refactor: select model type by env LLM_MODEL_USE

internal/version/VERSION

@@ -1 +1 @@
-v5.0.0-beta-2504292233
+v5.0.0-beta-2504292314

uixt/ai/ai.go

@@ -49,10 +49,9 @@ func WithCVService(service CVServiceType) AIServiceOption {
 type LLMServiceType string
 
 const (
-	LLMServiceTypeUITARS     LLMServiceType = "ui-tars"
+	LLMServiceTypeUITARS LLMServiceType = "ui-tars"
-	LLMServiceTypeGPT4o      LLMServiceType = "gpt-4o"
+	LLMServiceTypeGPT    LLMServiceType = "gpt"
-	LLMServiceTypeGPT4Vision LLMServiceType = "gpt-4-vision"
+	LLMServiceTypeQwenVL LLMServiceType = "qwen-vl"
-	LLMServiceTypeQwenVL     LLMServiceType = "qwen-vl"
 )
 
 // ILLMService 定义了 LLM 服务接口，包括规划和断言功能
@@ -67,7 +66,7 @@ func WithLLMService(modelType LLMServiceType) AIServiceOption {
 		var planner IPlanner
 		var err error
 		switch modelType {
-		case LLMServiceTypeGPT4o:
+		case LLMServiceTypeGPT:
 			// TODO: implement gpt-4o planner and asserter
 			planner, err = NewPlanner(context.Background())
 		case LLMServiceTypeUITARS:
@@ -79,9 +78,9 @@ func WithLLMService(modelType LLMServiceType) AIServiceOption {
 		}
 
 		// init asserter
-		asserter, err := NewAsserter(context.Background(), modelType)
+		asserter, err := NewAsserter(context.Background())
 		if err != nil {
-			log.Error().Err(err).Msgf("init %s asserter failed", modelType)
+			log.Error().Err(err).Msg("init asserter failed")
 			os.Exit(code.GetErrorCode(err))
 		}
 
@@ -115,11 +114,14 @@ const (
 	EnvModelName     = "LLM_MODEL_NAME"
 )
 
+var EnvModelUse string
+
 // GetOpenAIModelConfig get OpenAI config
 func GetOpenAIModelConfig() (*openai.ChatModelConfig, error) {
 	if err := config.LoadEnv(); err != nil {
 		return nil, errors.Wrap(code.LoadEnvError, err.Error())
 	}
+	EnvModelUse = os.Getenv("LLM_MODEL_USE")
 
 	openaiBaseURL := os.Getenv(EnvOpenAIBaseURL)
 	if openaiBaseURL == "" {

uixt/ai/asserter.go

@@ -43,14 +43,12 @@ type Asserter struct {
 	model        model.ToolCallingChatModel
 	systemPrompt string
 	history      ConversationHistory
-	modelType    LLMServiceType
 }
 
 // NewAsserter creates a new Asserter instance
-func NewAsserter(ctx context.Context, modelType LLMServiceType) (*Asserter, error) {
+func NewAsserter(ctx context.Context) (*Asserter, error) {
 	asserter := &Asserter{
 		ctx:          ctx,
-		modelType:    modelType,
 		systemPrompt: defaultAssertionPrompt,
 	}
 
@@ -59,12 +57,11 @@ func NewAsserter(ctx context.Context, modelType LLMServiceType) (*Asserter, erro
 		return nil, err
 	}
 
-	switch modelType {
+	if strings.Contains(EnvModelUse, string(LLMServiceTypeUITARS)) {
-	case LLMServiceTypeUITARS:
 		asserter.systemPrompt += "\n\n" + uiTarsAssertionResponseFormat
-	case LLMServiceTypeQwenVL:
+	} else if strings.Contains(EnvModelUse, string(LLMServiceTypeQwenVL)) {
 		asserter.systemPrompt += "\n\n" + defaultAssertionResponseJsonFormat
-	case LLMServiceTypeGPT4Vision, LLMServiceTypeGPT4o:
+	} else if strings.Contains(EnvModelUse, string(LLMServiceTypeGPT)) {
 		// define output format
 		type OutputFormat struct {
 			Thought string `json:"thought"`
@@ -86,8 +83,8 @@ func NewAsserter(ctx context.Context, modelType LLMServiceType) (*Asserter, erro
 				Strict:      false,
 			},
 		}
-	default:
+	} else {
-		return nil, errors.New("not supported model type for asserter")
+		return nil, fmt.Errorf("model type %s not supported for asserter", EnvModelUse)
 	}
 
 	asserter.model, err = openai.NewChatModel(ctx, config)
@@ -144,7 +141,7 @@ Here is the assertion. Please tell whether it is truthy according to the screens
 	startTime := time.Now()
 	resp, err := a.model.Generate(a.ctx, a.history)
 	log.Info().Float64("elapsed(s)", time.Since(startTime).Seconds()).
-		Str("model", string(a.modelType)).Msg("call model service for assertion")
+		Str("model", EnvModelUse).Msg("call model service for assertion")
 	if err != nil {
 		return nil, errors.Wrap(code.LLMRequestServiceError, err.Error())
 	}

uixt/ai/asserter_test.go

@@ -10,7 +10,7 @@ import (
 )
 
 func createAsserter(t *testing.T) *Asserter {
-	asserter, err := NewAsserter(context.Background(), LLMServiceTypeUITARS)
+	asserter, err := NewAsserter(context.Background())
 	require.NoError(t, err)
 	return asserter
 }

uixt/ai/planner_gpt.go

@@ -30,7 +30,7 @@ func NewPlanner(ctx context.Context) (*Planner, error) {
 	return &Planner{
 		ctx:          ctx,
 		model:        model,
-		modelType:    LLMServiceTypeGPT4o,
+		modelType:    LLMServiceTypeGPT,
 		systemPrompt: uiTarsPlanningPrompt, // TODO: change prompt with function calling
 	}, nil
 }