refactor: move cached screenResults from XTDriver to DriverSession

internal/version/VERSION

@@ -1 +1 @@
-v5.0.0-beta-2505051654
+v5.0.0-beta-2505060010

uixt/driver.go

@@ -85,7 +85,6 @@ type IDriver interface {
 func NewXTDriver(driver IDriver, opts ...option.AIServiceOption) (*XTDriver, error) {
 	driverExt := &XTDriver{
 		IDriver: driver,
-		screenResults: make([]*ScreenResult, 0),
 	}
 
 	services := option.NewAIServiceOptions(opts...)
@@ -114,7 +113,4 @@ type XTDriver struct {
 	IDriver
 	CVService  ai.ICVService  // OCR/CV
 	LLMService ai.ILLMService // LLM
-
-	// cache screenshot results
-	screenResults []*ScreenResult
 }

uixt/driver_ext_screenshot.go

@@ -133,8 +133,9 @@ func (dExt *XTDriver) GetScreenResult(opts ...option.ActionOption) (screenResult
 		}
 	}
 
-	// cache screen result
+	// save screen result to session
-	dExt.screenResults = append(dExt.screenResults, screenResult)
+	session := dExt.GetSession()
+	session.screenResults = append(session.screenResults, screenResult)
 
 	log.Debug().
 		Str("imagePath", imagePath).
@@ -313,7 +314,7 @@ func MarkUIOperation(driver IDriver, actionType ActionMethod, actionCoordinates
 	}
 
 	// create screenshot save path
-	timestamp := builtin.GenNameWithTimestamp("%d")
+	timestamp := builtin.GenNameWithTimestamp("action_%d")
 	var imagePath string
 
 	if actionType == ACTION_TapAbsXY {
@@ -322,18 +323,18 @@ func MarkUIOperation(driver IDriver, actionType ActionMethod, actionCoordinates
 		}
 		imagePath = filepath.Join(
 			config.GetConfig().ScreenShotsPath,
-			fmt.Sprintf("%s_tap_marked.png", timestamp),
+			fmt.Sprintf("%s_tap.png", timestamp),
 		)
 		x, y := actionCoordinates[0], actionCoordinates[1]
 		point := image.Point{X: int(x), Y: int(y)}
 		err = SaveImageWithCircleMarker(compressedBufSource, point, imagePath)
-	} else if actionType == ACTION_Swipe {
+	} else if actionType == ACTION_Swipe || actionType == ACTION_Drag {
 		if len(actionCoordinates) != 4 {
 			return fmt.Errorf("invalid swipe action coordinates: %v", actionCoordinates)
 		}
 		imagePath = filepath.Join(
 			config.GetConfig().ScreenShotsPath,
-			fmt.Sprintf("%s_swipe_marked.png", timestamp),
+			fmt.Sprintf("%s_%s.png", timestamp, actionType),
 		)
 		fromX, fromY := actionCoordinates[0], actionCoordinates[1]
 		toX, toY := actionCoordinates[2], actionCoordinates[3]
@@ -353,6 +354,13 @@ func MarkUIOperation(driver IDriver, actionType ActionMethod, actionCoordinates
 			Str("imagePath", imagePath).
 			Int64("duration(ms)", time.Since(start).Milliseconds()).
 			Msg("mark UI operation success")
+
+		// save screenshot to session
+		session := driver.GetSession()
+		session.screenResults = append(session.screenResults, &ScreenResult{
+			bufSource: compressedBufSource,
+			ImagePath: imagePath,
+		})
 	}
 
 	return nil

uixt/driver_session.go

@@ -47,9 +47,9 @@ func NewDriverSession() *DriverSession {
 		client: &http.Client{
 			Timeout: timeout,
 		},
-		requests: make([]*DriverRequests, 0),
 		maxRetry: 5,
 	}
+	session.Reset()
 	return session
 }
 
@@ -66,10 +66,14 @@ type DriverSession struct {
 
 	// cache driver request and response
 	requests []*DriverRequests
+
+	// cache screenshot results
+	screenResults []*ScreenResult
 }
 
 func (s *DriverSession) Reset() {
 	s.requests = make([]*DriverRequests, 0)
+	s.screenResults = make([]*ScreenResult, 0)
 }
 
 func (s *DriverSession) SetBaseURL(baseUrl string) {

uixt/driver_utils.go

@@ -117,11 +117,10 @@ func (dExt *XTDriver) GetData(withReset bool) map[string]interface{} {
 	session := dExt.GetSession()
 	data := map[string]interface{}{
 		"requests":       session.History(),
-		"screen_results": dExt.screenResults,
+		"screen_results": session.screenResults,
 	}
 	if withReset {
 		session.Reset()
-		dExt.screenResults = make([]*ScreenResult, 0)
 	}
 	return data
 }