added temperature pause

backend/performance.go

@@ -13,11 +13,13 @@ import (
 	"runtime"
 	"strconv"
 	"strings"
+	"sync"
 	"sync/atomic"
 	"time"
 
 	gocpu "github.com/shirou/gopsutil/v3/cpu"
 	godisk "github.com/shirou/gopsutil/v3/disk"
+	gohost "github.com/shirou/gopsutil/v3/host"
 )
 
 // Server-Startzeit für Perf/Uptime
@@ -25,6 +27,13 @@ var serverStartedAt = time.Now()
 
 var lastCPUUsageBits uint64 // atomic float64 bits
 
+var cpuTemperatureCache = struct {
+	sync.Mutex
+	value     float64
+	available bool
+	sampledAt time.Time
+}{}
+
 func setLastCPUUsage(v float64) {
 	atomic.StoreUint64(&lastCPUUsageBits, math.Float64bits(v))
 }
@@ -33,6 +42,71 @@ func getLastCPUUsage() float64 {
 	return math.Float64frombits(atomic.LoadUint64(&lastCPUUsageBits))
 }
 
+func getMaxCPUTemperatureC() (float64, bool) {
+	now := time.Now()
+
+	cpuTemperatureCache.Lock()
+	if now.Sub(cpuTemperatureCache.sampledAt) < 15*time.Second {
+		value := cpuTemperatureCache.value
+		available := cpuTemperatureCache.available
+		cpuTemperatureCache.Unlock()
+		return value, available
+	}
+	cpuTemperatureCache.Unlock()
+
+	ctx, cancel := context.WithTimeout(context.Background(), 2*time.Second)
+	defer cancel()
+
+	temps, err := gohost.SensorsTemperaturesWithContext(ctx)
+	if err != nil || len(temps) == 0 {
+		cpuTemperatureCache.Lock()
+		cpuTemperatureCache.value = 0
+		cpuTemperatureCache.available = false
+		cpuTemperatureCache.sampledAt = now
+		cpuTemperatureCache.Unlock()
+		return 0, false
+	}
+
+	maxCPU := 0.0
+	maxAny := 0.0
+	for _, temp := range temps {
+		v := temp.Temperature
+		if math.IsNaN(v) || math.IsInf(v, 0) || v <= 0 || v > 150 {
+			continue
+		}
+
+		if v > maxAny {
+			maxAny = v
+		}
+
+		key := strings.ToLower(strings.TrimSpace(temp.SensorKey))
+		if strings.Contains(key, "cpu") ||
+			strings.Contains(key, "core") ||
+			strings.Contains(key, "package") ||
+			strings.Contains(key, "tdie") ||
+			strings.Contains(key, "tctl") ||
+			strings.Contains(key, "k10temp") {
+			if v > maxCPU {
+				maxCPU = v
+			}
+		}
+	}
+
+	value := maxCPU
+	if value <= 0 {
+		value = maxAny
+	}
+	available := value > 0
+
+	cpuTemperatureCache.Lock()
+	cpuTemperatureCache.value = value
+	cpuTemperatureCache.available = available
+	cpuTemperatureCache.sampledAt = now
+	cpuTemperatureCache.Unlock()
+
+	return value, available
+}
+
 func startCPUUsageSampler(ctx context.Context) {
 	_, _ = gocpu.Percent(200*time.Millisecond, false)
 
@@ -194,6 +268,17 @@ func buildPerfSnapshot() map[string]any {
 			}
 			return v
 		}(),
+		"cpuTemperatureC": func() float64 {
+			v, ok := getMaxCPUTemperatureC()
+			if !ok {
+				return 0
+			}
+			return math.Round(v*10) / 10
+		}(),
+		"cpuTemperatureAvailable": func() bool {
+			_, ok := getMaxCPUTemperatureC()
+			return ok
+		}(),
 		"diskPath":        diskPath,
 		"diskFreeBytes":   diskFreeBytes,
 		"diskTotalBytes":  diskTotalBytes,

backend/settings.go

@@ -65,6 +65,9 @@ type RecorderSettings struct {
 	TrainingYoloBatchSize         int    `json:"trainingYoloBatchSize"`
 	TrainingLowPriority           bool   `json:"trainingLowPriority"`
 	TrainingVideoMAEEnabled       bool   `json:"trainingVideoMAEEnabled"`
+	TrainingAutoPauseEnabled      bool   `json:"trainingAutoPauseEnabled"`
+	TrainingAutoPauseCPUPercent   int    `json:"trainingAutoPauseCpuPercent"`
+	TrainingAutoPauseTemperatureC int    `json:"trainingAutoPauseTemperatureC"`
 
 	// EncryptedCookies contains base64(nonce+ciphertext) of a JSON cookie map.
 	EncryptedCookies string `json:"encryptedCookies"`
@@ -121,6 +124,9 @@ var (
 		TrainingYoloBatchSize:         0,
 		TrainingLowPriority:           false,
 		TrainingVideoMAEEnabled:       true,
+		TrainingAutoPauseEnabled:      false,
+		TrainingAutoPauseCPUPercent:   95,
+		TrainingAutoPauseTemperatureC: 85,
 
 		EncryptedCookies: "",
 	}
@@ -189,6 +195,20 @@ func normalizeTrainingSettings(s *RecorderSettings) {
 	if s.TrainingYoloBatchSize > 64 {
 		s.TrainingYoloBatchSize = 64
 	}
+
+	if s.TrainingAutoPauseCPUPercent < 1 {
+		s.TrainingAutoPauseCPUPercent = 95
+	}
+	if s.TrainingAutoPauseCPUPercent > 100 {
+		s.TrainingAutoPauseCPUPercent = 100
+	}
+
+	if s.TrainingAutoPauseTemperatureC < 1 {
+		s.TrainingAutoPauseTemperatureC = 85
+	}
+	if s.TrainingAutoPauseTemperatureC > 120 {
+		s.TrainingAutoPauseTemperatureC = 120
+	}
 }
 
 func settingsFilePath() string {
@@ -375,6 +395,9 @@ type RecorderSettingsPublic struct {
 	TrainingYoloBatchSize          int    `json:"trainingYoloBatchSize"`
 	TrainingLowPriority            bool   `json:"trainingLowPriority"`
 	TrainingVideoMAEEnabled        bool   `json:"trainingVideoMAEEnabled"`
+	TrainingAutoPauseEnabled       bool   `json:"trainingAutoPauseEnabled"`
+	TrainingAutoPauseCPUPercent    int    `json:"trainingAutoPauseCpuPercent"`
+	TrainingAutoPauseTemperatureC  int    `json:"trainingAutoPauseTemperatureC"`
 	TrainingCPUCoreCount           int    `json:"trainingCpuCoreCount"`
 	TrainingEffectiveMode          string `json:"trainingEffectiveMode"`
 	TrainingEffectiveCPUThreads    int    `json:"trainingEffectiveCpuThreads"`
@@ -436,6 +459,9 @@ func toPublicSettings(s RecorderSettings) RecorderSettingsPublic {
 		TrainingYoloBatchSize:          s.TrainingYoloBatchSize,
 		TrainingLowPriority:            s.TrainingLowPriority,
 		TrainingVideoMAEEnabled:        s.TrainingVideoMAEEnabled,
+		TrainingAutoPauseEnabled:       s.TrainingAutoPauseEnabled,
+		TrainingAutoPauseCPUPercent:    s.TrainingAutoPauseCPUPercent,
+		TrainingAutoPauseTemperatureC:  s.TrainingAutoPauseTemperatureC,
 		TrainingCPUCoreCount:           trainingOpts.CPUCoreCount,
 		TrainingEffectiveMode:          trainingOpts.PerformanceMode,
 		TrainingEffectiveCPUThreads:    trainingOpts.CPUThreads,
@@ -631,6 +657,9 @@ func recordSettingsHandler(w http.ResponseWriter, r *http.Request) {
 		next.TrainingYoloBatchSize = in.TrainingYoloBatchSize
 		next.TrainingLowPriority = in.TrainingLowPriority
 		next.TrainingVideoMAEEnabled = in.TrainingVideoMAEEnabled
+		next.TrainingAutoPauseEnabled = in.TrainingAutoPauseEnabled
+		next.TrainingAutoPauseCPUPercent = in.TrainingAutoPauseCPUPercent
+		next.TrainingAutoPauseTemperatureC = in.TrainingAutoPauseTemperatureC
 
 		dbChanged :=
 			normalizeDatabaseType(next.DatabaseType) != normalizeDatabaseType(current.DatabaseType) ||

backend/training.go

@@ -24,6 +24,8 @@ import (
 	"strings"
 	"sync"
 	"time"
+
+	goprocess "github.com/shirou/gopsutil/v3/process"
 )
 
 const trainingUncertainCandidateCount = 10
@@ -271,6 +273,10 @@ type TrainingJobStatus struct {
 	FinishedAt   string  `json:"finishedAt,omitempty"`
 	DurationMs   int64   `json:"durationMs,omitempty"`
 	PreviewURL   string  `json:"previewUrl,omitempty"`
+	Paused       bool    `json:"paused,omitempty"`
+	PauseReason  string  `json:"pauseReason,omitempty"`
+	CPUPercent   float64 `json:"cpuPercent,omitempty"`
+	TemperatureC float64 `json:"temperatureC,omitempty"`
 
 	Stage  string `json:"stage,omitempty"`
 	Epoch  int    `json:"epoch,omitempty"`
@@ -1181,6 +1187,8 @@ func trainingRunCommandStreaming(
 		return "", err
 	}
 	applyTrainingLowPriorityAfterStart(cmd, opts.LowPriority)
+	monitorCtx, stopResourcePauseMonitor := context.WithCancel(ctx)
+	resumeResourcePause := startTrainingResourcePauseMonitor(monitorCtx, cmd, opts)
 
 	var outMu sync.Mutex
 	var lines []string
@@ -1231,9 +1239,12 @@ func trainingRunCommandStreaming(
 	select {
 	case err = <-waitCh:
 	case <-ctx.Done():
+		resumeResourcePause()
 		terminateTrainingCommand(cmd)
 		err = <-waitCh
 	}
+	stopResourcePauseMonitor()
+	resumeResourcePause()
 	wg.Wait()
 
 	outMu.Lock()
@@ -1282,6 +1293,9 @@ type trainingRuntimeOptions struct {
 	YoloBatchSize         int
 	LowPriority           bool
 	VideoMAEEnabled       bool
+	AutoPauseEnabled      bool
+	AutoPauseCPUPercent   int
+	AutoPauseTemperatureC int
 }
 
 func normalizeTrainingPerformanceMode(raw string) string {
@@ -1389,6 +1403,9 @@ func trainingRuntimeOptionsFromRecorderSettings(s RecorderSettings) trainingRunt
 		YoloBatchSize:         presetBatch,
 		LowPriority:           presetLowPriority,
 		VideoMAEEnabled:       s.TrainingVideoMAEEnabled,
+		AutoPauseEnabled:      s.TrainingAutoPauseEnabled,
+		AutoPauseCPUPercent:   s.TrainingAutoPauseCPUPercent,
+		AutoPauseTemperatureC: s.TrainingAutoPauseTemperatureC,
 	}
 
 	if mode == trainingPerformanceCustom {
@@ -1448,6 +1465,331 @@ func trainingCommandEnv(opts trainingRuntimeOptions) []string {
 	return env
 }
 
+type trainingResourceSnapshot struct {
+	CPUPercent           float64
+	TemperatureC         float64
+	TemperatureAvailable bool
+}
+
+func trainingRoundMetric(v float64) float64 {
+	if math.IsNaN(v) || math.IsInf(v, 0) || v < 0 {
+		return 0
+	}
+	return math.Round(v*10) / 10
+}
+
+func trainingReadResourceSnapshot() trainingResourceSnapshot {
+	temp, tempOK := getMaxCPUTemperatureC()
+	return trainingResourceSnapshot{
+		CPUPercent:           trainingRoundMetric(getLastCPUUsage()),
+		TemperatureC:         trainingRoundMetric(temp),
+		TemperatureAvailable: tempOK,
+	}
+}
+
+func trainingPauseReasonForResources(opts trainingRuntimeOptions, snap trainingResourceSnapshot) (string, bool) {
+	if !opts.AutoPauseEnabled {
+		return "", false
+	}
+
+	reasons := []string{}
+	if opts.AutoPauseCPUPercent > 0 && snap.CPUPercent >= float64(opts.AutoPauseCPUPercent) {
+		reasons = append(
+			reasons,
+			fmt.Sprintf("CPU %.0f%% >= %d%%", snap.CPUPercent, opts.AutoPauseCPUPercent),
+		)
+	}
+	if opts.AutoPauseTemperatureC > 0 &&
+		snap.TemperatureAvailable &&
+		snap.TemperatureC >= float64(opts.AutoPauseTemperatureC) {
+		reasons = append(
+			reasons,
+			fmt.Sprintf("Temperatur %.0f°C >= %d°C", snap.TemperatureC, opts.AutoPauseTemperatureC),
+		)
+	}
+
+	if len(reasons) == 0 {
+		return "", false
+	}
+	return strings.Join(reasons, ", "), true
+}
+
+func trainingResourcesRecovered(opts trainingRuntimeOptions, snap trainingResourceSnapshot) bool {
+	if !opts.AutoPauseEnabled {
+		return true
+	}
+
+	if opts.AutoPauseCPUPercent > 0 {
+		resumeCPU := float64(opts.AutoPauseCPUPercent - 15)
+		if resumeCPU < 50 {
+			resumeCPU = 50
+		}
+		if snap.CPUPercent > resumeCPU {
+			return false
+		}
+	}
+
+	if opts.AutoPauseTemperatureC > 0 && snap.TemperatureAvailable {
+		resumeTemp := float64(opts.AutoPauseTemperatureC - 5)
+		if resumeTemp < 40 {
+			resumeTemp = 40
+		}
+		if snap.TemperatureC > resumeTemp {
+			return false
+		}
+	}
+
+	return true
+}
+
+func trainingProcessTree(pid int32) []*goprocess.Process {
+	if pid <= 0 {
+		return nil
+	}
+
+	root, err := goprocess.NewProcess(pid)
+	if err != nil || root == nil {
+		return nil
+	}
+
+	seen := map[int32]bool{}
+	var out []*goprocess.Process
+	var walk func(*goprocess.Process)
+	walk = func(p *goprocess.Process) {
+		if p == nil || seen[p.Pid] {
+			return
+		}
+
+		seen[p.Pid] = true
+		out = append(out, p)
+
+		children, err := p.Children()
+		if err != nil {
+			return
+		}
+		for _, child := range children {
+			walk(child)
+		}
+	}
+
+	walk(root)
+	return out
+}
+
+func trainingSuspendProcessTree(cmd *exec.Cmd) error {
+	if cmd == nil || cmd.Process == nil || cmd.Process.Pid <= 0 {
+		return nil
+	}
+
+	procs := trainingProcessTree(int32(cmd.Process.Pid))
+	if len(procs) == 0 {
+		return nil
+	}
+
+	var firstErr error
+	success := 0
+	for i := len(procs) - 1; i >= 0; i-- {
+		if err := procs[i].Suspend(); err != nil {
+			if firstErr == nil {
+				firstErr = err
+			}
+			continue
+		}
+		success++
+	}
+
+	if success > 0 {
+		return nil
+	}
+	return firstErr
+}
+
+func trainingResumeProcessTree(cmd *exec.Cmd) error {
+	if cmd == nil || cmd.Process == nil || cmd.Process.Pid <= 0 {
+		return nil
+	}
+
+	procs := trainingProcessTree(int32(cmd.Process.Pid))
+	if len(procs) == 0 {
+		return nil
+	}
+
+	var firstErr error
+	success := 0
+	for _, proc := range procs {
+		if err := proc.Resume(); err != nil {
+			if firstErr == nil {
+				firstErr = err
+			}
+			continue
+		}
+		success++
+	}
+
+	if success > 0 {
+		return nil
+	}
+	return firstErr
+}
+
+func startTrainingResourcePauseMonitor(ctx context.Context, cmd *exec.Cmd, opts trainingRuntimeOptions) func() {
+	if !opts.AutoPauseEnabled || cmd == nil || cmd.Process == nil {
+		return func() {}
+	}
+
+	const (
+		checkInterval       = 5 * time.Second
+		requiredHighSamples = 2
+		requiredCoolSamples = 2
+		minPauseDuration    = 45 * time.Second
+	)
+
+	var pauseMu sync.Mutex
+	paused := false
+	previousStep := ""
+	highSamples := 0
+	coolSamples := 0
+	pauseStartedAt := time.Time{}
+	loggedSuspendError := false
+
+	resumeNow := func(reason string) {
+		pauseMu.Lock()
+		if !paused {
+			pauseMu.Unlock()
+			return
+		}
+
+		err := trainingResumeProcessTree(cmd)
+		paused = false
+		restoreStep := strings.TrimSpace(previousStep)
+		previousStep = ""
+		pauseMu.Unlock()
+
+		if err != nil {
+			appLogln("⚠️ Training konnte nach Ressourcenpause nicht sauber fortgesetzt werden:", err)
+		}
+
+		if strings.TrimSpace(reason) != "" {
+			appLogln(reason)
+		}
+
+		trainingSetJobStatus(func(s *TrainingJobStatus) {
+			s.Paused = false
+			s.PauseReason = ""
+			s.CPUPercent = 0
+			s.TemperatureC = 0
+			s.Message = ""
+			if restoreStep != "" && s.Running {
+				s.Step = restoreStep
+			}
+		})
+	}
+
+	go func() {
+		ticker := time.NewTicker(checkInterval)
+		defer ticker.Stop()
+
+		for {
+			select {
+			case <-ctx.Done():
+				resumeNow("Training-Ressourcenpause wurde beendet.")
+				return
+			case <-ticker.C:
+			}
+
+			snap := trainingReadResourceSnapshot()
+
+			pauseMu.Lock()
+			isPaused := paused
+			pauseAge := time.Since(pauseStartedAt)
+			pauseMu.Unlock()
+
+			if !isPaused {
+				reason, shouldPause := trainingPauseReasonForResources(opts, snap)
+				if !shouldPause {
+					highSamples = 0
+					continue
+				}
+
+				highSamples++
+				if highSamples < requiredHighSamples {
+					continue
+				}
+
+				pauseMu.Lock()
+				if paused {
+					pauseMu.Unlock()
+					continue
+				}
+
+				currentStatus := trainingGetJobStatus()
+				previousStep = strings.TrimSpace(currentStatus.Step)
+				if previousStep == "" {
+					previousStep = "Training läuft..."
+				}
+
+				if err := trainingSuspendProcessTree(cmd); err != nil {
+					pauseMu.Unlock()
+					highSamples = 0
+					if !loggedSuspendError {
+						loggedSuspendError = true
+						appLogln("⚠️ Training-Ressourcenpause konnte Prozess nicht pausieren:", err)
+					}
+					continue
+				}
+
+				paused = true
+				pauseStartedAt = time.Now()
+				coolSamples = 0
+				pauseMu.Unlock()
+
+				appLogln("⏸ Training pausiert wegen Ressourcenlimit:", reason)
+				trainingSetJobStatus(func(s *TrainingJobStatus) {
+					s.Paused = true
+					s.PauseReason = reason
+					s.CPUPercent = snap.CPUPercent
+					if snap.TemperatureAvailable {
+						s.TemperatureC = snap.TemperatureC
+					} else {
+						s.TemperatureC = 0
+					}
+					s.Step = "Training pausiert: " + reason
+					s.Message = "Training wird automatisch fortgesetzt, sobald CPU/Temperatur wieder im grünen Bereich sind."
+				})
+				continue
+			}
+
+			trainingSetJobStatus(func(s *TrainingJobStatus) {
+				if !s.Running || !s.Paused {
+					return
+				}
+				s.CPUPercent = snap.CPUPercent
+				if snap.TemperatureAvailable {
+					s.TemperatureC = snap.TemperatureC
+				}
+			})
+
+			if pauseAge < minPauseDuration || !trainingResourcesRecovered(opts, snap) {
+				coolSamples = 0
+				continue
+			}
+
+			coolSamples++
+			if coolSamples < requiredCoolSamples {
+				continue
+			}
+
+			highSamples = 0
+			coolSamples = 0
+			resumeNow("▶ Training nach Ressourcenpause fortgesetzt.")
+		}
+	}()
+
+	return func() {
+		resumeNow("")
+	}
+}
+
 var trainingJob = struct {
 	mu     sync.Mutex
 	status TrainingJobStatus
@@ -1526,6 +1868,10 @@ func trainingFinishCancelled(root string) {
 		s.FinishedAt = finishedAt.Format(time.RFC3339)
 		s.DurationMs = durationMs
 		s.PreviewURL = ""
+		s.Paused = false
+		s.PauseReason = ""
+		s.CPUPercent = 0
+		s.TemperatureC = 0
 
 		if cleanupErr != nil {
 			s.Message = "Training wurde abgebrochen, aber temporäre Trainingsausgaben konnten nicht vollständig gelöscht werden."
@@ -3641,6 +3987,10 @@ func trainingRunJob(ctx context.Context, root string, count int, targets trainin
 			s.FinishedAt = finishedAt.Format(time.RFC3339)
 			s.DurationMs = durationMs
 			s.PreviewURL = ""
+			s.Paused = false
+			s.PauseReason = ""
+			s.CPUPercent = 0
+			s.TemperatureC = 0
 		})
 		trainingClearJobCancel()
 		return
@@ -3650,7 +4000,7 @@ func trainingRunJob(ctx context.Context, root string, count int, targets trainin
 	appLogln("ML-Python für Training:", python)
 	runtimeOpts := trainingRuntimeOptionsFromSettings()
 	appLogf(
-		"Training-Laufzeit: mode=%s cpuCores=%d schonmodus=%v threads=%d workers=%d yoloBatch=%d lowPriority=%v videoMAE=%v targets=%s",
+		"Training-Laufzeit: mode=%s cpuCores=%d schonmodus=%v threads=%d workers=%d yoloBatch=%d lowPriority=%v videoMAE=%v autoPause=%v cpuLimit=%d tempLimit=%d targets=%s",
 		runtimeOpts.PerformanceMode,
 		runtimeOpts.CPUCoreCount,
 		runtimeOpts.PowerSaveMode,
@@ -3659,6 +4009,9 @@ func trainingRunJob(ctx context.Context, root string, count int, targets trainin
 		runtimeOpts.YoloBatchSize,
 		runtimeOpts.LowPriority,
 		runtimeOpts.VideoMAEEnabled,
+		runtimeOpts.AutoPauseEnabled,
+		runtimeOpts.AutoPauseCPUPercent,
+		runtimeOpts.AutoPauseTemperatureC,
 		strings.Join(targets.list(), ","),
 	)
 
@@ -4209,6 +4562,10 @@ func trainingRunJob(ctx context.Context, root string, count int, targets trainin
 		s.FinishedAt = finishedAt.Format(time.RFC3339)
 		s.DurationMs = durationMs
 		s.PreviewURL = ""
+		s.Paused = false
+		s.PauseReason = ""
+		s.CPUPercent = 0
+		s.TemperatureC = 0
 	})
 
 	trainingClearJobCancel()
@@ -4783,6 +5140,9 @@ type TrainingHistoryEntry struct {
 	Workers               int     `json:"workers,omitempty"`
 	YoloBatchSize         int     `json:"yoloBatchSize,omitempty"`
 	LowPriority           bool    `json:"lowPriority,omitempty"`
+	AutoPauseEnabled      bool    `json:"autoPauseEnabled,omitempty"`
+	AutoPauseCPUPercent   int     `json:"autoPauseCpuPercent,omitempty"`
+	AutoPauseTemperatureC int     `json:"autoPauseTemperatureC,omitempty"`
 }
 
 type TrainingHistoryResponse struct {
@@ -4831,6 +5191,9 @@ func trainingAppendTargetHistory(root string, target string, status string, dura
 		Workers:               runtimeOpts.Workers,
 		YoloBatchSize:         runtimeOpts.YoloBatchSize,
 		LowPriority:           runtimeOpts.LowPriority,
+		AutoPauseEnabled:      runtimeOpts.AutoPauseEnabled,
+		AutoPauseCPUPercent:   runtimeOpts.AutoPauseCPUPercent,
+		AutoPauseTemperatureC: runtimeOpts.AutoPauseTemperatureC,
 	}
 
 	if info != nil {

frontend/src/App.tsx

@@ -1230,6 +1230,20 @@ export default function App() {
     }
   }, [authed])
 
+  useEffect(() => {
+    try {
+      const appWindow = window as any
+      appWindow.__nsfwappTrainingRunning = trainingTabRunning
+      window.dispatchEvent(
+        new CustomEvent('app:training-running', {
+          detail: { running: trainingTabRunning },
+        })
+      )
+    } catch {
+      // ignore
+    }
+  }, [trainingTabRunning])
+
   const loadPendingAutoStarts = useCallback(async (opts?: { force?: boolean }) => {
     const force = Boolean(opts?.force)
     const now = Date.now()

frontend/src/components/ui/RecordJobActions.tsx

@@ -76,6 +76,7 @@ type Props = {
     onAddToDownloads?: ActionFn
     onSplit?: ActionFn
     onAddToTraining?: ActionFn
+    trainingImportDisabled?: boolean
 
     order?: ActionKey[]
 
@@ -127,6 +128,7 @@ export default function RecordJobActions({
   onAddToDownloads,
   onSplit,
   onAddToTraining,
+  trainingImportDisabled = false,
   order,
   className,
 }: Props) {
@@ -205,6 +207,10 @@ export default function RecordJobActions({
 
   const [addState, setAddState] = useState<'idle' | 'busy' | 'ok'>('idle')
   const addTimerRef = useRef<number | null>(null)
+  const [appTrainingRunning, setAppTrainingRunning] = useState(() => {
+    if (typeof window === 'undefined') return false
+    return Boolean((window as any).__nsfwappTrainingRunning)
+  })
 
   useEffect(() => {
     return () => {
@@ -212,6 +218,19 @@ export default function RecordJobActions({
     }
   }, [])
 
+  useEffect(() => {
+    const onTrainingRunning = (event: Event) => {
+      const detail = (event as CustomEvent<{ running?: boolean }>).detail
+      setAppTrainingRunning(Boolean(detail?.running))
+    }
+
+    window.addEventListener('app:training-running', onTrainingRunning as EventListener)
+
+    return () => {
+      window.removeEventListener('app:training-running', onTrainingRunning as EventListener)
+    }
+  }, [])
+
   const flashAddOk = () => {
     setAddState('ok')
     if (addTimerRef.current) window.clearTimeout(addTimerRef.current)
@@ -249,8 +268,11 @@ export default function RecordJobActions({
     }
   }
 
+  const trainingImportBlocked = trainingImportDisabled || appTrainingRunning
+
   const doTrainingImport = async (): Promise<boolean> => {
     if (busy) return false
+    if (trainingImportBlocked) return false
 
     const output = String(job.output || '').trim()
     if (!output) return false
@@ -364,7 +386,7 @@ export default function RecordJobActions({
       className={btnBase}
       title="Video ins Training übernehmen"
       aria-label="Video ins Training übernehmen"
-      disabled={busy || (!onAddToTraining && !job.output)}
+      disabled={busy || trainingImportBlocked || (!onAddToTraining && !job.output)}
       onClick={async (e) => {
         e.preventDefault()
         e.stopPropagation()
@@ -774,7 +796,7 @@ export default function RecordJobActions({
           key="training"
           type="button"
           className="flex w-full items-center gap-2 rounded px-2 py-1.5 text-sm hover:bg-gray-100/70 dark:hover:bg-white/5 disabled:opacity-50"
-          disabled={busy || (!onAddToTraining && !job.output)}
+          disabled={busy || trainingImportBlocked || (!onAddToTraining && !job.output)}
           onClick={async (e) => {
             e.preventDefault()
             e.stopPropagation()

frontend/src/components/ui/Settings.tsx

@@ -55,6 +55,9 @@ type Settings = {
   trainingYoloBatchSize?: number
   trainingLowPriority?: boolean
   trainingVideoMAEEnabled?: boolean
+  trainingAutoPauseEnabled?: boolean
+  trainingAutoPauseCpuPercent?: number
+  trainingAutoPauseTemperatureC?: number
   trainingCpuCoreCount?: number
   trainingEffectiveMode?: TrainingPerformanceMode | string
   trainingEffectiveCpuThreads?: number
@@ -99,6 +102,10 @@ type TrainingJobStatus = {
   progress?: number
   step?: string
   message?: string
+  paused?: boolean
+  pauseReason?: string
+  cpuPercent?: number
+  temperatureC?: number
 }
 
 type AppLog = {
@@ -342,6 +349,9 @@ const DEFAULTS: Settings = {
   trainingYoloBatchSize: 0,
   trainingLowPriority: false,
   trainingVideoMAEEnabled: true,
+  trainingAutoPauseEnabled: false,
+  trainingAutoPauseCpuPercent: 95,
+  trainingAutoPauseTemperatureC: 85,
 }
 
 const TRAINING_PERFORMANCE_MODES: Array<{
@@ -1024,6 +1034,12 @@ export default function Settings({ onAssetsGenerated }: Props) {
             (data as any).trainingLowPriority ?? DEFAULTS.trainingLowPriority,
           trainingVideoMAEEnabled:
             (data as any).trainingVideoMAEEnabled ?? DEFAULTS.trainingVideoMAEEnabled,
+          trainingAutoPauseEnabled:
+            (data as any).trainingAutoPauseEnabled ?? DEFAULTS.trainingAutoPauseEnabled,
+          trainingAutoPauseCpuPercent:
+            (data as any).trainingAutoPauseCpuPercent ?? DEFAULTS.trainingAutoPauseCpuPercent,
+          trainingAutoPauseTemperatureC:
+            (data as any).trainingAutoPauseTemperatureC ?? DEFAULTS.trainingAutoPauseTemperatureC,
           trainingCpuCoreCount:
             (data as any).trainingCpuCoreCount ?? DEFAULTS.trainingCpuCoreCount,
           trainingEffectiveMode:
@@ -1073,6 +1089,10 @@ export default function Settings({ onAssetsGenerated }: Props) {
               progress: Number(job.progress || 0),
               step: String(job.step || ''),
               message: String(job.message || ''),
+              paused: Boolean(job.paused),
+              pauseReason: String(job.pauseReason || ''),
+              cpuPercent: Number(job.cpuPercent || 0),
+              temperatureC: Number(job.temperatureC || 0),
             }
           : null)
       } catch {
@@ -1611,6 +1631,15 @@ export default function Settings({ onAssetsGenerated }: Props) {
     )
     const trainingLowPriority = !!value.trainingLowPriority
     const trainingVideoMAEEnabled = value.trainingVideoMAEEnabled !== false
+    const trainingAutoPauseEnabled = !!value.trainingAutoPauseEnabled
+    const trainingAutoPauseCpuPercent = Math.max(
+      1,
+      Math.min(100, Math.floor(Number(value.trainingAutoPauseCpuPercent ?? DEFAULTS.trainingAutoPauseCpuPercent ?? 95)))
+    )
+    const trainingAutoPauseTemperatureC = Math.max(
+      1,
+      Math.min(120, Math.floor(Number(value.trainingAutoPauseTemperatureC ?? DEFAULTS.trainingAutoPauseTemperatureC ?? 85)))
+    )
 
     setSaving(true)
     try {
@@ -1656,6 +1685,9 @@ export default function Settings({ onAssetsGenerated }: Props) {
           trainingYoloBatchSize,
           trainingLowPriority,
           trainingVideoMAEEnabled,
+          trainingAutoPauseEnabled,
+          trainingAutoPauseCpuPercent,
+          trainingAutoPauseTemperatureC,
         }),
       })
       if (!res.ok) {
@@ -3089,6 +3121,100 @@ export default function Settings({ onAssetsGenerated }: Props) {
                     description="Deaktivieren spart viel CPU-Zeit. YOLO Detector und Pose können weiterhin trainiert werden."
                   />
                 </div>
+
+                <div className="mt-3 rounded-xl border border-gray-200 bg-white p-3 dark:border-white/10 dark:bg-gray-950/40">
+                  <LabeledSwitch
+                    checked={!!value.trainingAutoPauseEnabled}
+                    onChange={(checked) =>
+                      setValue((v) => ({
+                        ...v,
+                        trainingAutoPauseEnabled: checked,
+                      }))
+                    }
+                    disabled={trainingSettingsLocked}
+                    label="Training automatisch pausieren"
+                    description="Pausiert laufende Python-Trainingsprozesse bei dauerhaft hoher CPU-Last oder hoher CPU-Temperatur und setzt sie nach einer kurzen Abkühlphase fort."
+                  />
+
+                  <div
+                    className={[
+                      'mt-3 grid grid-cols-1 gap-3 sm:grid-cols-2',
+                      value.trainingAutoPauseEnabled && !trainingSettingsLocked ? '' : 'opacity-60',
+                    ].join(' ')}
+                  >
+                    <div>
+                      <label
+                        htmlFor="trainingAutoPauseCpuPercent"
+                        className="text-sm font-medium text-gray-900 dark:text-gray-200"
+                      >
+                        CPU-Grenze
+                      </label>
+                      <div className="mt-1 flex items-center gap-2">
+                        <input
+                          id="trainingAutoPauseCpuPercent"
+                          type="number"
+                          min={1}
+                          max={100}
+                          step={1}
+                          disabled={!value.trainingAutoPauseEnabled || trainingSettingsLocked}
+                          value={value.trainingAutoPauseCpuPercent ?? 95}
+                          onChange={(e) =>
+                            setValue((v) => ({
+                              ...v,
+                              trainingAutoPauseCpuPercent: Number(e.target.value || 95),
+                            }))
+                          }
+                          className="h-9 w-full rounded-md border border-gray-200 bg-white px-3 text-sm text-gray-900 shadow-sm disabled:cursor-not-allowed disabled:opacity-60 dark:border-white/10 dark:bg-gray-900 dark:text-gray-100"
+                        />
+                        <span className="shrink-0 text-xs text-gray-600 dark:text-gray-300">
+                          %
+                        </span>
+                      </div>
+                    </div>
+
+                    <div>
+                      <label
+                        htmlFor="trainingAutoPauseTemperatureC"
+                        className="text-sm font-medium text-gray-900 dark:text-gray-200"
+                      >
+                        Temperatur-Grenze
+                      </label>
+                      <div className="mt-1 flex items-center gap-2">
+                        <input
+                          id="trainingAutoPauseTemperatureC"
+                          type="number"
+                          min={1}
+                          max={120}
+                          step={1}
+                          disabled={!value.trainingAutoPauseEnabled || trainingSettingsLocked}
+                          value={value.trainingAutoPauseTemperatureC ?? 85}
+                          onChange={(e) =>
+                            setValue((v) => ({
+                              ...v,
+                              trainingAutoPauseTemperatureC: Number(e.target.value || 85),
+                            }))
+                          }
+                          className="h-9 w-full rounded-md border border-gray-200 bg-white px-3 text-sm text-gray-900 shadow-sm disabled:cursor-not-allowed disabled:opacity-60 dark:border-white/10 dark:bg-gray-900 dark:text-gray-100"
+                        />
+                        <span className="shrink-0 text-xs text-gray-600 dark:text-gray-300">
+                          °C
+                        </span>
+                      </div>
+                    </div>
+                  </div>
+
+                  {trainingJobStatus?.paused ? (
+                    <div className="mt-3 rounded-lg border border-amber-200 bg-amber-50 px-3 py-2 text-xs font-medium text-amber-900 dark:border-amber-400/30 dark:bg-amber-500/10 dark:text-amber-100">
+                      Training pausiert: {trainingJobStatus.pauseReason || 'Ressourcenlimit erreicht'}
+                      {trainingJobStatus.cpuPercent ? ` · CPU ${Math.round(trainingJobStatus.cpuPercent)}%` : ''}
+                      {trainingJobStatus.temperatureC ? ` · ${Math.round(trainingJobStatus.temperatureC)}°C` : ''}
+                    </div>
+                  ) : (
+                    <div className="mt-2 text-xs text-gray-500 dark:text-gray-400">
+                      Temperatur greift nur, wenn der Server passende Sensorwerte liefert.
+                    </div>
+                  )}
+                </div>
               </div>
             </div>
 

frontend/src/components/ui/TrainingTab.tsx

@@ -14,6 +14,7 @@ import {
   CheckIcon,
   ForwardIcon,
   InboxArrowDownIcon,
+  PauseIcon,
   RectangleGroupIcon,
   UserGroupIcon,
   TrashIcon,
@@ -106,6 +107,10 @@ type TrainingJobStatus = {
   epoch?: number
   epochs?: number
   previewUrl?: string
+  paused?: boolean
+  pauseReason?: string
+  cpuPercent?: number
+  temperatureC?: number
   map50?: number
   map5095?: number
 }
@@ -260,6 +265,9 @@ type TrainingHistoryEntry = {
   workers?: number
   yoloBatchSize?: number
   lowPriority?: boolean
+  autoPauseEnabled?: boolean
+  autoPauseCpuPercent?: number
+  autoPauseTemperatureC?: number
 }
 
 type TrainingStats = {
@@ -327,6 +335,7 @@ type TrainingEstimateRuntime = {
   workers: number
   yoloBatchLabel: string
   lowPriority: boolean
+  autoPause: boolean
   factor: number
   yoloFactor: number
 }
@@ -718,6 +727,7 @@ function trainingEstimateRuntimeFromSettings(settings?: RecorderSettingsState |
   const lowPriority = Boolean(
     settings?.trainingEffectiveLowPriority ?? settings?.trainingLowPriority
   )
+  const autoPause = Boolean(settings?.trainingAutoPauseEnabled)
   const autoThreads = cpuCores > 0
     ? Math.max(1, Math.min(16, Math.round(cpuCores * 0.75)))
     : 4
@@ -769,11 +779,12 @@ function trainingEstimateRuntimeFromSettings(settings?: RecorderSettingsState |
               ? 0.98
               : 1
   const lowPriorityFactor = lowPriority ? 1.12 : 1
+  const autoPauseFactor = autoPause ? 1.12 : 1
   const corePenalty =
     cpuCores > 0 && effectiveThreads > cpuCores
       ? 1 + Math.min(0.3, (effectiveThreads - cpuCores) * 0.04)
       : 1
-  const factor = modeFactor * threadFactor * workerFactor * lowPriorityFactor * corePenalty
+  const factor = modeFactor * threadFactor * workerFactor * lowPriorityFactor * autoPauseFactor * corePenalty
   const yoloFactor = factor * batchFactor
 
   return {
@@ -783,6 +794,7 @@ function trainingEstimateRuntimeFromSettings(settings?: RecorderSettingsState |
     workers,
     yoloBatchLabel: rawBatch > 0 ? String(rawBatch) : 'Auto',
     lowPriority,
+    autoPause,
     factor: Number.isFinite(factor) && factor > 0 ? factor : 1,
     yoloFactor: Number.isFinite(yoloFactor) && yoloFactor > 0 ? yoloFactor : 1,
   }
@@ -790,7 +802,8 @@ function trainingEstimateRuntimeFromSettings(settings?: RecorderSettingsState |
 
 function trainingEstimateRuntimeText(runtime: TrainingEstimateRuntime) {
   const priority = runtime.lowPriority ? ' · niedrige Priorität' : ''
-  return `${runtime.modeLabel} · ${runtime.threadsLabel} Threads · ${runtime.workers} Worker · Batch ${runtime.yoloBatchLabel}${priority}`
+  const autoPause = runtime.autoPause ? ' · Auto-Pause' : ''
+  return `${runtime.modeLabel} · ${runtime.threadsLabel} Threads · ${runtime.workers} Worker · Batch ${runtime.yoloBatchLabel}${priority}${autoPause}`
 }
 
 function trainingEstimateRuntimeFromHistory(entry: TrainingHistoryEntry) {
@@ -802,6 +815,7 @@ function trainingEstimateRuntimeFromHistory(entry: TrainingHistoryEntry) {
     trainingEffectiveWorkers: entry.workers,
     trainingEffectiveYoloBatchSize: entry.yoloBatchSize,
     trainingEffectiveLowPriority: entry.lowPriority,
+    trainingAutoPauseEnabled: entry.autoPauseEnabled,
   } as RecorderSettingsState)
 }
 
@@ -4644,6 +4658,10 @@ export default function TrainingTab(props: {
             epoch: Number(job.epoch ?? 0),
             epochs: Number(job.epochs ?? 0),
             previewUrl: String(job.previewUrl ?? ''),
+            paused: Boolean(job.paused),
+            pauseReason: String(job.pauseReason ?? ''),
+            cpuPercent: Number(job.cpuPercent ?? 0),
+            temperatureC: Number(job.temperatureC ?? 0),
             map50: Number(job.map50 ?? 0),
             map5095: Number(job.map5095 ?? 0),
           }
@@ -5125,6 +5143,10 @@ export default function TrainingTab(props: {
   const importVideoIntoTraining = useCallback(async (raw: any) => {
     const output = String(raw?.output || '').trim()
     if (!output) return false
+    if (trainingRunning) {
+      setMessage('Während das Training läuft, kann kein Video ins Training übernommen werden.')
+      return false
+    }
 
     setLoadingPreviewCandidate('')
 
@@ -5277,6 +5299,7 @@ export default function TrainingTab(props: {
     loadPriorityTrainingSamples,
     loadTrainingStatus,
     setLoadingPreviewCandidate,
+    trainingRunning,
     waitForVideoImportResult,
   ])
 
@@ -7243,13 +7266,25 @@ export default function TrainingTab(props: {
       },
     ]
 
+    const trainingPaused = Boolean(trainingStatus?.training?.paused)
+    const pauseReason = String(trainingStatus?.training?.pauseReason || '').trim()
+    const pauseMetrics = [
+      trainingStatus?.training?.cpuPercent
+        ? `CPU ${Math.round(Number(trainingStatus.training.cpuPercent))}%`
+        : '',
+      trainingStatus?.training?.temperatureC
+        ? `${Math.round(Number(trainingStatus.training.temperatureC))}°C`
+        : '',
+    ].filter(Boolean).join(' · ')
     const progress = clampPercent(
       trainingRunning
         ? shownTrainingProgress
         : readinessProgress * 100
     )
     const statusText = trainingRunning
-      ? shownTrainingStep || 'Training läuft…'
+      ? trainingPaused
+        ? `Pausiert${pauseReason ? `: ${pauseReason}` : ''}${pauseMetrics ? ` · ${pauseMetrics}` : ''}`
+        : shownTrainingStep || 'Training läuft…'
       : !feedbackReady
         ? `${Math.max(0, requiredCount - feedbackCount)} Feedback fehlen noch`
         : canStartTraining
@@ -7280,14 +7315,18 @@ export default function TrainingTab(props: {
               <div
                 className={[
                   'flex h-8 w-8 shrink-0 items-center justify-center rounded-lg ring-1',
-                  trainingRunning
+                  trainingPaused
+                    ? 'bg-amber-50 text-amber-700 ring-amber-200 dark:bg-amber-500/15 dark:text-amber-200 dark:ring-amber-400/30'
+                    : trainingRunning
                     ? 'bg-indigo-50 text-indigo-700 ring-indigo-200 dark:bg-indigo-500/15 dark:text-indigo-200 dark:ring-indigo-400/30'
                     : canStartTraining
                       ? 'bg-emerald-50 text-emerald-700 ring-emerald-200 dark:bg-emerald-500/15 dark:text-emerald-200 dark:ring-emerald-400/30'
                       : 'bg-gray-50 text-gray-600 ring-gray-200 dark:bg-white/5 dark:text-gray-300 dark:ring-white/10',
                 ].join(' ')}
               >
-                {trainingRunning ? (
+                {trainingPaused ? (
+                  <PauseIcon className="h-4 w-4" aria-hidden="true" />
+                ) : trainingRunning ? (
                   <ArrowPathIcon className="h-4 w-4 animate-spin" aria-hidden="true" />
                 ) : (
                   <BoltIcon className="h-4 w-4" aria-hidden="true" />
@@ -7394,7 +7433,9 @@ export default function TrainingTab(props: {
               <div
                 className={[
                   'h-full rounded-full transition-all duration-500',
-                  trainingRunning
+                  trainingPaused
+                    ? 'bg-amber-500'
+                    : trainingRunning
                     ? 'bg-indigo-500'
                     : canStartTraining
                       ? 'bg-emerald-500'

frontend/src/types.ts

@@ -289,6 +289,9 @@ export type RecorderSettingsState = {
   trainingYoloBatchSize?: number
   trainingLowPriority?: boolean
   trainingVideoMAEEnabled?: boolean
+  trainingAutoPauseEnabled?: boolean
+  trainingAutoPauseCpuPercent?: number
+  trainingAutoPauseTemperatureC?: number
   trainingCpuCoreCount?: number
   trainingEffectiveMode?: string
   trainingEffectiveCpuThreads?: number