Refresh agent topology state after task restart

pkg/agent/loop.go

@@ -9,6 +9,7 @@ package agent
 import (
 	"context"
 	"encoding/json"
+	"errors"
 	"fmt"
 	"hash/fnv"
 	"math"
@@ -544,6 +545,11 @@ func (al *AgentLoop) processInbound(ctx context.Context, msg bus.InboundMessage)
 
 	response, err := al.processPlannedMessage(ctx, msg)
 	if err != nil {
+		if errors.Is(err, context.Canceled) {
+			al.audit.Record(al.getTrigger(msg), msg.Channel, msg.SessionKey, true, err)
+			al.appendTaskAudit(taskID, msg, started, err, true)
+			return
+		}
 		response = fmt.Sprintf("Error processing message: %v", err)
 	}
 

pkg/agent/session_planner.go

@@ -4,7 +4,9 @@ import (
 	"bufio"
 	"context"
 	"encoding/json"
+	"errors"
 	"fmt"
+	"math"
 	"os"
 	"path/filepath"
 	"regexp"
@@ -26,6 +28,7 @@ type plannedTaskResult struct {
 	Index   int
 	Task    plannedTask
 	Output  string
+	Err     error
 	ErrText string
 }
 
@@ -123,6 +126,11 @@ func splitPlannedSegments(content string) []string {
 func (al *AgentLoop) runPlannedTasks(ctx context.Context, msg bus.InboundMessage, tasks []plannedTask) (string, error) {
 	results := make([]plannedTaskResult, len(tasks))
 	var wg sync.WaitGroup
+	var progressMu sync.Mutex
+	completed := 0
+	failed := 0
+	milestones := plannedProgressMilestones(len(tasks))
+	notified := make(map[int]struct{}, len(milestones))
 	for i, task := range tasks {
 		wg.Add(1)
 		go func(index int, t plannedTask) {
@@ -137,15 +145,33 @@ func (al *AgentLoop) runPlannedTasks(ctx context.Context, msg bus.InboundMessage
 			subMsg.Metadata["planned_task_index"] = fmt.Sprintf("%d", t.Index)
 			subMsg.Metadata["planned_task_total"] = fmt.Sprintf("%d", len(tasks))
 			out, err := al.processMessage(ctx, subMsg)
-			res := plannedTaskResult{Index: index, Task: t, Output: strings.TrimSpace(out)}
+			res := plannedTaskResult{Index: index, Task: t, Output: strings.TrimSpace(out), Err: err}
 			if err != nil {
 				res.ErrText = err.Error()
 			}
 			results[index] = res
-			al.publishPlannedTaskProgress(msg, len(tasks), res)
+			progressMu.Lock()
+			completed++
+			if res.ErrText != "" && !isPlannedTaskCancellation(ctx, res) {
+				failed++
+			}
+			snapshotCompleted := completed
+			snapshotFailed := failed
+			shouldNotify := shouldPublishPlannedTaskProgress(ctx, len(tasks), snapshotCompleted, res, milestones, notified)
+			if shouldNotify && res.ErrText == "" {
+				notified[snapshotCompleted] = struct{}{}
+			}
+			progressMu.Unlock()
+
+			if shouldNotify {
+				al.publishPlannedTaskProgress(msg, len(tasks), snapshotCompleted, snapshotFailed, res)
+			}
 		}(i, task)
 	}
 	wg.Wait()
+	if err := ctx.Err(); err != nil {
+		return "", err
+	}
 	var b strings.Builder
 	b.WriteString(fmt.Sprintf("已自动拆解为 %d 个任务并执行：\n\n", len(results)))
 	for _, r := range results {
@@ -163,7 +189,63 @@ func (al *AgentLoop) runPlannedTasks(ctx context.Context, msg bus.InboundMessage
 	return strings.TrimSpace(b.String()), nil
 }
 
-func (al *AgentLoop) publishPlannedTaskProgress(msg bus.InboundMessage, total int, res plannedTaskResult) {
+func plannedProgressMilestones(total int) []int {
+	if total <= 3 {
+		return nil
+	}
+	points := []float64{0.33, 0.66}
+	out := make([]int, 0, len(points))
+	seen := map[int]struct{}{}
+	for _, p := range points {
+		step := int(math.Round(float64(total) * p))
+		if step <= 0 || step >= total {
+			continue
+		}
+		if _, ok := seen[step]; ok {
+			continue
+		}
+		seen[step] = struct{}{}
+		out = append(out, step)
+	}
+	return out
+}
+
+func shouldPublishPlannedTaskProgress(ctx context.Context, total, completed int, res plannedTaskResult, milestones []int, notified map[int]struct{}) bool {
+	if total <= 1 {
+		return false
+	}
+	if isPlannedTaskCancellation(ctx, res) {
+		return false
+	}
+	if strings.TrimSpace(res.ErrText) != "" {
+		return true
+	}
+	if completed >= total {
+		return false
+	}
+	for _, step := range milestones {
+		if completed != step {
+			continue
+		}
+		if _, ok := notified[step]; ok {
+			return false
+		}
+		return true
+	}
+	return false
+}
+
+func isPlannedTaskCancellation(ctx context.Context, res plannedTaskResult) bool {
+	if res.Err != nil && errors.Is(res.Err, context.Canceled) {
+		return true
+	}
+	if strings.EqualFold(strings.TrimSpace(res.ErrText), context.Canceled.Error()) {
+		return true
+	}
+	return ctx != nil && errors.Is(ctx.Err(), context.Canceled)
+}
+
+func (al *AgentLoop) publishPlannedTaskProgress(msg bus.InboundMessage, total, completed, failed int, res plannedTaskResult) {
 	if al == nil || al.bus == nil || total <= 1 {
 		return
 	}
@@ -184,7 +266,7 @@ func (al *AgentLoop) publishPlannedTaskProgress(msg bus.InboundMessage, total in
 		body = "(无输出)"
 	}
 	body = summarizePlannedTaskProgressBody(body, 6, 320)
-	content := fmt.Sprintf("进度 %d/%d：任务%d已%s\n%s", idx, total, idx, status, body)
+	content := fmt.Sprintf("阶段进度 %d/%d（失败 %d）\n最近任务：%d 已%s\n%s", completed, total, failed, idx, status, body)
 	al.bus.PublishOutbound(bus.OutboundMessage{
 		Channel: msg.Channel,
 		ChatID:  msg.ChatID,

pkg/agent/session_planner_progress_test.go

@@ -0,0 +1,63 @@
+package agent
+
+import (
+	"context"
+	"errors"
+	"testing"
+)
+
+func TestPlannedProgressMilestones(t *testing.T) {
+	t.Parallel()
+
+	got := plannedProgressMilestones(12)
+	if len(got) != 2 || got[0] != 4 || got[1] != 8 {
+		t.Fatalf("unexpected milestones: %#v", got)
+	}
+}
+
+func TestShouldPublishPlannedTaskProgress(t *testing.T) {
+	t.Parallel()
+
+	milestones := plannedProgressMilestones(12)
+	notified := map[int]struct{}{}
+	if shouldPublishPlannedTaskProgress(context.Background(), 12, 1, plannedTaskResult{}, milestones, notified) {
+		t.Fatalf("did not expect early success notification")
+	}
+	if !shouldPublishPlannedTaskProgress(context.Background(), 12, 4, plannedTaskResult{}, milestones, notified) {
+		t.Fatalf("expected milestone notification")
+	}
+	notified[4] = struct{}{}
+	if shouldPublishPlannedTaskProgress(context.Background(), 12, 4, plannedTaskResult{}, milestones, notified) {
+		t.Fatalf("did not expect duplicate milestone notification")
+	}
+	if !shouldPublishPlannedTaskProgress(context.Background(), 12, 5, plannedTaskResult{ErrText: "boom"}, milestones, notified) {
+		t.Fatalf("expected failure notification")
+	}
+	if shouldPublishPlannedTaskProgress(context.Background(), 3, 3, plannedTaskResult{}, plannedProgressMilestones(3), map[int]struct{}{}) {
+		t.Fatalf("did not expect final success notification")
+	}
+	if shouldPublishPlannedTaskProgress(context.Background(), 12, 5, plannedTaskResult{Err: context.Canceled, ErrText: context.Canceled.Error()}, milestones, notified) {
+		t.Fatalf("did not expect cancellation notification")
+	}
+	ctx, cancel := context.WithCancel(context.Background())
+	cancel()
+	if shouldPublishPlannedTaskProgress(ctx, 12, 5, plannedTaskResult{Err: errors.New("worker exited after parent stop"), ErrText: "worker exited after parent stop"}, milestones, notified) {
+		t.Fatalf("did not expect notification after parent cancellation")
+	}
+}
+
+func TestIsPlannedTaskCancellation(t *testing.T) {
+	t.Parallel()
+
+	if !isPlannedTaskCancellation(context.Background(), plannedTaskResult{Err: context.Canceled, ErrText: context.Canceled.Error()}) {
+		t.Fatalf("expected direct context cancellation to be detected")
+	}
+	ctx, cancel := context.WithCancel(context.Background())
+	cancel()
+	if !isPlannedTaskCancellation(ctx, plannedTaskResult{Err: errors.New("worker exited after parent stop"), ErrText: "worker exited after parent stop"}) {
+		t.Fatalf("expected canceled parent context to suppress planned task result")
+	}
+	if isPlannedTaskCancellation(context.Background(), plannedTaskResult{Err: errors.New("boom"), ErrText: "boom"}) {
+		t.Fatalf("did not expect non-cancellation error to be suppressed")
+	}
+}

pkg/tools/subagent.go

@@ -51,6 +51,7 @@ type SubagentTask struct {
 type SubagentManager struct {
 	tasks              map[string]*SubagentTask
 	cancelFuncs        map[string]context.CancelFunc
+	recoverableTaskIDs []string
 	archiveAfterMinute int64
 	mu                 sync.RWMutex
 	provider           providers.LLMProvider
@@ -99,9 +100,13 @@ func NewSubagentManager(provider providers.LLMProvider, workspace string, bus *b
 	if runStore != nil {
 		for _, task := range runStore.List() {
 			mgr.tasks[task.ID] = task
+			if task.Status == "running" {
+				mgr.recoverableTaskIDs = append(mgr.recoverableTaskIDs, task.ID)
+			}
 		}
 		mgr.nextID = runStore.NextIDSeed()
 	}
+	go mgr.resumeRecoveredTasks()
 	return mgr
 }
 
@@ -534,6 +539,7 @@ func (sm *SubagentManager) SetRunFunc(f SubagentRunFunc) {
 	sm.mu.Lock()
 	defer sm.mu.Unlock()
 	sm.runFunc = f
+	go sm.resumeRecoveredTasks()
 }
 
 func (sm *SubagentManager) ProfileStore() *SubagentProfileStore {
@@ -542,6 +548,38 @@ func (sm *SubagentManager) ProfileStore() *SubagentProfileStore {
 	return sm.profileStore
 }
 
+func (sm *SubagentManager) resumeRecoveredTasks() {
+	if sm == nil {
+		return
+	}
+	sm.mu.Lock()
+	if sm.runFunc == nil && sm.provider == nil {
+		sm.mu.Unlock()
+		return
+	}
+	taskIDs := append([]string(nil), sm.recoverableTaskIDs...)
+	sm.recoverableTaskIDs = nil
+	toResume := make([]*SubagentTask, 0, len(taskIDs))
+	for _, taskID := range taskIDs {
+		task, ok := sm.tasks[taskID]
+		if !ok || task == nil || task.Status != "running" {
+			continue
+		}
+		task.Updated = time.Now().UnixMilli()
+		sm.persistTaskLocked(task, "recovered", "auto-resumed after restart")
+		toResume = append(toResume, task)
+	}
+	sm.mu.Unlock()
+
+	for _, task := range toResume {
+		taskCtx, cancel := context.WithCancel(context.Background())
+		sm.mu.Lock()
+		sm.cancelFuncs[task.ID] = cancel
+		sm.mu.Unlock()
+		go sm.runTask(taskCtx, task)
+	}
+}
+
 func (sm *SubagentManager) NextTaskSequence() int {
 	sm.mu.RLock()
 	defer sm.mu.RUnlock()

pkg/tools/subagent_runtime_control_test.go

@@ -205,6 +205,55 @@ func TestSubagentManagerRestoresPersistedRuns(t *testing.T) {
 	time.Sleep(100 * time.Millisecond)
 }
 
+func TestSubagentManagerAutoRecoversRunningTaskAfterRestart(t *testing.T) {
+	workspace := t.TempDir()
+	block := make(chan struct{})
+	manager := NewSubagentManager(nil, workspace, nil)
+	manager.SetRunFunc(func(ctx context.Context, task *SubagentTask) (string, error) {
+		<-block
+		return "should-not-complete-here", nil
+	})
+
+	_, err := manager.Spawn(context.Background(), SubagentSpawnOptions{
+		Task:          "recover me",
+		AgentID:       "coder",
+		OriginChannel: "cli",
+		OriginChatID:  "direct",
+	})
+	if err != nil {
+		t.Fatalf("spawn failed: %v", err)
+	}
+	time.Sleep(80 * time.Millisecond)
+
+	recovered := make(chan string, 1)
+	reloaded := NewSubagentManager(nil, workspace, nil)
+	reloaded.SetRunFunc(func(ctx context.Context, task *SubagentTask) (string, error) {
+		recovered <- task.ID
+		return "recovered-ok", nil
+	})
+
+	select {
+	case taskID := <-recovered:
+		if taskID != "subagent-1" {
+			t.Fatalf("expected recovered task id subagent-1, got %s", taskID)
+		}
+	case <-time.After(2 * time.Second):
+		t.Fatalf("expected running task to auto-recover after restart")
+	}
+
+	got, ok := reloaded.GetTask("subagent-1")
+	if !ok {
+		t.Fatalf("expected recovered task to exist")
+	}
+	if got.Status != "completed" || got.Result != "recovered-ok" {
+		t.Fatalf("unexpected recovered task: %+v", got)
+	}
+
+	close(block)
+	_ = waitSubagentDone(t, manager, 4*time.Second)
+	time.Sleep(100 * time.Millisecond)
+}
+
 func TestSubagentManagerPersistsEvents(t *testing.T) {
 	workspace := t.TempDir()
 	manager := NewSubagentManager(nil, workspace, nil)

webui/src/pages/Subagents.tsx

@@ -115,6 +115,8 @@ type AgentTaskStats = {
   running: number;
   failed: number;
   waiting: number;
+  latestStatus: string;
+  latestUpdated: number;
   active: Array<{ id: string; status: string; title: string }>;
 };
 
@@ -199,13 +201,18 @@ function buildTaskStats(tasks: SubagentTask[]): Record<string, AgentTaskStats> {
     const agentID = normalizeTitle(task.agent_id, '');
     if (!agentID) return acc;
     if (!acc[agentID]) {
-      acc[agentID] = { total: 0, running: 0, failed: 0, waiting: 0, active: [] };
+      acc[agentID] = { total: 0, running: 0, failed: 0, waiting: 0, latestStatus: '', latestUpdated: 0, active: [] };
     }
     const item = acc[agentID];
     item.total += 1;
     if (task.status === 'running') item.running += 1;
-    if (task.status === 'failed') item.failed += 1;
     if (task.waiting_for_reply) item.waiting += 1;
+    const updatedAt = Math.max(task.updated || 0, task.created || 0);
+    if (updatedAt >= item.latestUpdated) {
+      item.latestUpdated = updatedAt;
+      item.latestStatus = normalizeTitle(task.status, '');
+      item.failed = task.status === 'failed' ? 1 : 0;
+    }
     if (task.status === 'running' || task.waiting_for_reply) {
       item.active.push({
         id: task.id,
@@ -400,6 +407,13 @@ const Subagents: React.FC = () => {
     load().catch(() => { });
   }, [q, selectedAgentID]);
 
+  useEffect(() => {
+    const timer = window.setInterval(() => {
+      load().catch(() => { });
+    }, 5000);
+    return () => window.clearInterval(timer);
+  }, [q, selectedAgentID]);
+
   const selected = useMemo(() => items.find((x) => x.id === selectedId) || null, [items, selectedId]);
   const selectedRegistryItem = useMemo(
     () => registryItems.find((x) => x.agent_id === selectedAgentID) || null,
@@ -489,7 +503,7 @@ const Subagents: React.FC = () => {
       failed: 0,
     };
 
-    const localMainStats = taskStats[normalizeTitle(localRoot.agent_id, 'main')] || { total: 0, running: 0, failed: 0, waiting: 0, active: [] };
+    const localMainStats = taskStats[normalizeTitle(localRoot.agent_id, 'main')] || { total: 0, running: 0, failed: 0, waiting: 0, latestStatus: '', latestUpdated: 0, active: [] };
     const localMainTask = recentTaskByAgent[normalizeTitle(localRoot.agent_id, 'main')];
     localBranchStats.running += localMainStats.running;
     localBranchStats.failed += localMainStats.failed;
@@ -512,7 +526,7 @@ const Subagents: React.FC = () => {
         `transport=${normalizeTitle(localRoot.transport, 'local')} type=${normalizeTitle(localRoot.type, 'router')}`,
         localMainStats.active[0] ? `task: ${localMainStats.active[0].title}` : t('noLiveTasks'),
       ],
-      accent: 'bg-amber-400',
+      accent: localMainStats.running > 0 ? 'bg-emerald-500' : localMainStats.latestStatus === 'failed' ? 'bg-red-500' : 'bg-amber-400',
       clickable: true,
       scale,
       onClick: () => {
@@ -525,7 +539,7 @@ const Subagents: React.FC = () => {
     localChildren.forEach((child, idx) => {
       const childX = localOriginX + idx * (cardWidth + clusterGap);
       const childY = childStartY;
-      const stats = taskStats[normalizeTitle(child.agent_id, '')] || { total: 0, running: 0, failed: 0, waiting: 0, active: [] };
+      const stats = taskStats[normalizeTitle(child.agent_id, '')] || { total: 0, running: 0, failed: 0, waiting: 0, latestStatus: '', latestUpdated: 0, active: [] };
       const task = recentTaskByAgent[normalizeTitle(child.agent_id, '')];
       localBranchStats.running += stats.running;
       localBranchStats.failed += stats.failed;
@@ -547,7 +561,7 @@ const Subagents: React.FC = () => {
           `transport=${normalizeTitle(child.transport, 'local')} type=${normalizeTitle(child.type, 'worker')}`,
           stats.active[0] ? `task: ${stats.active[0].title}` : task ? `last: ${summarizeTask(task.task, task.label)}` : t('noLiveTasks'),
         ],
-        accent: stats.running > 0 ? 'bg-emerald-500' : stats.failed > 0 ? 'bg-red-500' : 'bg-sky-400',
+        accent: stats.running > 0 ? 'bg-emerald-500' : stats.latestStatus === 'failed' ? 'bg-red-500' : 'bg-sky-400',
         clickable: true,
         scale,
         onClick: () => {