Let reviews be stopped, and fail stuck reviewers automatically

internal/agents/review.go

@@ -525,6 +525,51 @@ func (m *Manager) cancelReviews(ctx context.Context, task api.Task, project api.
 	}
 }
 
+func (m *Manager) StopReview(ctx context.Context, reviewID, reason string) (api.Review, error) {
+	m.reviewMu.Lock()
+	defer m.reviewMu.Unlock()
+	r, err := m.store.GetReview(ctx, reviewID)
+	if err != nil {
+		return api.Review{}, err
+	}
+	if r.State != api.ReviewStateRunning {
+		return api.Review{}, fmt.Errorf("%w: review %s is %s, not running", ErrConflict, r.Id, r.State)
+	}
+	task, err := m.store.GetTask(ctx, r.TaskId)
+	if err != nil {
+		return api.Review{}, err
+	}
+	project, err := m.store.GetProject(ctx, task.ProjectId)
+	if err != nil {
+		return api.Review{}, err
+	}
+	m.endReviewer(ctx, r.AgentSessionId, project, r.Id)
+	m.failReview(ctx, r.Id, reason)
+	return m.store.GetReview(ctx, r.Id)
+}
+
+func (m *Manager) stopStuckReviewers(ctx context.Context, h Health) {
+	for _, p := range h.Processes {
+		if p.Health != api.HealthStuck {
+			continue
+		}
+		m.mu.Lock()
+		l := m.live[p.SessionId]
+		reviewID := ""
+		if l != nil {
+			reviewID = l.reviewID
+		}
+		m.mu.Unlock()
+		if reviewID == "" {
+			continue
+		}
+		_, err := m.StopReview(ctx, reviewID, fmt.Sprintf("the reviewer went %d minutes without an event", int(StuckAfter/time.Minute)))
+		if err != nil && !errors.Is(err, ErrConflict) {
+			slog.Error("stop stuck reviewer", "review", reviewID, "session", p.SessionId, "err", err)
+		}
+	}
+}
+
 func (m *Manager) recoverReviews(ctx context.Context) error {
 	err := m.store.ClearReviewRequests(ctx)
 	if err != nil {