@ -0,0 +1,118 @@
+import * as React from 'react';
+import { cn } from '../utils/cn';
+import { Send } from 'lucide-react';
+
+export interface ChatInputProps {
+  placeholder?: string;
+  maxLength?: number;
+  disabled?: boolean;
+  /** Called when the user submits a message. Return false to indicate failure. */
+  onSubmit: (text: string) => boolean | void;
+  /** Called when send fails (onSubmit returns false). */
+  onSendError?: (text: string) => void;
+  className?: string;
+}
+
+const ChatInput = React.forwardRef<HTMLDivElement, ChatInputProps>(
+  ({ placeholder = 'Type a message...', maxLength = 4000, disabled, onSubmit, onSendError, className }, ref) => {
+    const [value, setValue] = React.useState('');
+    const textareaRef = React.useRef<HTMLTextAreaElement>(null);
+
+    const adjustHeight = React.useCallback(() => {
+      const textarea = textareaRef.current;
+      if (textarea) {
+        textarea.style.height = 'auto';
+        textarea.style.height = `${Math.min(textarea.scrollHeight, 200)}px`;
+      }
+    }, []);
+
+    React.useEffect(() => {
+      adjustHeight();
+    }, [value, adjustHeight]);
+
+    const handleSubmit = React.useCallback(() => {
+      const trimmed = value.trim();
+      if (trimmed && !disabled) {
+        const result = onSubmit(trimmed);
+        if (result === false) {
+          // Send failed - notify via callback
+          onSendError?.(trimmed);
+          return;
+        }
+        setValue('');
+        if (textareaRef.current) {
+          textareaRef.current.style.height = 'auto';
+        }
+      }
+    }, [value, disabled, onSubmit, onSendError]);
+
+    const handleKeyDown = React.useCallback(
+      (e: React.KeyboardEvent<HTMLTextAreaElement>) => {
+        if (e.key === 'Enter' && (e.metaKey || e.ctrlKey)) {
+          e.preventDefault();
+          handleSubmit();
+        }
+      },
+      [handleSubmit]
+    );
+
+    const charCount = value.length;
+    const isNearLimit = charCount > maxLength * 0.9;
+    const isOverLimit = charCount > maxLength;
+
+    return (
+      <div
+        ref={ref}
+        className={cn(
+          'relative flex items-end gap-2 rounded-xl border border-[var(--border)] bg-[var(--surface-100)] p-2 transition-colors focus-within:border-[var(--accent)]',
+          disabled && 'opacity-50 cursor-not-allowed',
+          className
+        )}
+      >
+        <textarea
+          ref={textareaRef}
+          value={value}
+          onChange={(e) => setValue(e.target.value)}
+          onKeyDown={handleKeyDown}
+          placeholder={placeholder}
+          disabled={disabled}
+          rows={1}
+          className={cn(
+            'flex-1 resize-none bg-transparent text-sm text-[var(--text-primary)] placeholder:text-[var(--text-muted)] focus:outline-none disabled:cursor-not-allowed',
+            'min-h-[24px] max-h-[200px] py-1 px-2'
+          )}
+        />
+
+        <div className="flex items-center gap-2 flex-shrink-0">
+          {isNearLimit && (
+            <span
+              className={cn(
+                'text-xs tabular-nums',
+                isOverLimit ? 'text-[var(--error)]' : 'text-[var(--text-muted)]'
+              )}
+            >
+              {charCount}/{maxLength}
+            </span>
+          )}
+
+          <button
+            type="button"
+            onClick={handleSubmit}
+            disabled={disabled || !value.trim() || isOverLimit}
+            className={cn(
+              'flex items-center justify-center w-8 h-8 rounded-lg transition-colors',
+              'bg-[var(--accent)] text-[var(--accent-foreground)]',
+              'hover:bg-[var(--accent-hover)]',
+              'disabled:opacity-50 disabled:cursor-not-allowed disabled:hover:bg-[var(--accent)]'
+            )}
+          >
+            <Send className="w-4 h-4" />
+          </button>
+        </div>
+      </div>
+    );
+  }
+);
+ChatInput.displayName = 'ChatInput';
+
+export { ChatInput };

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

@@ -0,0 +1,43 @@
+import * as React from 'react';
+import { cn } from '../utils/cn';
+import { AlertCircle } from 'lucide-react';
+
+export interface ErrorMessageProps extends React.HTMLAttributes<HTMLParagraphElement> {
+  /** The error message to display */
+  message?: string;
+  /** Show an icon alongside the message */
+  showIcon?: boolean;
+}
+
+/**
+ * ErrorMessage displays a field-level error message.
+ *
+ * @example
+ * <ErrorMessage message="Email is required" />
+ *
+ * @example
+ * <ErrorMessage message={errors.email} showIcon />
+ */
+const ErrorMessage = React.forwardRef<HTMLParagraphElement, ErrorMessageProps>(
+  ({ className, message, showIcon = false, ...props }, ref) => {
+    if (!message) return null;
+
+    return (
+      <p
+        ref={ref}
+        role="alert"
+        className={cn(
+          'flex items-center gap-1.5 text-sm text-[var(--error)] mt-1.5',
+          className
+        )}
+        {...props}
+      >
+        {showIcon && <AlertCircle className="h-3.5 w-3.5 flex-shrink-0" />}
+        <span>{message}</span>
+      </p>
+    );
+  }
+);
+ErrorMessage.displayName = 'ErrorMessage';
+
+export { ErrorMessage };

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

@@ -0,0 +1,93 @@
+import * as React from 'react';
+import { cn } from '../utils/cn';
+import { Label } from './Label';
+import { Input, type InputProps } from './Input';
+import { ErrorMessage } from './ErrorMessage';
+
+export interface FormFieldProps extends Omit<InputProps, 'id' | 'error'> {
+  /** Field label text */
+  label: string;
+  /** Field name (used for id and error lookup) */
+  name: string;
+  /** Error message for this field */
+  error?: string;
+  /** Additional description text */
+  description?: string;
+  /** Whether the field is required */
+  required?: boolean;
+  /** Container className */
+  containerClassName?: string;
+}
+
+/**
+ * FormField is a compound component that combines Label, Input, and ErrorMessage.
+ *
+ * @example
+ * <FormField
+ *   label="Email"
+ *   name="email"
+ *   type="email"
+ *   error={errors.email}
+ *   required
+ * />
+ *
+ * @example
+ * <FormField
+ *   label="Password"
+ *   name="password"
+ *   type="password"
+ *   description="Must be at least 8 characters"
+ *   error={errors.password}
+ * />
+ */
+const FormField = React.forwardRef<HTMLInputElement, FormFieldProps>(
+  (
+    {
+      label,
+      name,
+      error,
+      description,
+      required,
+      containerClassName,
+      className,
+      ...inputProps
+    },
+    ref
+  ) => {
+    const id = `field-${name}`;
+    const descriptionId = description ? `${id}-description` : undefined;
+    const errorId = error ? `${id}-error` : undefined;
+
+    return (
+      <div className={cn('space-y-1.5', containerClassName)}>
+        <Label htmlFor={id} className="flex items-center gap-1">
+          {label}
+          {required && <span className="text-[var(--error)]">*</span>}
+        </Label>
+
+        {description && (
+          <p id={descriptionId} className="text-sm text-[var(--text-muted)]">
+            {description}
+          </p>
+        )}
+
+        <Input
+          ref={ref}
+          id={id}
+          name={name}
+          error={!!error}
+          aria-describedby={descriptionId}
+          aria-invalid={!!error}
+          aria-errormessage={errorId}
+          className={className}
+          {...inputProps}
+        />
+
+        <ErrorMessage id={errorId} message={error} />
+      </div>
+    );
+  }
+);
+FormField.displayName = 'FormField';
+
+export { FormField };

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

@@ -0,0 +1,45 @@
+import * as React from 'react';
+import { cn } from '../utils/cn';
+
+export interface GenerationProgressProps extends React.HTMLAttributes<HTMLDivElement> {
+  /** Optional: 0-100. When provided, shows determinate progress. Otherwise indeterminate. */
+  percent?: number;
+}
+
+const GenerationProgress = React.forwardRef<HTMLDivElement, GenerationProgressProps>(
+  ({ percent, className, ...props }, ref) => {
+    const isDeterminate = percent != null && percent > 0;
+    const clampedPercent = isDeterminate ? Math.max(0, Math.min(100, percent)) : 0;
+
+    return (
+      <div
+        ref={ref}
+        className={cn('relative h-1 rounded-full bg-[var(--surface-200)] overflow-hidden', className)}
+        role="progressbar"
+        aria-valuenow={isDeterminate ? clampedPercent : undefined}
+        aria-valuemin={0}
+        aria-valuemax={100}
+        {...props}
+      >
+        {isDeterminate ? (
+          <div
+            className="absolute inset-y-0 left-0 rounded-full bg-[var(--accent)] transition-all duration-300 ease-out"
+            style={{ width: `${clampedPercent}%` }}
+          />
+        ) : (
+          <div className="absolute inset-y-0 w-1/3 rounded-full bg-[var(--accent)] animate-[indeterminate_1.4s_ease-in-out_infinite]" />
+        )}
+
+        <style>{`
+          @keyframes indeterminate {
+            0% { left: -33%; }
+            100% { left: 100%; }
+          }
+        `}</style>
+      </div>
+    );
+  }
+);
+GenerationProgress.displayName = 'GenerationProgress';
+
+export { GenerationProgress };

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

@@ -0,0 +1,90 @@
+import * as React from 'react';
+import { cn } from '../utils/cn';
+
+export interface ImageGridImage {
+  src: string;
+  alt?: string;
+}
+
+export interface ImageGridProps extends React.HTMLAttributes<HTMLDivElement> {
+  images: ImageGridImage[];
+  columns?: 1 | 2 | 3 | 4;
+  onImageClick?: (index: number) => void;
+  loading?: boolean;
+}
+
+const columnClasses = {
+  1: 'grid-cols-1',
+  2: 'grid-cols-2',
+  3: 'grid-cols-3',
+  4: 'grid-cols-4',
+} as const;
+
+const ImageGrid = React.forwardRef<HTMLDivElement, ImageGridProps>(
+  ({ images, columns = 2, onImageClick, loading, className, ...props }, ref) => {
+    const [loadedImages, setLoadedImages] = React.useState<Set<number>>(new Set());
+
+    const handleImageLoad = React.useCallback((index: number) => {
+      setLoadedImages((prev) => new Set(prev).add(index));
+    }, []);
+
+    if (loading) {
+      return (
+        <div
+          ref={ref}
+          className={cn('grid gap-2', columnClasses[columns], className)}
+          {...props}
+        >
+          {Array.from({ length: columns * 2 }).map((_, i) => (
+            <div
+              key={i}
+              className="aspect-square rounded-lg bg-[var(--surface-200)] animate-pulse"
+            />
+          ))}
+        </div>
+      );
+    }
+
+    if (images.length === 0) {
+      return null;
+    }
+
+    return (
+      <div
+        ref={ref}
+        className={cn('grid gap-2', columnClasses[columns], className)}
+        {...props}
+      >
+        {images.map((image, index) => (
+          <button
+            key={index}
+            type="button"
+            onClick={() => onImageClick?.(index)}
+            disabled={!onImageClick}
+            className={cn(
+              'relative aspect-square rounded-lg overflow-hidden bg-[var(--surface-200)]',
+              'focus:outline-none focus:ring-2 focus:ring-[var(--accent)] focus:ring-offset-2 focus:ring-offset-[var(--background)]',
+              onImageClick && 'cursor-pointer hover:opacity-90 transition-opacity'
+            )}
+          >
+            {!loadedImages.has(index) && (
+              <div className="absolute inset-0 animate-pulse bg-[var(--surface-300)]" />
+            )}
+            <img
+              src={image.src}
+              alt={image.alt || `Generated image ${index + 1}`}
+              className={cn(
+                'w-full h-full object-cover transition-opacity duration-300',
+                loadedImages.has(index) ? 'opacity-100' : 'opacity-0'
+              )}
+              onLoad={() => handleImageLoad(index)}
+            />
+          </button>
+        ))}
+      </div>
+    );
+  }
+);
+ImageGrid.displayName = 'ImageGrid';
+
+export { ImageGrid };

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

@@ -0,0 +1,144 @@
+import { useState } from 'react';
+import { Trash2, Image, Video, ExternalLink } from 'lucide-react';
+
+export interface MediaItem {
+  path: string;
+  url: string;
+  contentType: string;
+  size: number;
+  createdAt: string;
+}
+
+export interface MediaLibraryProps {
+  /** Media items to display */
+  items: MediaItem[];
+  /** Called when a media item is deleted */
+  onDelete?: (path: string) => void;
+  /** Called when a media item is selected */
+  onSelect?: (item: MediaItem) => void;
+  /** Whether delete operations are in progress */
+  isDeleting?: boolean;
+  /** Show empty state */
+  emptyMessage?: string;
+}
+
+function formatSize(bytes: number): string {
+  if (bytes < 1024) return `${bytes} B`;
+  if (bytes < 1024 * 1024) return `${(bytes / 1024).toFixed(1)} KB`;
+  return `${(bytes / (1024 * 1024)).toFixed(1)} MB`;
+}
+
+function isImage(contentType: string): boolean {
+  return contentType.startsWith('image/');
+}
+
+function isVideo(contentType: string): boolean {
+  return contentType.startsWith('video/');
+}
+
+export function MediaLibrary({
+  items,
+  onDelete,
+  onSelect,
+  isDeleting = false,
+  emptyMessage = 'No media files yet. Upload or generate some!',
+}: MediaLibraryProps) {
+  const [selectedPath, setSelectedPath] = useState<string | null>(null);
+
+  if (items.length === 0) {
+    return (
+      <div className="text-center py-12 text-[var(--text-muted)]">
+        <Image className="h-12 w-12 mx-auto mb-3 opacity-50" />
+        <p className="text-sm">{emptyMessage}</p>
+      </div>
+    );
+  }
+
+  return (
+    <div className="grid grid-cols-2 md:grid-cols-3 lg:grid-cols-4 gap-4">
+      {items.map((item) => (
+        <div
+          key={item.path}
+          className={`
+            group relative rounded-lg overflow-hidden border transition-colors cursor-pointer
+            ${selectedPath === item.path
+              ? 'border-[var(--accent)] ring-2 ring-[var(--accent)]/20'
+              : 'border-[var(--border-muted)] hover:border-[var(--border-default)]'}
+          `}
+          onClick={() => {
+            setSelectedPath(item.path);
+            onSelect?.(item);
+          }}
+        >
+          {/* Preview */}
+          <div className="aspect-square bg-[var(--surface-200)] flex items-center justify-center overflow-hidden">
+            {isImage(item.contentType) ? (
+              <img
+                src={item.url}
+                alt={item.path.split('/').pop() || ''}
+                className="w-full h-full object-cover"
+                loading="lazy"
+              />
+            ) : isVideo(item.contentType) ? (
+              <video
+                src={item.url}
+                className="w-full h-full object-cover"
+                muted
+                playsInline
+                onMouseOver={(e) => (e.target as HTMLVideoElement).play()}
+                onMouseOut={(e) => { const v = e.target as HTMLVideoElement; v.pause(); v.currentTime = 0; }}
+              />
+            ) : (
+              <div className="text-[var(--text-muted)]">
+                <Image className="h-8 w-8" />
+              </div>
+            )}
+          </div>
+
+          {/* Info bar */}
+          <div className="p-2 bg-[var(--surface-100)]">
+            <p className="text-xs text-[var(--text-primary)] truncate">
+              {item.path.split('/').pop()}
+            </p>
+            <p className="text-xs text-[var(--text-muted)]">
+              {formatSize(item.size)}
+            </p>
+          </div>
+
+          {/* Overlay actions */}
+          <div className="absolute top-2 right-2 flex gap-1 opacity-0 group-hover:opacity-100 transition-opacity">
+            <a
+              href={item.url}
+              target="_blank"
+              rel="noopener noreferrer"
+              onClick={(e) => e.stopPropagation()}
+              className="p-1.5 rounded bg-black/60 text-white hover:bg-black/80"
+              title="Open in new tab"
+            >
+              <ExternalLink className="h-3.5 w-3.5" />
+            </a>
+            {onDelete && (
+              <button
+                onClick={(e) => { e.stopPropagation(); onDelete(item.path); }}
+                disabled={isDeleting}
+                className="p-1.5 rounded bg-black/60 text-white hover:bg-red-600"
+                title="Delete"
+              >
+                <Trash2 className="h-3.5 w-3.5" />
+              </button>
+            )}
+          </div>
+
+          {/* Type badge */}
+          <div className="absolute top-2 left-2">
+            {isVideo(item.contentType) ? (
+              <span className="flex items-center gap-1 px-1.5 py-0.5 rounded bg-black/60 text-white text-xs">
+                <Video className="h-3 w-3" /> Video
+              </span>
+            ) : null}
+          </div>
+        </div>
+      ))}
+    </div>
+  );
+}

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

@@ -0,0 +1,143 @@
+import { useCallback, useState, useRef } from 'react';
+import { Upload, Loader2, Check } from 'lucide-react';
+
+export interface MediaUploaderProps {
+  /** Called when upload completes successfully */
+  onUploadComplete?: (result: { url: string; path: string }) => void;
+  /** Called on error */
+  onError?: (error: string) => void;
+  /** Upload function from useMediaUpload hook */
+  upload: (file: File) => Promise<{ url: string; path: string }>;
+  /** Whether an upload is currently in progress */
+  isUploading?: boolean;
+  /** Upload progress percentage (0-100) */
+  progress?: number;
+  /** Accepted file types (e.g., "image/*,video/*") */
+  accept?: string;
+  /** Max file size in MB */
+  maxSizeMB?: number;
+}
+
+export function MediaUploader({
+  onUploadComplete,
+  onError,
+  upload,
+  isUploading = false,
+  progress = 0,
+  accept = 'image/*,video/*',
+  maxSizeMB = 50,
+}: MediaUploaderProps) {
+  const [isDragOver, setIsDragOver] = useState(false);
+  const [lastResult, setLastResult] = useState<{ url: string; path: string } | null>(null);
+  const fileInputRef = useRef<HTMLInputElement>(null);
+
+  const handleFile = useCallback(async (file: File) => {
+    // Validate file type (accept prop only restricts file picker, not drag-and-drop)
+    if (accept) {
+      const acceptedTypes = accept.split(',').map((t) => t.trim());
+      const isAccepted = acceptedTypes.some((type) => {
+        if (type.endsWith('/*')) return file.type.startsWith(type.replace('/*', '/'));
+        return file.type === type;
+      });
+      if (!isAccepted) {
+        onError?.(`File type "${file.type}" is not accepted. Allowed: ${accept}`);
+        return;
+      }
+    }
+
+    if (file.size > maxSizeMB * 1024 * 1024) {
+      onError?.(`File too large. Maximum size is ${maxSizeMB}MB.`);
+      return;
+    }
+
+    try {
+      const result = await upload(file);
+      setLastResult(result);
+      onUploadComplete?.(result);
+    } catch (err) {
+      onError?.(err instanceof Error ? err.message : 'Upload failed');
+    }
+  }, [upload, accept, maxSizeMB, onUploadComplete, onError]);
+
+  const handleDrop = useCallback((e: React.DragEvent) => {
+    e.preventDefault();
+    setIsDragOver(false);
+    const file = e.dataTransfer.files[0];
+    if (file) handleFile(file);
+  }, [handleFile]);
+
+  const handleDragOver = useCallback((e: React.DragEvent) => {
+    e.preventDefault();
+    setIsDragOver(true);
+  }, []);
+
+  const handleDragLeave = useCallback(() => {
+    setIsDragOver(false);
+  }, []);
+
+  const handleClick = () => fileInputRef.current?.click();
+
+  const handleInputChange = (e: React.ChangeEvent<HTMLInputElement>) => {
+    const file = e.target.files?.[0];
+    if (file) handleFile(file);
+    if (fileInputRef.current) fileInputRef.current.value = '';
+  };
+
+  return (
+    <div
+      onDrop={handleDrop}
+      onDragOver={handleDragOver}
+      onDragLeave={handleDragLeave}
+      onClick={isUploading ? undefined : handleClick}
+      className={`
+        relative border-2 border-dashed rounded-lg p-8 text-center transition-colors
+        ${isDragOver
+          ? 'border-[var(--accent)] bg-[var(--accent)]/5'
+          : 'border-[var(--border-muted)] hover:border-[var(--border-default)]'}
+        ${isUploading ? 'pointer-events-none opacity-75' : 'cursor-pointer'}
+      `}
+    >
+      <input
+        ref={fileInputRef}
+        type="file"
+        accept={accept}
+        onChange={handleInputChange}
+        className="hidden"
+      />
+
+      {isUploading ? (
+        <div className="space-y-3">
+          <Loader2 className="h-8 w-8 animate-spin text-[var(--accent)] mx-auto" />
+          <p className="text-sm text-[var(--text-muted)]">Uploading... {progress}%</p>
+          <div className="w-full max-w-xs mx-auto h-2 rounded-full bg-[var(--surface-200)] overflow-hidden">
+            <div
+              className="h-full rounded-full bg-[var(--accent)] transition-all duration-300"
+              style={{ width: `${progress}%` }}
+            />
+          </div>
+        </div>
+      ) : lastResult ? (
+        <div className="space-y-2">
+          <Check className="h-8 w-8 text-green-500 mx-auto" />
+          <p className="text-sm text-[var(--text-primary)]">Upload complete</p>
+          <button
+            onClick={(e) => { e.stopPropagation(); setLastResult(null); }}
+            className="text-xs text-[var(--text-muted)] hover:text-[var(--text-primary)] underline"
+          >
+            Upload another
+          </button>
+        </div>
+      ) : (
+        <div className="space-y-2">
+          <Upload className="h-8 w-8 text-[var(--text-muted)] mx-auto" />
+          <p className="text-sm text-[var(--text-primary)]">
+            Drop a file here or click to browse
+          </p>
+          <p className="text-xs text-[var(--text-muted)]">
+            Max {maxSizeMB}MB
+          </p>
+        </div>
+      )}
+    </div>
+  );
+}

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

@@ -0,0 +1,70 @@
+import * as React from 'react';
+import { cva, type VariantProps } from 'class-variance-authority';
+import { cn } from '../utils/cn';
+
+const providerBadgeVariants = cva(
+  'inline-flex items-center gap-1.5 rounded-full px-2 py-0.5 text-xs font-medium',
+  {
+    variants: {
+      provider: {
+        gemini: 'bg-blue-500/15 text-blue-400 border border-blue-500/30',
+        laozhang: 'bg-emerald-500/15 text-emerald-400 border border-emerald-500/30',
+        default: 'bg-[var(--surface-200)] text-[var(--text-secondary)] border border-[var(--border)]',
+      },
+    },
+    defaultVariants: {
+      provider: 'default',
+    },
+  }
+);
+
+type ProviderVariant = 'gemini' | 'laozhang' | 'default';
+
+export interface ProviderBadgeProps
+  extends React.HTMLAttributes<HTMLSpanElement>,
+    Omit<VariantProps<typeof providerBadgeVariants>, 'provider'> {
+  provider: string;
+  latency?: number;
+}
+
+const providerIcons: Record<string, string> = {
+  gemini: '✦',
+  laozhang: '◆',
+};
+
+const providerLabels: Record<string, string> = {
+  gemini: 'Gemini',
+  laozhang: 'LaoZhang',
+};
+
+function getProviderVariant(provider: string): ProviderVariant {
+  const normalized = provider.toLowerCase();
+  if (normalized.includes('gemini')) return 'gemini';
+  if (normalized.includes('laozhang')) return 'laozhang';
+  return 'default';
+}
+
+const ProviderBadge = React.forwardRef<HTMLSpanElement, ProviderBadgeProps>(
+  ({ provider, latency, className, ...props }, ref) => {
+    const variant = getProviderVariant(provider);
+    const icon = providerIcons[variant] || '●';
+    const label = providerLabels[variant] || provider;
+
+    return (
+      <span
+        ref={ref}
+        className={cn(providerBadgeVariants({ provider: variant }), className)}
+        {...props}
+      >
+        <span className="opacity-70">{icon}</span>
+        <span>{label}</span>
+        {latency !== undefined && (
+          <span className="opacity-60 tabular-nums">{latency}ms</span>
+        )}
+      </span>
+    );
+  }
+);
+ProviderBadge.displayName = 'ProviderBadge';
+
+export { ProviderBadge, providerBadgeVariants };

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

@@ -0,0 +1,35 @@
+import * as React from 'react';
+import { cn } from '../utils/cn';
+
+export interface StreamingTextProps extends React.HTMLAttributes<HTMLSpanElement> {
+  text: string;
+  isStreaming: boolean;
+}
+
+const StreamingText = React.forwardRef<HTMLSpanElement, StreamingTextProps>(
+  ({ text, isStreaming, className, ...props }, ref) => {
+    return (
+      <span ref={ref} className={cn('whitespace-pre-wrap', className)} {...props}>
+        {text}
+        {isStreaming && (
+          <span
+            className="inline-block w-[2px] h-[1.1em] ml-0.5 bg-current align-text-bottom animate-[blink_1s_step-end_infinite]"
+            style={{
+              // Inline keyframes for cursor blink
+              animation: 'blink 1s step-end infinite',
+            }}
+          />
+        )}
+        <style>{`
+          @keyframes blink {
+            0%, 100% { opacity: 1; }
+            50% { opacity: 0; }
+          }
+        `}</style>
+      </span>
+    );
+  }
+);
+StreamingText.displayName = 'StreamingText';
+
+export { StreamingText };

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

@@ -0,0 +1,103 @@
+'use client';
+
+import { useRef, useState } from 'react';
+import { cn } from '../utils/cn';
+
+export interface VideoPlayerProps {
+  /** Video source URL or base64 data URI. */
+  src: string;
+  /** MIME type for the source. */
+  mimeType?: string;
+  /** Alt text / accessible label. */
+  alt?: string;
+  /** Additional class names for the container. */
+  className?: string;
+  /** Aspect ratio CSS class (default: aspect-video). */
+  aspectRatio?: string;
+}
+
+/**
+ * VideoPlayer renders an HTML5 video with controls, loading state,
+ * and error handling. Supports both URL and base64 data URIs.
+ */
+export function VideoPlayer({
+  src,
+  mimeType = 'video/mp4',
+  alt = 'Generated video',
+  className,
+  aspectRatio = 'aspect-video',
+}: VideoPlayerProps) {
+  const videoRef = useRef<HTMLVideoElement>(null);
+  const [isLoading, setIsLoading] = useState(true);
+  const [error, setError] = useState<string | null>(null);
+
+  return (
+    <div
+      className={cn(
+        'relative rounded-lg overflow-hidden bg-[var(--surface)]',
+        aspectRatio,
+        className,
+      )}
+    >
+      {isLoading && !error && (
+        <div className="absolute inset-0 flex items-center justify-center bg-[var(--surface)]">
+          <div className="h-8 w-8 animate-spin rounded-full border-2 border-[var(--text-muted)] border-t-[var(--accent)]" />
+        </div>
+      )}
+
+      {error && (
+        <div className="absolute inset-0 flex items-center justify-center bg-[var(--surface)]">
+          <p className="text-sm text-[var(--error)]">{error}</p>
+        </div>
+      )}
+
+      <video
+        ref={videoRef}
+        className={cn('h-full w-full object-contain', isLoading && 'invisible')}
+        controls
+        playsInline
+        aria-label={alt}
+        onLoadedData={() => setIsLoading(false)}
+        onError={() => {
+          setIsLoading(false);
+          setError('Failed to load video');
+        }}
+      >
+        <source src={src} type={mimeType} />
+      </video>
+    </div>
+  );
+}
+
+export interface VideoGridProps {
+  /** Videos to display. */
+  videos: Array<{ src: string; mimeType?: string; alt?: string }>;
+  /** Additional class names. */
+  className?: string;
+}
+
+/**
+ * VideoGrid displays one or more videos in a responsive grid.
+ */
+export function VideoGrid({ videos, className }: VideoGridProps) {
+  if (videos.length === 0) return null;
+
+  return (
+    <div
+      className={cn(
+        'grid gap-4',
+        videos.length === 1 ? 'grid-cols-1' : 'grid-cols-1 md:grid-cols-2',
+        className,
+      )}
+    >
+      {videos.map((video, i) => (
+        <VideoPlayer
+          key={`video-${i}`}
+          src={video.src}
+          mimeType={video.mimeType}
+          alt={video.alt}
+        />
+      ))}
+    </div>
+  );
+}

internal/adapter/templates/templates/skeleton/packages/ui/src/hooks/useFormErrors.ts

@@ -0,0 +1,103 @@
+'use client';
+
+import { useState, useCallback } from 'react';
+
+export interface FieldErrors {
+  [field: string]: string;
+}
+
+export interface UseFormErrorsReturn {
+  /** Current field errors */
+  errors: FieldErrors;
+  /** Set error for a specific field */
+  setFieldError: (field: string, message: string) => void;
+  /** Clear error for a specific field */
+  clearFieldError: (field: string) => void;
+  /** Set multiple field errors at once */
+  setErrors: (errors: FieldErrors) => void;
+  /** Clear all errors */
+  clearErrors: () => void;
+  /** Check if a field has an error */
+  hasError: (field: string) => boolean;
+  /** Get error message for a field */
+  getError: (field: string) => string | undefined;
+  /** Check if any errors exist */
+  hasAnyErrors: boolean;
+}
+
+/**
+ * Hook for managing form field errors.
+ * Works well with ApiClientError.getFieldErrors().
+ *
+ * @example
+ * const { errors, setErrors, clearErrors, hasError, getError } = useFormErrors();
+ *
+ * const handleSubmit = async (data) => {
+ *   clearErrors();
+ *   try {
+ *     await api.post('/users', data);
+ *   } catch (error) {
+ *     if (isApiClientError(error) && error.isValidationError()) {
+ *       setErrors(error.getFieldErrors());
+ *     }
+ *   }
+ * };
+ *
+ * return (
+ *   <FormField
+ *     name="email"
+ *     label="Email"
+ *     error={getError('email')}
+ *   />
+ * );
+ */
+export function useFormErrors(initialErrors: FieldErrors = {}): UseFormErrorsReturn {
+  const [errors, setErrorsState] = useState<FieldErrors>(initialErrors);
+
+  const setFieldError = useCallback((field: string, message: string) => {
+    setErrorsState((prev) => ({ ...prev, [field]: message }));
+  }, []);
+
+  const clearFieldError = useCallback((field: string) => {
+    setErrorsState((prev) => {
+      const next = { ...prev };
+      delete next[field];
+      return next;
+    });
+  }, []);
+
+  const setErrors = useCallback((newErrors: FieldErrors) => {
+    setErrorsState(newErrors);
+  }, []);
+
+  const clearErrors = useCallback(() => {
+    setErrorsState({});
+  }, []);
+
+  const hasError = useCallback(
+    (field: string) => {
+      return !!errors[field];
+    },
+    [errors]
+  );
+
+  const getError = useCallback(
+    (field: string) => {
+      return errors[field];
+    },
+    [errors]
+  );
+
+  const hasAnyErrors = Object.keys(errors).length > 0;
+
+  return {
+    errors,
+    setFieldError,
+    clearFieldError,
+    setErrors,
+    clearErrors,
+    hasError,
+    getError,
+    hasAnyErrors,
+  };
+}