Initial commit: notifications, monitoring, on-call and incidents

One binary and one SQLite file. Topics with ntfy-style publish and
subscribe, HTTP/TCP/DNS/TLS/heartbeat monitors with a public status page,
schedules and escalation policies, and incidents with a timeline and
post-mortem. The UI follows the Lightwell design language.

Builds against an unreleased Reef (secretbox) through a local replace line.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

internal/sandbar/server.go

@@ -0,0 +1,119 @@
+package sandbar
+
+import (
+	_ "embed"
+	"log"
+	"net/http"
+	"sync"
+	"time"
+)
+
+//go:embed web/index.html
+var indexHTML []byte
+
+//go:embed web/status.html
+var statusHTML []byte
+
+type SMTPConfig struct{ Host, Port, User, Password, From string }
+
+// Options configures a Server. Zero values are sensible defaults.
+type Options struct {
+	DataDir      string
+	URL          string // public URL, used for links in notifications
+	MasterKey    string
+	MetricsToken string
+	SMTP         SMTPConfig
+	MessageDays  int // how long topic messages are kept (default 30)
+	CheckDays    int // how long monitor results are kept (default 30)
+
+	// Tests replace these.
+	Now        func() time.Time
+	HTTPClient *http.Client
+	Mailer     func(to, subject, body string) error
+}
+
+type Server struct {
+	opts Options
+	st   *Store
+	hub  *hub
+
+	// monitor runtime state
+	mu      sync.Mutex
+	running map[string]bool // monitor ids with a check in flight
+
+	loginMu  sync.Mutex
+	loginTry map[string][]time.Time
+
+	stop chan struct{}
+	wg   sync.WaitGroup
+}
+
+func New(opts Options, st *Store) *Server {
+	if opts.MessageDays <= 0 {
+		opts.MessageDays = 30
+	}
+	if opts.CheckDays <= 0 {
+		opts.CheckDays = 30
+	}
+	return &Server{opts: opts, st: st, hub: newHub(), running: map[string]bool{}, loginTry: map[string][]time.Time{}, stop: make(chan struct{})}
+}
+
+func (s *Server) now() time.Time {
+	if s.opts.Now != nil {
+		return s.opts.Now()
+	}
+	return time.Now()
+}
+
+// Start runs the background workers: monitor scheduler, escalation, pruning.
+func (s *Server) Start() {
+	s.every(time.Second, func() { s.runDueMonitors(s.now()) })
+	s.every(5*time.Second, func() { s.escalate(s.now()) })
+	s.every(time.Hour, func() { s.st.prune(s.now(), s.opts.MessageDays, s.opts.CheckDays) })
+}
+
+func (s *Server) Close() {
+	close(s.stop)
+	s.wg.Wait()
+}
+
+func (s *Server) every(d time.Duration, f func()) {
+	s.wg.Add(1)
+	go func() {
+		defer s.wg.Done()
+		t := time.NewTicker(d)
+		defer t.Stop()
+		for {
+			select {
+			case <-s.stop:
+				return
+			case <-t.C:
+				func() {
+					defer func() {
+						if r := recover(); r != nil {
+							log.Printf("worker panic: %v", r)
+						}
+					}()
+					f()
+				}()
+			}
+		}
+	}()
+}
+
+func (s *Server) link(path string) string {
+	if s.opts.URL == "" {
+		return ""
+	}
+	return s.opts.URL + path
+}
+
+// BootstrapAdmin creates the first administrator from the environment when there are no users yet.
+func (s *Server) BootstrapAdmin(email, password string) {
+	if email == "" || password == "" || s.countUsers() > 0 {
+		return
+	}
+	if _, err := s.createUser("Admin", email, password, "admin", ""); err != nil {
+		log.Printf("SANDBAR_ADMIN_EMAIL: %v", err)
+	}
+}