refactor: add AIAction

internal/version/VERSION

@@ -1 +1 @@
-v5.0.0-beta-2503220119
+v5.0.0-beta-2503221208

uixt/driver_ext_ai.go

@@ -21,24 +21,8 @@ func (dExt *XTDriver) StartToGoal(text string, opts ...option.ActionOption) erro
 	for {
 		attempt++
 		log.Info().Int("attempt", attempt).Msg("planning attempt")
-
+		if err := dExt.AIAction(text, opts...); err != nil {
-		// plan next action
+			return err
-		result, err := dExt.PlanNextAction(text, opts...)
-		if err != nil {
-			return errors.Wrap(err, "failed to get next action from planner")
-		}
-
-		// do actions
-		for _, action := range result.NextActions {
-			switch action.ActionType {
-			case ai.ActionTypeClick:
-				point := action.ActionInputs["startBox"].([]float64)
-				if err := dExt.TapAbsXY(point[0], point[1], opts...); err != nil {
-					return err
-				}
-			case ai.ActionTypeFinished:
-				return nil
-			}
 		}
 
 		if options.MaxRetryTimes > 1 && attempt >= options.MaxRetryTimes {
@@ -47,6 +31,30 @@ func (dExt *XTDriver) StartToGoal(text string, opts ...option.ActionOption) erro
 	}
 }
 
+func (dExt *XTDriver) AIAction(text string, opts ...option.ActionOption) error {
+	// plan next action
+	result, err := dExt.PlanNextAction(text, opts...)
+	if err != nil {
+		return err
+	}
+
+	// do actions
+	for _, action := range result.NextActions {
+		switch action.ActionType {
+		case ai.ActionTypeClick:
+			point := action.ActionInputs["startBox"].([]float64)
+			if err := dExt.TapAbsXY(point[0], point[1], opts...); err != nil {
+				return err
+			}
+		case ai.ActionTypeFinished:
+			log.Info().Msg("ai action done")
+			return nil
+		}
+	}
+
+	return nil
+}
+
 func (dExt *XTDriver) PlanNextAction(text string, opts ...option.ActionOption) (*ai.PlanningResult, error) {
 	if dExt.LLMService == nil {
 		return nil, errors.New("LLM service is not initialized")

uixt/driver_ext_tap.go

@@ -8,22 +8,6 @@ import (
 	"github.com/rs/zerolog/log"
 )
 
-func (dExt *XTDriver) TapByLLM(text string, opts ...option.ActionOption) error {
-	text = "[click] " + text
-	result, err := dExt.PlanNextAction(text, opts...)
-	if err != nil {
-		return err
-	}
-
-	action := result.NextActions[0]
-	if action.ActionType != ai.ActionTypeClick {
-		return fmt.Errorf("expected click action, got: %s", action.ActionType)
-	}
-
-	point := action.ActionInputs["startBox"].([]float64)
-	return dExt.TapAbsXY(point[0], point[1], opts...)
-}
-
 func (dExt *XTDriver) TapByOCR(text string, opts ...option.ActionOption) error {
 	actionOptions := option.NewActionOptions(opts...)
 	if actionOptions.ScreenShotFileName == "" {

uixt/driver_ext_test.go

@@ -125,7 +125,7 @@ func TestDriverExt_TapByOCR(t *testing.T) {
 
 func TestDriverExt_TapByLLM(t *testing.T) {
 	driver := setupDriverExt(t)
-	err := driver.TapByLLM("点击第一个帖子的作者头像")
+	err := driver.AIAction("点击第一个帖子的作者头像")
 	assert.Nil(t, err)
 }