rename /src to /internal (sorry i'll fix the prs)

internal/hub/agent_connect.go

@@ -0,0 +1,321 @@
+package hub
+
+import (
+	"errors"
+	"net"
+	"net/http"
+	"strings"
+	"sync"
+	"time"
+
+	"github.com/henrygd/beszel/src/common"
+	"github.com/henrygd/beszel/src/hub/expirymap"
+	"github.com/henrygd/beszel/src/hub/ws"
+
+	"github.com/blang/semver"
+	"github.com/lxzan/gws"
+	"github.com/pocketbase/dbx"
+	"github.com/pocketbase/pocketbase/core"
+)
+
+// agentConnectRequest holds information related to an agent's connection attempt.
+type agentConnectRequest struct {
+	hub         *Hub
+	req         *http.Request
+	res         http.ResponseWriter
+	token       string
+	agentSemVer semver.Version
+	// isUniversalToken is true if the token is a universal token.
+	isUniversalToken bool
+	// userId is the user ID associated with the universal token.
+	userId string
+}
+
+// universalTokenMap stores active universal tokens and their associated user IDs.
+var universalTokenMap tokenMap
+
+type tokenMap struct {
+	store *expirymap.ExpiryMap[string]
+	once  sync.Once
+}
+
+// getMap returns the expirymap, creating it if necessary.
+func (tm *tokenMap) GetMap() *expirymap.ExpiryMap[string] {
+	tm.once.Do(func() {
+		tm.store = expirymap.New[string](time.Hour)
+	})
+	return tm.store
+}
+
+// handleAgentConnect is the HTTP handler for an agent's connection request.
+func (h *Hub) handleAgentConnect(e *core.RequestEvent) error {
+	agentRequest := agentConnectRequest{req: e.Request, res: e.Response, hub: h}
+	_ = agentRequest.agentConnect()
+	return nil
+}
+
+// agentConnect validates agent credentials and upgrades the connection to a WebSocket.
+func (acr *agentConnectRequest) agentConnect() (err error) {
+	var agentVersion string
+
+	acr.token, agentVersion, err = acr.validateAgentHeaders(acr.req.Header)
+	if err != nil {
+		return acr.sendResponseError(acr.res, http.StatusBadRequest, "")
+	}
+
+	// Check if token is an active universal token
+	acr.userId, acr.isUniversalToken = universalTokenMap.GetMap().GetOk(acr.token)
+
+	// Find matching fingerprint records for this token
+	fpRecords := getFingerprintRecordsByToken(acr.token, acr.hub)
+	if len(fpRecords) == 0 && !acr.isUniversalToken {
+		// Invalid token - no records found and not a universal token
+		return acr.sendResponseError(acr.res, http.StatusUnauthorized, "Invalid token")
+	}
+
+	// Validate agent version
+	acr.agentSemVer, err = semver.Parse(agentVersion)
+	if err != nil {
+		return acr.sendResponseError(acr.res, http.StatusUnauthorized, "Invalid agent version")
+	}
+
+	// Upgrade connection to WebSocket
+	conn, err := ws.GetUpgrader().Upgrade(acr.res, acr.req)
+	if err != nil {
+		return acr.sendResponseError(acr.res, http.StatusInternalServerError, "WebSocket upgrade failed")
+	}
+
+	go acr.verifyWsConn(conn, fpRecords)
+
+	return nil
+}
+
+// verifyWsConn verifies the WebSocket connection using the agent's fingerprint and
+// SSH key signature, then adds the system to the system manager.
+func (acr *agentConnectRequest) verifyWsConn(conn *gws.Conn, fpRecords []ws.FingerprintRecord) (err error) {
+	wsConn := ws.NewWsConnection(conn)
+
+	// must set wsConn in connection store before the read loop
+	conn.Session().Store("wsConn", wsConn)
+
+	// make sure connection is closed if there is an error
+	defer func() {
+		if err != nil {
+			wsConn.Close([]byte(err.Error()))
+		}
+	}()
+
+	go conn.ReadLoop()
+
+	signer, err := acr.hub.GetSSHKey("")
+	if err != nil {
+		return err
+	}
+
+	agentFingerprint, err := wsConn.GetFingerprint(acr.token, signer, acr.isUniversalToken)
+	if err != nil {
+		return err
+	}
+
+	// Find or create the appropriate system for this token and fingerprint
+	fpRecord, err := acr.findOrCreateSystemForToken(fpRecords, agentFingerprint)
+	if err != nil {
+		return err
+	}
+
+	return acr.hub.sm.AddWebSocketSystem(fpRecord.SystemId, acr.agentSemVer, wsConn)
+}
+
+// validateAgentHeaders extracts and validates the token and agent version from HTTP headers.
+func (acr *agentConnectRequest) validateAgentHeaders(headers http.Header) (string, string, error) {
+	token := headers.Get("X-Token")
+	agentVersion := headers.Get("X-Beszel")
+
+	if agentVersion == "" || token == "" || len(token) > 64 {
+		return "", "", errors.New("")
+	}
+	return token, agentVersion, nil
+}
+
+// sendResponseError writes an HTTP error response.
+func (acr *agentConnectRequest) sendResponseError(res http.ResponseWriter, code int, message string) error {
+	res.WriteHeader(code)
+	if message != "" {
+		res.Write([]byte(message))
+	}
+	return nil
+}
+
+// getFingerprintRecordsByToken retrieves all fingerprint records associated with a given token.
+func getFingerprintRecordsByToken(token string, h *Hub) []ws.FingerprintRecord {
+	var records []ws.FingerprintRecord
+	// All will populate empty slice even on error
+	_ = h.DB().NewQuery("SELECT id, system, fingerprint, token FROM fingerprints WHERE token = {:token}").
+		Bind(dbx.Params{
+			"token": token,
+		}).
+		All(&records)
+	return records
+}
+
+// findOrCreateSystemForToken finds an existing system matching the token and fingerprint,
+// or creates a new one for a universal token.
+func (acr *agentConnectRequest) findOrCreateSystemForToken(fpRecords []ws.FingerprintRecord, agentFingerprint common.FingerprintResponse) (ws.FingerprintRecord, error) {
+	// No records - only valid for active universal tokens
+	if len(fpRecords) == 0 {
+		return acr.handleNoRecords(agentFingerprint)
+	}
+
+	// Single record - handle as regular token
+	if len(fpRecords) == 1 && !acr.isUniversalToken {
+		return acr.handleSingleRecord(fpRecords[0], agentFingerprint)
+	}
+
+	// Multiple records or universal token - look for matching fingerprint
+	return acr.handleMultipleRecordsOrUniversalToken(fpRecords, agentFingerprint)
+}
+
+// handleNoRecords handles the case where no fingerprint records are found for a token.
+// A new system is created if the token is a valid universal token.
+func (acr *agentConnectRequest) handleNoRecords(agentFingerprint common.FingerprintResponse) (ws.FingerprintRecord, error) {
+	var fpRecord ws.FingerprintRecord
+
+	if !acr.isUniversalToken || acr.userId == "" {
+		return fpRecord, errors.New("no matching fingerprints")
+	}
+
+	return acr.createNewSystemForUniversalToken(agentFingerprint)
+}
+
+// handleSingleRecord handles the case with a single fingerprint record. It validates
+// the agent's fingerprint against the stored one, or sets it on first connect.
+func (acr *agentConnectRequest) handleSingleRecord(fpRecord ws.FingerprintRecord, agentFingerprint common.FingerprintResponse) (ws.FingerprintRecord, error) {
+	// If no current fingerprint, update with new fingerprint (first time connecting)
+	if fpRecord.Fingerprint == "" {
+		if err := acr.hub.SetFingerprint(&fpRecord, agentFingerprint.Fingerprint); err != nil {
+			return fpRecord, err
+		}
+		// Update the record with the fingerprint that was set
+		fpRecord.Fingerprint = agentFingerprint.Fingerprint
+		return fpRecord, nil
+	}
+
+	// Abort if fingerprint exists but doesn't match (different machine)
+	if fpRecord.Fingerprint != agentFingerprint.Fingerprint {
+		return fpRecord, errors.New("fingerprint mismatch")
+	}
+
+	return fpRecord, nil
+}
+
+// handleMultipleRecordsOrUniversalToken finds a matching fingerprint from multiple records.
+// If no match is found and the token is a universal token, a new system is created.
+func (acr *agentConnectRequest) handleMultipleRecordsOrUniversalToken(fpRecords []ws.FingerprintRecord, agentFingerprint common.FingerprintResponse) (ws.FingerprintRecord, error) {
+	// Return existing record with matching fingerprint if found
+	for i := range fpRecords {
+		if fpRecords[i].Fingerprint == agentFingerprint.Fingerprint {
+			return fpRecords[i], nil
+		}
+	}
+
+	// No matching fingerprint record found, but it's
+	// an active universal token so create a new system
+	if acr.isUniversalToken {
+		return acr.createNewSystemForUniversalToken(agentFingerprint)
+	}
+
+	return ws.FingerprintRecord{}, errors.New("fingerprint mismatch")
+}
+
+// createNewSystemForUniversalToken creates a new system and fingerprint record for a universal token.
+func (acr *agentConnectRequest) createNewSystemForUniversalToken(agentFingerprint common.FingerprintResponse) (ws.FingerprintRecord, error) {
+	var fpRecord ws.FingerprintRecord
+	if !acr.isUniversalToken || acr.userId == "" {
+		return fpRecord, errors.New("invalid token")
+	}
+
+	fpRecord.Token = acr.token
+
+	systemId, err := acr.createSystem(agentFingerprint)
+	if err != nil {
+		return fpRecord, err
+	}
+	fpRecord.SystemId = systemId
+
+	// Set the fingerprint for the new system
+	if err := acr.hub.SetFingerprint(&fpRecord, agentFingerprint.Fingerprint); err != nil {
+		return fpRecord, err
+	}
+
+	// Update the record with the fingerprint that was set
+	fpRecord.Fingerprint = agentFingerprint.Fingerprint
+
+	return fpRecord, nil
+}
+
+// createSystem creates a new system record in the database using details from the agent.
+func (acr *agentConnectRequest) createSystem(agentFingerprint common.FingerprintResponse) (recordId string, err error) {
+	systemsCollection, err := acr.hub.FindCachedCollectionByNameOrId("systems")
+	if err != nil {
+		return "", err
+	}
+	remoteAddr := getRealIP(acr.req)
+	// separate port from address
+	if agentFingerprint.Hostname == "" {
+		agentFingerprint.Hostname = remoteAddr
+	}
+	if agentFingerprint.Port == "" {
+		agentFingerprint.Port = "45876"
+	}
+	// create new record
+	systemRecord := core.NewRecord(systemsCollection)
+	systemRecord.Set("name", agentFingerprint.Hostname)
+	systemRecord.Set("host", remoteAddr)
+	systemRecord.Set("port", agentFingerprint.Port)
+	systemRecord.Set("users", []string{acr.userId})
+
+	return systemRecord.Id, acr.hub.Save(systemRecord)
+}
+
+// SetFingerprint creates or updates a fingerprint record in the database.
+func (h *Hub) SetFingerprint(fpRecord *ws.FingerprintRecord, fingerprint string) (err error) {
+	// // can't use raw query here because it doesn't trigger SSE
+	var record *core.Record
+	switch fpRecord.Id {
+	case "":
+		// create new record for universal token
+		collection, _ := h.FindCachedCollectionByNameOrId("fingerprints")
+		record = core.NewRecord(collection)
+		record.Set("system", fpRecord.SystemId)
+	default:
+		record, err = h.FindRecordById("fingerprints", fpRecord.Id)
+	}
+	if err != nil {
+		return err
+	}
+	record.Set("token", fpRecord.Token)
+	record.Set("fingerprint", fingerprint)
+	return h.SaveNoValidate(record)
+}
+
+// getRealIP extracts the client's real IP address from request headers,
+// checking common proxy headers before falling back to the remote address.
+func getRealIP(r *http.Request) string {
+	if ip := r.Header.Get("CF-Connecting-IP"); ip != "" {
+		return ip
+	}
+	if ip := r.Header.Get("X-Forwarded-For"); ip != "" {
+		// X-Forwarded-For can contain a comma-separated list: "client_ip, proxy1, proxy2"
+		// Take the first one
+		ips := strings.Split(ip, ",")
+		if len(ips) > 0 {
+			return strings.TrimSpace(ips[0])
+		}
+	}
+	// Fallback to RemoteAddr
+	ip, _, err := net.SplitHostPort(r.RemoteAddr)
+	if err != nil {
+		return r.RemoteAddr
+	}
+	return ip
+}

internal/hub/config/config.go

@@ -0,0 +1,290 @@
+// Package config provides functions for syncing systems with the config.yml file
+package config
+
+import (
+	"fmt"
+	"log"
+	"os"
+	"path/filepath"
+
+	"github.com/google/uuid"
+	"github.com/henrygd/beszel/src/entities/system"
+
+	"github.com/pocketbase/dbx"
+	"github.com/pocketbase/pocketbase/core"
+	"github.com/spf13/cast"
+	"gopkg.in/yaml.v3"
+)
+
+type config struct {
+	Systems []systemConfig `yaml:"systems"`
+}
+
+type systemConfig struct {
+	Name  string   `yaml:"name"`
+	Host  string   `yaml:"host"`
+	Port  uint16   `yaml:"port,omitempty"`
+	Token string   `yaml:"token,omitempty"`
+	Users []string `yaml:"users"`
+}
+
+// Syncs systems with the config.yml file
+func SyncSystems(e *core.ServeEvent) error {
+	h := e.App
+	configPath := filepath.Join(h.DataDir(), "config.yml")
+	configData, err := os.ReadFile(configPath)
+	if err != nil {
+		return nil
+	}
+
+	var config config
+	err = yaml.Unmarshal(configData, &config)
+	if err != nil {
+		return fmt.Errorf("failed to parse config.yml: %v", err)
+	}
+
+	if len(config.Systems) == 0 {
+		log.Println("No systems defined in config.yml.")
+		return nil
+	}
+
+	var firstUser *core.Record
+
+	// Create a map of email to user ID
+	userEmailToID := make(map[string]string)
+	users, err := h.FindAllRecords("users", dbx.NewExp("id != ''"))
+	if err != nil {
+		return err
+	}
+	if len(users) > 0 {
+		firstUser = users[0]
+		for _, user := range users {
+			userEmailToID[user.GetString("email")] = user.Id
+		}
+	}
+
+	// add default settings for systems if not defined in config
+	for i := range config.Systems {
+		system := &config.Systems[i]
+		if system.Port == 0 {
+			system.Port = 45876
+		}
+		if len(users) > 0 && len(system.Users) == 0 {
+			// default to first user if none are defined
+			system.Users = []string{firstUser.Id}
+		} else {
+			// Convert email addresses to user IDs
+			userIDs := make([]string, 0, len(system.Users))
+			for _, email := range system.Users {
+				if id, ok := userEmailToID[email]; ok {
+					userIDs = append(userIDs, id)
+				} else {
+					log.Printf("User %s not found", email)
+				}
+			}
+			system.Users = userIDs
+		}
+	}
+
+	// Get existing systems
+	existingSystems, err := h.FindAllRecords("systems", dbx.NewExp("id != ''"))
+	if err != nil {
+		return err
+	}
+
+	// Create a map of existing systems
+	existingSystemsMap := make(map[string]*core.Record)
+	for _, system := range existingSystems {
+		key := system.GetString("name") + system.GetString("host") + system.GetString("port")
+		existingSystemsMap[key] = system
+	}
+
+	// Process systems from config
+	for _, sysConfig := range config.Systems {
+		key := sysConfig.Name + sysConfig.Host + cast.ToString(sysConfig.Port)
+		if existingSystem, ok := existingSystemsMap[key]; ok {
+			// Update existing system
+			existingSystem.Set("name", sysConfig.Name)
+			existingSystem.Set("users", sysConfig.Users)
+			existingSystem.Set("port", sysConfig.Port)
+			if err := h.Save(existingSystem); err != nil {
+				return err
+			}
+
+			// Only update token if one is specified in config, otherwise preserve existing token
+			if sysConfig.Token != "" {
+				if err := updateFingerprintToken(h, existingSystem.Id, sysConfig.Token); err != nil {
+					return err
+				}
+			}
+
+			delete(existingSystemsMap, key)
+		} else {
+			// Create new system
+			systemsCollection, err := h.FindCollectionByNameOrId("systems")
+			if err != nil {
+				return fmt.Errorf("failed to find systems collection: %v", err)
+			}
+			newSystem := core.NewRecord(systemsCollection)
+			newSystem.Set("name", sysConfig.Name)
+			newSystem.Set("host", sysConfig.Host)
+			newSystem.Set("port", sysConfig.Port)
+			newSystem.Set("users", sysConfig.Users)
+			newSystem.Set("info", system.Info{})
+			newSystem.Set("status", "pending")
+			if err := h.Save(newSystem); err != nil {
+				return fmt.Errorf("failed to create new system: %v", err)
+			}
+
+			// For new systems, generate token if not provided
+			token := sysConfig.Token
+			if token == "" {
+				token = uuid.New().String()
+			}
+
+			// Create fingerprint record for new system
+			if err := createFingerprintRecord(h, newSystem.Id, token); err != nil {
+				return err
+			}
+		}
+	}
+
+	// Delete systems not in config (and their fingerprint records will cascade delete)
+	for _, system := range existingSystemsMap {
+		if err := h.Delete(system); err != nil {
+			return err
+		}
+	}
+
+	log.Println("Systems synced with config.yml")
+	return nil
+}
+
+// Generates content for the config.yml file as a YAML string
+func generateYAML(h core.App) (string, error) {
+	// Fetch all systems from the database
+	systems, err := h.FindRecordsByFilter("systems", "id != ''", "name", -1, 0)
+	if err != nil {
+		return "", err
+	}
+
+	// Create a Config struct to hold the data
+	config := config{
+		Systems: make([]systemConfig, 0, len(systems)),
+	}
+
+	// Fetch all users at once
+	allUserIDs := make([]string, 0)
+	for _, system := range systems {
+		allUserIDs = append(allUserIDs, system.GetStringSlice("users")...)
+	}
+	userEmailMap, err := getUserEmailMap(h, allUserIDs)
+	if err != nil {
+		return "", err
+	}
+
+	// Fetch all fingerprint records to get tokens
+	type fingerprintData struct {
+		ID     string `db:"id"`
+		System string `db:"system"`
+		Token  string `db:"token"`
+	}
+	var fingerprints []fingerprintData
+	err = h.DB().NewQuery("SELECT id, system, token FROM fingerprints").All(&fingerprints)
+	if err != nil {
+		return "", err
+	}
+
+	// Create a map of system ID to token
+	systemTokenMap := make(map[string]string)
+	for _, fingerprint := range fingerprints {
+		systemTokenMap[fingerprint.System] = fingerprint.Token
+	}
+
+	// Populate the Config struct with system data
+	for _, system := range systems {
+		userIDs := system.GetStringSlice("users")
+		userEmails := make([]string, 0, len(userIDs))
+		for _, userID := range userIDs {
+			if email, ok := userEmailMap[userID]; ok {
+				userEmails = append(userEmails, email)
+			}
+		}
+
+		sysConfig := systemConfig{
+			Name:  system.GetString("name"),
+			Host:  system.GetString("host"),
+			Port:  cast.ToUint16(system.Get("port")),
+			Users: userEmails,
+			Token: systemTokenMap[system.Id],
+		}
+		config.Systems = append(config.Systems, sysConfig)
+	}
+
+	// Marshal the Config struct to YAML
+	yamlData, err := yaml.Marshal(&config)
+	if err != nil {
+		return "", err
+	}
+
+	// Add a header to the YAML
+	yamlData = append([]byte("# Values for port, users, and token are optional.\n# Defaults are port 45876, the first created user, and a generated UUID token.\n\n"), yamlData...)
+
+	return string(yamlData), nil
+}
+
+// New helper function to get a map of user IDs to emails
+func getUserEmailMap(h core.App, userIDs []string) (map[string]string, error) {
+	users, err := h.FindRecordsByIds("users", userIDs)
+	if err != nil {
+		return nil, err
+	}
+
+	userEmailMap := make(map[string]string, len(users))
+	for _, user := range users {
+		userEmailMap[user.Id] = user.GetString("email")
+	}
+
+	return userEmailMap, nil
+}
+
+// Helper function to update or create fingerprint token for an existing system
+func updateFingerprintToken(app core.App, systemID, token string) error {
+	// Try to find existing fingerprint record
+	fingerprint, err := app.FindFirstRecordByFilter("fingerprints", "system = {:system}", dbx.Params{"system": systemID})
+	if err != nil {
+		// If no fingerprint record exists, create one
+		return createFingerprintRecord(app, systemID, token)
+	}
+
+	// Update existing fingerprint record with new token (keep existing fingerprint)
+	fingerprint.Set("token", token)
+	return app.Save(fingerprint)
+}
+
+// Helper function to create a new fingerprint record for a system
+func createFingerprintRecord(app core.App, systemID, token string) error {
+	fingerprintsCollection, err := app.FindCollectionByNameOrId("fingerprints")
+	if err != nil {
+		return fmt.Errorf("failed to find fingerprints collection: %v", err)
+	}
+
+	newFingerprint := core.NewRecord(fingerprintsCollection)
+	newFingerprint.Set("system", systemID)
+	newFingerprint.Set("token", token)
+	newFingerprint.Set("fingerprint", "") // Empty fingerprint, will be set on first connection
+
+	return app.Save(newFingerprint)
+}
+
+// Returns the current config.yml file as a JSON object
+func GetYamlConfig(e *core.RequestEvent) error {
+	if e.Auth.GetString("role") != "admin" {
+		return e.ForbiddenError("Requires admin role", nil)
+	}
+	configContent, err := generateYAML(e.App)
+	if err != nil {
+		return err
+	}
+	return e.JSON(200, map[string]string{"config": configContent})
+}

internal/hub/config/config_test.go

@@ -0,0 +1,247 @@
+//go:build testing
+// +build testing
+
+package config_test
+
+import (
+	"os"
+	"path/filepath"
+	"testing"
+
+	"github.com/henrygd/beszel/src/tests"
+
+	"github.com/henrygd/beszel/src/hub/config"
+
+	"github.com/pocketbase/pocketbase/core"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+	"gopkg.in/yaml.v3"
+)
+
+// Config struct for testing (copied from config package since it's not exported)
+type testConfig struct {
+	Systems []testSystemConfig `yaml:"systems"`
+}
+
+type testSystemConfig struct {
+	Name  string   `yaml:"name"`
+	Host  string   `yaml:"host"`
+	Port  uint16   `yaml:"port,omitempty"`
+	Users []string `yaml:"users"`
+	Token string   `yaml:"token,omitempty"`
+}
+
+// Helper function to create a test system for config tests
+// func createConfigTestSystem(app core.App, name, host string, port uint16, userIDs []string) (*core.Record, error) {
+// 	systemCollection, err := app.FindCollectionByNameOrId("systems")
+// 	if err != nil {
+// 		return nil, err
+// 	}
+
+// 	system := core.NewRecord(systemCollection)
+// 	system.Set("name", name)
+// 	system.Set("host", host)
+// 	system.Set("port", port)
+// 	system.Set("users", userIDs)
+// 	system.Set("status", "pending")
+
+// 	return system, app.Save(system)
+// }
+
+// Helper function to create a fingerprint record
+func createConfigTestFingerprint(app core.App, systemID, token, fingerprint string) (*core.Record, error) {
+	fingerprintCollection, err := app.FindCollectionByNameOrId("fingerprints")
+	if err != nil {
+		return nil, err
+	}
+
+	fp := core.NewRecord(fingerprintCollection)
+	fp.Set("system", systemID)
+	fp.Set("token", token)
+	fp.Set("fingerprint", fingerprint)
+
+	return fp, app.Save(fp)
+}
+
+// TestConfigSyncWithTokens tests the config.SyncSystems function with various token scenarios
+func TestConfigSyncWithTokens(t *testing.T) {
+	testHub, err := tests.NewTestHub()
+	require.NoError(t, err)
+	defer testHub.Cleanup()
+
+	// Create test user
+	user, err := tests.CreateUser(testHub.App, "admin@example.com", "testtesttest")
+	require.NoError(t, err)
+
+	testCases := []struct {
+		name        string
+		setupFunc   func() (string, *core.Record, *core.Record) // Returns: existing token, system record, fingerprint record
+		configYAML  string
+		expectToken string // Expected token after sync
+		description string
+	}{
+		{
+			name: "new system with token in config",
+			setupFunc: func() (string, *core.Record, *core.Record) {
+				return "", nil, nil // No existing system
+			},
+			configYAML: `systems:
+  - name: "new-server"
+    host: "new.example.com"
+    port: 45876
+    users:
+      - "admin@example.com"
+    token: "explicit-token-123"`,
+			expectToken: "explicit-token-123",
+			description: "New system should use token from config",
+		},
+		{
+			name: "existing system without token in config (preserve existing)",
+			setupFunc: func() (string, *core.Record, *core.Record) {
+				// Create existing system and fingerprint
+				system, err := tests.CreateRecord(testHub.App, "systems", map[string]any{
+					"name":  "preserve-server",
+					"host":  "preserve.example.com",
+					"port":  45876,
+					"users": []string{user.Id},
+				})
+				require.NoError(t, err)
+
+				fingerprint, err := createConfigTestFingerprint(testHub.App, system.Id, "preserve-token-999", "preserve-fingerprint")
+				require.NoError(t, err)
+
+				return "preserve-token-999", system, fingerprint
+			},
+			configYAML: `systems:
+  - name: "preserve-server"
+    host: "preserve.example.com"
+    port: 45876
+    users:
+      - "admin@example.com"`,
+			expectToken: "preserve-token-999",
+			description: "Existing system should preserve original token when config doesn't specify one",
+		},
+	}
+
+	for _, tc := range testCases {
+		t.Run(tc.name, func(t *testing.T) {
+			// Setup test data
+			_, existingSystem, existingFingerprint := tc.setupFunc()
+
+			// Write config file
+			configPath := filepath.Join(testHub.DataDir(), "config.yml")
+			err := os.WriteFile(configPath, []byte(tc.configYAML), 0644)
+			require.NoError(t, err)
+
+			// Create serve event and sync
+			event := &core.ServeEvent{App: testHub.App}
+			err = config.SyncSystems(event)
+			require.NoError(t, err)
+
+			// Parse the config to get the system name for verification
+			var configData testConfig
+			err = yaml.Unmarshal([]byte(tc.configYAML), &configData)
+			require.NoError(t, err)
+			require.Len(t, configData.Systems, 1)
+			systemName := configData.Systems[0].Name
+
+			// Find the system after sync
+			systems, err := testHub.FindRecordsByFilter("systems", "name = {:name}", "", -1, 0, map[string]any{"name": systemName})
+			require.NoError(t, err)
+			require.Len(t, systems, 1)
+			system := systems[0]
+
+			// Find the fingerprint record
+			fingerprints, err := testHub.FindRecordsByFilter("fingerprints", "system = {:system}", "", -1, 0, map[string]any{"system": system.Id})
+			require.NoError(t, err)
+			require.Len(t, fingerprints, 1)
+			fingerprint := fingerprints[0]
+
+			// Verify token
+			actualToken := fingerprint.GetString("token")
+			if tc.expectToken == "" {
+				// For generated tokens, just verify it's not empty and is a valid UUID format
+				assert.NotEmpty(t, actualToken, tc.description)
+				assert.Len(t, actualToken, 36, "Generated token should be UUID format") // UUID length
+			} else {
+				assert.Equal(t, tc.expectToken, actualToken, tc.description)
+			}
+
+			// For existing systems, verify fingerprint is preserved
+			if existingFingerprint != nil {
+				actualFingerprint := fingerprint.GetString("fingerprint")
+				expectedFingerprint := existingFingerprint.GetString("fingerprint")
+				assert.Equal(t, expectedFingerprint, actualFingerprint, "Fingerprint should be preserved")
+			}
+
+			// Cleanup for next test
+			if existingSystem != nil {
+				testHub.Delete(existingSystem)
+			}
+			if existingFingerprint != nil {
+				testHub.Delete(existingFingerprint)
+			}
+			// Clean up the new records
+			testHub.Delete(system)
+			testHub.Delete(fingerprint)
+		})
+	}
+}
+
+// TestConfigMigrationScenario tests the specific migration scenario mentioned in the discussion
+func TestConfigMigrationScenario(t *testing.T) {
+	testHub, err := tests.NewTestHub(t.TempDir())
+	require.NoError(t, err)
+	defer testHub.Cleanup()
+
+	// Create test user
+	user, err := tests.CreateUser(testHub.App, "admin@example.com", "testtesttest")
+	require.NoError(t, err)
+
+	// Simulate migration scenario: system exists with token from migration
+	existingSystem, err := tests.CreateRecord(testHub.App, "systems", map[string]any{
+		"name":  "migrated-server",
+		"host":  "migrated.example.com",
+		"port":  45876,
+		"users": []string{user.Id},
+	})
+	require.NoError(t, err)
+
+	migrationToken := "migration-generated-token-123"
+	existingFingerprint, err := createConfigTestFingerprint(testHub.App, existingSystem.Id, migrationToken, "existing-fingerprint-from-agent")
+	require.NoError(t, err)
+
+	// User exports config BEFORE this update (so no token field in YAML)
+	oldConfigYAML := `systems:
+  - name: "migrated-server"
+    host: "migrated.example.com"
+    port: 45876
+    users:
+      - "admin@example.com"`
+
+	// Write old config file and import
+	configPath := filepath.Join(testHub.DataDir(), "config.yml")
+	err = os.WriteFile(configPath, []byte(oldConfigYAML), 0644)
+	require.NoError(t, err)
+
+	event := &core.ServeEvent{App: testHub.App}
+	err = config.SyncSystems(event)
+	require.NoError(t, err)
+
+	// Verify the original token is preserved
+	updatedFingerprint, err := testHub.FindRecordById("fingerprints", existingFingerprint.Id)
+	require.NoError(t, err)
+
+	actualToken := updatedFingerprint.GetString("token")
+	assert.Equal(t, migrationToken, actualToken, "Migration token should be preserved when config doesn't specify a token")
+
+	// Verify fingerprint is also preserved
+	actualFingerprint := updatedFingerprint.GetString("fingerprint")
+	assert.Equal(t, "existing-fingerprint-from-agent", actualFingerprint, "Existing fingerprint should be preserved")
+
+	// Verify system still exists and is updated correctly
+	updatedSystem, err := testHub.FindRecordById("systems", existingSystem.Id)
+	require.NoError(t, err)
+	assert.Equal(t, "migrated-server", updatedSystem.GetString("name"))
+	assert.Equal(t, "migrated.example.com", updatedSystem.GetString("host"))
+}

internal/hub/expirymap/expirymap.go

@@ -0,0 +1,104 @@
+package expirymap
+
+import (
+	"reflect"
+	"time"
+
+	"github.com/pocketbase/pocketbase/tools/store"
+)
+
+type val[T any] struct {
+	value   T
+	expires time.Time
+}
+
+type ExpiryMap[T any] struct {
+	store           *store.Store[string, *val[T]]
+	cleanupInterval time.Duration
+}
+
+// New creates a new expiry map with custom cleanup interval
+func New[T any](cleanupInterval time.Duration) *ExpiryMap[T] {
+	m := &ExpiryMap[T]{
+		store:           store.New(map[string]*val[T]{}),
+		cleanupInterval: cleanupInterval,
+	}
+	m.startCleaner()
+	return m
+}
+
+// Set stores a value with the given TTL
+func (m *ExpiryMap[T]) Set(key string, value T, ttl time.Duration) {
+	m.store.Set(key, &val[T]{
+		value:   value,
+		expires: time.Now().Add(ttl),
+	})
+}
+
+// GetOk retrieves a value and checks if it exists and hasn't expired
+// Performs lazy cleanup of expired entries on access
+func (m *ExpiryMap[T]) GetOk(key string) (T, bool) {
+	value, ok := m.store.GetOk(key)
+	if !ok {
+		return *new(T), false
+	}
+
+	// Check if expired and perform lazy cleanup
+	if value.expires.Before(time.Now()) {
+		m.store.Remove(key)
+		return *new(T), false
+	}
+
+	return value.value, true
+}
+
+// GetByValue retrieves a value by value
+func (m *ExpiryMap[T]) GetByValue(val T) (key string, value T, ok bool) {
+	for key, v := range m.store.GetAll() {
+		if reflect.DeepEqual(v.value, val) {
+			// check if expired
+			if v.expires.Before(time.Now()) {
+				m.store.Remove(key)
+				break
+			}
+			return key, v.value, true
+		}
+	}
+	return "", *new(T), false
+}
+
+// Remove explicitly removes a key
+func (m *ExpiryMap[T]) Remove(key string) {
+	m.store.Remove(key)
+}
+
+// RemovebyValue removes a value by value
+func (m *ExpiryMap[T]) RemovebyValue(value T) (T, bool) {
+	for key, val := range m.store.GetAll() {
+		if reflect.DeepEqual(val.value, value) {
+			m.store.Remove(key)
+			return val.value, true
+		}
+	}
+	return *new(T), false
+}
+
+// startCleaner runs the background cleanup process
+func (m *ExpiryMap[T]) startCleaner() {
+	go func() {
+		tick := time.Tick(m.cleanupInterval)
+		for range tick {
+			m.cleanup()
+		}
+	}()
+}
+
+// cleanup removes all expired entries
+func (m *ExpiryMap[T]) cleanup() {
+	now := time.Now()
+	for key, val := range m.store.GetAll() {
+		if val.expires.Before(now) {
+			m.store.Remove(key)
+		}
+	}
+}

internal/hub/expirymap/expirymap_test.go

@@ -0,0 +1,477 @@
+//go:build testing
+// +build testing
+
+package expirymap
+
+import (
+	"testing"
+	"time"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+// Not using the following methods but are useful for testing
+
+// TESTING: Has checks if a key exists and hasn't expired
+func (m *ExpiryMap[T]) Has(key string) bool {
+	_, ok := m.GetOk(key)
+	return ok
+}
+
+// TESTING: Get retrieves a value, returns zero value if not found or expired
+func (m *ExpiryMap[T]) Get(key string) T {
+	value, _ := m.GetOk(key)
+	return value
+}
+
+// TESTING: Len returns the number of non-expired entries
+func (m *ExpiryMap[T]) Len() int {
+	count := 0
+	now := time.Now()
+	for _, val := range m.store.Values() {
+		if val.expires.After(now) {
+			count++
+		}
+	}
+	return count
+}
+
+func TestExpiryMap_BasicOperations(t *testing.T) {
+	em := New[string](time.Hour)
+
+	// Test Set and GetOk
+	em.Set("key1", "value1", time.Hour)
+	value, ok := em.GetOk("key1")
+	assert.True(t, ok)
+	assert.Equal(t, "value1", value)
+
+	// Test Get
+	value = em.Get("key1")
+	assert.Equal(t, "value1", value)
+
+	// Test Has
+	assert.True(t, em.Has("key1"))
+	assert.False(t, em.Has("nonexistent"))
+
+	// Test Remove
+	em.Remove("key1")
+	assert.False(t, em.Has("key1"))
+}
+
+func TestExpiryMap_Expiration(t *testing.T) {
+	em := New[string](time.Hour)
+
+	// Set a value with very short TTL
+	em.Set("shortlived", "value", time.Millisecond*10)
+
+	// Should exist immediately
+	assert.True(t, em.Has("shortlived"))
+
+	// Wait for expiration
+	time.Sleep(time.Millisecond * 20)
+
+	// Should be expired and automatically cleaned up on access
+	assert.False(t, em.Has("shortlived"))
+	value, ok := em.GetOk("shortlived")
+	assert.False(t, ok)
+	assert.Equal(t, "", value) // zero value for string
+}
+
+func TestExpiryMap_LazyCleanup(t *testing.T) {
+	em := New[int](time.Hour)
+
+	// Set multiple values with short TTL
+	em.Set("key1", 1, time.Millisecond*10)
+	em.Set("key2", 2, time.Millisecond*10)
+	em.Set("key3", 3, time.Hour) // This one won't expire
+
+	// Wait for expiration
+	time.Sleep(time.Millisecond * 20)
+
+	// Access expired keys should trigger lazy cleanup
+	_, ok := em.GetOk("key1")
+	assert.False(t, ok)
+
+	// Non-expired key should still exist
+	value, ok := em.GetOk("key3")
+	assert.True(t, ok)
+	assert.Equal(t, 3, value)
+}
+
+func TestExpiryMap_Len(t *testing.T) {
+	em := New[string](time.Hour)
+
+	// Initially empty
+	assert.Equal(t, 0, em.Len())
+
+	// Add some values
+	em.Set("key1", "value1", time.Hour)
+	em.Set("key2", "value2", time.Hour)
+	em.Set("key3", "value3", time.Millisecond*10) // Will expire soon
+
+	// Should count all initially
+	assert.Equal(t, 3, em.Len())
+
+	// Wait for one to expire
+	time.Sleep(time.Millisecond * 20)
+
+	// Len should reflect only non-expired entries
+	assert.Equal(t, 2, em.Len())
+}
+
+func TestExpiryMap_CustomInterval(t *testing.T) {
+	// Create with very short cleanup interval for testing
+	em := New[string](time.Millisecond * 50)
+
+	// Set a value that expires quickly
+	em.Set("test", "value", time.Millisecond*10)
+
+	// Should exist initially
+	assert.True(t, em.Has("test"))
+
+	// Wait for expiration + cleanup cycle
+	time.Sleep(time.Millisecond * 100)
+
+	// Should be cleaned up by background process
+	// Note: This test might be flaky due to timing, but demonstrates the concept
+	assert.False(t, em.Has("test"))
+}
+
+func TestExpiryMap_GenericTypes(t *testing.T) {
+	// Test with different types
+	t.Run("Int", func(t *testing.T) {
+		em := New[int](time.Hour)
+
+		em.Set("num", 42, time.Hour)
+		value, ok := em.GetOk("num")
+		assert.True(t, ok)
+		assert.Equal(t, 42, value)
+	})
+
+	t.Run("Struct", func(t *testing.T) {
+		type TestStruct struct {
+			Name string
+			Age  int
+		}
+
+		em := New[TestStruct](time.Hour)
+
+		expected := TestStruct{Name: "John", Age: 30}
+		em.Set("person", expected, time.Hour)
+
+		value, ok := em.GetOk("person")
+		assert.True(t, ok)
+		assert.Equal(t, expected, value)
+	})
+
+	t.Run("Pointer", func(t *testing.T) {
+		em := New[*string](time.Hour)
+
+		str := "hello"
+		em.Set("ptr", &str, time.Hour)
+
+		value, ok := em.GetOk("ptr")
+		assert.True(t, ok)
+		require.NotNil(t, value)
+		assert.Equal(t, "hello", *value)
+	})
+}
+
+func TestExpiryMap_ZeroValues(t *testing.T) {
+	em := New[string](time.Hour)
+
+	// Test getting non-existent key returns zero value
+	value := em.Get("nonexistent")
+	assert.Equal(t, "", value)
+
+	// Test getting expired key returns zero value
+	em.Set("expired", "value", time.Millisecond*10)
+	time.Sleep(time.Millisecond * 20)
+
+	value = em.Get("expired")
+	assert.Equal(t, "", value)
+}
+
+func TestExpiryMap_Concurrent(t *testing.T) {
+	em := New[int](time.Hour)
+
+	// Simple concurrent access test
+	done := make(chan bool, 2)
+
+	// Writer goroutine
+	go func() {
+		for i := 0; i < 100; i++ {
+			em.Set("key", i, time.Hour)
+			time.Sleep(time.Microsecond)
+		}
+		done <- true
+	}()
+
+	// Reader goroutine
+	go func() {
+		for i := 0; i < 100; i++ {
+			_ = em.Get("key")
+			time.Sleep(time.Microsecond)
+		}
+		done <- true
+	}()
+
+	// Wait for both to complete
+	<-done
+	<-done
+
+	// Should not panic and should have some value
+	assert.True(t, em.Has("key"))
+}
+
+func TestExpiryMap_GetByValue(t *testing.T) {
+	em := New[string](time.Hour)
+
+	// Test getting by value when value exists
+	em.Set("key1", "value1", time.Hour)
+	em.Set("key2", "value2", time.Hour)
+	em.Set("key3", "value1", time.Hour) // Duplicate value - should return first match
+
+	// Test successful retrieval
+	key, value, ok := em.GetByValue("value1")
+	assert.True(t, ok)
+	assert.Equal(t, "value1", value)
+	assert.Contains(t, []string{"key1", "key3"}, key) // Should be one of the keys with this value
+
+	// Test retrieval of unique value
+	key, value, ok = em.GetByValue("value2")
+	assert.True(t, ok)
+	assert.Equal(t, "value2", value)
+	assert.Equal(t, "key2", key)
+
+	// Test getting non-existent value
+	key, value, ok = em.GetByValue("nonexistent")
+	assert.False(t, ok)
+	assert.Equal(t, "", value) // zero value for string
+	assert.Equal(t, "", key)   // zero value for string
+}
+
+func TestExpiryMap_GetByValue_Expiration(t *testing.T) {
+	em := New[string](time.Hour)
+
+	// Set a value with short TTL
+	em.Set("shortkey", "shortvalue", time.Millisecond*10)
+	em.Set("longkey", "longvalue", time.Hour)
+
+	// Should find the short-lived value initially
+	key, value, ok := em.GetByValue("shortvalue")
+	assert.True(t, ok)
+	assert.Equal(t, "shortvalue", value)
+	assert.Equal(t, "shortkey", key)
+
+	// Wait for expiration
+	time.Sleep(time.Millisecond * 20)
+
+	// Should not find expired value and should trigger lazy cleanup
+	key, value, ok = em.GetByValue("shortvalue")
+	assert.False(t, ok)
+	assert.Equal(t, "", value)
+	assert.Equal(t, "", key)
+
+	// Should still find non-expired value
+	key, value, ok = em.GetByValue("longvalue")
+	assert.True(t, ok)
+	assert.Equal(t, "longvalue", value)
+	assert.Equal(t, "longkey", key)
+}
+
+func TestExpiryMap_GetByValue_GenericTypes(t *testing.T) {
+	t.Run("Int", func(t *testing.T) {
+		em := New[int](time.Hour)
+
+		em.Set("num1", 42, time.Hour)
+		em.Set("num2", 84, time.Hour)
+
+		key, value, ok := em.GetByValue(42)
+		assert.True(t, ok)
+		assert.Equal(t, 42, value)
+		assert.Equal(t, "num1", key)
+
+		key, value, ok = em.GetByValue(99)
+		assert.False(t, ok)
+		assert.Equal(t, 0, value)
+		assert.Equal(t, "", key)
+	})
+
+	t.Run("Struct", func(t *testing.T) {
+		type TestStruct struct {
+			Name string
+			Age  int
+		}
+
+		em := New[TestStruct](time.Hour)
+
+		person1 := TestStruct{Name: "John", Age: 30}
+		person2 := TestStruct{Name: "Jane", Age: 25}
+
+		em.Set("person1", person1, time.Hour)
+		em.Set("person2", person2, time.Hour)
+
+		key, value, ok := em.GetByValue(person1)
+		assert.True(t, ok)
+		assert.Equal(t, person1, value)
+		assert.Equal(t, "person1", key)
+
+		nonexistent := TestStruct{Name: "Bob", Age: 40}
+		key, value, ok = em.GetByValue(nonexistent)
+		assert.False(t, ok)
+		assert.Equal(t, TestStruct{}, value)
+		assert.Equal(t, "", key)
+	})
+}
+
+func TestExpiryMap_RemoveValue(t *testing.T) {
+	em := New[string](time.Hour)
+
+	// Test removing existing value
+	em.Set("key1", "value1", time.Hour)
+	em.Set("key2", "value2", time.Hour)
+	em.Set("key3", "value1", time.Hour) // Duplicate value
+
+	// Remove by value should remove one instance
+	removedValue, ok := em.RemovebyValue("value1")
+	assert.True(t, ok)
+	assert.Equal(t, "value1", removedValue)
+
+	// Should still have the other instance or value2
+	assert.True(t, em.Has("key2")) // value2 should still exist
+
+	// Check if one of the duplicate values was removed
+	// At least one key with "value1" should be gone
+	key1Exists := em.Has("key1")
+	key3Exists := em.Has("key3")
+	assert.False(t, key1Exists && key3Exists) // Both shouldn't exist
+	assert.True(t, key1Exists || key3Exists)  // At least one should be gone
+
+	// Test removing non-existent value
+	removedValue, ok = em.RemovebyValue("nonexistent")
+	assert.False(t, ok)
+	assert.Equal(t, "", removedValue) // zero value for string
+}
+
+func TestExpiryMap_RemoveValue_GenericTypes(t *testing.T) {
+	t.Run("Int", func(t *testing.T) {
+		em := New[int](time.Hour)
+
+		em.Set("num1", 42, time.Hour)
+		em.Set("num2", 84, time.Hour)
+
+		// Remove existing value
+		removedValue, ok := em.RemovebyValue(42)
+		assert.True(t, ok)
+		assert.Equal(t, 42, removedValue)
+		assert.False(t, em.Has("num1"))
+		assert.True(t, em.Has("num2"))
+
+		// Remove non-existent value
+		removedValue, ok = em.RemovebyValue(99)
+		assert.False(t, ok)
+		assert.Equal(t, 0, removedValue)
+	})
+
+	t.Run("Struct", func(t *testing.T) {
+		type TestStruct struct {
+			Name string
+			Age  int
+		}
+
+		em := New[TestStruct](time.Hour)
+
+		person1 := TestStruct{Name: "John", Age: 30}
+		person2 := TestStruct{Name: "Jane", Age: 25}
+
+		em.Set("person1", person1, time.Hour)
+		em.Set("person2", person2, time.Hour)
+
+		// Remove existing struct
+		removedValue, ok := em.RemovebyValue(person1)
+		assert.True(t, ok)
+		assert.Equal(t, person1, removedValue)
+		assert.False(t, em.Has("person1"))
+		assert.True(t, em.Has("person2"))
+
+		// Remove non-existent struct
+		nonexistent := TestStruct{Name: "Bob", Age: 40}
+		removedValue, ok = em.RemovebyValue(nonexistent)
+		assert.False(t, ok)
+		assert.Equal(t, TestStruct{}, removedValue)
+	})
+}
+
+func TestExpiryMap_RemoveValue_WithExpiration(t *testing.T) {
+	em := New[string](time.Hour)
+
+	// Set values with different TTLs
+	em.Set("key1", "value1", time.Millisecond*10) // Will expire
+	em.Set("key2", "value2", time.Hour)           // Won't expire
+	em.Set("key3", "value1", time.Hour)           // Won't expire, duplicate value
+
+	// Wait for first value to expire
+	time.Sleep(time.Millisecond * 20)
+
+	// Try to remove the expired value - should remove one of the "value1" entries
+	removedValue, ok := em.RemovebyValue("value1")
+	assert.True(t, ok)
+	assert.Equal(t, "value1", removedValue)
+
+	// Should still have key2 (different value)
+	assert.True(t, em.Has("key2"))
+
+	// Should have removed one of the "value1" entries (either key1 or key3)
+	// But we can't predict which one due to map iteration order
+	key1Exists := em.Has("key1")
+	key3Exists := em.Has("key3")
+
+	// Exactly one of key1 or key3 should be gone
+	assert.False(t, key1Exists && key3Exists) // Both shouldn't exist
+	assert.True(t, key1Exists || key3Exists)  // At least one should still exist
+}
+
+func TestExpiryMap_ValueOperations_Integration(t *testing.T) {
+	em := New[string](time.Hour)
+
+	// Test integration of GetByValue and RemoveValue
+	em.Set("key1", "shared", time.Hour)
+	em.Set("key2", "unique", time.Hour)
+	em.Set("key3", "shared", time.Hour)
+
+	// Find shared value
+	key, value, ok := em.GetByValue("shared")
+	assert.True(t, ok)
+	assert.Equal(t, "shared", value)
+	assert.Contains(t, []string{"key1", "key3"}, key)
+
+	// Remove shared value
+	removedValue, ok := em.RemovebyValue("shared")
+	assert.True(t, ok)
+	assert.Equal(t, "shared", removedValue)
+
+	// Should still be able to find the other shared value
+	key, value, ok = em.GetByValue("shared")
+	assert.True(t, ok)
+	assert.Equal(t, "shared", value)
+	assert.Contains(t, []string{"key1", "key3"}, key)
+
+	// Remove the other shared value
+	removedValue, ok = em.RemovebyValue("shared")
+	assert.True(t, ok)
+	assert.Equal(t, "shared", removedValue)
+
+	// Should not find shared value anymore
+	key, value, ok = em.GetByValue("shared")
+	assert.False(t, ok)
+	assert.Equal(t, "", value)
+	assert.Equal(t, "", key)
+
+	// Unique value should still exist
+	key, value, ok = em.GetByValue("unique")
+	assert.True(t, ok)
+	assert.Equal(t, "unique", value)
+	assert.Equal(t, "key2", key)
+}

internal/hub/hub.go

@@ -0,0 +1,300 @@
+// Package hub handles updating systems and serving the web UI.
+package hub
+
+import (
+	"crypto/ed25519"
+	"encoding/pem"
+	"fmt"
+	"net/http"
+	"net/url"
+	"os"
+	"path"
+	"strings"
+	"time"
+
+	"github.com/henrygd/beszel"
+	"github.com/henrygd/beszel/src/alerts"
+	"github.com/henrygd/beszel/src/hub/config"
+	"github.com/henrygd/beszel/src/hub/systems"
+	"github.com/henrygd/beszel/src/records"
+	"github.com/henrygd/beszel/src/users"
+
+	"github.com/google/uuid"
+	"github.com/pocketbase/pocketbase"
+	"github.com/pocketbase/pocketbase/apis"
+	"github.com/pocketbase/pocketbase/core"
+	"golang.org/x/crypto/ssh"
+)
+
+type Hub struct {
+	core.App
+	*alerts.AlertManager
+	um     *users.UserManager
+	rm     *records.RecordManager
+	sm     *systems.SystemManager
+	pubKey string
+	signer ssh.Signer
+	appURL string
+}
+
+// NewHub creates a new Hub instance with default configuration
+func NewHub(app core.App) *Hub {
+	hub := &Hub{}
+	hub.App = app
+
+	hub.AlertManager = alerts.NewAlertManager(hub)
+	hub.um = users.NewUserManager(hub)
+	hub.rm = records.NewRecordManager(hub)
+	hub.sm = systems.NewSystemManager(hub)
+	hub.appURL, _ = GetEnv("APP_URL")
+	return hub
+}
+
+// GetEnv retrieves an environment variable with a "BESZEL_HUB_" prefix, or falls back to the unprefixed key.
+func GetEnv(key string) (value string, exists bool) {
+	if value, exists = os.LookupEnv("BESZEL_HUB_" + key); exists {
+		return value, exists
+	}
+	// Fallback to the old unprefixed key
+	return os.LookupEnv(key)
+}
+
+func (h *Hub) StartHub() error {
+	h.App.OnServe().BindFunc(func(e *core.ServeEvent) error {
+		// initialize settings / collections
+		if err := h.initialize(e); err != nil {
+			return err
+		}
+		// sync systems with config
+		if err := config.SyncSystems(e); err != nil {
+			return err
+		}
+		// register api routes
+		if err := h.registerApiRoutes(e); err != nil {
+			return err
+		}
+		// register cron jobs
+		if err := h.registerCronJobs(e); err != nil {
+			return err
+		}
+		// start server
+		if err := h.startServer(e); err != nil {
+			return err
+		}
+		// start system updates
+		if err := h.sm.Initialize(); err != nil {
+			return err
+		}
+		return e.Next()
+	})
+
+	// TODO: move to users package
+	// handle default values for user / user_settings creation
+	h.App.OnRecordCreate("users").BindFunc(h.um.InitializeUserRole)
+	h.App.OnRecordCreate("user_settings").BindFunc(h.um.InitializeUserSettings)
+
+	if pb, ok := h.App.(*pocketbase.PocketBase); ok {
+		// log.Println("Starting pocketbase")
+		err := pb.Start()
+		if err != nil {
+			return err
+		}
+	}
+
+	return nil
+}
+
+// initialize sets up initial configuration (collections, settings, etc.)
+func (h *Hub) initialize(e *core.ServeEvent) error {
+	// set general settings
+	settings := e.App.Settings()
+	// batch requests (for global alerts)
+	settings.Batch.Enabled = true
+	// set URL if BASE_URL env is set
+	if h.appURL != "" {
+		settings.Meta.AppURL = h.appURL
+	}
+	if err := e.App.Save(settings); err != nil {
+		return err
+	}
+	// set auth settings
+	usersCollection, err := e.App.FindCollectionByNameOrId("users")
+	if err != nil {
+		return err
+	}
+	// disable email auth if DISABLE_PASSWORD_AUTH env var is set
+	disablePasswordAuth, _ := GetEnv("DISABLE_PASSWORD_AUTH")
+	usersCollection.PasswordAuth.Enabled = disablePasswordAuth != "true"
+	usersCollection.PasswordAuth.IdentityFields = []string{"email"}
+	// disable oauth if no providers are configured (todo: remove this in post 0.9.0 release)
+	if usersCollection.OAuth2.Enabled {
+		usersCollection.OAuth2.Enabled = len(usersCollection.OAuth2.Providers) > 0
+	}
+	// allow oauth user creation if USER_CREATION is set
+	if userCreation, _ := GetEnv("USER_CREATION"); userCreation == "true" {
+		cr := "@request.context = 'oauth2'"
+		usersCollection.CreateRule = &cr
+	} else {
+		usersCollection.CreateRule = nil
+	}
+	if err := e.App.Save(usersCollection); err != nil {
+		return err
+	}
+	// allow all users to access systems if SHARE_ALL_SYSTEMS is set
+	systemsCollection, err := e.App.FindCachedCollectionByNameOrId("systems")
+	if err != nil {
+		return err
+	}
+	shareAllSystems, _ := GetEnv("SHARE_ALL_SYSTEMS")
+	systemsReadRule := "@request.auth.id != \"\""
+	if shareAllSystems != "true" {
+		// default is to only show systems that the user id is assigned to
+		systemsReadRule += " && users.id ?= @request.auth.id"
+	}
+	updateDeleteRule := systemsReadRule + " && @request.auth.role != \"readonly\""
+	systemsCollection.ListRule = &systemsReadRule
+	systemsCollection.ViewRule = &systemsReadRule
+	systemsCollection.UpdateRule = &updateDeleteRule
+	systemsCollection.DeleteRule = &updateDeleteRule
+	if err := e.App.Save(systemsCollection); err != nil {
+		return err
+	}
+	return nil
+}
+
+// registerCronJobs sets up scheduled tasks
+func (h *Hub) registerCronJobs(_ *core.ServeEvent) error {
+	// delete old system_stats and alerts_history records once every hour
+	h.Cron().MustAdd("delete old records", "8 * * * *", h.rm.DeleteOldRecords)
+	// create longer records every 10 minutes
+	h.Cron().MustAdd("create longer records", "*/10 * * * *", h.rm.CreateLongerRecords)
+	return nil
+}
+
+// custom api routes
+func (h *Hub) registerApiRoutes(se *core.ServeEvent) error {
+	// auth protected routes
+	apiAuth := se.Router.Group("/api/beszel")
+	apiAuth.Bind(apis.RequireAuth())
+	// auth optional routes
+	apiNoAuth := se.Router.Group("/api/beszel")
+
+	// create first user endpoint only needed if no users exist
+	if totalUsers, _ := se.App.CountRecords("users"); totalUsers == 0 {
+		apiNoAuth.POST("/create-user", h.um.CreateFirstUser)
+	}
+	// check if first time setup on login page
+	apiNoAuth.GET("/first-run", func(e *core.RequestEvent) error {
+		total, err := e.App.CountRecords("users")
+		return e.JSON(http.StatusOK, map[string]bool{"firstRun": err == nil && total == 0})
+	})
+	// get public key and version
+	apiAuth.GET("/getkey", func(e *core.RequestEvent) error {
+		return e.JSON(http.StatusOK, map[string]string{"key": h.pubKey, "v": beszel.Version})
+	})
+	// send test notification
+	apiAuth.POST("/test-notification", h.SendTestNotification)
+	// get config.yml content
+	apiAuth.GET("/config-yaml", config.GetYamlConfig)
+	// handle agent websocket connection
+	apiNoAuth.GET("/agent-connect", h.handleAgentConnect)
+	// get or create universal tokens
+	apiAuth.GET("/universal-token", h.getUniversalToken)
+	// update / delete user alerts
+	apiAuth.POST("/user-alerts", alerts.UpsertUserAlerts)
+	apiAuth.DELETE("/user-alerts", alerts.DeleteUserAlerts)
+
+	return nil
+}
+
+// Handler for universal token API endpoint (create, read, delete)
+func (h *Hub) getUniversalToken(e *core.RequestEvent) error {
+	tokenMap := universalTokenMap.GetMap()
+	userID := e.Auth.Id
+	query := e.Request.URL.Query()
+	token := query.Get("token")
+
+	if token == "" {
+		// return existing token if it exists
+		if token, _, ok := tokenMap.GetByValue(userID); ok {
+			return e.JSON(http.StatusOK, map[string]any{"token": token, "active": true})
+		}
+		// if no token is provided, generate a new one
+		token = uuid.New().String()
+	}
+	response := map[string]any{"token": token}
+
+	switch query.Get("enable") {
+	case "1":
+		tokenMap.Set(token, userID, time.Hour)
+	case "0":
+		tokenMap.RemovebyValue(userID)
+	}
+	_, response["active"] = tokenMap.GetOk(token)
+	return e.JSON(http.StatusOK, response)
+}
+
+// generates key pair if it doesn't exist and returns signer
+func (h *Hub) GetSSHKey(dataDir string) (ssh.Signer, error) {
+	if h.signer != nil {
+		return h.signer, nil
+	}
+
+	if dataDir == "" {
+		dataDir = h.DataDir()
+	}
+
+	privateKeyPath := path.Join(dataDir, "id_ed25519")
+
+	// check if the key pair already exists
+	existingKey, err := os.ReadFile(privateKeyPath)
+	if err == nil {
+		private, err := ssh.ParsePrivateKey(existingKey)
+		if err != nil {
+			return nil, fmt.Errorf("failed to parse private key: %s", err)
+		}
+		pubKeyBytes := ssh.MarshalAuthorizedKey(private.PublicKey())
+		h.pubKey = strings.TrimSuffix(string(pubKeyBytes), "\n")
+		return private, nil
+	} else if !os.IsNotExist(err) {
+		// File exists but couldn't be read for some other reason
+		return nil, fmt.Errorf("failed to read %s: %w", privateKeyPath, err)
+	}
+
+	// Generate the Ed25519 key pair
+	_, privKey, err := ed25519.GenerateKey(nil)
+	if err != nil {
+		return nil, err
+	}
+	privKeyPem, err := ssh.MarshalPrivateKey(privKey, "")
+	if err != nil {
+		return nil, err
+	}
+
+	if err := os.WriteFile(privateKeyPath, pem.EncodeToMemory(privKeyPem), 0600); err != nil {
+		return nil, fmt.Errorf("failed to write private key to %q: err: %w", privateKeyPath, err)
+	}
+
+	// These are fine to ignore the errors on, as we've literally just created a crypto.PublicKey | crypto.Signer
+	sshPrivate, _ := ssh.NewSignerFromSigner(privKey)
+	pubKeyBytes := ssh.MarshalAuthorizedKey(sshPrivate.PublicKey())
+	h.pubKey = strings.TrimSuffix(string(pubKeyBytes), "\n")
+
+	h.Logger().Info("ed25519 key pair generated successfully.")
+	h.Logger().Info("Saved to: " + privateKeyPath)
+
+	return sshPrivate, err
+}
+
+// MakeLink formats a link with the app URL and path segments.
+// Only path segments should be provided.
+func (h *Hub) MakeLink(parts ...string) string {
+	base := strings.TrimSuffix(h.Settings().Meta.AppURL, "/")
+	for _, part := range parts {
+		if part == "" {
+			continue
+		}
+		base = fmt.Sprintf("%s/%s", base, url.PathEscape(part))
+	}
+	return base
+}

internal/hub/hub_test.go

@@ -0,0 +1,713 @@
+//go:build testing
+// +build testing
+
+package hub_test
+
+import (
+	"bytes"
+	"crypto/ed25519"
+	"encoding/json"
+	"encoding/pem"
+	"io"
+	"net/http"
+	"os"
+	"path/filepath"
+	"strings"
+	"testing"
+
+	"github.com/henrygd/beszel/src/migrations"
+	beszelTests "github.com/henrygd/beszel/src/tests"
+
+	"github.com/pocketbase/pocketbase/core"
+	pbTests "github.com/pocketbase/pocketbase/tests"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+	"golang.org/x/crypto/ssh"
+)
+
+// marshal to json and return an io.Reader (for use in ApiScenario.Body)
+func jsonReader(v any) io.Reader {
+	data, err := json.Marshal(v)
+	if err != nil {
+		panic(err)
+	}
+	return bytes.NewReader(data)
+}
+
+func TestMakeLink(t *testing.T) {
+	hub, _ := beszelTests.NewTestHub(t.TempDir())
+
+	tests := []struct {
+		name     string
+		appURL   string
+		parts    []string
+		expected string
+	}{
+		{
+			name:     "no parts, no trailing slash in AppURL",
+			appURL:   "http://localhost:8090",
+			parts:    []string{},
+			expected: "http://localhost:8090",
+		},
+		{
+			name:     "no parts, with trailing slash in AppURL",
+			appURL:   "http://localhost:8090/",
+			parts:    []string{},
+			expected: "http://localhost:8090", // TrimSuffix should handle the trailing slash
+		},
+		{
+			name:     "one part",
+			appURL:   "http://example.com",
+			parts:    []string{"one"},
+			expected: "http://example.com/one",
+		},
+		{
+			name:     "multiple parts",
+			appURL:   "http://example.com",
+			parts:    []string{"alpha", "beta", "gamma"},
+			expected: "http://example.com/alpha/beta/gamma",
+		},
+		{
+			name:     "parts with spaces needing escaping",
+			appURL:   "http://example.com",
+			parts:    []string{"path with spaces", "another part"},
+			expected: "http://example.com/path%20with%20spaces/another%20part",
+		},
+		{
+			name:     "parts with slashes needing escaping",
+			appURL:   "http://example.com",
+			parts:    []string{"a/b", "c"},
+			expected: "http://example.com/a%2Fb/c", // url.PathEscape escapes '/'
+		},
+		{
+			name:     "AppURL with subpath, no trailing slash",
+			appURL:   "http://localhost/sub",
+			parts:    []string{"resource"},
+			expected: "http://localhost/sub/resource",
+		},
+		{
+			name:     "AppURL with subpath, with trailing slash",
+			appURL:   "http://localhost/sub/",
+			parts:    []string{"item"},
+			expected: "http://localhost/sub/item",
+		},
+		{
+			name:     "empty parts in the middle",
+			appURL:   "http://localhost",
+			parts:    []string{"first", "", "third"},
+			expected: "http://localhost/first/third",
+		},
+		{
+			name:     "leading and trailing empty parts",
+			appURL:   "http://localhost",
+			parts:    []string{"", "path", ""},
+			expected: "http://localhost/path",
+		},
+		{
+			name:     "parts with various special characters",
+			appURL:   "https://test.dev/",
+			parts:    []string{"p@th?", "key=value&"},
+			expected: "https://test.dev/p@th%3F/key=value&",
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			// Store original app URL and restore it after the test
+			originalAppURL := hub.Settings().Meta.AppURL
+			hub.Settings().Meta.AppURL = tt.appURL
+			defer func() { hub.Settings().Meta.AppURL = originalAppURL }()
+
+			got := hub.MakeLink(tt.parts...)
+			assert.Equal(t, tt.expected, got, "MakeLink generated URL does not match expected")
+		})
+	}
+}
+
+func TestGetSSHKey(t *testing.T) {
+	hub, _ := beszelTests.NewTestHub(t.TempDir())
+
+	// Test Case 1: Key generation (no existing key)
+	t.Run("KeyGeneration", func(t *testing.T) {
+		tempDir := t.TempDir()
+
+		// Ensure pubKey is initially empty or different to ensure GetSSHKey sets it
+		hub.SetPubkey("")
+
+		signer, err := hub.GetSSHKey(tempDir)
+		assert.NoError(t, err, "GetSSHKey should not error when generating a new key")
+		assert.NotNil(t, signer, "GetSSHKey should return a non-nil signer")
+
+		// Check if private key file was created
+		privateKeyPath := filepath.Join(tempDir, "id_ed25519")
+		info, err := os.Stat(privateKeyPath)
+		assert.NoError(t, err, "Private key file should be created")
+		assert.False(t, info.IsDir(), "Private key path should be a file, not a directory")
+
+		// Check if h.pubKey was set
+		assert.NotEmpty(t, hub.GetPubkey(), "h.pubKey should be set after key generation")
+		assert.True(t, strings.HasPrefix(hub.GetPubkey(), "ssh-ed25519 "), "h.pubKey should start with 'ssh-ed25519 '")
+
+		// Verify the generated private key is parsable
+		keyData, err := os.ReadFile(privateKeyPath)
+		require.NoError(t, err)
+		_, err = ssh.ParsePrivateKey(keyData)
+		assert.NoError(t, err, "Generated private key should be parsable by ssh.ParsePrivateKey")
+	})
+
+	// Test Case 2: Existing key
+	t.Run("ExistingKey", func(t *testing.T) {
+		tempDir := t.TempDir()
+
+		// Manually create a valid key pair for the test
+		rawPubKey, rawPrivKey, err := ed25519.GenerateKey(nil)
+		require.NoError(t, err, "Failed to generate raw ed25519 key pair for pre-existing key test")
+
+		// Marshal the private key into OpenSSH PEM format
+		pemBlock, err := ssh.MarshalPrivateKey(rawPrivKey, "")
+		require.NoError(t, err, "Failed to marshal private key to PEM block for pre-existing key test")
+
+		privateKeyBytes := pem.EncodeToMemory(pemBlock)
+		require.NotNil(t, privateKeyBytes, "PEM encoded private key bytes should not be nil")
+
+		privateKeyPath := filepath.Join(tempDir, "id_ed25519")
+		err = os.WriteFile(privateKeyPath, privateKeyBytes, 0600)
+		require.NoError(t, err, "Failed to write pre-existing private key")
+
+		// Determine the expected public key string
+		sshPubKey, err := ssh.NewPublicKey(rawPubKey)
+		require.NoError(t, err)
+		expectedPubKeyStr := strings.TrimSpace(string(ssh.MarshalAuthorizedKey(sshPubKey)))
+
+		// Reset h.pubKey to ensure it's set by GetSSHKey from the file
+		hub.SetPubkey("")
+
+		signer, err := hub.GetSSHKey(tempDir)
+		assert.NoError(t, err, "GetSSHKey should not error when reading an existing key")
+		assert.NotNil(t, signer, "GetSSHKey should return a non-nil signer for an existing key")
+
+		// Check if h.pubKey was set correctly to the public key from the file
+		assert.Equal(t, expectedPubKeyStr, hub.GetPubkey(), "h.pubKey should match the existing public key")
+
+		// Verify the signer's public key matches the original public key
+		signerPubKey := signer.PublicKey()
+		marshaledSignerPubKey := strings.TrimSpace(string(ssh.MarshalAuthorizedKey(signerPubKey)))
+		assert.Equal(t, expectedPubKeyStr, marshaledSignerPubKey, "Signer's public key should match the existing public key")
+	})
+
+	// Test Case 3: Error cases
+	t.Run("ErrorCases", func(t *testing.T) {
+		tests := []struct {
+			name       string
+			setupFunc  func(dir string) error
+			errorCheck func(t *testing.T, err error)
+		}{
+			{
+				name: "CorruptedKey",
+				setupFunc: func(dir string) error {
+					return os.WriteFile(filepath.Join(dir, "id_ed25519"), []byte("this is not a valid SSH key"), 0600)
+				},
+				errorCheck: func(t *testing.T, err error) {
+					assert.Error(t, err)
+					assert.Contains(t, err.Error(), "ssh: no key found")
+				},
+			},
+			{
+				name: "PermissionDenied",
+				setupFunc: func(dir string) error {
+					// Create the key file
+					keyPath := filepath.Join(dir, "id_ed25519")
+					if err := os.WriteFile(keyPath, []byte("dummy content"), 0600); err != nil {
+						return err
+					}
+					// Make it read-only (can't be opened for writing in case a new key needs to be written)
+					return os.Chmod(keyPath, 0400)
+				},
+				errorCheck: func(t *testing.T, err error) {
+					// On read-only key, the parser will attempt to parse it and fail with "ssh: no key found"
+					assert.Error(t, err)
+				},
+			},
+			{
+				name: "EmptyFile",
+				setupFunc: func(dir string) error {
+					// Create an empty file
+					return os.WriteFile(filepath.Join(dir, "id_ed25519"), []byte{}, 0600)
+				},
+				errorCheck: func(t *testing.T, err error) {
+					assert.Error(t, err)
+					// The error from attempting to parse an empty file
+					assert.Contains(t, err.Error(), "ssh: no key found")
+				},
+			},
+		}
+
+		for _, tc := range tests {
+			t.Run(tc.name, func(t *testing.T) {
+				tempDir := t.TempDir()
+
+				// Setup the test case
+				err := tc.setupFunc(tempDir)
+				require.NoError(t, err, "Setup failed")
+
+				// Reset h.pubKey before each test case
+				hub.SetPubkey("")
+
+				// Attempt to get SSH key
+				_, err = hub.GetSSHKey(tempDir)
+
+				// Verify the error
+				tc.errorCheck(t, err)
+
+				// Check that pubKey was not set in error cases
+				assert.Empty(t, hub.GetPubkey(), "h.pubKey should not be set if there was an error")
+			})
+		}
+	})
+}
+
+func TestApiRoutesAuthentication(t *testing.T) {
+	hub, _ := beszelTests.NewTestHub(t.TempDir())
+	defer hub.Cleanup()
+
+	hub.StartHub()
+
+	// Create test user and get auth token
+	user, err := beszelTests.CreateUser(hub, "testuser@example.com", "password123")
+	require.NoError(t, err, "Failed to create test user")
+
+	adminUser, err := beszelTests.CreateRecord(hub, "users", map[string]any{
+		"email":    "admin@example.com",
+		"password": "password123",
+		"role":     "admin",
+	})
+	require.NoError(t, err, "Failed to create admin user")
+	adminUserToken, err := adminUser.NewAuthToken()
+
+	// superUser, err := beszelTests.CreateRecord(hub, core.CollectionNameSuperusers, map[string]any{
+	// 	"email":    "superuser@example.com",
+	// 	"password": "password123",
+	// })
+	// require.NoError(t, err, "Failed to create superuser")
+
+	userToken, err := user.NewAuthToken()
+	require.NoError(t, err, "Failed to create auth token")
+
+	// Create test system for user-alerts endpoints
+	system, err := beszelTests.CreateRecord(hub, "systems", map[string]any{
+		"name":  "test-system",
+		"users": []string{user.Id},
+		"host":  "127.0.0.1",
+	})
+	require.NoError(t, err, "Failed to create test system")
+
+	testAppFactory := func(t testing.TB) *pbTests.TestApp {
+		return hub.TestApp
+	}
+
+	scenarios := []beszelTests.ApiScenario{
+		// Auth Protected Routes - Should require authentication
+		{
+			Name:            "POST /test-notification - no auth should fail",
+			Method:          http.MethodPost,
+			URL:             "/api/beszel/test-notification",
+			ExpectedStatus:  401,
+			ExpectedContent: []string{"requires valid"},
+			TestAppFactory:  testAppFactory,
+			Body: jsonReader(map[string]any{
+				"url": "generic://127.0.0.1",
+			}),
+		},
+		{
+			Name:           "POST /test-notification - with auth should succeed",
+			Method:         http.MethodPost,
+			URL:            "/api/beszel/test-notification",
+			TestAppFactory: testAppFactory,
+			Headers: map[string]string{
+				"Authorization": userToken,
+			},
+			Body: jsonReader(map[string]any{
+				"url": "generic://127.0.0.1",
+			}),
+			ExpectedStatus:  200,
+			ExpectedContent: []string{"sending message"},
+		},
+		{
+			Name:            "GET /config-yaml - no auth should fail",
+			Method:          http.MethodGet,
+			URL:             "/api/beszel/config-yaml",
+			ExpectedStatus:  401,
+			ExpectedContent: []string{"requires valid"},
+			TestAppFactory:  testAppFactory,
+		},
+		{
+			Name:   "GET /config-yaml - with user auth should fail",
+			Method: http.MethodGet,
+			URL:    "/api/beszel/config-yaml",
+			Headers: map[string]string{
+				"Authorization": userToken,
+			},
+			ExpectedStatus:  403,
+			ExpectedContent: []string{"Requires admin"},
+			TestAppFactory:  testAppFactory,
+		},
+		{
+			Name:   "GET /config-yaml - with admin auth should succeed",
+			Method: http.MethodGet,
+			URL:    "/api/beszel/config-yaml",
+			Headers: map[string]string{
+				"Authorization": adminUserToken,
+			},
+			ExpectedStatus:  200,
+			ExpectedContent: []string{"test-system"},
+			TestAppFactory:  testAppFactory,
+		},
+		{
+			Name:            "GET /universal-token - no auth should fail",
+			Method:          http.MethodGet,
+			URL:             "/api/beszel/universal-token",
+			ExpectedStatus:  401,
+			ExpectedContent: []string{"requires valid"},
+			TestAppFactory:  testAppFactory,
+		},
+		{
+			Name:   "GET /universal-token - with auth should succeed",
+			Method: http.MethodGet,
+			URL:    "/api/beszel/universal-token",
+			Headers: map[string]string{
+				"Authorization": userToken,
+			},
+			ExpectedStatus:  200,
+			ExpectedContent: []string{"active", "token"},
+			TestAppFactory:  testAppFactory,
+		},
+		{
+			Name:            "POST /user-alerts - no auth should fail",
+			Method:          http.MethodPost,
+			URL:             "/api/beszel/user-alerts",
+			ExpectedStatus:  401,
+			ExpectedContent: []string{"requires valid"},
+			TestAppFactory:  testAppFactory,
+			Body: jsonReader(map[string]any{
+				"name":    "CPU",
+				"value":   80,
+				"min":     10,
+				"systems": []string{system.Id},
+			}),
+		},
+		{
+			Name:   "POST /user-alerts - with auth should succeed",
+			Method: http.MethodPost,
+			URL:    "/api/beszel/user-alerts",
+			Headers: map[string]string{
+				"Authorization": userToken,
+			},
+			ExpectedStatus:  200,
+			ExpectedContent: []string{"\"success\":true"},
+			TestAppFactory:  testAppFactory,
+			Body: jsonReader(map[string]any{
+				"name":    "CPU",
+				"value":   80,
+				"min":     10,
+				"systems": []string{system.Id},
+			}),
+		},
+		{
+			Name:            "DELETE /user-alerts - no auth should fail",
+			Method:          http.MethodDelete,
+			URL:             "/api/beszel/user-alerts",
+			ExpectedStatus:  401,
+			ExpectedContent: []string{"requires valid"},
+			TestAppFactory:  testAppFactory,
+			Body: jsonReader(map[string]any{
+				"name":    "CPU",
+				"systems": []string{system.Id},
+			}),
+		},
+		{
+			Name:   "DELETE /user-alerts - with auth should succeed",
+			Method: http.MethodDelete,
+			URL:    "/api/beszel/user-alerts",
+			Headers: map[string]string{
+				"Authorization": userToken,
+			},
+			ExpectedStatus:  200,
+			ExpectedContent: []string{"\"success\":true"},
+			TestAppFactory:  testAppFactory,
+			Body: jsonReader(map[string]any{
+				"name":    "CPU",
+				"systems": []string{system.Id},
+			}),
+			BeforeTestFunc: func(t testing.TB, app *pbTests.TestApp, e *core.ServeEvent) {
+				// Create an alert to delete
+				beszelTests.CreateRecord(app, "alerts", map[string]any{
+					"name":   "CPU",
+					"system": system.Id,
+					"user":   user.Id,
+					"value":  80,
+					"min":    10,
+				})
+			},
+		},
+
+		// Auth Optional Routes - Should work without authentication
+		{
+			Name:            "GET /getkey - no auth should fail",
+			Method:          http.MethodGet,
+			URL:             "/api/beszel/getkey",
+			ExpectedStatus:  401,
+			ExpectedContent: []string{"requires valid"},
+			TestAppFactory:  testAppFactory,
+		},
+		{
+			Name:   "GET /getkey - with auth should also succeed",
+			Method: http.MethodGet,
+			URL:    "/api/beszel/getkey",
+			Headers: map[string]string{
+				"Authorization": userToken,
+			},
+			ExpectedStatus:  200,
+			ExpectedContent: []string{"\"key\":", "\"v\":"},
+			TestAppFactory:  testAppFactory,
+		},
+		{
+			Name:            "GET /first-run - no auth should succeed",
+			Method:          http.MethodGet,
+			URL:             "/api/beszel/first-run",
+			ExpectedStatus:  200,
+			ExpectedContent: []string{"\"firstRun\":false"},
+			TestAppFactory:  testAppFactory,
+		},
+		{
+			Name:   "GET /first-run - with auth should also succeed",
+			Method: http.MethodGet,
+			URL:    "/api/beszel/first-run",
+			Headers: map[string]string{
+				"Authorization": userToken,
+			},
+			ExpectedStatus:  200,
+			ExpectedContent: []string{"\"firstRun\":false"},
+			TestAppFactory:  testAppFactory,
+		},
+		{
+			Name:            "GET /agent-connect - no auth should succeed (websocket upgrade fails but route is accessible)",
+			Method:          http.MethodGet,
+			URL:             "/api/beszel/agent-connect",
+			ExpectedStatus:  400,
+			ExpectedContent: []string{},
+			TestAppFactory:  testAppFactory,
+		},
+		{
+			Name:   "POST /test-notification - invalid auth token should fail",
+			Method: http.MethodPost,
+			URL:    "/api/beszel/test-notification",
+			Body: jsonReader(map[string]any{
+				"url": "generic://127.0.0.1",
+			}),
+			Headers: map[string]string{
+				"Authorization": "invalid-token",
+			},
+			ExpectedStatus:  401,
+			ExpectedContent: []string{"requires valid"},
+			TestAppFactory:  testAppFactory,
+		},
+		{
+			Name:   "POST /user-alerts - invalid auth token should fail",
+			Method: http.MethodPost,
+			URL:    "/api/beszel/user-alerts",
+			Headers: map[string]string{
+				"Authorization": "invalid-token",
+			},
+			ExpectedStatus:  401,
+			ExpectedContent: []string{"requires valid"},
+			TestAppFactory:  testAppFactory,
+			Body: jsonReader(map[string]any{
+				"name":    "CPU",
+				"value":   80,
+				"min":     10,
+				"systems": []string{system.Id},
+			}),
+		},
+	}
+
+	for _, scenario := range scenarios {
+		scenario.Test(t)
+	}
+}
+
+func TestFirstUserCreation(t *testing.T) {
+	t.Run("CreateUserEndpoint available when no users exist", func(t *testing.T) {
+		hub, _ := beszelTests.NewTestHub(t.TempDir())
+		defer hub.Cleanup()
+
+		hub.StartHub()
+
+		testAppFactoryExisting := func(t testing.TB) *pbTests.TestApp {
+			return hub.TestApp
+		}
+
+		scenarios := []beszelTests.ApiScenario{
+			{
+				Name:   "POST /create-user - should be available when no users exist",
+				Method: http.MethodPost,
+				URL:    "/api/beszel/create-user",
+				Body: jsonReader(map[string]any{
+					"email":    "firstuser@example.com",
+					"password": "password123",
+				}),
+				ExpectedStatus:  200,
+				ExpectedContent: []string{"User created"},
+				TestAppFactory:  testAppFactoryExisting,
+				BeforeTestFunc: func(t testing.TB, app *pbTests.TestApp, e *core.ServeEvent) {
+					userCount, err := hub.CountRecords("users")
+					require.NoError(t, err)
+					require.Zero(t, userCount, "Should start with no users")
+					superusers, err := hub.FindAllRecords(core.CollectionNameSuperusers)
+					require.NoError(t, err)
+					require.EqualValues(t, 1, len(superusers), "Should start with one temporary superuser")
+					require.EqualValues(t, migrations.TempAdminEmail, superusers[0].GetString("email"), "Should have created one temporary superuser")
+				},
+				AfterTestFunc: func(t testing.TB, app *pbTests.TestApp, res *http.Response) {
+					userCount, err := hub.CountRecords("users")
+					require.NoError(t, err)
+					require.EqualValues(t, 1, userCount, "Should have created one user")
+					superusers, err := hub.FindAllRecords(core.CollectionNameSuperusers)
+					require.NoError(t, err)
+					require.EqualValues(t, 1, len(superusers), "Should have created one superuser")
+					require.EqualValues(t, "firstuser@example.com", superusers[0].GetString("email"), "Should have created one superuser")
+				},
+			},
+			{
+				Name:   "POST /create-user - should not be available when users exist",
+				Method: http.MethodPost,
+				URL:    "/api/beszel/create-user",
+				Body: jsonReader(map[string]any{
+					"email":    "firstuser@example.com",
+					"password": "password123",
+				}),
+				ExpectedStatus:  404,
+				ExpectedContent: []string{"wasn't found"},
+				TestAppFactory:  testAppFactoryExisting,
+			},
+		}
+
+		for _, scenario := range scenarios {
+			scenario.Test(t)
+		}
+	})
+
+	t.Run("CreateUserEndpoint not available when USER_EMAIL, USER_PASSWORD are set", func(t *testing.T) {
+		os.Setenv("BESZEL_HUB_USER_EMAIL", "me@example.com")
+		os.Setenv("BESZEL_HUB_USER_PASSWORD", "password123")
+		defer os.Unsetenv("BESZEL_HUB_USER_EMAIL")
+		defer os.Unsetenv("BESZEL_HUB_USER_PASSWORD")
+
+		hub, _ := beszelTests.NewTestHub(t.TempDir())
+		defer hub.Cleanup()
+
+		hub.StartHub()
+
+		testAppFactory := func(t testing.TB) *pbTests.TestApp {
+			return hub.TestApp
+		}
+
+		scenario := beszelTests.ApiScenario{
+			Name:            "POST /create-user - should not be available when USER_EMAIL, USER_PASSWORD are set",
+			Method:          http.MethodPost,
+			URL:             "/api/beszel/create-user",
+			ExpectedStatus:  404,
+			ExpectedContent: []string{"wasn't found"},
+			TestAppFactory:  testAppFactory,
+			BeforeTestFunc: func(t testing.TB, app *pbTests.TestApp, e *core.ServeEvent) {
+				users, err := hub.FindAllRecords("users")
+				require.NoError(t, err)
+				require.EqualValues(t, 1, len(users), "Should start with one user")
+				require.EqualValues(t, "me@example.com", users[0].GetString("email"), "Should have created one user")
+				superusers, err := hub.FindAllRecords(core.CollectionNameSuperusers)
+				require.NoError(t, err)
+				require.EqualValues(t, 1, len(superusers), "Should start with one superuser")
+				require.EqualValues(t, "me@example.com", superusers[0].GetString("email"), "Should have created one superuser")
+			},
+			AfterTestFunc: func(t testing.TB, app *pbTests.TestApp, res *http.Response) {
+				users, err := hub.FindAllRecords("users")
+				require.NoError(t, err)
+				require.EqualValues(t, 1, len(users), "Should still have one user")
+				require.EqualValues(t, "me@example.com", users[0].GetString("email"), "Should have created one user")
+				superusers, err := hub.FindAllRecords(core.CollectionNameSuperusers)
+				require.NoError(t, err)
+				require.EqualValues(t, 1, len(superusers), "Should still have one superuser")
+				require.EqualValues(t, "me@example.com", superusers[0].GetString("email"), "Should have created one superuser")
+			},
+		}
+
+		scenario.Test(t)
+	})
+}
+
+func TestCreateUserEndpointAvailability(t *testing.T) {
+	t.Run("CreateUserEndpoint available when no users exist", func(t *testing.T) {
+		hub, _ := beszelTests.NewTestHub(t.TempDir())
+		defer hub.Cleanup()
+
+		// Ensure no users exist
+		userCount, err := hub.CountRecords("users")
+		require.NoError(t, err)
+		require.Zero(t, userCount, "Should start with no users")
+
+		hub.StartHub()
+
+		testAppFactory := func(t testing.TB) *pbTests.TestApp {
+			return hub.TestApp
+		}
+
+		scenario := beszelTests.ApiScenario{
+			Name:   "POST /create-user - should be available when no users exist",
+			Method: http.MethodPost,
+			URL:    "/api/beszel/create-user",
+			Body: jsonReader(map[string]any{
+				"email":    "firstuser@example.com",
+				"password": "password123",
+			}),
+			ExpectedStatus:  200,
+			ExpectedContent: []string{"User created"},
+			TestAppFactory:  testAppFactory,
+		}
+
+		scenario.Test(t)
+
+		// Verify user was created
+		userCount, err = hub.CountRecords("users")
+		require.NoError(t, err)
+		require.EqualValues(t, 1, userCount, "Should have created one user")
+	})
+
+	t.Run("CreateUserEndpoint not available when users exist", func(t *testing.T) {
+		hub, _ := beszelTests.NewTestHub(t.TempDir())
+		defer hub.Cleanup()
+
+		// Create a user first
+		_, err := beszelTests.CreateUser(hub, "existing@example.com", "password")
+		require.NoError(t, err)
+
+		hub.StartHub()
+
+		testAppFactory := func(t testing.TB) *pbTests.TestApp {
+			return hub.TestApp
+		}
+
+		scenario := beszelTests.ApiScenario{
+			Name:   "POST /create-user - should not be available when users exist",
+			Method: http.MethodPost,
+			URL:    "/api/beszel/create-user",
+			Body: jsonReader(map[string]any{
+				"email":    "another@example.com",
+				"password": "password123",
+			}),
+			ExpectedStatus:  404,
+			ExpectedContent: []string{"wasn't found"},
+			TestAppFactory:  testAppFactory,
+		}
+
+		scenario.Test(t)
+	})
+}

internal/hub/hub_test_helpers.go

@@ -0,0 +1,21 @@
+//go:build testing
+// +build testing
+
+package hub
+
+import "github.com/henrygd/beszel/src/hub/systems"
+
+// TESTING ONLY: GetSystemManager returns the system manager
+func (h *Hub) GetSystemManager() *systems.SystemManager {
+	return h.sm
+}
+
+// TESTING ONLY: GetPubkey returns the public key
+func (h *Hub) GetPubkey() string {
+	return h.pubKey
+}
+
+// TESTING ONLY: SetPubkey sets the public key
+func (h *Hub) SetPubkey(pubkey string) {
+	h.pubKey = pubkey
+}

internal/hub/server_development.go

@@ -0,0 +1,82 @@
+//go:build development
+
+package hub
+
+import (
+	"fmt"
+	"io"
+	"log/slog"
+	"net/http"
+	"net/http/httputil"
+	"net/url"
+	"strings"
+
+	"github.com/henrygd/beszel"
+
+	"github.com/pocketbase/pocketbase/core"
+	"github.com/pocketbase/pocketbase/tools/osutils"
+)
+
+// Wraps http.RoundTripper to modify dev proxy HTML responses
+type responseModifier struct {
+	transport http.RoundTripper
+	hub       *Hub
+}
+
+func (rm *responseModifier) RoundTrip(req *http.Request) (*http.Response, error) {
+	resp, err := rm.transport.RoundTrip(req)
+	if err != nil {
+		return resp, err
+	}
+	// Only modify HTML responses
+	contentType := resp.Header.Get("Content-Type")
+	if !strings.Contains(contentType, "text/html") {
+		return resp, nil
+	}
+	body, err := io.ReadAll(resp.Body)
+	if err != nil {
+		return resp, err
+	}
+	resp.Body.Close()
+	// Create a new response with the modified body
+	modifiedBody := rm.modifyHTML(string(body))
+	resp.Body = io.NopCloser(strings.NewReader(modifiedBody))
+	resp.ContentLength = int64(len(modifiedBody))
+	resp.Header.Set("Content-Length", fmt.Sprintf("%d", len(modifiedBody)))
+
+	return resp, nil
+}
+
+func (rm *responseModifier) modifyHTML(html string) string {
+	parsedURL, err := url.Parse(rm.hub.appURL)
+	if err != nil {
+		return html
+	}
+	// fix base paths in html if using subpath
+	basePath := strings.TrimSuffix(parsedURL.Path, "/") + "/"
+	html = strings.ReplaceAll(html, "./", basePath)
+	html = strings.Replace(html, "{{V}}", beszel.Version, 1)
+	html = strings.Replace(html, "{{HUB_URL}}", rm.hub.appURL, 1)
+	return html
+}
+
+// startServer sets up the development server for Beszel
+func (h *Hub) startServer(se *core.ServeEvent) error {
+	slog.Info("starting server", "appURL", h.appURL)
+	proxy := httputil.NewSingleHostReverseProxy(&url.URL{
+		Scheme: "http",
+		Host:   "localhost:5173",
+	})
+
+	proxy.Transport = &responseModifier{
+		transport: http.DefaultTransport,
+		hub:       h,
+	}
+
+	se.Router.GET("/{path...}", func(e *core.RequestEvent) error {
+		proxy.ServeHTTP(e.Response, e.Request)
+		return nil
+	})
+	_ = osutils.LaunchURL(h.appURL)
+	return nil
+}

internal/hub/server_production.go

@@ -0,0 +1,52 @@
+//go:build !development
+
+package hub
+
+import (
+	"io/fs"
+	"net/http"
+	"net/url"
+	"strings"
+
+	"github.com/henrygd/beszel"
+	"github.com/henrygd/beszel/src/site"
+
+	"github.com/pocketbase/pocketbase/apis"
+	"github.com/pocketbase/pocketbase/core"
+)
+
+// startServer sets up the production server for Beszel
+func (h *Hub) startServer(se *core.ServeEvent) error {
+	// parse app url
+	parsedURL, err := url.Parse(h.appURL)
+	if err != nil {
+		return err
+	}
+	// fix base paths in html if using subpath
+	basePath := strings.TrimSuffix(parsedURL.Path, "/") + "/"
+	indexFile, _ := fs.ReadFile(site.DistDirFS, "index.html")
+	html := strings.ReplaceAll(string(indexFile), "./", basePath)
+	html = strings.Replace(html, "{{V}}", beszel.Version, 1)
+	html = strings.Replace(html, "{{HUB_URL}}", h.appURL, 1)
+	// set up static asset serving
+	staticPaths := [2]string{"/static/", "/assets/"}
+	serveStatic := apis.Static(site.DistDirFS, false)
+	// get CSP configuration
+	csp, cspExists := GetEnv("CSP")
+	// add route
+	se.Router.GET("/{path...}", func(e *core.RequestEvent) error {
+		// serve static assets if path is in staticPaths
+		for i := range staticPaths {
+			if strings.Contains(e.Request.URL.Path, staticPaths[i]) {
+				e.Response.Header().Set("Cache-Control", "public, max-age=2592000")
+				return serveStatic(e)
+			}
+		}
+		if cspExists {
+			e.Response.Header().Del("X-Frame-Options")
+			e.Response.Header().Set("Content-Security-Policy", csp)
+		}
+		return e.HTML(http.StatusOK, html)
+	})
+	return nil
+}

internal/hub/systems/system.go

@@ -0,0 +1,390 @@
+package systems
+
+import (
+	"context"
+	"encoding/json"
+	"errors"
+	"fmt"
+	"math/rand"
+	"net"
+	"strings"
+	"time"
+
+	"github.com/henrygd/beszel/src/hub/ws"
+
+	"github.com/henrygd/beszel/src/entities/system"
+
+	"github.com/henrygd/beszel"
+
+	"github.com/blang/semver"
+	"github.com/fxamacker/cbor/v2"
+	"github.com/pocketbase/pocketbase/core"
+	"golang.org/x/crypto/ssh"
+)
+
+type System struct {
+	Id           string               `db:"id"`
+	Host         string               `db:"host"`
+	Port         string               `db:"port"`
+	Status       string               `db:"status"`
+	manager      *SystemManager       // Manager that this system belongs to
+	client       *ssh.Client          // SSH client for fetching data
+	data         *system.CombinedData // system data from agent
+	ctx          context.Context      // Context for stopping the updater
+	cancel       context.CancelFunc   // Stops and removes system from updater
+	WsConn       *ws.WsConn           // Handler for agent WebSocket connection
+	agentVersion semver.Version       // Agent version
+	updateTicker *time.Ticker         // Ticker for updating the system
+}
+
+func (sm *SystemManager) NewSystem(systemId string) *System {
+	system := &System{
+		Id:   systemId,
+		data: &system.CombinedData{},
+	}
+	system.ctx, system.cancel = system.getContext()
+	return system
+}
+
+// StartUpdater starts the system updater.
+// It first fetches the data from the agent then updates the records.
+// If the data is not found or the system is down, it sets the system down.
+func (sys *System) StartUpdater() {
+	// Channel that can be used to set the system down. Currently only used to
+	// allow a short delay for reconnection after websocket connection is closed.
+	var downChan chan struct{}
+
+	// Add random jitter to first WebSocket connection to prevent
+	// clustering if all agents are started at the same time.
+	// SSH connections during hub startup are already staggered.
+	var jitter <-chan time.Time
+	if sys.WsConn != nil {
+		jitter = getJitter()
+		// use the websocket connection's down channel to set the system down
+		downChan = sys.WsConn.DownChan
+	} else {
+		// if the system does not have a websocket connection, wait before updating
+		// to allow the agent to connect via websocket (makes sure fingerprint is set).
+		time.Sleep(11 * time.Second)
+	}
+
+	// update immediately if system is not paused (only for ws connections)
+	// we'll wait a minute before connecting via SSH to prioritize ws connections
+	if sys.Status != paused && sys.ctx.Err() == nil {
+		if err := sys.update(); err != nil {
+			_ = sys.setDown(err)
+		}
+	}
+
+	sys.updateTicker = time.NewTicker(time.Duration(interval) * time.Millisecond)
+	// Go 1.23+ will automatically stop the ticker when the system is garbage collected, however we seem to need this or testing/synctest will block even if calling runtime.GC()
+	defer sys.updateTicker.Stop()
+
+	for {
+		select {
+		case <-sys.ctx.Done():
+			return
+		case <-sys.updateTicker.C:
+			if err := sys.update(); err != nil {
+				_ = sys.setDown(err)
+			}
+		case <-downChan:
+			sys.WsConn = nil
+			downChan = nil
+			_ = sys.setDown(nil)
+		case <-jitter:
+			sys.updateTicker.Reset(time.Duration(interval) * time.Millisecond)
+			if err := sys.update(); err != nil {
+				_ = sys.setDown(err)
+			}
+		}
+	}
+}
+
+// update updates the system data and records.
+func (sys *System) update() error {
+	if sys.Status == paused {
+		sys.handlePaused()
+		return nil
+	}
+	data, err := sys.fetchDataFromAgent()
+	if err == nil {
+		_, err = sys.createRecords(data)
+	}
+	return err
+}
+
+func (sys *System) handlePaused() {
+	if sys.WsConn == nil {
+		// if the system is paused and there's no websocket connection, remove the system
+		_ = sys.manager.RemoveSystem(sys.Id)
+	} else {
+		// Send a ping to the agent to keep the connection alive if the system is paused
+		if err := sys.WsConn.Ping(); err != nil {
+			sys.manager.hub.Logger().Warn("Failed to ping agent", "system", sys.Id, "err", err)
+			_ = sys.manager.RemoveSystem(sys.Id)
+		}
+	}
+}
+
+// createRecords updates the system record and adds system_stats and container_stats records
+func (sys *System) createRecords(data *system.CombinedData) (*core.Record, error) {
+	systemRecord, err := sys.getRecord()
+	if err != nil {
+		return nil, err
+	}
+	hub := sys.manager.hub
+	// add system_stats and container_stats records
+	systemStatsCollection, err := hub.FindCachedCollectionByNameOrId("system_stats")
+	if err != nil {
+		return nil, err
+	}
+
+	systemStatsRecord := core.NewRecord(systemStatsCollection)
+	systemStatsRecord.Set("system", systemRecord.Id)
+	systemStatsRecord.Set("stats", data.Stats)
+	systemStatsRecord.Set("type", "1m")
+	if err := hub.SaveNoValidate(systemStatsRecord); err != nil {
+		return nil, err
+	}
+	// add new container_stats record
+	if len(data.Containers) > 0 {
+		containerStatsCollection, err := hub.FindCachedCollectionByNameOrId("container_stats")
+		if err != nil {
+			return nil, err
+		}
+		containerStatsRecord := core.NewRecord(containerStatsCollection)
+		containerStatsRecord.Set("system", systemRecord.Id)
+		containerStatsRecord.Set("stats", data.Containers)
+		containerStatsRecord.Set("type", "1m")
+		if err := hub.SaveNoValidate(containerStatsRecord); err != nil {
+			return nil, err
+		}
+	}
+	// update system record (do this last because it triggers alerts and we need above records to be inserted first)
+	systemRecord.Set("status", up)
+
+	systemRecord.Set("info", data.Info)
+	if err := hub.SaveNoValidate(systemRecord); err != nil {
+		return nil, err
+	}
+	return systemRecord, nil
+}
+
+// getRecord retrieves the system record from the database.
+// If the record is not found, it removes the system from the manager.
+func (sys *System) getRecord() (*core.Record, error) {
+	record, err := sys.manager.hub.FindRecordById("systems", sys.Id)
+	if err != nil || record == nil {
+		_ = sys.manager.RemoveSystem(sys.Id)
+		return nil, err
+	}
+	return record, nil
+}
+
+// setDown marks a system as down in the database.
+// It takes the original error that caused the system to go down and returns any error
+// encountered during the process of updating the system status.
+func (sys *System) setDown(originalError error) error {
+	if sys.Status == down || sys.Status == paused {
+		return nil
+	}
+	record, err := sys.getRecord()
+	if err != nil {
+		return err
+	}
+	if originalError != nil {
+		sys.manager.hub.Logger().Error("System down", "system", record.GetString("name"), "err", originalError)
+	}
+	record.Set("status", down)
+	return sys.manager.hub.SaveNoValidate(record)
+}
+
+func (sys *System) getContext() (context.Context, context.CancelFunc) {
+	if sys.ctx == nil {
+		sys.ctx, sys.cancel = context.WithCancel(context.Background())
+	}
+	return sys.ctx, sys.cancel
+}
+
+// fetchDataFromAgent attempts to fetch data from the agent,
+// prioritizing WebSocket if available.
+func (sys *System) fetchDataFromAgent() (*system.CombinedData, error) {
+	if sys.data == nil {
+		sys.data = &system.CombinedData{}
+	}
+
+	if sys.WsConn != nil && sys.WsConn.IsConnected() {
+		wsData, err := sys.fetchDataViaWebSocket()
+		if err == nil {
+			return wsData, nil
+		}
+		// close the WebSocket connection if error and try SSH
+		sys.closeWebSocketConnection()
+	}
+
+	sshData, err := sys.fetchDataViaSSH()
+	if err != nil {
+		return nil, err
+	}
+	return sshData, nil
+}
+
+func (sys *System) fetchDataViaWebSocket() (*system.CombinedData, error) {
+	if sys.WsConn == nil || !sys.WsConn.IsConnected() {
+		return nil, errors.New("no websocket connection")
+	}
+	err := sys.WsConn.RequestSystemData(sys.data)
+	if err != nil {
+		return nil, err
+	}
+	return sys.data, nil
+}
+
+// fetchDataViaSSH handles fetching data using SSH.
+// This function encapsulates the original SSH logic.
+// It updates sys.data directly upon successful fetch.
+func (sys *System) fetchDataViaSSH() (*system.CombinedData, error) {
+	maxRetries := 1
+	for attempt := 0; attempt <= maxRetries; attempt++ {
+		if sys.client == nil || sys.Status == down {
+			if err := sys.createSSHClient(); err != nil {
+				return nil, err
+			}
+		}
+
+		session, err := sys.createSessionWithTimeout(4 * time.Second)
+		if err != nil {
+			if attempt >= maxRetries {
+				return nil, err
+			}
+			sys.manager.hub.Logger().Warn("Session closed. Retrying...", "host", sys.Host, "port", sys.Port, "err", err)
+			sys.closeSSHConnection()
+			// Reset format detection on connection failure - agent might have been upgraded
+			continue
+		}
+		defer session.Close()
+
+		stdout, err := session.StdoutPipe()
+		if err != nil {
+			return nil, err
+		}
+		if err := session.Shell(); err != nil {
+			return nil, err
+		}
+
+		*sys.data = system.CombinedData{}
+
+		if sys.agentVersion.GTE(beszel.MinVersionCbor) {
+			err = cbor.NewDecoder(stdout).Decode(sys.data)
+		} else {
+			err = json.NewDecoder(stdout).Decode(sys.data)
+		}
+
+		if err != nil {
+			sys.closeSSHConnection()
+			if attempt < maxRetries {
+				continue
+			}
+			return nil, err
+		}
+
+		// wait for the session to complete
+		if err := session.Wait(); err != nil {
+			return nil, err
+		}
+
+		return sys.data, nil
+	}
+
+	// this should never be reached due to the return in the loop
+	return nil, fmt.Errorf("failed to fetch data")
+}
+
+// createSSHClient creates a new SSH client for the system
+func (s *System) createSSHClient() error {
+	if s.manager.sshConfig == nil {
+		if err := s.manager.createSSHClientConfig(); err != nil {
+			return err
+		}
+	}
+	network := "tcp"
+	host := s.Host
+	if strings.HasPrefix(host, "/") {
+		network = "unix"
+	} else {
+		host = net.JoinHostPort(host, s.Port)
+	}
+	var err error
+	s.client, err = ssh.Dial(network, host, s.manager.sshConfig)
+	if err != nil {
+		return err
+	}
+	s.agentVersion, _ = extractAgentVersion(string(s.client.Conn.ServerVersion()))
+	return nil
+}
+
+// createSessionWithTimeout creates a new SSH session with a timeout to avoid hanging
+// in case of network issues
+func (sys *System) createSessionWithTimeout(timeout time.Duration) (*ssh.Session, error) {
+	if sys.client == nil {
+		return nil, fmt.Errorf("client not initialized")
+	}
+
+	ctx, cancel := context.WithTimeout(sys.ctx, timeout)
+	defer cancel()
+
+	sessionChan := make(chan *ssh.Session, 1)
+	errChan := make(chan error, 1)
+
+	go func() {
+		if session, err := sys.client.NewSession(); err != nil {
+			errChan <- err
+		} else {
+			sessionChan <- session
+		}
+	}()
+
+	select {
+	case session := <-sessionChan:
+		return session, nil
+	case err := <-errChan:
+		return nil, err
+	case <-ctx.Done():
+		return nil, fmt.Errorf("timeout")
+	}
+}
+
+// closeSSHConnection closes the SSH connection but keeps the system in the manager
+func (sys *System) closeSSHConnection() {
+	if sys.client != nil {
+		sys.client.Close()
+		sys.client = nil
+	}
+}
+
+// closeWebSocketConnection closes the WebSocket connection but keeps the system in the manager
+// to allow updating via SSH. It will be removed if the WS connection is re-established.
+// The system will be set as down a few seconds later if the connection is not re-established.
+func (sys *System) closeWebSocketConnection() {
+	if sys.WsConn != nil {
+		sys.WsConn.Close(nil)
+	}
+}
+
+// extractAgentVersion extracts the beszel version from SSH server version string
+func extractAgentVersion(versionString string) (semver.Version, error) {
+	_, after, _ := strings.Cut(versionString, "_")
+	return semver.Parse(after)
+}
+
+// getJitter returns a channel that will be triggered after a random delay
+// between 40% and 90% of the interval.
+// This is used to stagger the initial WebSocket connections to prevent clustering.
+func getJitter() <-chan time.Time {
+	minPercent := 40
+	maxPercent := 90
+	jitterRange := maxPercent - minPercent
+	msDelay := (interval * minPercent / 100) + rand.Intn(interval*jitterRange/100)
+	return time.After(time.Duration(msDelay) * time.Millisecond)
+}

internal/hub/systems/system_manager.go

@@ -0,0 +1,348 @@
+package systems
+
+import (
+	"errors"
+	"fmt"
+	"time"
+
+	"github.com/henrygd/beszel/src/hub/ws"
+
+	"github.com/henrygd/beszel/src/entities/system"
+
+	"github.com/henrygd/beszel/src/common"
+
+	"github.com/henrygd/beszel"
+
+	"github.com/blang/semver"
+	"github.com/pocketbase/pocketbase/core"
+	"github.com/pocketbase/pocketbase/tools/store"
+	"golang.org/x/crypto/ssh"
+)
+
+// System status constants
+const (
+	up      string = "up"      // System is online and responding
+	down    string = "down"    // System is offline or not responding
+	paused  string = "paused"  // System monitoring is paused
+	pending string = "pending" // System is waiting on initial connection result
+
+	// interval is the default update interval in milliseconds (60 seconds)
+	interval int = 60_000
+	// interval int = 10_000 // Debug interval for faster updates
+
+	// sessionTimeout is the maximum time to wait for SSH connections
+	sessionTimeout = 4 * time.Second
+)
+
+// errSystemExists is returned when attempting to add a system that already exists
+var errSystemExists = errors.New("system exists")
+
+// SystemManager manages a collection of monitored systems and their connections.
+// It handles system lifecycle, status updates, and maintains both SSH and WebSocket connections.
+type SystemManager struct {
+	hub       hubLike                       // Hub interface for database and alert operations
+	systems   *store.Store[string, *System] // Thread-safe store of active systems
+	sshConfig *ssh.ClientConfig             // SSH client configuration for system connections
+}
+
+// hubLike defines the interface requirements for the hub dependency.
+// It extends core.App with system-specific functionality.
+type hubLike interface {
+	core.App
+	GetSSHKey(dataDir string) (ssh.Signer, error)
+	HandleSystemAlerts(systemRecord *core.Record, data *system.CombinedData) error
+	HandleStatusAlerts(status string, systemRecord *core.Record) error
+}
+
+// NewSystemManager creates a new SystemManager instance with the provided hub.
+// The hub must implement the hubLike interface to provide database and alert functionality.
+func NewSystemManager(hub hubLike) *SystemManager {
+	return &SystemManager{
+		systems: store.New(map[string]*System{}),
+		hub:     hub,
+	}
+}
+
+// Initialize sets up the system manager by binding event hooks and starting existing systems.
+// It configures SSH client settings and begins monitoring all non-paused systems from the database.
+// Systems are started with staggered delays to prevent overwhelming the hub during startup.
+func (sm *SystemManager) Initialize() error {
+	sm.bindEventHooks()
+
+	// Initialize SSH client configuration
+	err := sm.createSSHClientConfig()
+	if err != nil {
+		return err
+	}
+
+	// Load existing systems from database (excluding paused ones)
+	var systems []*System
+	err = sm.hub.DB().NewQuery("SELECT id, host, port, status FROM systems WHERE status != 'paused'").All(&systems)
+	if err != nil || len(systems) == 0 {
+		return err
+	}
+
+	// Start systems in background with staggered timing
+	go func() {
+		// Calculate staggered delay between system starts (max 2 seconds per system)
+		delta := interval / max(1, len(systems))
+		delta = min(delta, 2_000)
+		sleepTime := time.Duration(delta) * time.Millisecond
+
+		for _, system := range systems {
+			time.Sleep(sleepTime)
+			_ = sm.AddSystem(system)
+		}
+	}()
+	return nil
+}
+
+// bindEventHooks registers event handlers for system and fingerprint record changes.
+// These hooks ensure the system manager stays synchronized with database changes.
+func (sm *SystemManager) bindEventHooks() {
+	sm.hub.OnRecordCreate("systems").BindFunc(sm.onRecordCreate)
+	sm.hub.OnRecordAfterCreateSuccess("systems").BindFunc(sm.onRecordAfterCreateSuccess)
+	sm.hub.OnRecordUpdate("systems").BindFunc(sm.onRecordUpdate)
+	sm.hub.OnRecordAfterUpdateSuccess("systems").BindFunc(sm.onRecordAfterUpdateSuccess)
+	sm.hub.OnRecordAfterDeleteSuccess("systems").BindFunc(sm.onRecordAfterDeleteSuccess)
+	sm.hub.OnRecordAfterUpdateSuccess("fingerprints").BindFunc(sm.onTokenRotated)
+}
+
+// onTokenRotated handles fingerprint token rotation events.
+// When a system's authentication token is rotated, any existing WebSocket connection
+// must be closed to force re-authentication with the new token.
+func (sm *SystemManager) onTokenRotated(e *core.RecordEvent) error {
+	systemID := e.Record.GetString("system")
+	system, ok := sm.systems.GetOk(systemID)
+	if !ok {
+		return e.Next()
+	}
+	// No need to close connection if not connected via websocket
+	if system.WsConn == nil {
+		return e.Next()
+	}
+	system.setDown(nil)
+	sm.RemoveSystem(systemID)
+	return e.Next()
+}
+
+// onRecordCreate is called before a new system record is committed to the database.
+// It initializes the record with default values: empty info and pending status.
+func (sm *SystemManager) onRecordCreate(e *core.RecordEvent) error {
+	e.Record.Set("info", system.Info{})
+	e.Record.Set("status", pending)
+	return e.Next()
+}
+
+// onRecordAfterCreateSuccess is called after a new system record is successfully created.
+// It adds the new system to the manager to begin monitoring.
+func (sm *SystemManager) onRecordAfterCreateSuccess(e *core.RecordEvent) error {
+	if err := sm.AddRecord(e.Record, nil); err != nil {
+		e.App.Logger().Error("Error adding record", "err", err)
+	}
+	return e.Next()
+}
+
+// onRecordUpdate is called before a system record is updated in the database.
+// It clears system info when the status is changed to paused.
+func (sm *SystemManager) onRecordUpdate(e *core.RecordEvent) error {
+	if e.Record.GetString("status") == paused {
+		e.Record.Set("info", system.Info{})
+	}
+	return e.Next()
+}
+
+// onRecordAfterUpdateSuccess handles system record updates after they're committed to the database.
+// It manages system lifecycle based on status changes and triggers appropriate alerts.
+// Status transitions are handled as follows:
+// - paused: Closes SSH connection and deactivates alerts
+// - pending: Starts monitoring (reuses WebSocket if available)
+// - up: Triggers system alerts
+// - down: Triggers status change alerts
+func (sm *SystemManager) onRecordAfterUpdateSuccess(e *core.RecordEvent) error {
+	newStatus := e.Record.GetString("status")
+	prevStatus := pending
+	system, ok := sm.systems.GetOk(e.Record.Id)
+	if ok {
+		prevStatus = system.Status
+		system.Status = newStatus
+	}
+
+	switch newStatus {
+	case paused:
+		if ok {
+			// Pause monitoring but keep system in manager for potential resume
+			system.closeSSHConnection()
+		}
+		_ = deactivateAlerts(e.App, e.Record.Id)
+		return e.Next()
+	case pending:
+		// Resume monitoring, preferring existing WebSocket connection
+		if ok && system.WsConn != nil {
+			go system.update()
+			return e.Next()
+		}
+		// Start new monitoring session
+		if err := sm.AddRecord(e.Record, nil); err != nil {
+			e.App.Logger().Error("Error adding record", "err", err)
+		}
+		_ = deactivateAlerts(e.App, e.Record.Id)
+		return e.Next()
+	}
+
+	// Handle systems not in manager
+	if !ok {
+		return sm.AddRecord(e.Record, nil)
+	}
+
+	// Trigger system alerts when system comes online
+	if newStatus == up {
+		if err := sm.hub.HandleSystemAlerts(e.Record, system.data); err != nil {
+			e.App.Logger().Error("Error handling system alerts", "err", err)
+		}
+	}
+
+	// Trigger status change alerts for up/down transitions
+	if (newStatus == down && prevStatus == up) || (newStatus == up && prevStatus == down) {
+		if err := sm.hub.HandleStatusAlerts(newStatus, e.Record); err != nil {
+			e.App.Logger().Error("Error handling status alerts", "err", err)
+		}
+	}
+	return e.Next()
+}
+
+// onRecordAfterDeleteSuccess is called after a system record is successfully deleted.
+// It removes the system from the manager and cleans up all associated resources.
+func (sm *SystemManager) onRecordAfterDeleteSuccess(e *core.RecordEvent) error {
+	sm.RemoveSystem(e.Record.Id)
+	return e.Next()
+}
+
+// AddSystem adds a system to the manager and starts monitoring it.
+// It validates required fields, initializes the system context, and starts the update goroutine.
+// Returns error if a system with the same ID already exists.
+func (sm *SystemManager) AddSystem(sys *System) error {
+	if sm.systems.Has(sys.Id) {
+		return errSystemExists
+	}
+	if sys.Id == "" || sys.Host == "" {
+		return errors.New("system missing required fields")
+	}
+
+	// Initialize system for monitoring
+	sys.manager = sm
+	sys.ctx, sys.cancel = sys.getContext()
+	sys.data = &system.CombinedData{}
+	sm.systems.Set(sys.Id, sys)
+
+	// Start monitoring in background
+	go sys.StartUpdater()
+	return nil
+}
+
+// RemoveSystem removes a system from the manager and cleans up all associated resources.
+// It cancels the system's context, closes all connections, and removes it from the store.
+// Returns an error if the system is not found.
+func (sm *SystemManager) RemoveSystem(systemID string) error {
+	system, ok := sm.systems.GetOk(systemID)
+	if !ok {
+		return errors.New("system not found")
+	}
+
+	// Stop the update goroutine
+	if system.cancel != nil {
+		system.cancel()
+	}
+
+	// Clean up all connections
+	system.closeSSHConnection()
+	system.closeWebSocketConnection()
+	sm.systems.Remove(systemID)
+	return nil
+}
+
+// AddRecord creates a System instance from a database record and adds it to the manager.
+// If a system with the same ID already exists, it's removed first to ensure clean state.
+// If no system instance is provided, a new one is created.
+// This method is typically called when systems are created or their status changes to pending.
+func (sm *SystemManager) AddRecord(record *core.Record, system *System) (err error) {
+	// Remove existing system to ensure clean state
+	if sm.systems.Has(record.Id) {
+		_ = sm.RemoveSystem(record.Id)
+	}
+
+	// Create new system if none provided
+	if system == nil {
+		system = sm.NewSystem(record.Id)
+	}
+
+	// Populate system from record
+	system.Status = record.GetString("status")
+	system.Host = record.GetString("host")
+	system.Port = record.GetString("port")
+
+	return sm.AddSystem(system)
+}
+
+// AddWebSocketSystem creates and adds a system with an established WebSocket connection.
+// This method is called when an agent connects via WebSocket with valid authentication.
+// The system is immediately added to monitoring with the provided connection and version info.
+func (sm *SystemManager) AddWebSocketSystem(systemId string, agentVersion semver.Version, wsConn *ws.WsConn) error {
+	systemRecord, err := sm.hub.FindRecordById("systems", systemId)
+	if err != nil {
+		return err
+	}
+
+	system := sm.NewSystem(systemId)
+	system.WsConn = wsConn
+	system.agentVersion = agentVersion
+
+	if err := sm.AddRecord(systemRecord, system); err != nil {
+		return err
+	}
+	return nil
+}
+
+// createSSHClientConfig initializes the SSH client configuration for connecting to an agent's server
+func (sm *SystemManager) createSSHClientConfig() error {
+	privateKey, err := sm.hub.GetSSHKey("")
+	if err != nil {
+		return err
+	}
+
+	sm.sshConfig = &ssh.ClientConfig{
+		User: "u",
+		Auth: []ssh.AuthMethod{
+			ssh.PublicKeys(privateKey),
+		},
+		Config: ssh.Config{
+			Ciphers:      common.DefaultCiphers,
+			KeyExchanges: common.DefaultKeyExchanges,
+			MACs:         common.DefaultMACs,
+		},
+		HostKeyCallback: ssh.InsecureIgnoreHostKey(),
+		ClientVersion:   fmt.Sprintf("SSH-2.0-%s_%s", beszel.AppName, beszel.Version),
+		Timeout:         sessionTimeout,
+	}
+	return nil
+}
+
+// deactivateAlerts finds all triggered alerts for a system and sets them to inactive.
+// This is called when a system is paused or goes offline to prevent continued alerts.
+func deactivateAlerts(app core.App, systemID string) error {
+	// Note: Direct SQL updates don't trigger SSE, so we use the PocketBase API
+	// _, err := app.DB().NewQuery(fmt.Sprintf("UPDATE alerts SET triggered = false WHERE system = '%s'", systemID)).Execute()
+
+	alerts, err := app.FindRecordsByFilter("alerts", fmt.Sprintf("system = '%s' && triggered = 1", systemID), "", -1, 0)
+	if err != nil {
+		return err
+	}
+
+	for _, alert := range alerts {
+		alert.Set("triggered", false)
+		if err := app.SaveNoValidate(alert); err != nil {
+			return err
+		}
+	}
+	return nil
+}

internal/hub/systems/systems_test.go

@@ -0,0 +1,422 @@
+//go:build testing
+// +build testing
+
+package systems_test
+
+import (
+	"fmt"
+	"sync"
+	"testing"
+	"testing/synctest"
+	"time"
+
+	"github.com/henrygd/beszel/src/entities/container"
+	"github.com/henrygd/beszel/src/entities/system"
+	"github.com/henrygd/beszel/src/hub/systems"
+	"github.com/henrygd/beszel/src/tests"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+func TestSystemManagerNew(t *testing.T) {
+	hub, err := tests.NewTestHub(t.TempDir())
+	if err != nil {
+		t.Fatal(err)
+	}
+	defer hub.Cleanup()
+	sm := hub.GetSystemManager()
+
+	user, err := tests.CreateUser(hub, "test@test.com", "testtesttest")
+	require.NoError(t, err)
+
+	synctest.Test(t, func(t *testing.T) {
+		sm.Initialize()
+
+		record, err := tests.CreateRecord(hub, "systems", map[string]any{
+			"name":  "it-was-coney-island",
+			"host":  "the-playground-of-the-world",
+			"port":  "33914",
+			"users": []string{user.Id},
+		})
+		require.NoError(t, err)
+
+		assert.Equal(t, "pending", record.GetString("status"), "System status should be 'pending'")
+		assert.Equal(t, "pending", sm.GetSystemStatusFromStore(record.Id), "System status should be 'pending'")
+
+		// Verify the system host and port
+		host, port := sm.GetSystemHostPort(record.Id)
+		assert.Equal(t, record.GetString("host"), host, "System host should match")
+		assert.Equal(t, record.GetString("port"), port, "System port should match")
+
+		time.Sleep(13 * time.Second)
+		synctest.Wait()
+
+		assert.Equal(t, "pending", record.Fresh().GetString("status"), "System status should be 'pending'")
+		// Verify the system was added by checking if it exists
+		assert.True(t, sm.HasSystem(record.Id), "System should exist in the store")
+
+		time.Sleep(10 * time.Second)
+		synctest.Wait()
+
+		// system should be set to down after 15 seconds (no websocket connection)
+		assert.Equal(t, "down", sm.GetSystemStatusFromStore(record.Id), "System status should be 'down'")
+		// make sure the system is down in the db
+		record, err = hub.FindRecordById("systems", record.Id)
+		require.NoError(t, err)
+		assert.Equal(t, "down", record.GetString("status"), "System status should be 'down'")
+
+		assert.Equal(t, 1, sm.GetSystemCount(), "System count should be 1")
+
+		err = sm.RemoveSystem(record.Id)
+		assert.NoError(t, err)
+
+		assert.Equal(t, 0, sm.GetSystemCount(), "System count should be 0")
+		assert.False(t, sm.HasSystem(record.Id), "System should not exist in the store after removal")
+
+		// let's also make sure a system is removed from the store when the record is deleted
+		record, err = tests.CreateRecord(hub, "systems", map[string]any{
+			"name":  "there-was-no-place-like-it",
+			"host":  "in-the-whole-world",
+			"port":  "33914",
+			"users": []string{user.Id},
+		})
+		require.NoError(t, err)
+
+		assert.True(t, sm.HasSystem(record.Id), "System should exist in the store after creation")
+
+		time.Sleep(8 * time.Second)
+		synctest.Wait()
+		assert.Equal(t, "pending", sm.GetSystemStatusFromStore(record.Id), "System status should be 'pending'")
+
+		sm.SetSystemStatusInDB(record.Id, "up")
+		time.Sleep(time.Second)
+		synctest.Wait()
+		assert.Equal(t, "up", sm.GetSystemStatusFromStore(record.Id), "System status should be 'up'")
+
+		// make sure the system switches to down after 11 seconds
+		sm.RemoveSystem(record.Id)
+		sm.AddRecord(record, nil)
+		assert.Equal(t, "pending", sm.GetSystemStatusFromStore(record.Id), "System status should be 'pending'")
+		time.Sleep(12 * time.Second)
+		synctest.Wait()
+		assert.Equal(t, "down", sm.GetSystemStatusFromStore(record.Id), "System status should be 'down'")
+
+		// sm.SetSystemStatusInDB(record.Id, "paused")
+		// time.Sleep(time.Second)
+		// synctest.Wait()
+		// assert.Equal(t, "paused", sm.GetSystemStatusFromStore(record.Id), "System status should be 'paused'")
+
+		// delete the record
+		err = hub.Delete(record)
+		require.NoError(t, err)
+		assert.False(t, sm.HasSystem(record.Id), "System should not exist in the store after deletion")
+	})
+
+	testOld(t, hub)
+
+	synctest.Test(t, func(t *testing.T) {
+		time.Sleep(time.Second)
+		synctest.Wait()
+
+		for _, systemId := range sm.GetAllSystemIDs() {
+			err = sm.RemoveSystem(systemId)
+			require.NoError(t, err)
+			assert.False(t, sm.HasSystem(systemId), "System should not exist in the store after deletion")
+		}
+
+		assert.Equal(t, 0, sm.GetSystemCount(), "System count should be 0")
+
+		// TODO: test with websocket client
+	})
+}
+
+func testOld(t *testing.T, hub *tests.TestHub) {
+	user, err := tests.CreateUser(hub, "test@testy.com", "testtesttest")
+	require.NoError(t, err)
+
+	sm := hub.GetSystemManager()
+	assert.NotNil(t, sm)
+
+	// error expected when creating a user with a duplicate email
+	_, err = tests.CreateUser(hub, "test@test.com", "testtesttest")
+	require.Error(t, err)
+
+	// Test collection existence. todo: move to hub package tests
+	t.Run("CollectionExistence", func(t *testing.T) {
+		// Verify that required collections exist
+		systems, err := hub.FindCachedCollectionByNameOrId("systems")
+		require.NoError(t, err)
+		assert.NotNil(t, systems)
+
+		systemStats, err := hub.FindCachedCollectionByNameOrId("system_stats")
+		require.NoError(t, err)
+		assert.NotNil(t, systemStats)
+
+		containerStats, err := hub.FindCachedCollectionByNameOrId("container_stats")
+		require.NoError(t, err)
+		assert.NotNil(t, containerStats)
+	})
+
+	t.Run("RemoveSystem", func(t *testing.T) {
+		// Get the count before adding the system
+		countBefore := sm.GetSystemCount()
+
+		// Create a test system record
+		record, err := tests.CreateRecord(hub, "systems", map[string]any{
+			"name":  "i-even-got-lost-at-coney-island",
+			"host":  "but-they-found-me",
+			"port":  "33914",
+			"users": []string{user.Id},
+		})
+		require.NoError(t, err)
+
+		// Verify the system count increased
+		countAfterAdd := sm.GetSystemCount()
+		assert.Equal(t, countBefore+1, countAfterAdd, "System count should increase after adding a system via event hook")
+
+		// Verify the system exists
+		assert.True(t, sm.HasSystem(record.Id), "System should exist in the store")
+
+		// Remove the system
+		err = sm.RemoveSystem(record.Id)
+		assert.NoError(t, err)
+
+		// Check that the system count decreased
+		countAfterRemove := sm.GetSystemCount()
+		assert.Equal(t, countAfterAdd-1, countAfterRemove, "System count should decrease after removing a system")
+
+		// Verify the system no longer exists
+		assert.False(t, sm.HasSystem(record.Id), "System should not exist in the store after removal")
+
+		// Verify the system is not in the list of all system IDs
+		ids := sm.GetAllSystemIDs()
+		assert.NotContains(t, ids, record.Id, "System ID should not be in the list of all system IDs after removal")
+
+		// Verify the system status is empty
+		status := sm.GetSystemStatusFromStore(record.Id)
+		assert.Equal(t, "", status, "System status should be empty after removal")
+
+		// Try to remove it again - should return an error since it's already removed
+		err = sm.RemoveSystem(record.Id)
+		assert.Error(t, err)
+	})
+
+	t.Run("NewRecordPending", func(t *testing.T) {
+		// Create a test system
+		record, err := tests.CreateRecord(hub, "systems", map[string]any{
+			"name":  "and-you-know",
+			"host":  "i-feel-very-bad",
+			"port":  "33914",
+			"users": []string{user.Id},
+		})
+		require.NoError(t, err)
+
+		// Add the record to the system manager
+		err = sm.AddRecord(record, nil)
+		require.NoError(t, err)
+
+		// Test filtering records by status - should be "pending" now
+		filter := "status = 'pending'"
+		pendingSystems, err := hub.FindRecordsByFilter("systems", filter, "-created", 0, 0, nil)
+		require.NoError(t, err)
+		assert.GreaterOrEqual(t, len(pendingSystems), 1)
+	})
+
+	t.Run("SystemStatusUpdate", func(t *testing.T) {
+		// Create a test system record
+		record, err := tests.CreateRecord(hub, "systems", map[string]any{
+			"name":  "we-used-to-sleep-on-the-beach",
+			"host":  "sleep-overnight-here",
+			"port":  "33914",
+			"users": []string{user.Id},
+		})
+		require.NoError(t, err)
+
+		// Add the record to the system manager
+		err = sm.AddRecord(record, nil)
+		require.NoError(t, err)
+
+		// Test status changes
+		initialStatus := sm.GetSystemStatusFromStore(record.Id)
+
+		// Set a new status
+		sm.SetSystemStatusInDB(record.Id, "up")
+
+		// Verify status was updated
+		newStatus := sm.GetSystemStatusFromStore(record.Id)
+		assert.Equal(t, "up", newStatus, "System status should be updated to 'up'")
+		assert.NotEqual(t, initialStatus, newStatus, "Status should have changed")
+
+		// Verify the database was updated
+		updatedRecord, err := hub.FindRecordById("systems", record.Id)
+		require.NoError(t, err)
+		assert.Equal(t, "up", updatedRecord.Get("status"), "Database status should match")
+	})
+
+	t.Run("HandleSystemData", func(t *testing.T) {
+		// Create a test system record
+		record, err := tests.CreateRecord(hub, "systems", map[string]any{
+			"name":  "things-changed-you-know",
+			"host":  "they-dont-sleep-anymore-on-the-beach",
+			"port":  "33914",
+			"users": []string{user.Id},
+		})
+		require.NoError(t, err)
+
+		// Create test system data
+		testData := &system.CombinedData{
+			Info: system.Info{
+				Hostname:      "data-test.example.com",
+				KernelVersion: "5.15.0-generic",
+				Cores:         4,
+				Threads:       8,
+				CpuModel:      "Test CPU",
+				Uptime:        3600,
+				Cpu:           25.5,
+				MemPct:        40.2,
+				DiskPct:       60.0,
+				Bandwidth:     100.0,
+				AgentVersion:  "1.0.0",
+			},
+			Stats: system.Stats{
+				Cpu:         25.5,
+				Mem:         16384.0,
+				MemUsed:     6553.6,
+				MemPct:      40.0,
+				DiskTotal:   1024000.0,
+				DiskUsed:    614400.0,
+				DiskPct:     60.0,
+				NetworkSent: 1024.0,
+				NetworkRecv: 2048.0,
+			},
+			Containers: []*container.Stats{},
+		}
+
+		// Test handling system data. todo: move to hub/alerts package tests
+		err = hub.HandleSystemAlerts(record, testData)
+		assert.NoError(t, err)
+	})
+
+	t.Run("ErrorHandling", func(t *testing.T) {
+		// Try to add a non-existent record
+		nonExistentId := "non_existent_id"
+		err := sm.RemoveSystem(nonExistentId)
+		assert.Error(t, err)
+
+		// Try to add a system with invalid host
+		system := &systems.System{
+			Host: "",
+		}
+		err = sm.AddSystem(system)
+		assert.Error(t, err)
+	})
+
+	t.Run("ConcurrentOperations", func(t *testing.T) {
+		// Create a test system
+		record, err := tests.CreateRecord(hub, "systems", map[string]any{
+			"name":  "jfkjahkfajs",
+			"host":  "localhost",
+			"port":  "33914",
+			"users": []string{user.Id},
+		})
+		require.NoError(t, err)
+
+		// Run concurrent operations
+		const goroutines = 5
+		var wg sync.WaitGroup
+		wg.Add(goroutines)
+
+		for i := range goroutines {
+			go func(i int) {
+				defer wg.Done()
+
+				// Alternate between different operations
+				switch i % 3 {
+				case 0:
+					status := fmt.Sprintf("status-%d", i)
+					sm.SetSystemStatusInDB(record.Id, status)
+				case 1:
+					_ = sm.GetSystemStatusFromStore(record.Id)
+				case 2:
+					_, _ = sm.GetSystemHostPort(record.Id)
+				}
+			}(i)
+		}
+
+		wg.Wait()
+
+		// Verify system still exists and is in a valid state
+		assert.True(t, sm.HasSystem(record.Id), "System should still exist after concurrent operations")
+		status := sm.GetSystemStatusFromStore(record.Id)
+		assert.NotEmpty(t, status, "System should have a status after concurrent operations")
+	})
+
+	t.Run("ContextCancellation", func(t *testing.T) {
+		// Create a test system record
+		record, err := tests.CreateRecord(hub, "systems", map[string]any{
+			"name":  "lkhsdfsjf",
+			"host":  "localhost",
+			"port":  "33914",
+			"users": []string{user.Id},
+		})
+		require.NoError(t, err)
+
+		// Verify the system exists in the store
+		assert.True(t, sm.HasSystem(record.Id), "System should exist in the store")
+
+		// Store the original context and cancel function
+		originalCtx, originalCancel, err := sm.GetSystemContextFromStore(record.Id)
+		assert.NoError(t, err)
+
+		// Ensure the context is not nil
+		assert.NotNil(t, originalCtx, "System context should not be nil")
+		assert.NotNil(t, originalCancel, "System cancel function should not be nil")
+
+		// Cancel the context
+		originalCancel()
+
+		// Wait a short time for cancellation to propagate
+		time.Sleep(10 * time.Millisecond)
+
+		// Verify the context is done
+		select {
+		case <-originalCtx.Done():
+			// Context was properly cancelled
+		default:
+			t.Fatal("Context was not cancelled")
+		}
+
+		// Verify the system is still in the store (cancellation shouldn't remove it)
+		assert.True(t, sm.HasSystem(record.Id), "System should still exist after context cancellation")
+
+		// Explicitly remove the system
+		err = sm.RemoveSystem(record.Id)
+		assert.NoError(t, err, "RemoveSystem should succeed")
+
+		// Verify the system is removed
+		assert.False(t, sm.HasSystem(record.Id), "System should be removed after RemoveSystem")
+
+		// Try to remove it again - should return an error
+		err = sm.RemoveSystem(record.Id)
+		assert.Error(t, err, "RemoveSystem should fail for non-existent system")
+
+		// Add the system back
+		err = sm.AddRecord(record, nil)
+		require.NoError(t, err, "AddRecord should succeed")
+
+		// Verify the system is back in the store
+		assert.True(t, sm.HasSystem(record.Id), "System should exist after re-adding")
+
+		// Verify a new context was created
+		newCtx, newCancel, err := sm.GetSystemContextFromStore(record.Id)
+		assert.NoError(t, err)
+		assert.NotNil(t, newCtx, "New system context should not be nil")
+		assert.NotNil(t, newCancel, "New system cancel function should not be nil")
+		assert.NotEqual(t, originalCtx, newCtx, "New context should be different from original")
+
+		// Clean up
+		err = sm.RemoveSystem(record.Id)
+		assert.NoError(t, err)
+	})
+}

internal/hub/systems/systems_test_helpers.go

@@ -0,0 +1,110 @@
+//go:build testing
+// +build testing
+
+package systems
+
+import (
+	"context"
+	"fmt"
+
+	entities "github.com/henrygd/beszel/src/entities/system"
+)
+
+// TESTING ONLY: GetSystemCount returns the number of systems in the store
+func (sm *SystemManager) GetSystemCount() int {
+	return sm.systems.Length()
+}
+
+// TESTING ONLY: HasSystem checks if a system with the given ID exists in the store
+func (sm *SystemManager) HasSystem(systemID string) bool {
+	return sm.systems.Has(systemID)
+}
+
+// TESTING ONLY: GetSystemStatusFromStore returns the status of a system with the given ID
+// Returns an empty string if the system doesn't exist
+func (sm *SystemManager) GetSystemStatusFromStore(systemID string) string {
+	sys, ok := sm.systems.GetOk(systemID)
+	if !ok {
+		return ""
+	}
+	return sys.Status
+}
+
+// TESTING ONLY: GetSystemContextFromStore returns the context and cancel function for a system
+func (sm *SystemManager) GetSystemContextFromStore(systemID string) (context.Context, context.CancelFunc, error) {
+	sys, ok := sm.systems.GetOk(systemID)
+	if !ok {
+		return nil, nil, fmt.Errorf("no system")
+	}
+	return sys.ctx, sys.cancel, nil
+}
+
+// TESTING ONLY: GetSystemFromStore returns a store from the system
+func (sm *SystemManager) GetSystemFromStore(systemID string) (*System, error) {
+	sys, ok := sm.systems.GetOk(systemID)
+	if !ok {
+		return nil, fmt.Errorf("no system")
+	}
+	return sys, nil
+}
+
+// TESTING ONLY: GetAllSystemIDs returns a slice of all system IDs in the store
+func (sm *SystemManager) GetAllSystemIDs() []string {
+	data := sm.systems.GetAll()
+	ids := make([]string, 0, len(data))
+	for id := range data {
+		ids = append(ids, id)
+	}
+	return ids
+}
+
+// TESTING ONLY: GetSystemData returns the combined data for a system with the given ID
+// Returns nil if the system doesn't exist
+// This method is intended for testing
+func (sm *SystemManager) GetSystemData(systemID string) *entities.CombinedData {
+	sys, ok := sm.systems.GetOk(systemID)
+	if !ok {
+		return nil
+	}
+	return sys.data
+}
+
+// TESTING ONLY: GetSystemHostPort returns the host and port for a system with the given ID
+// Returns empty strings if the system doesn't exist
+func (sm *SystemManager) GetSystemHostPort(systemID string) (string, string) {
+	sys, ok := sm.systems.GetOk(systemID)
+	if !ok {
+		return "", ""
+	}
+	return sys.Host, sys.Port
+}
+
+// TESTING ONLY: SetSystemStatusInDB sets the status of a system directly and updates the database record
+// This is intended for testing
+// Returns false if the system doesn't exist
+func (sm *SystemManager) SetSystemStatusInDB(systemID string, status string) bool {
+	if !sm.HasSystem(systemID) {
+		return false
+	}
+
+	// Update the database record
+	record, err := sm.hub.FindRecordById("systems", systemID)
+	if err != nil {
+		return false
+	}
+
+	record.Set("status", status)
+	err = sm.hub.Save(record)
+	if err != nil {
+		return false
+	}
+
+	return true
+}
+
+// TESTING ONLY: RemoveAllSystems removes all systems from the store
+func (sm *SystemManager) RemoveAllSystems() {
+	for _, system := range sm.systems.GetAll() {
+		sm.RemoveSystem(system.Id)
+	}
+}

internal/hub/update.go

@@ -0,0 +1,85 @@
+package hub
+
+import (
+	"fmt"
+	"log"
+	"os"
+	"os/exec"
+
+	"github.com/henrygd/beszel/src/ghupdate"
+	"github.com/spf13/cobra"
+)
+
+// Update updates beszel to the latest version
+func Update(cmd *cobra.Command, _ []string) {
+	dataDir := os.TempDir()
+
+	// set dataDir to ./beszel_data if it exists
+	if _, err := os.Stat("./beszel_data"); err == nil {
+		dataDir = "./beszel_data"
+	}
+
+	// Check if china-mirrors flag is set
+	useMirror, _ := cmd.Flags().GetBool("china-mirrors")
+
+	updated, err := ghupdate.Update(ghupdate.Config{
+		ArchiveExecutable: "beszel",
+		DataDir:           dataDir,
+		UseMirror:         useMirror,
+	})
+	if err != nil {
+		log.Fatal(err)
+	}
+	if !updated {
+		return
+	}
+
+	// make sure the file is executable
+	exePath, err := os.Executable()
+	if err == nil {
+		if err := os.Chmod(exePath, 0755); err != nil {
+			fmt.Printf("Warning: failed to set executable permissions: %v\n", err)
+		}
+	}
+
+	// Try to restart the service if it's running
+	restartService()
+}
+
+// restartService attempts to restart the beszel service
+func restartService() {
+	// Check if we're running as a service by looking for systemd
+	if _, err := exec.LookPath("systemctl"); err == nil {
+		// Check if beszel service exists and is active
+		cmd := exec.Command("systemctl", "is-active", "beszel.service")
+		if err := cmd.Run(); err == nil {
+			ghupdate.ColorPrint(ghupdate.ColorYellow, "Restarting beszel service...")
+			restartCmd := exec.Command("systemctl", "restart", "beszel.service")
+			if err := restartCmd.Run(); err != nil {
+				ghupdate.ColorPrintf(ghupdate.ColorYellow, "Warning: Failed to restart service: %v\n", err)
+				ghupdate.ColorPrint(ghupdate.ColorYellow, "Please restart the service manually: sudo systemctl restart beszel")
+			} else {
+				ghupdate.ColorPrint(ghupdate.ColorGreen, "Service restarted successfully")
+			}
+			return
+		}
+	}
+
+	// Check for OpenRC (Alpine Linux)
+	if _, err := exec.LookPath("rc-service"); err == nil {
+		cmd := exec.Command("rc-service", "beszel", "status")
+		if err := cmd.Run(); err == nil {
+			ghupdate.ColorPrint(ghupdate.ColorYellow, "Restarting beszel service...")
+			restartCmd := exec.Command("rc-service", "beszel", "restart")
+			if err := restartCmd.Run(); err != nil {
+				ghupdate.ColorPrintf(ghupdate.ColorYellow, "Warning: Failed to restart service: %v\n", err)
+				ghupdate.ColorPrint(ghupdate.ColorYellow, "Please restart the service manually: sudo rc-service beszel restart")
+			} else {
+				ghupdate.ColorPrint(ghupdate.ColorGreen, "Service restarted successfully")
+			}
+			return
+		}
+	}
+
+	ghupdate.ColorPrint(ghupdate.ColorYellow, "Service restart not attempted. If running as a service, restart manually.")
+}

internal/hub/ws/ws.go

@@ -0,0 +1,182 @@
+package ws
+
+import (
+	"errors"
+	"time"
+	"weak"
+
+	"github.com/henrygd/beszel/src/entities/system"
+
+	"github.com/henrygd/beszel/src/common"
+
+	"github.com/fxamacker/cbor/v2"
+	"github.com/lxzan/gws"
+	"golang.org/x/crypto/ssh"
+)
+
+const (
+	deadline = 70 * time.Second
+)
+
+// Handler implements the WebSocket event handler for agent connections.
+type Handler struct {
+	gws.BuiltinEventHandler
+}
+
+// WsConn represents a WebSocket connection to an agent.
+type WsConn struct {
+	conn         *gws.Conn
+	responseChan chan *gws.Message
+	DownChan     chan struct{}
+}
+
+// FingerprintRecord is fingerprints collection record data in the hub
+type FingerprintRecord struct {
+	Id          string `db:"id"`
+	SystemId    string `db:"system"`
+	Fingerprint string `db:"fingerprint"`
+	Token       string `db:"token"`
+}
+
+var upgrader *gws.Upgrader
+
+// GetUpgrader returns a singleton WebSocket upgrader instance.
+func GetUpgrader() *gws.Upgrader {
+	if upgrader != nil {
+		return upgrader
+	}
+	handler := &Handler{}
+	upgrader = gws.NewUpgrader(handler, &gws.ServerOption{})
+	return upgrader
+}
+
+// NewWsConnection creates a new WebSocket connection wrapper.
+func NewWsConnection(conn *gws.Conn) *WsConn {
+	return &WsConn{
+		conn:         conn,
+		responseChan: make(chan *gws.Message, 1),
+		DownChan:     make(chan struct{}, 1),
+	}
+}
+
+// OnOpen sets a deadline for the WebSocket connection.
+func (h *Handler) OnOpen(conn *gws.Conn) {
+	conn.SetDeadline(time.Now().Add(deadline))
+}
+
+// OnMessage routes incoming WebSocket messages to the response channel.
+func (h *Handler) OnMessage(conn *gws.Conn, message *gws.Message) {
+	conn.SetDeadline(time.Now().Add(deadline))
+	if message.Opcode != gws.OpcodeBinary || message.Data.Len() == 0 {
+		return
+	}
+	wsConn, ok := conn.Session().Load("wsConn")
+	if !ok {
+		_ = conn.WriteClose(1000, nil)
+		return
+	}
+	select {
+	case wsConn.(*WsConn).responseChan <- message:
+	default:
+		// close if the connection is not expecting a response
+		wsConn.(*WsConn).Close(nil)
+	}
+}
+
+// OnClose handles WebSocket connection closures and triggers system down status after delay.
+func (h *Handler) OnClose(conn *gws.Conn, err error) {
+	wsConn, ok := conn.Session().Load("wsConn")
+	if !ok {
+		return
+	}
+	wsConn.(*WsConn).conn = nil
+	// wait 5 seconds to allow reconnection before setting system down
+	// use a weak pointer to avoid keeping references if the system is removed
+	go func(downChan weak.Pointer[chan struct{}]) {
+		time.Sleep(5 * time.Second)
+		downChanValue := downChan.Value()
+		if downChanValue != nil {
+			*downChanValue <- struct{}{}
+		}
+	}(weak.Make(&wsConn.(*WsConn).DownChan))
+}
+
+// Close terminates the WebSocket connection gracefully.
+func (ws *WsConn) Close(msg []byte) {
+	if ws.IsConnected() {
+		ws.conn.WriteClose(1000, msg)
+	}
+}
+
+// Ping sends a ping frame to keep the connection alive.
+func (ws *WsConn) Ping() error {
+	ws.conn.SetDeadline(time.Now().Add(deadline))
+	return ws.conn.WritePing(nil)
+}
+
+// sendMessage encodes data to CBOR and sends it as a binary message to the agent.
+func (ws *WsConn) sendMessage(data common.HubRequest[any]) error {
+	if ws.conn == nil {
+		return gws.ErrConnClosed
+	}
+	bytes, err := cbor.Marshal(data)
+	if err != nil {
+		return err
+	}
+	return ws.conn.WriteMessage(gws.OpcodeBinary, bytes)
+}
+
+// RequestSystemData requests system metrics from the agent and unmarshals the response.
+func (ws *WsConn) RequestSystemData(data *system.CombinedData) error {
+	var message *gws.Message
+
+	ws.sendMessage(common.HubRequest[any]{
+		Action: common.GetData,
+	})
+	select {
+	case <-time.After(10 * time.Second):
+		ws.Close(nil)
+		return gws.ErrConnClosed
+	case message = <-ws.responseChan:
+	}
+	defer message.Close()
+	return cbor.Unmarshal(message.Data.Bytes(), data)
+}
+
+// GetFingerprint authenticates with the agent using SSH signature and returns the agent's fingerprint.
+func (ws *WsConn) GetFingerprint(token string, signer ssh.Signer, needSysInfo bool) (common.FingerprintResponse, error) {
+	var clientFingerprint common.FingerprintResponse
+	challenge := []byte(token)
+
+	signature, err := signer.Sign(nil, challenge)
+	if err != nil {
+		return clientFingerprint, err
+	}
+
+	err = ws.sendMessage(common.HubRequest[any]{
+		Action: common.CheckFingerprint,
+		Data: common.FingerprintRequest{
+			Signature:   signature.Blob,
+			NeedSysInfo: needSysInfo,
+		},
+	})
+	if err != nil {
+		return clientFingerprint, err
+	}
+
+	var message *gws.Message
+	select {
+	case message = <-ws.responseChan:
+	case <-time.After(10 * time.Second):
+		return clientFingerprint, errors.New("request expired")
+	}
+	defer message.Close()
+
+	err = cbor.Unmarshal(message.Data.Bytes(), &clientFingerprint)
+	return clientFingerprint, err
+}
+
+// IsConnected returns true if the WebSocket connection is active.
+func (ws *WsConn) IsConnected() bool {
+	return ws.conn != nil
+}

internal/hub/ws/ws_test.go

@@ -0,0 +1,222 @@
+//go:build testing
+// +build testing
+
+package ws
+
+import (
+	"crypto/ed25519"
+	"testing"
+	"time"
+
+	"github.com/henrygd/beszel/src/common"
+
+	"github.com/fxamacker/cbor/v2"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+	"golang.org/x/crypto/ssh"
+)
+
+// TestGetUpgrader tests the singleton upgrader
+func TestGetUpgrader(t *testing.T) {
+	// Reset the global upgrader to test singleton behavior
+	upgrader = nil
+
+	// First call should create the upgrader
+	upgrader1 := GetUpgrader()
+	assert.NotNil(t, upgrader1, "Upgrader should not be nil")
+
+	// Second call should return the same instance
+	upgrader2 := GetUpgrader()
+	assert.Same(t, upgrader1, upgrader2, "Should return the same upgrader instance")
+
+	// Verify it's properly configured
+	assert.NotNil(t, upgrader1, "Upgrader should be configured")
+}
+
+// TestNewWsConnection tests WebSocket connection creation
+func TestNewWsConnection(t *testing.T) {
+	// We can't easily mock gws.Conn, so we'll pass nil and test the structure
+	wsConn := NewWsConnection(nil)
+
+	assert.NotNil(t, wsConn, "WebSocket connection should not be nil")
+	assert.Nil(t, wsConn.conn, "Connection should be nil as passed")
+	assert.NotNil(t, wsConn.responseChan, "Response channel should be initialized")
+	assert.NotNil(t, wsConn.DownChan, "Down channel should be initialized")
+	assert.Equal(t, 1, cap(wsConn.responseChan), "Response channel should have capacity of 1")
+	assert.Equal(t, 1, cap(wsConn.DownChan), "Down channel should have capacity of 1")
+}
+
+// TestWsConn_IsConnected tests the connection status check
+func TestWsConn_IsConnected(t *testing.T) {
+	// Test with nil connection
+	wsConn := NewWsConnection(nil)
+	assert.False(t, wsConn.IsConnected(), "Should not be connected when conn is nil")
+}
+
+// TestWsConn_Close tests the connection closing with nil connection
+func TestWsConn_Close(t *testing.T) {
+	wsConn := NewWsConnection(nil)
+
+	// Should handle nil connection gracefully
+	assert.NotPanics(t, func() {
+		wsConn.Close([]byte("test message"))
+	}, "Should not panic when closing nil connection")
+}
+
+// TestWsConn_SendMessage_CBOR tests CBOR encoding in sendMessage
+func TestWsConn_SendMessage_CBOR(t *testing.T) {
+	wsConn := NewWsConnection(nil)
+
+	testData := common.HubRequest[any]{
+		Action: common.GetData,
+		Data:   "test data",
+	}
+
+	// This will fail because conn is nil, but we can test the CBOR encoding logic
+	// by checking that the function properly encodes to CBOR before failing
+	err := wsConn.sendMessage(testData)
+	assert.Error(t, err, "Should error with nil connection")
+
+	// Test CBOR encoding separately
+	bytes, err := cbor.Marshal(testData)
+	assert.NoError(t, err, "Should encode to CBOR successfully")
+
+	// Verify we can decode it back
+	var decodedData common.HubRequest[any]
+	err = cbor.Unmarshal(bytes, &decodedData)
+	assert.NoError(t, err, "Should decode from CBOR successfully")
+	assert.Equal(t, testData.Action, decodedData.Action, "Action should match")
+}
+
+// TestWsConn_GetFingerprint_SignatureGeneration tests signature creation logic
+func TestWsConn_GetFingerprint_SignatureGeneration(t *testing.T) {
+	// Generate test key pair
+	_, privKey, err := ed25519.GenerateKey(nil)
+	require.NoError(t, err)
+
+	signer, err := ssh.NewSignerFromKey(privKey)
+	require.NoError(t, err)
+
+	token := "test-token"
+
+	// This will timeout since conn is nil, but we can verify the signature logic
+	// We can't test the full flow, but we can test that the signature is created properly
+	challenge := []byte(token)
+	signature, err := signer.Sign(nil, challenge)
+	assert.NoError(t, err, "Should create signature successfully")
+	assert.NotEmpty(t, signature.Blob, "Signature blob should not be empty")
+	assert.Equal(t, signer.PublicKey().Type(), signature.Format, "Signature format should match key type")
+
+	// Test the fingerprint request structure
+	fpRequest := common.FingerprintRequest{
+		Signature:   signature.Blob,
+		NeedSysInfo: true,
+	}
+
+	// Test CBOR encoding of fingerprint request
+	fpData, err := cbor.Marshal(fpRequest)
+	assert.NoError(t, err, "Should encode fingerprint request to CBOR")
+
+	var decodedFpRequest common.FingerprintRequest
+	err = cbor.Unmarshal(fpData, &decodedFpRequest)
+	assert.NoError(t, err, "Should decode fingerprint request from CBOR")
+	assert.Equal(t, fpRequest.Signature, decodedFpRequest.Signature, "Signature should match")
+	assert.Equal(t, fpRequest.NeedSysInfo, decodedFpRequest.NeedSysInfo, "NeedSysInfo should match")
+
+	// Test the full hub request structure
+	hubRequest := common.HubRequest[any]{
+		Action: common.CheckFingerprint,
+		Data:   fpRequest,
+	}
+
+	hubData, err := cbor.Marshal(hubRequest)
+	assert.NoError(t, err, "Should encode hub request to CBOR")
+
+	var decodedHubRequest common.HubRequest[cbor.RawMessage]
+	err = cbor.Unmarshal(hubData, &decodedHubRequest)
+	assert.NoError(t, err, "Should decode hub request from CBOR")
+	assert.Equal(t, common.CheckFingerprint, decodedHubRequest.Action, "Action should be CheckFingerprint")
+}
+
+// TestWsConn_RequestSystemData_RequestFormat tests system data request format
+func TestWsConn_RequestSystemData_RequestFormat(t *testing.T) {
+	// Test the request format that would be sent
+	request := common.HubRequest[any]{
+		Action: common.GetData,
+	}
+
+	// Test CBOR encoding
+	data, err := cbor.Marshal(request)
+	assert.NoError(t, err, "Should encode request to CBOR")
+
+	// Test decoding
+	var decodedRequest common.HubRequest[any]
+	err = cbor.Unmarshal(data, &decodedRequest)
+	assert.NoError(t, err, "Should decode request from CBOR")
+	assert.Equal(t, common.GetData, decodedRequest.Action, "Should have GetData action")
+}
+
+// TestFingerprintRecord tests the FingerprintRecord struct
+func TestFingerprintRecord(t *testing.T) {
+	record := FingerprintRecord{
+		Id:          "test-id",
+		SystemId:    "system-123",
+		Fingerprint: "test-fingerprint",
+		Token:       "test-token",
+	}
+
+	assert.Equal(t, "test-id", record.Id)
+	assert.Equal(t, "system-123", record.SystemId)
+	assert.Equal(t, "test-fingerprint", record.Fingerprint)
+	assert.Equal(t, "test-token", record.Token)
+}
+
+// TestDeadlineConstant tests that the deadline constant is reasonable
+func TestDeadlineConstant(t *testing.T) {
+	assert.Equal(t, 70*time.Second, deadline, "Deadline should be 70 seconds")
+}
+
+// TestCommonActions tests that the common actions are properly defined
+func TestCommonActions(t *testing.T) {
+	// Test that the actions we use exist and have expected values
+	assert.Equal(t, common.WebSocketAction(0), common.GetData, "GetData should be action 0")
+	assert.Equal(t, common.WebSocketAction(1), common.CheckFingerprint, "CheckFingerprint should be action 1")
+}
+
+// TestHandler tests that we can create a Handler
+func TestHandler(t *testing.T) {
+	handler := &Handler{}
+	assert.NotNil(t, handler, "Handler should be created successfully")
+
+	// The Handler embeds gws.BuiltinEventHandler, so it should have the embedded type
+	assert.NotNil(t, handler.BuiltinEventHandler, "Should have embedded BuiltinEventHandler")
+}
+
+// TestWsConnChannelBehavior tests channel behavior without WebSocket connections
+func TestWsConnChannelBehavior(t *testing.T) {
+	wsConn := NewWsConnection(nil)
+
+	// Test that channels are properly initialized and can be used
+	select {
+	case wsConn.DownChan <- struct{}{}:
+		// Should be able to write to channel
+	default:
+		t.Error("Should be able to write to DownChan")
+	}
+
+	// Test reading from DownChan
+	select {
+	case <-wsConn.DownChan:
+		// Should be able to read from channel
+	case <-time.After(10 * time.Millisecond):
+		t.Error("Should be able to read from DownChan")
+	}
+
+	// Response channel should be empty initially
+	select {
+	case <-wsConn.responseChan:
+		t.Error("Response channel should be empty initially")
+	default:
+		// Expected - channel should be empty
+	}
+}