fix(agent): let stopped threads start a new task and scope thread state to the latest task

Replace the stored cancel func with a handle so a stale deferred cleanup
can't delete a newer task's cancellation, and clear a thread's cancelled
flag when it submits fresh work instead of refusing to restart. Thread
progress/finished state now derives from the latest user task in the
thread rather than the whole history. Adds coverage for continuing a
stopped thread and for sanitized generation-failure messages.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

server/internal/application/generation_failure_test.go

@@ -64,6 +64,42 @@ func TestMarkGenerationFailedRemovesCanvasPlaceholders(t *testing.T) {
 	}
 }
 
+func TestMarkGenerationFailedDoesNotExposeInternalEndpoint(t *testing.T) {
+	ctx := context.Background()
+	repo := memory.NewProjectRepository()
+	service := NewDesignService(repo, nil, nil, nil, nil)
+	project := design.Project{
+		ID:        "project-safe-error",
+		Title:     "Safe failure",
+		Brief:     "prompt",
+		Status:    design.StatusExploring,
+		UpdatedAt: time.Now(),
+		Nodes: []design.Node{
+			{ID: "placeholder", Type: design.NodeTypeImage, Title: "生成中", Content: "prompt", Status: "generating"},
+		},
+	}
+	if err := repo.Save(ctx, project); err != nil {
+		t.Fatal(err)
+	}
+
+	failure := errors.New(`gpt image 2 generation failed: Post "http://154.21.87.3:8080/v1/images/generations": context deadline exceeded (Client.Timeout exceeded while awaiting headers)`)
+	if err := service.markGenerationFailed(ctx, project.ID, failure); err != nil {
+		t.Fatal(err)
+	}
+
+	saved, err := repo.Get(ctx, project.ID)
+	if err != nil {
+		t.Fatal(err)
+	}
+	content := saved.Messages[len(saved.Messages)-1].Content
+	if strings.Contains(content, "154.21.87.3") || strings.Contains(content, "/v1/images/generations") || strings.Contains(content, "Client.Timeout") {
+		t.Fatalf("expected sanitized user-facing failure, got %q", content)
+	}
+	if !strings.Contains(content, "图片生成耗时较久") {
+		t.Fatalf("expected friendly timeout message, got %q", content)
+	}
+}
+
 func TestMarkGenerationFailedKeepsRealImageNodeActionTarget(t *testing.T) {
 	ctx := context.Background()
 	repo := memory.NewProjectRepository()
@@ -895,7 +931,7 @@ func TestCreateProjectGenerationPartialFailurePreservesSuccessfulImagesAndCloses
 		if message.Name == "generate_media" && message.Status == "success" {
 			foundArtifact = true
 		}
-		if message.Role == "error" && message.Title == "图片生成部分失败" && strings.Contains(message.Content, "image 2") {
+		if message.Role == "error" && message.Title == "图片生成部分失败" && strings.Contains(message.Content, "已保留成功结果") {
 			foundError = true
 		}
 	}
@@ -1154,8 +1190,8 @@ func TestCreateProjectGenerationFailureUsesJobUserContext(t *testing.T) {
 		t.Fatalf("expected failed project, got %s", saved.Status)
 	}
 	last := saved.Messages[len(saved.Messages)-1]
-	if last.Role != "error" || last.Content != "upstream image response timed out" {
-		t.Fatalf("expected original failure message for job user, got %#v", last)
+	if last.Role != "error" || strings.Contains(last.Content, "upstream") || !strings.Contains(last.Content, "画布内容已保留") {
+		t.Fatalf("expected safe failure message for job user, got %#v", last)
 	}
 }
 
@@ -1486,9 +1522,10 @@ func TestStopAgentTaskClearsAgentPlaceholdersWithoutStoppingImageGenerator(t *te
 	}
 }
 
-func TestAgentChatDoesNotRestartAlreadyCancelledThread(t *testing.T) {
+func TestAgentChatContinuesStoppedThreadWithNewTask(t *testing.T) {
 	ctx := context.Background()
 	repo := memory.NewProjectRepository()
+	queue := &fakeJobQueue{}
 	service := NewDesignService(repo, nil, nil, nil, nil, fakeCreativeChatAgent{
 		decision: design.AgentDecision{Intent: "image"},
 		taskPlan: design.AgentTaskPlan{
@@ -1498,6 +1535,7 @@ func TestAgentChatDoesNotRestartAlreadyCancelledThread(t *testing.T) {
 			ImageTasks:         []design.AgentImageTask{{Title: "视觉方案", Brief: "视觉方案"}},
 		},
 	})
+	service.SetJobQueue(queue)
 	project := design.Project{
 		ID:           "project-cancelled-thread",
 		Title:        "Cancelled",
@@ -1511,6 +1549,7 @@ func TestAgentChatDoesNotRestartAlreadyCancelledThread(t *testing.T) {
 	if err := repo.Save(ctx, project); err != nil {
 		t.Fatal(err)
 	}
+	service.cancelAgentTask(project.ID, "thread-cancelled")
 
 	thread, err := service.AgentChat(ctx, project.ID, design.AgentChatRequest{
 		ThreadID: "thread-cancelled",
@@ -1526,14 +1565,20 @@ func TestAgentChatDoesNotRestartAlreadyCancelledThread(t *testing.T) {
 	if err != nil {
 		t.Fatal(err)
 	}
-	if thread.Status != "cancelled" {
-		t.Fatalf("expected cancelled thread response, got %s", thread.Status)
+	if thread.Status != "running" {
+		t.Fatalf("expected stopped thread to accept a new running task, got %s", thread.Status)
 	}
 	saved, err := repo.Get(ctx, project.ID)
 	if err != nil {
 		t.Fatal(err)
 	}
-	if len(saved.Nodes) != 0 || len(saved.Messages) != 1 {
-		t.Fatalf("expected cancelled thread not to create new work, got nodes=%#v messages=%#v", saved.Nodes, saved.Messages)
+	if len(saved.Nodes) != 1 || saved.Nodes[0].Status != "generating" {
+		t.Fatalf("expected new image task placeholder, got %#v", saved.Nodes)
+	}
+	if len(saved.Messages) <= 1 || saved.Messages[len(saved.Messages)-1].Status != "success" {
+		t.Fatalf("expected new thread messages after cancelled task, got %#v", saved.Messages)
+	}
+	if len(queue.jobs) != 1 || queue.jobs[0].Kind != design.JobFollowupGeneration {
+		t.Fatalf("expected new followup generation job, got %#v", queue.jobs)
 	}
 }