feat(gateway): support hot reload and empty startup

- extract gateway runtime into pkg/gateway
- add gateway.hot_reload config with default and example values
- allow starting the gateway without a default model via --allow-empty
- stop treating missing enabled channels as a startup error
- update related tests

cmd/picoclaw/internal/gateway/command.go

@@ -5,6 +5,8 @@ import (
 
 	"github.com/spf13/cobra"
 
+	"github.com/sipeed/picoclaw/cmd/picoclaw/internal"
+	"github.com/sipeed/picoclaw/pkg/gateway"
 	"github.com/sipeed/picoclaw/pkg/logger"
 	"github.com/sipeed/picoclaw/pkg/utils"
 )
@@ -12,6 +14,7 @@ import (
 func NewGatewayCommand() *cobra.Command {
 	var debug bool
 	var noTruncate bool
+	var allowEmpty bool
 
 	cmd := &cobra.Command{
 		Use:     "gateway",
@@ -31,12 +34,19 @@ func NewGatewayCommand() *cobra.Command {
 			return nil
 		},
 		RunE: func(_ *cobra.Command, _ []string) error {
-			return gatewayCmd(debug)
+			return gateway.Run(debug, internal.GetConfigPath(), allowEmpty)
 		},
 	}
 
 	cmd.Flags().BoolVarP(&debug, "debug", "d", false, "Enable debug logging")
 	cmd.Flags().BoolVarP(&noTruncate, "no-truncate", "T", false, "Disable string truncation in debug logs")
+	cmd.Flags().BoolVarP(
+		&allowEmpty,
+		"allow-empty",
+		"E",
+		false,
+		"Continue starting even when no default model is configured",
+	)
 
 	return cmd
 }

cmd/picoclaw/internal/gateway/command_test.go

@@ -28,4 +28,5 @@ func TestNewGatewayCommand(t *testing.T) {
 
 	assert.True(t, cmd.HasFlags())
 	assert.NotNil(t, cmd.Flags().Lookup("debug"))
+	assert.NotNil(t, cmd.Flags().Lookup("allow-empty"))
 }

config/config.example.json

@@ -518,6 +518,7 @@
   },
   "gateway": {
     "host": "127.0.0.1",
-    "port": 18790
+    "port": 18790,
+    "hot_reload": false
   }
 }

pkg/channels/manager.go

@@ -357,7 +357,6 @@ func (m *Manager) StartAll(ctx context.Context) error {
 
 	if len(m.channels) == 0 {
 		logger.WarnC("channels", "No channels enabled")
-		return errors.New("no channels enabled")
 	}
 
 	logger.InfoC("channels", "Starting all channels")

pkg/config/config.go

@@ -625,8 +625,9 @@ func (c *ModelConfig) Validate() error {
 }
 
 type GatewayConfig struct {
-	Host string `json:"host" env:"PICOCLAW_GATEWAY_HOST"`
+	Host      string `json:"host"       env:"PICOCLAW_GATEWAY_HOST"`
-	Port int    `json:"port" env:"PICOCLAW_GATEWAY_PORT"`
+	Port      int    `json:"port"       env:"PICOCLAW_GATEWAY_PORT"`
+	HotReload bool   `json:"hot_reload" env:"PICOCLAW_GATEWAY_HOT_RELOAD"`
 }
 
 type ToolDiscoveryConfig struct {

pkg/config/config_test.go

@@ -267,6 +267,9 @@ func TestDefaultConfig_Gateway(t *testing.T) {
 	if cfg.Gateway.Port == 0 {
 		t.Error("Gateway port should have default value")
 	}
+	if cfg.Gateway.HotReload {
+		t.Error("Gateway hot reload should be disabled by default")
+	}
 }
 
 // TestDefaultConfig_Providers verifies provider structure

pkg/config/defaults.go

@@ -395,8 +395,9 @@ func DefaultConfig() *Config {
 			},
 		},
 		Gateway: GatewayConfig{
-			Host: "127.0.0.1",
+			Host:      "127.0.0.1",
-			Port: 18790,
+			Port:      18790,
+			HotReload: false,
 		},
 		Tools: ToolsConfig{
 			MediaCleanup: MediaCleanupConfig{

pkg/gateway/gateway.go

@@ -10,7 +10,6 @@ import (
 	"syscall"
 	"time"
 
-	"github.com/sipeed/picoclaw/cmd/picoclaw/internal"
 	"github.com/sipeed/picoclaw/pkg/agent"
 	"github.com/sipeed/picoclaw/pkg/bus"
 	"github.com/sipeed/picoclaw/pkg/channels"
@@ -42,15 +41,13 @@ import (
 	"github.com/sipeed/picoclaw/pkg/voice"
 )
 
-// Timeout constants for service operations
 const (
 	serviceShutdownTimeout  = 30 * time.Second
 	providerReloadTimeout   = 30 * time.Second
 	gracefulShutdownTimeout = 15 * time.Second
 )
 
-// gatewayServices holds references to all running services
+type services struct {
-type gatewayServices struct {
 	CronService      *cron.CronService
 	HeartbeatService *heartbeat.HeartbeatService
 	MediaStore       media.MediaStore
@@ -59,24 +56,41 @@ type gatewayServices struct {
 	HealthServer     *health.Server
 }
 
-func gatewayCmd(debug bool) error {
+type startupBlockedProvider struct {
+	reason string
+}
+
+func (p *startupBlockedProvider) Chat(
+	_ context.Context,
+	_ []providers.Message,
+	_ []providers.ToolDefinition,
+	_ string,
+	_ map[string]any,
+) (*providers.LLMResponse, error) {
+	return nil, fmt.Errorf("%s", p.reason)
+}
+
+func (p *startupBlockedProvider) GetDefaultModel() string {
+	return ""
+}
+
+// Run starts the gateway runtime using the configuration loaded from configPath.
+func Run(debug bool, configPath string, allowEmptyStartup bool) error {
 	if debug {
 		logger.SetLevel(logger.DEBUG)
 		fmt.Println("🔍 Debug mode enabled")
 	}
 
-	configPath := internal.GetConfigPath()
+	cfg, err := config.LoadConfig(configPath)
-	cfg, err := internal.LoadConfig()
 	if err != nil {
 		return fmt.Errorf("error loading config: %w", err)
 	}
 
-	provider, modelID, err := providers.CreateProvider(cfg)
+	provider, modelID, err := createStartupProvider(cfg, allowEmptyStartup)
 	if err != nil {
 		return fmt.Errorf("error creating provider: %w", err)
 	}
 
-	// Use the resolved model ID from provider creation
 	if modelID != "" {
 		cfg.Agents.Defaults.ModelName = modelID
 	}
@@ -84,17 +98,13 @@ func gatewayCmd(debug bool) error {
 	msgBus := bus.NewMessageBus()
 	agentLoop := agent.NewAgentLoop(cfg, msgBus, provider)
 
-	// Print agent startup info
 	fmt.Println("\n📦 Agent Status:")
 	startupInfo := agentLoop.GetStartupInfo()
 	toolsInfo := startupInfo["tools"].(map[string]any)
 	skillsInfo := startupInfo["skills"].(map[string]any)
 	fmt.Printf("  • Tools: %d loaded\n", toolsInfo["count"])
-	fmt.Printf("  • Skills: %d/%d available\n",
+	fmt.Printf("  • Skills: %d/%d available\n", skillsInfo["available"], skillsInfo["total"])
-		skillsInfo["available"],
-		skillsInfo["total"])
 
-	// Log to file as well
 	logger.InfoCF("agent", "Agent initialized",
 		map[string]any{
 			"tools_count":      toolsInfo["count"],
@@ -102,8 +112,7 @@ func gatewayCmd(debug bool) error {
 			"skills_available": skillsInfo["available"],
 		})
 
-	// Setup and start all services
+	runningServices, err := setupAndStartServices(cfg, agentLoop, msgBus)
-	services, err := setupAndStartServices(cfg, agentLoop, msgBus)
 	if err != nil {
 		return err
 	}
@@ -116,23 +125,25 @@ func gatewayCmd(debug bool) error {
 
 	go agentLoop.Run(ctx)
 
-	// Setup config file watcher for hot reload
+	var configReloadChan <-chan *config.Config
-	configReloadChan, stopWatch := setupConfigWatcherPolling(configPath, debug)
+	stopWatch := func() {}
+	if cfg.Gateway.HotReload {
+		configReloadChan, stopWatch = setupConfigWatcherPolling(configPath, debug)
+		logger.Info("Config hot reload enabled")
+	}
 	defer stopWatch()
 
 	sigChan := make(chan os.Signal, 1)
 	signal.Notify(sigChan, os.Interrupt, syscall.SIGTERM)
 
-	// Main event loop - wait for signals or config changes
 	for {
 		select {
 		case <-sigChan:
 			logger.Info("Shutting down...")
-			shutdownGateway(services, agentLoop, provider, true)
+			shutdownGateway(runningServices, agentLoop, provider, true)
 			return nil
-
 		case newCfg := <-configReloadChan:
-			err := handleConfigReload(ctx, agentLoop, newCfg, &provider, services, msgBus)
+			err := handleConfigReload(ctx, agentLoop, newCfg, &provider, runningServices, msgBus, allowEmptyStartup)
 			if err != nil {
 				logger.Errorf("Config reload failed: %v", err)
 			}
@@ -140,18 +151,33 @@ func gatewayCmd(debug bool) error {
 	}
 }
 
-// setupAndStartServices initializes and starts all services
+func createStartupProvider(
+	cfg *config.Config,
+	allowEmptyStartup bool,
+) (providers.LLMProvider, string, error) {
+	modelName := cfg.Agents.Defaults.GetModelName()
+	if modelName == "" && allowEmptyStartup {
+		reason := "no default model configured; gateway started in limited mode"
+		fmt.Printf("⚠ Warning: %s\n", reason)
+		logger.WarnCF("gateway", "Gateway started without default model", map[string]any{
+			"limited_mode": true,
+		})
+		return &startupBlockedProvider{reason: reason}, "", nil
+	}
+
+	return providers.CreateProvider(cfg)
+}
+
 func setupAndStartServices(
 	cfg *config.Config,
 	agentLoop *agent.AgentLoop,
 	msgBus *bus.MessageBus,
-) (*gatewayServices, error) {
+) (*services, error) {
-	services := &gatewayServices{}
+	runningServices := &services{}
 
-	// Setup cron tool and service
 	execTimeout := time.Duration(cfg.Tools.Cron.ExecTimeoutMinutes) * time.Minute
 	var err error
-	services.CronService, err = setupCronTool(
+	runningServices.CronService, err = setupCronTool(
 		agentLoop,
 		msgBus,
 		cfg.WorkspacePath(),
@@ -162,120 +188,105 @@ func setupAndStartServices(
 	if err != nil {
 		return nil, fmt.Errorf("error setting up cron service: %w", err)
 	}
-	if err = services.CronService.Start(); err != nil {
+	if err = runningServices.CronService.Start(); err != nil {
 		return nil, fmt.Errorf("error starting cron service: %w", err)
 	}
 	fmt.Println("✓ Cron service started")
 
-	// Setup heartbeat service
+	runningServices.HeartbeatService = heartbeat.NewHeartbeatService(
-	services.HeartbeatService = heartbeat.NewHeartbeatService(
 		cfg.WorkspacePath(),
 		cfg.Heartbeat.Interval,
 		cfg.Heartbeat.Enabled,
 	)
-	services.HeartbeatService.SetBus(msgBus)
+	runningServices.HeartbeatService.SetBus(msgBus)
-	services.HeartbeatService.SetHandler(createHeartbeatHandler(agentLoop))
+	runningServices.HeartbeatService.SetHandler(createHeartbeatHandler(agentLoop))
-	if err = services.HeartbeatService.Start(); err != nil {
+	if err = runningServices.HeartbeatService.Start(); err != nil {
 		return nil, fmt.Errorf("error starting heartbeat service: %w", err)
 	}
 	fmt.Println("✓ Heartbeat service started")
 
-	// Create media store for file lifecycle management with TTL cleanup
+	runningServices.MediaStore = media.NewFileMediaStoreWithCleanup(media.MediaCleanerConfig{
-	services.MediaStore = media.NewFileMediaStoreWithCleanup(media.MediaCleanerConfig{
 		Enabled:  cfg.Tools.MediaCleanup.Enabled,
 		MaxAge:   time.Duration(cfg.Tools.MediaCleanup.MaxAge) * time.Minute,
 		Interval: time.Duration(cfg.Tools.MediaCleanup.Interval) * time.Minute,
 	})
-	// Start the media store if it's a FileMediaStore with cleanup
+	if fms, ok := runningServices.MediaStore.(*media.FileMediaStore); ok {
-	if fms, ok := services.MediaStore.(*media.FileMediaStore); ok {
 		fms.Start()
 	}
 
-	// Create channel manager
+	runningServices.ChannelManager, err = channels.NewManager(cfg, msgBus, runningServices.MediaStore)
-	services.ChannelManager, err = channels.NewManager(cfg, msgBus, services.MediaStore)
 	if err != nil {
-		// Stop the media store if it's a FileMediaStore with cleanup
+		if fms, ok := runningServices.MediaStore.(*media.FileMediaStore); ok {
-		if fms, ok := services.MediaStore.(*media.FileMediaStore); ok {
 			fms.Stop()
 		}
 		return nil, fmt.Errorf("error creating channel manager: %w", err)
 	}
 
-	// Inject channel manager and media store into agent loop
+	agentLoop.SetChannelManager(runningServices.ChannelManager)
-	agentLoop.SetChannelManager(services.ChannelManager)
+	agentLoop.SetMediaStore(runningServices.MediaStore)
-	agentLoop.SetMediaStore(services.MediaStore)
 
-	// Wire up voice transcription if a supported provider is configured.
 	if transcriber := voice.DetectTranscriber(cfg); transcriber != nil {
 		agentLoop.SetTranscriber(transcriber)
 		logger.InfoCF("voice", "Transcription enabled (agent-level)", map[string]any{"provider": transcriber.Name()})
 	}
 
-	enabledChannels := services.ChannelManager.GetEnabledChannels()
+	enabledChannels := runningServices.ChannelManager.GetEnabledChannels()
 	if len(enabledChannels) > 0 {
 		fmt.Printf("✓ Channels enabled: %s\n", enabledChannels)
 	} else {
 		fmt.Println("⚠ Warning: No channels enabled")
 	}
 
-	// Setup shared HTTP server with health endpoints and webhook handlers
 	addr := fmt.Sprintf("%s:%d", cfg.Gateway.Host, cfg.Gateway.Port)
-	services.HealthServer = health.NewServer(cfg.Gateway.Host, cfg.Gateway.Port)
+	runningServices.HealthServer = health.NewServer(cfg.Gateway.Host, cfg.Gateway.Port)
-	services.ChannelManager.SetupHTTPServer(addr, services.HealthServer)
+	runningServices.ChannelManager.SetupHTTPServer(addr, runningServices.HealthServer)
 
-	if err = services.ChannelManager.StartAll(context.Background()); err != nil {
+	if err = runningServices.ChannelManager.StartAll(context.Background()); err != nil {
 		return nil, fmt.Errorf("error starting channels: %w", err)
 	}
 
 	fmt.Printf("✓ Health endpoints available at http://%s:%d/health and /ready\n", cfg.Gateway.Host, cfg.Gateway.Port)
 
-	// Setup state manager and device service
 	stateManager := state.NewManager(cfg.WorkspacePath())
-	services.DeviceService = devices.NewService(devices.Config{
+	runningServices.DeviceService = devices.NewService(devices.Config{
 		Enabled:    cfg.Devices.Enabled,
 		MonitorUSB: cfg.Devices.MonitorUSB,
 	}, stateManager)
-	services.DeviceService.SetBus(msgBus)
+	runningServices.DeviceService.SetBus(msgBus)
-	if err = services.DeviceService.Start(context.Background()); err != nil {
+	if err = runningServices.DeviceService.Start(context.Background()); err != nil {
 		logger.ErrorCF("device", "Error starting device service", map[string]any{"error": err.Error()})
 	} else if cfg.Devices.Enabled {
 		fmt.Println("✓ Device event service started")
 	}
 
-	return services, nil
+	return runningServices, nil
 }
 
-// stopAndCleanupServices stops all services and cleans up resources
+func stopAndCleanupServices(runningServices *services, shutdownTimeout time.Duration) {
-func stopAndCleanupServices(
-	services *gatewayServices,
-	shutdownTimeout time.Duration,
-) {
 	shutdownCtx, shutdownCancel := context.WithTimeout(context.Background(), shutdownTimeout)
 	defer shutdownCancel()
 
-	if services.ChannelManager != nil {
+	if runningServices.ChannelManager != nil {
-		services.ChannelManager.StopAll(shutdownCtx)
+		runningServices.ChannelManager.StopAll(shutdownCtx)
 	}
-	if services.DeviceService != nil {
+	if runningServices.DeviceService != nil {
-		services.DeviceService.Stop()
+		runningServices.DeviceService.Stop()
 	}
-	if services.HeartbeatService != nil {
+	if runningServices.HeartbeatService != nil {
-		services.HeartbeatService.Stop()
+		runningServices.HeartbeatService.Stop()
 	}
-	if services.CronService != nil {
+	if runningServices.CronService != nil {
-		services.CronService.Stop()
+		runningServices.CronService.Stop()
 	}
-	if services.MediaStore != nil {
+	if runningServices.MediaStore != nil {
-		// Stop the media store if it's a FileMediaStore with cleanup
+		if fms, ok := runningServices.MediaStore.(*media.FileMediaStore); ok {
-		if fms, ok := services.MediaStore.(*media.FileMediaStore); ok {
 			fms.Stop()
 		}
 	}
 }
 
-// shutdownGateway performs a complete gateway shutdown
 func shutdownGateway(
-	services *gatewayServices,
+	runningServices *services,
 	agentLoop *agent.AgentLoop,
 	provider providers.LLMProvider,
 	fullShutdown bool,
@@ -284,7 +295,7 @@ func shutdownGateway(
 		cp.Close()
 	}
 
-	stopAndCleanupServices(services, gracefulShutdownTimeout)
+	stopAndCleanupServices(runningServices, gracefulShutdownTimeout)
 
 	agentLoop.Stop()
 	agentLoop.Close()
@@ -292,15 +303,14 @@ func shutdownGateway(
 	logger.Info("✓ Gateway stopped")
 }
 
-// handleConfigReload handles config file reload by stopping all services,
-// reloading the provider and config, and restarting services with the new config.
 func handleConfigReload(
 	ctx context.Context,
 	al *agent.AgentLoop,
 	newCfg *config.Config,
 	providerRef *providers.LLMProvider,
-	services *gatewayServices,
+	runningServices *services,
 	msgBus *bus.MessageBus,
+	allowEmptyStartup bool,
 ) error {
 	logger.Info("🔄 Config file changed, reloading...")
 
@@ -311,18 +321,14 @@ func handleConfigReload(
 
 	logger.Infof(" New model is '%s', recreating provider...", newModel)
 
-	// Stop all services before reloading
 	logger.Info("  Stopping all services...")
-	stopAndCleanupServices(services, serviceShutdownTimeout)
+	stopAndCleanupServices(runningServices, serviceShutdownTimeout)
 
-	// Create new provider from updated config first to ensure validity
+	newProvider, newModelID, err := createStartupProvider(newCfg, allowEmptyStartup)
-	// This will use the correct API key and settings from newCfg.ModelList
-	newProvider, newModelID, err := providers.CreateProvider(newCfg)
 	if err != nil {
 		logger.Errorf("  ⚠ Error creating new provider: %v", err)
 		logger.Warn("  Attempting to restart services with old provider and config...")
-		// Try to restart services with old configuration
+		if restartErr := restartServices(al, runningServices, msgBus); restartErr != nil {
-		if restartErr := restartServices(al, services, msgBus); restartErr != nil {
 			logger.Errorf("  ⚠ Failed to restart services: %v", restartErr)
 		}
 		return fmt.Errorf("error creating new provider: %w", err)
@@ -332,31 +338,25 @@ func handleConfigReload(
 		newCfg.Agents.Defaults.ModelName = newModelID
 	}
 
-	// Use the atomic reload method on AgentLoop to safely swap provider and config.
-	// This handles locking internally to prevent races with in-flight LLM calls
-	// and concurrent reads of registry/config while the swap occurs.
 	reloadCtx, reloadCancel := context.WithTimeout(context.Background(), providerReloadTimeout)
 	defer reloadCancel()
 
 	if err := al.ReloadProviderAndConfig(reloadCtx, newProvider, newCfg); err != nil {
 		logger.Errorf("  ⚠ Error reloading agent loop: %v", err)
-		// Close the newly created provider since it wasn't adopted
 		if cp, ok := newProvider.(providers.StatefulProvider); ok {
 			cp.Close()
 		}
 		logger.Warn("  Attempting to restart services with old provider and config...")
-		if restartErr := restartServices(al, services, msgBus); restartErr != nil {
+		if restartErr := restartServices(al, runningServices, msgBus); restartErr != nil {
 			logger.Errorf("  ⚠ Failed to restart services: %v", restartErr)
 		}
 		return fmt.Errorf("error reloading agent loop: %w", err)
 	}
 
-	// Update local provider reference only after successful atomic reload
 	*providerRef = newProvider
 
-	// Restart all services with new config
 	logger.Info("  Restarting all services with new configuration...")
-	if err := restartServices(al, services, msgBus); err != nil {
+	if err := restartServices(al, runningServices, msgBus); err != nil {
 		logger.Errorf("  ⚠ Error restarting services: %v", err)
 		return fmt.Errorf("error restarting services: %w", err)
 	}
@@ -365,19 +365,16 @@ func handleConfigReload(
 	return nil
 }
 
-// restartServices restarts all services after a config reload
 func restartServices(
 	al *agent.AgentLoop,
-	services *gatewayServices,
+	runningServices *services,
 	msgBus *bus.MessageBus,
 ) error {
-	// Get current config from agent loop (which has been updated if this is a reload)
 	cfg := al.GetConfig()
 
-	// Re-create and start cron service with new config
 	execTimeout := time.Duration(cfg.Tools.Cron.ExecTimeoutMinutes) * time.Minute
 	var err error
-	services.CronService, err = setupCronTool(
+	runningServices.CronService, err = setupCronTool(
 		al,
 		msgBus,
 		cfg.WorkspacePath(),
@@ -388,57 +385,51 @@ func restartServices(
 	if err != nil {
 		return fmt.Errorf("error restarting cron service: %w", err)
 	}
-	if err = services.CronService.Start(); err != nil {
+	if err = runningServices.CronService.Start(); err != nil {
 		return fmt.Errorf("error restarting cron service: %w", err)
 	}
 	fmt.Println("  ✓ Cron service restarted")
 
-	// Re-create and start heartbeat service with new config
+	runningServices.HeartbeatService = heartbeat.NewHeartbeatService(
-	services.HeartbeatService = heartbeat.NewHeartbeatService(
 		cfg.WorkspacePath(),
 		cfg.Heartbeat.Interval,
 		cfg.Heartbeat.Enabled,
 	)
-	services.HeartbeatService.SetBus(msgBus)
+	runningServices.HeartbeatService.SetBus(msgBus)
-	services.HeartbeatService.SetHandler(createHeartbeatHandler(al))
+	runningServices.HeartbeatService.SetHandler(createHeartbeatHandler(al))
-	if err = services.HeartbeatService.Start(); err != nil {
+	if err = runningServices.HeartbeatService.Start(); err != nil {
 		return fmt.Errorf("error restarting heartbeat service: %w", err)
 	}
 	fmt.Println("  ✓ Heartbeat service restarted")
 
-	// Re-create media store with new config
+	runningServices.MediaStore = media.NewFileMediaStoreWithCleanup(media.MediaCleanerConfig{
-	services.MediaStore = media.NewFileMediaStoreWithCleanup(media.MediaCleanerConfig{
 		Enabled:  cfg.Tools.MediaCleanup.Enabled,
 		MaxAge:   time.Duration(cfg.Tools.MediaCleanup.MaxAge) * time.Minute,
 		Interval: time.Duration(cfg.Tools.MediaCleanup.Interval) * time.Minute,
 	})
-	// Start the media store if it's a FileMediaStore with cleanup
+	if fms, ok := runningServices.MediaStore.(*media.FileMediaStore); ok {
-	if fms, ok := services.MediaStore.(*media.FileMediaStore); ok {
 		fms.Start()
 	}
-	al.SetMediaStore(services.MediaStore)
+	al.SetMediaStore(runningServices.MediaStore)
 
-	// Re-create channel manager with new config
+	runningServices.ChannelManager, err = channels.NewManager(cfg, msgBus, runningServices.MediaStore)
-	services.ChannelManager, err = channels.NewManager(cfg, msgBus, services.MediaStore)
 	if err != nil {
 		return fmt.Errorf("error recreating channel manager: %w", err)
 	}
-	al.SetChannelManager(services.ChannelManager)
+	al.SetChannelManager(runningServices.ChannelManager)
 
-	enabledChannels := services.ChannelManager.GetEnabledChannels()
+	enabledChannels := runningServices.ChannelManager.GetEnabledChannels()
 	if len(enabledChannels) > 0 {
 		fmt.Printf("  ✓ Channels enabled: %s\n", enabledChannels)
 	} else {
 		fmt.Println("  ⚠ Warning: No channels enabled")
 	}
 
-	// Setup HTTP server with new config
 	addr := fmt.Sprintf("%s:%d", cfg.Gateway.Host, cfg.Gateway.Port)
-	services.HealthServer = health.NewServer(cfg.Gateway.Host, cfg.Gateway.Port)
+	runningServices.HealthServer = health.NewServer(cfg.Gateway.Host, cfg.Gateway.Port)
-	services.ChannelManager.SetupHTTPServer(addr, services.HealthServer)
+	runningServices.ChannelManager.SetupHTTPServer(addr, runningServices.HealthServer)
 
-	// Use background context for lifecycle to ensure services persist after restartServices returns
+	if err = runningServices.ChannelManager.StartAll(context.Background()); err != nil {
-	if err = services.ChannelManager.StartAll(context.Background()); err != nil {
 		return fmt.Errorf("error restarting channels: %w", err)
 	}
 	fmt.Printf(
@@ -447,22 +438,20 @@ func restartServices(
 		cfg.Gateway.Port,
 	)
 
-	// Re-create device service with new config
 	stateManager := state.NewManager(cfg.WorkspacePath())
-	services.DeviceService = devices.NewService(devices.Config{
+	runningServices.DeviceService = devices.NewService(devices.Config{
 		Enabled:    cfg.Devices.Enabled,
 		MonitorUSB: cfg.Devices.MonitorUSB,
 	}, stateManager)
-	services.DeviceService.SetBus(msgBus)
+	runningServices.DeviceService.SetBus(msgBus)
-	if err := services.DeviceService.Start(context.Background()); err != nil {
+	if err := runningServices.DeviceService.Start(context.Background()); err != nil {
 		logger.WarnCF("device", "Failed to restart device service", map[string]any{"error": err.Error()})
 	} else if cfg.Devices.Enabled {
 		fmt.Println("  ✓ Device event service restarted")
 	}
 
-	// Wire up voice transcription with new config
 	transcriber := voice.DetectTranscriber(cfg)
-	al.SetTranscriber(transcriber) // This will set it to nil if disabled
+	al.SetTranscriber(transcriber)
 	if transcriber != nil {
 		logger.InfoCF("voice", "Transcription re-enabled (agent-level)", map[string]any{"provider": transcriber.Name()})
 	} else {
@@ -472,8 +461,6 @@ func restartServices(
 	return nil
 }
 
-// setupConfigWatcherPolling sets up a simple polling-based config file watcher
-// Returns a channel for config updates and a stop function
 func setupConfigWatcherPolling(configPath string, debug bool) (chan *config.Config, func()) {
 	configChan := make(chan *config.Config, 1)
 	stop := make(chan struct{})
@@ -483,11 +470,10 @@ func setupConfigWatcherPolling(configPath string, debug bool) (chan *config.Conf
 	go func() {
 		defer wg.Done()
 
-		// Get initial file info
 		lastModTime := getFileModTime(configPath)
 		lastSize := getFileSize(configPath)
 
-		ticker := time.NewTicker(2 * time.Second) // Check every 2 seconds
+		ticker := time.NewTicker(2 * time.Second)
 		defer ticker.Stop()
 
 		for {
@@ -496,20 +482,16 @@ func setupConfigWatcherPolling(configPath string, debug bool) (chan *config.Conf
 				currentModTime := getFileModTime(configPath)
 				currentSize := getFileSize(configPath)
 
-				// Check if file changed (modification time or size changed)
 				if currentModTime.After(lastModTime) || currentSize != lastSize {
 					if debug {
 						logger.Debugf("🔍 Config file change detected")
 					}
 
-					// Debounce - wait a bit to ensure file write is complete
 					time.Sleep(500 * time.Millisecond)
 
-					// Update last known state to prevent repeated reload attempts on failure
 					lastModTime = currentModTime
 					lastSize = currentSize
 
-					// Validate and load new config
 					newCfg, err := config.LoadConfig(configPath)
 					if err != nil {
 						logger.Errorf("⚠ Error loading new config: %v", err)
@@ -517,7 +499,6 @@ func setupConfigWatcherPolling(configPath string, debug bool) (chan *config.Conf
 						continue
 					}
 
-					// Validate the new config
 					if err := newCfg.ValidateModelList(); err != nil {
 						logger.Errorf("  ⚠ New config validation failed: %v", err)
 						logger.Warn("  Using previous valid config")
@@ -526,15 +507,12 @@ func setupConfigWatcherPolling(configPath string, debug bool) (chan *config.Conf
 
 					logger.Info("✓ Config file validated and loaded")
 
-					// Send new config to main loop (non-blocking)
 					select {
 					case configChan <- newCfg:
 					default:
-						// Channel full, skip this update
 						logger.Warn("⚠ Previous config reload still in progress, skipping")
 					}
 				}
-
 			case <-stop:
 				return
 			}
@@ -549,7 +527,6 @@ func setupConfigWatcherPolling(configPath string, debug bool) (chan *config.Conf
 	return configChan, stopFunc
 }
 
-// getFileModTime returns the modification time of a file, or zero time if file doesn't exist
 func getFileModTime(path string) time.Time {
 	info, err := os.Stat(path)
 	if err != nil {
@@ -558,7 +535,6 @@ func getFileModTime(path string) time.Time {
 	return info.ModTime()
 }
 
-// getFileSize returns the size of a file, or 0 if file doesn't exist
 func getFileSize(path string) int64 {
 	info, err := os.Stat(path)
 	if err != nil {
@@ -577,10 +553,8 @@ func setupCronTool(
 ) (*cron.CronService, error) {
 	cronStorePath := filepath.Join(workspace, "cron", "jobs.json")
 
-	// Create cron service
 	cronService := cron.NewCronService(cronStorePath, nil)
 
-	// Create and register CronTool if enabled
 	var cronTool *tools.CronTool
 	if cfg.Tools.IsToolEnabled("cron") {
 		var err error
@@ -592,7 +566,6 @@ func setupCronTool(
 		agentLoop.RegisterTool(cronTool)
 	}
 
-	// Set onJob handler
 	if cronTool != nil {
 		cronService.SetOnJob(func(job *cron.CronJob) (string, error) {
 			result := cronTool.ExecuteJob(context.Background(), job)
@@ -605,22 +578,17 @@ func setupCronTool(
 
 func createHeartbeatHandler(agentLoop *agent.AgentLoop) func(prompt, channel, chatID string) *tools.ToolResult {
 	return func(prompt, channel, chatID string) *tools.ToolResult {
-		// Use cli:direct as fallback if no valid channel
 		if channel == "" || chatID == "" {
 			channel, chatID = "cli", "direct"
 		}
-		// Use ProcessHeartbeat - no session history, each heartbeat is independent
+
-		var response string
+		response, err := agentLoop.ProcessHeartbeat(context.Background(), prompt, channel, chatID)
-		var err error
-		response, err = agentLoop.ProcessHeartbeat(context.Background(), prompt, channel, chatID)
 		if err != nil {
 			return tools.ErrorResult(fmt.Sprintf("Heartbeat error: %v", err))
 		}
 		if response == "HEARTBEAT_OK" {
 			return tools.SilentResult("Heartbeat OK")
 		}
-		// For heartbeat, always return silent - the subagent result will be
-		// sent to user via processSystemMessage when the async task completes
 		return tools.SilentResult(response)
 	}
 }