home-services/ha-gateway/cmd/gateway/main.go

package main

import (
	"context"
	"crypto/tls"
	"crypto/x509"
	"fmt"
	"net"
	"os"
	"os/signal"
	"path/filepath"
	"syscall"

	"github.com/joho/godotenv"
	"go.opentelemetry.io/contrib/instrumentation/google.golang.org/grpc/otelgrpc"
	"google.golang.org/grpc"
	"google.golang.org/grpc/credentials"
	"google.golang.org/grpc/health"
	grpc_health_v1 "google.golang.org/grpc/health/grpc_health_v1"
	"google.golang.org/grpc/reflection"

	hav1 "gitea.nik4nao.com/nik/home-services/gen/ha/v1"
	grpcadapter "gitea.nik4nao.com/nik/home-services/ha-gateway/internal/adapters/primary/grpc"
	"gitea.nik4nao.com/nik/home-services/ha-gateway/internal/adapters/secondary/ha"
	"gitea.nik4nao.com/nik/home-services/ha-gateway/internal/adapters/secondary/switchbot"
	"gitea.nik4nao.com/nik/home-services/ha-gateway/internal/app"
	"gitea.nik4nao.com/nik/home-services/ha-gateway/internal/config"
	"gitea.nik4nao.com/nik/home-services/ha-gateway/internal/logger"
	"gitea.nik4nao.com/nik/home-services/ha-gateway/internal/telemetry"
)

// MEMO: auth is not implemented.
// Add one of the following before exposing this service to any untrusted network:
//   Option A — shared API key per client: unary + stream interceptors read
//              "authorization" from gRPC metadata and compare to a secret
//              from config. Good for small number of known clients.
//   Option B — mTLS (recommended): tls.Config with ClientAuth: RequireAndVerifyClientCert,
//              cert pool from the internal CA. Each client gets a cert from
//              cert-manager. No runtime auth dependency, identity in the cert CN/SAN.

// version is set at build time via -ldflags "-X main.version=<tag>".
var version = "dev"

func main() {
	_ = godotenv.Load()

	// Config is loaded before logger setup so fatal startup errors still stop early.
	cfg, err := config.Load()
	if err != nil {
		os.Stderr.WriteString("config error: " + err.Error() + "\n")
		os.Exit(1)
	}

	log := logger.New(cfg.LogFormat, cfg.LogLevel)
	log.Info("starting ha-gateway",
		"version", version,
		"grpc_port", cfg.GRPCPort,
		"ha_base_url", cfg.HABaseURL,
		"ha_token", redactToken(cfg.HAToken),
		"tls_dir", cfg.TLSDir,
		"otel_endpoint", cfg.OTELEndpoint,
		"log_level", cfg.LogLevel,
		"log_format", cfg.LogFormat,
		"switchbot_configured", cfg.SwitchBotToken != "" && cfg.SwitchBotSecret != "",
	)

	ctx, stop := signal.NotifyContext(context.Background(), syscall.SIGTERM, syscall.SIGINT)
	defer stop()
	ctx = logger.WithLogger(ctx, log)

	// Telemetry is optional; an empty OTEL endpoint installs no-op providers.
	shutdown, err := telemetry.Setup(ctx, "ha-gateway", version, cfg)
	if err != nil {
		log.Error("telemetry setup failed", "err", err)
		os.Exit(1)
	}
	if cfg.OTELEndpoint != "" {
		log.Info("telemetry enabled", "endpoint", cfg.OTELEndpoint)
	} else {
		log.Debug("telemetry disabled")
	}

	haClient := ha.NewClient(cfg, log)
	switchBotClient := switchbot.NewClient(cfg, log)

	// App services stay free of gRPC and HTTP details; adapters are wired here.
	entityApp := app.NewEntityApp(haClient)
	lightApp := app.NewLightApp(haClient)
	switchApp := app.NewSwitchApp(haClient)
	climateApp := app.NewClimateApp(haClient)
	remoteApp := app.NewRemoteApp(switchBotClient)

	if err := lightApp.Refresh(ctx); err != nil {
		log.Warn("initial light discovery failed, will retry on first request", "err", err)
	}
	if err := switchApp.Refresh(ctx); err != nil {
		log.Warn("initial switch discovery failed, will retry on first request", "err", err)
	}
	if err := climateApp.Refresh(ctx); err != nil {
		log.Warn("initial climate discovery failed, will retry on first request", "err", err)
	}

	serverOpts := []grpc.ServerOption{
		grpc.StatsHandler(otelgrpc.NewServerHandler()),
		grpc.ChainUnaryInterceptor(grpcadapter.LoggingUnaryInterceptor(log)),
		grpc.ChainStreamInterceptor(grpcadapter.LoggingStreamInterceptor(log)),
	}
	if cfg.TLSDir != "" {
		creds, err := loadServerCredentials(cfg.TLSDir)
		if err != nil {
			log.Error("load mTLS credentials failed", "tls_dir", cfg.TLSDir, "err", err)
			os.Exit(1)
		}
		serverOpts = append(serverOpts, grpc.Creds(creds))
		log.Info("mTLS enabled", "tls_dir", cfg.TLSDir)
	} else {
		log.Info("mTLS disabled")
	}

	srv := grpc.NewServer(serverOpts...)
	healthSrv := health.NewServer()
	healthSrv.SetServingStatus("", grpc_health_v1.HealthCheckResponse_SERVING)

	hav1.RegisterEntityServiceServer(srv, grpcadapter.NewEntityGRPC(entityApp))
	hav1.RegisterLightServiceServer(srv, grpcadapter.NewLightGRPC(lightApp))
	hav1.RegisterSwitchServiceServer(srv, grpcadapter.NewSwitchGRPC(switchApp))
	hav1.RegisterClimateServiceServer(srv, grpcadapter.NewClimateGRPC(climateApp))
	hav1.RegisterRemoteServiceServer(srv, grpcadapter.NewRemoteGRPC(remoteApp))
	hav1.RegisterEventServiceServer(srv, &grpcadapter.EventGRPC{})
	grpc_health_v1.RegisterHealthServer(srv, healthSrv)
	reflection.Register(srv)

	lis, err := net.Listen("tcp", ":"+cfg.GRPCPort)
	if err != nil {
		log.Error("listen failed", "err", err)
		os.Exit(1)
	}

	go func() {
		log.Info("ha-gateway listening", "addr", lis.Addr().String())
		if err := srv.Serve(lis); err != nil {
			log.Error("serve failed", "err", err)
		}
	}()

	<-ctx.Done()
	log.Info("shutdown signal received, draining")

	healthSrv.SetServingStatus("", grpc_health_v1.HealthCheckResponse_NOT_SERVING)
	srv.GracefulStop()
	log.Info("shutdown complete")

	if err := shutdown(context.Background()); err != nil {
		log.Error("telemetry shutdown error", "err", err)
	}
}

// redactToken logs only a short prefix so startup logs remain useful without
// leaking a full Home Assistant credential.
func redactToken(token string) string {
	if token == "" {
		return "[not set]"
	}
	if len(token) <= 8 {
		return token + "..."
	}
	return token[:8] + "..."
}

func loadServerCredentials(tlsDir string) (credentials.TransportCredentials, error) {
	cert, err := tls.LoadX509KeyPair(
		filepath.Join(tlsDir, "tls.crt"),
		filepath.Join(tlsDir, "tls.key"),
	)
	if err != nil {
		return nil, fmt.Errorf("load server key pair: %w", err)
	}

	caPEM, err := os.ReadFile(filepath.Join(tlsDir, "ca.crt"))
	if err != nil {
		return nil, fmt.Errorf("read client CA: %w", err)
	}

	clientCAs := x509.NewCertPool()
	if !clientCAs.AppendCertsFromPEM(caPEM) {
		return nil, fmt.Errorf("append client CA: invalid PEM")
	}

	return credentials.NewTLS(&tls.Config{
		Certificates: []tls.Certificate{cert},
		ClientCAs:    clientCAs,
		ClientAuth:   tls.RequireAndVerifyClientCert,
		MinVersion:   tls.VersionTLS13,
	}), nil
}