fix: check LLM env

code/code.go

@@ -104,7 +104,8 @@ var (
 	CVResponseError          = errors.New("CV parse response error")  // 83
 	CVResultNotFoundError    = errors.New("CV result not found")      // 84
 
-	StateUnknowError = errors.New("detect state failed") // 85
+	LLMEnvMissedError = errors.New("LLM env missed error") // 85
+	StateUnknowError  = errors.New("detect state failed")  // 89
 )
 
 // trackings related: [90, 100)
@@ -199,7 +200,8 @@ var errorsMap = map[error]int{
 	CVServiceConnectionError: 82,
 	CVResponseError:          83,
 	CVResultNotFoundError:    84,
-	StateUnknowError:         85,
+	LLMEnvMissedError:        85,
+	StateUnknowError:         89,
 
 	// trackings related
 	TrackingGetError:   90,

internal/version/VERSION

@@ -1 +1 @@
-v5.0.0-beta-2503192116
+v5.0.0-beta-2503192228

uixt/ai/ai.go

@@ -53,6 +53,11 @@ const (
 )
 
 func WithLLMService(service LLMServiceType) AIServiceOption {
+	if err := checkEnvLLM(); err != nil {
+		log.Error().Err(err).Msg("check LLM env failed")
+		os.Exit(code.GetErrorCode(err))
+	}
+
 	return func(opts *AIServices) {
 		if service == LLMServiceTypeGPT4o {
 			var err error

uixt/ai/ai_test.go

@@ -1,6 +1,8 @@
 package ai
 
-import "testing"
+import (
+	"testing"
+)
 
 func TestOption(t *testing.T) {
 	options := NewAIService(

uixt/ai/env.go

@@ -9,7 +9,9 @@ import (
 	"time"
 
 	"github.com/cloudwego/eino-ext/components/model/openai"
+	"github.com/httprunner/httprunner/v5/code"
 	"github.com/joho/godotenv"
+	"github.com/pkg/errors"
 	"github.com/rs/zerolog/log"
 )
 
@@ -25,11 +27,29 @@ type OpenAIInitConfig struct {
 const (
 	EnvOpenAIBaseURL        = "OPENAI_BASE_URL"
 	EnvOpenAIAPIKey         = "OPENAI_API_KEY"
-	EnvModelName            = "MIDSCENE_MODEL_NAME"
+	EnvModelName            = "LLM_MODEL_NAME"
-	EnvOpenAIInitConfigJSON = "MIDSCENE_OPENAI_INIT_CONFIG_JSON"
+	EnvOpenAIInitConfigJSON = "OPENAI_INIT_CONFIG_JSON"
-	EnvUseVLMUITars         = "MIDSCENE_USE_VLM_UI_TARS"
 )
 
+func checkEnvLLM() error {
+	openaiBaseURL := os.Getenv("OPENAI_BASE_URL")
+	if openaiBaseURL == "" {
+		return errors.Wrap(code.LLMEnvMissedError, "OPENAI_BASE_URL missed")
+	}
+	log.Info().Str("OPENAI_BASE_URL", openaiBaseURL).Msg("get env")
+	openaiAPIKey := os.Getenv("OPENAI_API_KEY")
+	if openaiAPIKey == "" {
+		return errors.Wrap(code.LLMEnvMissedError, "OPENAI_API_KEY missed")
+	}
+	log.Info().Str("OPENAI_API_KEY", maskAPIKey(openaiAPIKey)).Msg("get env")
+	modelName := os.Getenv("LLM_MODEL_NAME")
+	if modelName == "" {
+		return errors.Wrap(code.LLMEnvMissedError, "LLM_MODEL_NAME missed")
+	}
+	log.Info().Str("LLM_MODEL_NAME", modelName).Msg("get env")
+	return nil
+}
+
 // loadEnv loads environment variables from a file
 func loadEnv(envPath string) error {
 	err := godotenv.Load(envPath)
@@ -164,7 +184,3 @@ func maskAPIKey(key string) string {
 
 	return key[:4] + "******" + key[len(key)-4:]
 }
-
-func IsUseVLMUITars() bool {
-	return GetEnvConfigInBool(EnvUseVLMUITars)
-}

uixt/ai/llm.go

@@ -1,6 +1,8 @@
 package ai
 
-import "github.com/cloudwego/eino/schema"
+import (
+	"github.com/cloudwego/eino/schema"
+)
 
 type ILLMService interface {
 	Call(opts *PlanningOptions) (*PlanningResult, error)

uixt/ai/planner_test.go

@@ -11,7 +11,7 @@ import (
 )
 
 func TestVLMPlanning(t *testing.T) {
-	imageBase64, err := loadImage("testdata/llk_3.jpg")
+	imageBase64, err := loadImage("testdata/llk_1.png")
 	require.NoError(t, err)
 
 	userInstruction := `连连看是一款经典的益智消除类小游戏，通常以图案或图标为主要元素。以下是连连看的基本规则说明：
@@ -25,9 +25,7 @@ func TestVLMPlanning(t *testing.T) {
 	5. 得分机制: 每成功连接并消除一对图案，玩家会获得相应的分数。完成游戏后，根据剩余时间和消除效率计算总分。
 	6. 关卡设计: 游戏可能包含多个关卡，随着关卡的推进，图案的复杂度和数量会增加。`
 
-	// userInstruction += "\n\n请基于以上游戏规则，请先点击第一个图标"
+	userInstruction += "\n\n请基于以上游戏规则，给出消除所有图案的行动序列"
-
-	userInstruction += "\n\n点击[3排1列]果酱图案"
 
 	planner, err := NewPlanner(context.Background())
 	require.NoError(t, err)

uixt/driver_ext_screenshot.go

@@ -69,8 +69,11 @@ func (dExt *XTDriver) GetScreenShotBase64() (base64Str string, err error) {
 	if err != nil {
 		return "", err
 	}
+
+	// convert buffer to base64 string
 	base64Str = "data:image/jpeg;base64," +
 		base64.StdEncoding.EncodeToString(compressedBufSource.Bytes())
+
 	return base64Str, nil
 }