Function RunServer

cmd/server.go:62–274

RunServer executes the 'chatcli server' subcommand.

(args []string, llmMgr manager.LLMManager, logger *zap.Logger)

// RunServer executes the 'chatcli server' subcommand.
func RunServer(args []string, llmMgr manager.LLMManager, logger *zap.Logger) error {
	fs := flag.NewFlagSet("server", flag.ContinueOnError)

	opts := &ServerOptions{}
	fs.IntVar(&opts.Port, "port", getEnvInt("CHATCLI_SERVER_PORT", 50051), "gRPC server port")
	fs.StringVar(&opts.Token, "token", os.Getenv("CHATCLI_SERVER_TOKEN"), "Authentication token (empty = no auth)")
	fs.StringVar(&opts.CertFile, "tls-cert", os.Getenv("CHATCLI_SERVER_TLS_CERT"), "TLS certificate file path")
	fs.StringVar(&opts.KeyFile, "tls-key", os.Getenv("CHATCLI_SERVER_TLS_KEY"), "TLS key file path")
	fs.StringVar(&opts.Provider, "provider", os.Getenv("LLM_PROVIDER"), "Default LLM provider")
	fs.StringVar(&opts.Model, "model", "", "Default LLM model")
	fs.IntVar(&opts.MetricsPort, "metrics-port", getEnvInt("CHATCLI_METRICS_PORT", 9090), "Prometheus metrics HTTP port (0 = disabled)")

	// Fallback chain flags
	fs.StringVar(&opts.FallbackProviders, "fallback-providers", os.Getenv("CHATCLI_FALLBACK_PROVIDERS"), "Comma-separated fallback providers (e.g. OPENAI,CLAUDEAI,GOOGLEAI,ZAI,MINIMAX)")
	fs.IntVar(&opts.FallbackMaxRetries, "fallback-max-retries", getEnvInt("CHATCLI_FALLBACK_MAX_RETRIES", 2), "Max retries per provider before fallback")
	fs.DurationVar(&opts.FallbackCooldownBase, "fallback-cooldown-base", getEnvDuration("CHATCLI_FALLBACK_COOLDOWN_BASE", 30*time.Second), "Base cooldown duration after provider failure")
	fs.DurationVar(&opts.FallbackCooldownMax, "fallback-cooldown-max", getEnvDuration("CHATCLI_FALLBACK_COOLDOWN_MAX", 5*time.Minute), "Maximum cooldown duration")

	// MCP flags
	fs.StringVar(&opts.MCPConfigPath, "mcp-config", os.Getenv("CHATCLI_MCP_CONFIG"), "Path to MCP servers config JSON")

	// K8s watcher flags
	fs.StringVar(&opts.WatchDeployment, "watch-deployment", os.Getenv("CHATCLI_WATCH_DEPLOYMENT"), "K8s deployment to monitor (enables watcher)")
	fs.StringVar(&opts.WatchNamespace, "watch-namespace", getEnvOrDefault("CHATCLI_WATCH_NAMESPACE", "default"), "K8s namespace for watcher")
	fs.DurationVar(&opts.WatchInterval, "watch-interval", getEnvDuration("CHATCLI_WATCH_INTERVAL", 30*time.Second), "Watcher collection interval")
	fs.DurationVar(&opts.WatchWindow, "watch-window", getEnvDuration("CHATCLI_WATCH_WINDOW", 2*time.Hour), "Watcher observation window")
	fs.IntVar(&opts.WatchMaxLogs, "watch-max-log-lines", getEnvInt("CHATCLI_WATCH_MAX_LOG_LINES", 100), "Max log lines per pod")
	fs.StringVar(&opts.WatchKubeconfig, "watch-kubeconfig", os.Getenv("CHATCLI_KUBECONFIG"), "Path to kubeconfig for watcher")
	fs.StringVar(&opts.WatchConfig, "watch-config", os.Getenv("CHATCLI_WATCH_CONFIG"), "Path to multi-target watch config YAML")

	if err := fs.Parse(args); err != nil {
		return err
	}

	// Resolve provider if not set
	if opts.Provider == "" {
		available := llmMgr.GetAvailableProviders()
		if len(available) > 0 {
			opts.Provider = available[0]
		}
	}

	// Resolve model name from the client if possible
	if opts.Model == "" && opts.Provider != "" {
		if c, err := llmMgr.GetClient(opts.Provider, ""); err == nil {
			opts.Model = c.GetModelName()
		}
	}

	// Create session manager for server-side session persistence
	sessionMgr, err := cli.NewSessionManager(logger)
	if err != nil {
		logger.Warn(i18n.T("cmd.server.session_init_failed"), zap.Error(err))
	}

	cfg := server.Config{
		Port:        opts.Port,
		Token:       opts.Token,