feat：语音对话优化

backend/internal/ai/stt/mimo.go

@@ -158,7 +158,8 @@ func (m *MiMoService) Recognize(ctx context.Context, audio []byte, opts Options)
 	}
 
 	if len(mResp.Choices) == 0 {
-		return "", fmt.Errorf("stt: mimo returned empty choices")
+		// MiMo 返回空结果，视为无法识别（非错误），返回空文本
+		return "", nil
 	}
 
 	text := strings.TrimSpace(mResp.Choices[0].Message.Content)

backend/internal/ai/stt/mimo_test.go

@@ -103,9 +103,12 @@ func TestMiMoService_Recognize_EmptyChoices(t *testing.T) {
 	defer srv.Close()
 
 	wav := makeValidWAV([]byte{0x00, 0x00})
-	_, err := s.Recognize(context.Background(), wav, Options{})
-	if err == nil {
-		t.Fatal("expected error for empty choices")
+	text, err := s.Recognize(context.Background(), wav, Options{})
+	if err != nil {
+		t.Fatalf("unexpected error for empty choices: %v", err)
+	}
+	if text != "" {
+		t.Errorf("expected empty string for empty choices, got %q", text)
 	}
 }