@midscene/shared 0.30.10 → 1.0.0

package/dist/types/env/types.d.ts

@@ -1,8 +1,7 @@
-export declare const MIDSCENE_OPENAI_INIT_CONFIG_JSON = "MIDSCENE_OPENAI_INIT_CONFIG_JSON";
+export declare const MIDSCENE_MODEL_INIT_CONFIG_JSON = "MIDSCENE_MODEL_INIT_CONFIG_JSON";
 export declare const MIDSCENE_MODEL_NAME = "MIDSCENE_MODEL_NAME";
-export declare const MIDSCENE_LANGSMITH_DEBUG = "MIDSCENE_LANGSMITH_DEBUG";
-export declare const MIDSCENE_DEBUG_AI_PROFILE = "MIDSCENE_DEBUG_AI_PROFILE";
-export declare const MIDSCENE_DEBUG_AI_RESPONSE = "MIDSCENE_DEBUG_AI_RESPONSE";
+export declare const MIDSCENE_DEBUG_MODEL_PROFILE = "MIDSCENE_DEBUG_MODEL_PROFILE";
+export declare const MIDSCENE_DEBUG_MODEL_RESPONSE = "MIDSCENE_DEBUG_MODEL_RESPONSE";
 export declare const MIDSCENE_DANGEROUSLY_PRINT_ALL_CONFIG = "MIDSCENE_DANGEROUSLY_PRINT_ALL_CONFIG";
 export declare const MIDSCENE_DEBUG_MODE = "MIDSCENE_DEBUG_MODE";
 export declare const MIDSCENE_MCP_USE_PUPPETEER_MODE = "MIDSCENE_MCP_USE_PUPPETEER_MODE";
@@ -10,10 +9,37 @@ export declare const MIDSCENE_MCP_CHROME_PATH = "MIDSCENE_MCP_CHROME_PATH";
 export declare const MIDSCENE_MCP_ANDROID_MODE = "MIDSCENE_MCP_ANDROID_MODE";
 export declare const DOCKER_CONTAINER = "DOCKER_CONTAINER";
 export declare const MIDSCENE_FORCE_DEEP_THINK = "MIDSCENE_FORCE_DEEP_THINK";
-export declare const MIDSCENE_OPENAI_SOCKS_PROXY = "MIDSCENE_OPENAI_SOCKS_PROXY";
-export declare const MIDSCENE_OPENAI_HTTP_PROXY = "MIDSCENE_OPENAI_HTTP_PROXY";
+export declare const MIDSCENE_LANGSMITH_DEBUG = "MIDSCENE_LANGSMITH_DEBUG";
+export declare const MIDSCENE_LANGFUSE_DEBUG = "MIDSCENE_LANGFUSE_DEBUG";
+export declare const MIDSCENE_MODEL_SOCKS_PROXY = "MIDSCENE_MODEL_SOCKS_PROXY";
+export declare const MIDSCENE_MODEL_HTTP_PROXY = "MIDSCENE_MODEL_HTTP_PROXY";
+export declare const MIDSCENE_MODEL_API_KEY = "MIDSCENE_MODEL_API_KEY";
+export declare const MIDSCENE_MODEL_BASE_URL = "MIDSCENE_MODEL_BASE_URL";
+export declare const MIDSCENE_MODEL_MAX_TOKENS = "MIDSCENE_MODEL_MAX_TOKENS";
+export declare const MIDSCENE_MODEL_TIMEOUT = "MIDSCENE_MODEL_TIMEOUT";
+/**
+ * @deprecated Use MIDSCENE_MODEL_API_KEY instead. This is kept for backward compatibility.
+ */
 export declare const OPENAI_API_KEY = "OPENAI_API_KEY";
+/**
+ * @deprecated Use MIDSCENE_MODEL_BASE_URL instead. This is kept for backward compatibility.
+ */
 export declare const OPENAI_BASE_URL = "OPENAI_BASE_URL";
+/**
+ * @deprecated Use MIDSCENE_MODEL_INIT_CONFIG_JSON instead. This is kept for backward compatibility.
+ */
+export declare const MIDSCENE_OPENAI_INIT_CONFIG_JSON = "MIDSCENE_OPENAI_INIT_CONFIG_JSON";
+/**
+ * @deprecated Use MIDSCENE_MODEL_HTTP_PROXY instead. This is kept for backward compatibility.
+ */
+export declare const MIDSCENE_OPENAI_HTTP_PROXY = "MIDSCENE_OPENAI_HTTP_PROXY";
+/**
+ * @deprecated Use MIDSCENE_MODEL_SOCKS_PROXY instead. This is kept for backward compatibility.
+ */
+export declare const MIDSCENE_OPENAI_SOCKS_PROXY = "MIDSCENE_OPENAI_SOCKS_PROXY";
+/**
+ * @deprecated Use MIDSCENE_MODEL_MAX_TOKENS instead. This is kept for backward compatibility.
+ */
 export declare const OPENAI_MAX_TOKENS = "OPENAI_MAX_TOKENS";
 export declare const MIDSCENE_ADB_PATH = "MIDSCENE_ADB_PATH";
 export declare const MIDSCENE_ADB_REMOTE_HOST = "MIDSCENE_ADB_REMOTE_HOST";
@@ -29,82 +55,26 @@ export declare const MIDSCENE_USE_DOUBAO_VISION = "MIDSCENE_USE_DOUBAO_VISION";
 export declare const MIDSCENE_USE_GEMINI = "MIDSCENE_USE_GEMINI";
 export declare const MIDSCENE_USE_VL_MODEL = "MIDSCENE_USE_VL_MODEL";
 export declare const MATCH_BY_POSITION = "MATCH_BY_POSITION";
-export declare const MIDSCENE_API_TYPE = "MIDSCENE-API-TYPE";
 export declare const MIDSCENE_REPORT_TAG_NAME = "MIDSCENE_REPORT_TAG_NAME";
-export declare const MIDSCENE_REPLANNING_CYCLE_LIMIT = "MIDSCENE_REPLANNING_CYCLE_LIMIT";
 export declare const MIDSCENE_PREFERRED_LANGUAGE = "MIDSCENE_PREFERRED_LANGUAGE";
-export declare const MIDSCENE_USE_AZURE_OPENAI = "MIDSCENE_USE_AZURE_OPENAI";
-export declare const MIDSCENE_AZURE_OPENAI_SCOPE = "MIDSCENE_AZURE_OPENAI_SCOPE";
-export declare const MIDSCENE_AZURE_OPENAI_INIT_CONFIG_JSON = "MIDSCENE_AZURE_OPENAI_INIT_CONFIG_JSON";
 export declare const MIDSCENE_CACHE_MAX_FILENAME_LENGTH = "MIDSCENE_CACHE_MAX_FILENAME_LENGTH";
-export declare const AZURE_OPENAI_ENDPOINT = "AZURE_OPENAI_ENDPOINT";
-export declare const AZURE_OPENAI_KEY = "AZURE_OPENAI_KEY";
-export declare const AZURE_OPENAI_API_VERSION = "AZURE_OPENAI_API_VERSION";
-export declare const AZURE_OPENAI_DEPLOYMENT = "AZURE_OPENAI_DEPLOYMENT";
-export declare const MIDSCENE_USE_ANTHROPIC_SDK = "MIDSCENE_USE_ANTHROPIC_SDK";
-export declare const ANTHROPIC_API_KEY = "ANTHROPIC_API_KEY";
+export declare const MIDSCENE_REPLANNING_CYCLE_LIMIT = "MIDSCENE_REPLANNING_CYCLE_LIMIT";
 export declare const MIDSCENE_RUN_DIR = "MIDSCENE_RUN_DIR";
-export declare const MIDSCENE_OPENAI_BASE_URL = "MIDSCENE_OPENAI_BASE_URL";
-export declare const MIDSCENE_OPENAI_API_KEY = "MIDSCENE_OPENAI_API_KEY";
-export declare const MIDSCENE_OPENAI_USE_AZURE = "MIDSCENE_OPENAI_USE_AZURE";
-export declare const MIDSCENE_AZURE_OPENAI_KEY = "MIDSCENE_AZURE_OPENAI_KEY";
-export declare const MIDSCENE_AZURE_OPENAI_ENDPOINT = "MIDSCENE_AZURE_OPENAI_ENDPOINT";
-export declare const MIDSCENE_AZURE_OPENAI_API_VERSION = "MIDSCENE_AZURE_OPENAI_API_VERSION";
-export declare const MIDSCENE_AZURE_OPENAI_DEPLOYMENT = "MIDSCENE_AZURE_OPENAI_DEPLOYMENT";
-export declare const MIDSCENE_ANTHROPIC_API_KEY = "MIDSCENE_ANTHROPIC_API_KEY";
-export declare const MIDSCENE_VL_MODE = "MIDSCENE_VL_MODE";
-export declare const MIDSCENE_VQA_MODEL_NAME = "MIDSCENE_VQA_MODEL_NAME";
-export declare const MIDSCENE_VQA_OPENAI_SOCKS_PROXY = "MIDSCENE_VQA_OPENAI_SOCKS_PROXY";
-export declare const MIDSCENE_VQA_OPENAI_HTTP_PROXY = "MIDSCENE_VQA_OPENAI_HTTP_PROXY";
-export declare const MIDSCENE_VQA_OPENAI_BASE_URL = "MIDSCENE_VQA_OPENAI_BASE_URL";
-export declare const MIDSCENE_VQA_OPENAI_API_KEY = "MIDSCENE_VQA_OPENAI_API_KEY";
-export declare const MIDSCENE_VQA_OPENAI_INIT_CONFIG_JSON = "MIDSCENE_VQA_OPENAI_INIT_CONFIG_JSON";
-export declare const MIDSCENE_VQA_OPENAI_USE_AZURE = "MIDSCENE_VQA_OPENAI_USE_AZURE";
-export declare const MIDSCENE_VQA_USE_AZURE_OPENAI = "MIDSCENE_VQA_USE_AZURE_OPENAI";
-export declare const MIDSCENE_VQA_AZURE_OPENAI_SCOPE = "MIDSCENE_VQA_AZURE_OPENAI_SCOPE";
-export declare const MIDSCENE_VQA_AZURE_OPENAI_KEY = "MIDSCENE_VQA_AZURE_OPENAI_KEY";
-export declare const MIDSCENE_VQA_AZURE_OPENAI_ENDPOINT = "MIDSCENE_VQA_AZURE_OPENAI_ENDPOINT";
-export declare const MIDSCENE_VQA_AZURE_OPENAI_API_VERSION = "MIDSCENE_VQA_AZURE_OPENAI_API_VERSION";
-export declare const MIDSCENE_VQA_AZURE_OPENAI_DEPLOYMENT = "MIDSCENE_VQA_AZURE_OPENAI_DEPLOYMENT";
-export declare const MIDSCENE_VQA_AZURE_OPENAI_INIT_CONFIG_JSON = "MIDSCENE_VQA_AZURE_OPENAI_INIT_CONFIG_JSON";
-export declare const MIDSCENE_VQA_USE_ANTHROPIC_SDK = "MIDSCENE_VQA_USE_ANTHROPIC_SDK";
-export declare const MIDSCENE_VQA_ANTHROPIC_API_KEY = "MIDSCENE_VQA_ANTHROPIC_API_KEY";
-export declare const MIDSCENE_VQA_VL_MODE = "MIDSCENE_VQA_VL_MODE";
+export declare const MIDSCENE_INSIGHT_MODEL_NAME = "MIDSCENE_INSIGHT_MODEL_NAME";
+export declare const MIDSCENE_INSIGHT_MODEL_SOCKS_PROXY = "MIDSCENE_INSIGHT_MODEL_SOCKS_PROXY";
+export declare const MIDSCENE_INSIGHT_MODEL_HTTP_PROXY = "MIDSCENE_INSIGHT_MODEL_HTTP_PROXY";
+export declare const MIDSCENE_INSIGHT_MODEL_BASE_URL = "MIDSCENE_INSIGHT_MODEL_BASE_URL";
+export declare const MIDSCENE_INSIGHT_MODEL_API_KEY = "MIDSCENE_INSIGHT_MODEL_API_KEY";
+export declare const MIDSCENE_INSIGHT_MODEL_INIT_CONFIG_JSON = "MIDSCENE_INSIGHT_MODEL_INIT_CONFIG_JSON";
+export declare const MIDSCENE_INSIGHT_MODEL_TIMEOUT = "MIDSCENE_INSIGHT_MODEL_TIMEOUT";
 export declare const MIDSCENE_PLANNING_MODEL_NAME = "MIDSCENE_PLANNING_MODEL_NAME";
-export declare const MIDSCENE_PLANNING_OPENAI_SOCKS_PROXY = "MIDSCENE_PLANNING_OPENAI_SOCKS_PROXY";
-export declare const MIDSCENE_PLANNING_OPENAI_HTTP_PROXY = "MIDSCENE_PLANNING_OPENAI_HTTP_PROXY";
-export declare const MIDSCENE_PLANNING_OPENAI_BASE_URL = "MIDSCENE_PLANNING_OPENAI_BASE_URL";
-export declare const MIDSCENE_PLANNING_OPENAI_API_KEY = "MIDSCENE_PLANNING_OPENAI_API_KEY";
-export declare const MIDSCENE_PLANNING_OPENAI_INIT_CONFIG_JSON = "MIDSCENE_PLANNING_OPENAI_INIT_CONFIG_JSON";
-export declare const MIDSCENE_PLANNING_OPENAI_USE_AZURE = "MIDSCENE_PLANNING_OPENAI_USE_AZURE";
-export declare const MIDSCENE_PLANNING_USE_AZURE_OPENAI = "MIDSCENE_PLANNING_USE_AZURE_OPENAI";
-export declare const MIDSCENE_PLANNING_AZURE_OPENAI_SCOPE = "MIDSCENE_PLANNING_AZURE_OPENAI_SCOPE";
-export declare const MIDSCENE_PLANNING_AZURE_OPENAI_KEY = "MIDSCENE_PLANNING_AZURE_OPENAI_KEY";
-export declare const MIDSCENE_PLANNING_AZURE_OPENAI_ENDPOINT = "MIDSCENE_PLANNING_AZURE_OPENAI_ENDPOINT";
-export declare const MIDSCENE_PLANNING_AZURE_OPENAI_API_VERSION = "MIDSCENE_PLANNING_AZURE_OPENAI_API_VERSION";
-export declare const MIDSCENE_PLANNING_AZURE_OPENAI_DEPLOYMENT = "MIDSCENE_PLANNING_AZURE_OPENAI_DEPLOYMENT";
-export declare const MIDSCENE_PLANNING_AZURE_OPENAI_INIT_CONFIG_JSON = "MIDSCENE_PLANNING_AZURE_OPENAI_INIT_CONFIG_JSON";
-export declare const MIDSCENE_PLANNING_USE_ANTHROPIC_SDK = "MIDSCENE_PLANNING_USE_ANTHROPIC_SDK";
-export declare const MIDSCENE_PLANNING_ANTHROPIC_API_KEY = "MIDSCENE_PLANNING_ANTHROPIC_API_KEY";
-export declare const MIDSCENE_PLANNING_VL_MODE = "MIDSCENE_PLANNING_VL_MODE";
-export declare const MIDSCENE_GROUNDING_MODEL_NAME = "MIDSCENE_GROUNDING_MODEL_NAME";
-export declare const MIDSCENE_GROUNDING_OPENAI_SOCKS_PROXY = "MIDSCENE_GROUNDING_OPENAI_SOCKS_PROXY";
-export declare const MIDSCENE_GROUNDING_OPENAI_HTTP_PROXY = "MIDSCENE_GROUNDING_OPENAI_HTTP_PROXY";
-export declare const MIDSCENE_GROUNDING_OPENAI_BASE_URL = "MIDSCENE_GROUNDING_OPENAI_BASE_URL";
-export declare const MIDSCENE_GROUNDING_OPENAI_API_KEY = "MIDSCENE_GROUNDING_OPENAI_API_KEY";
-export declare const MIDSCENE_GROUNDING_OPENAI_INIT_CONFIG_JSON = "MIDSCENE_GROUNDING_OPENAI_INIT_CONFIG_JSON";
-export declare const MIDSCENE_GROUNDING_OPENAI_USE_AZURE = "MIDSCENE_GROUNDING_OPENAI_USE_AZURE";
-export declare const MIDSCENE_GROUNDING_USE_AZURE_OPENAI = "MIDSCENE_GROUNDING_USE_AZURE_OPENAI";
-export declare const MIDSCENE_GROUNDING_AZURE_OPENAI_SCOPE = "MIDSCENE_GROUNDING_AZURE_OPENAI_SCOPE";
-export declare const MIDSCENE_GROUNDING_AZURE_OPENAI_KEY = "MIDSCENE_GROUNDING_AZURE_OPENAI_KEY";
-export declare const MIDSCENE_GROUNDING_AZURE_OPENAI_ENDPOINT = "MIDSCENE_GROUNDING_AZURE_OPENAI_ENDPOINT";
-export declare const MIDSCENE_GROUNDING_AZURE_OPENAI_API_VERSION = "MIDSCENE_GROUNDING_AZURE_OPENAI_API_VERSION";
-export declare const MIDSCENE_GROUNDING_AZURE_OPENAI_DEPLOYMENT = "MIDSCENE_GROUNDING_AZURE_OPENAI_DEPLOYMENT";
-export declare const MIDSCENE_GROUNDING_AZURE_OPENAI_INIT_CONFIG_JSON = "MIDSCENE_GROUNDING_AZURE_OPENAI_INIT_CONFIG_JSON";
-export declare const MIDSCENE_GROUNDING_USE_ANTHROPIC_SDK = "MIDSCENE_GROUNDING_USE_ANTHROPIC_SDK";
-export declare const MIDSCENE_GROUNDING_ANTHROPIC_API_KEY = "MIDSCENE_GROUNDING_ANTHROPIC_API_KEY";
-export declare const MIDSCENE_GROUNDING_VL_MODE = "MIDSCENE_GROUNDING_VL_MODE";
-export declare const OPENAI_USE_AZURE = "OPENAI_USE_AZURE";
+export declare const MIDSCENE_PLANNING_MODEL_SOCKS_PROXY = "MIDSCENE_PLANNING_MODEL_SOCKS_PROXY";
+export declare const MIDSCENE_PLANNING_MODEL_HTTP_PROXY = "MIDSCENE_PLANNING_MODEL_HTTP_PROXY";
+export declare const MIDSCENE_PLANNING_MODEL_BASE_URL = "MIDSCENE_PLANNING_MODEL_BASE_URL";
+export declare const MIDSCENE_PLANNING_MODEL_API_KEY = "MIDSCENE_PLANNING_MODEL_API_KEY";
+export declare const MIDSCENE_PLANNING_MODEL_INIT_CONFIG_JSON = "MIDSCENE_PLANNING_MODEL_INIT_CONFIG_JSON";
+export declare const MIDSCENE_PLANNING_MODEL_TIMEOUT = "MIDSCENE_PLANNING_MODEL_TIMEOUT";
+export declare const MIDSCENE_MODEL_FAMILY = "MIDSCENE_MODEL_FAMILY";
 /**
  * env keys declared but unused
  */
@@ -113,102 +83,79 @@ export declare const UNUSED_ENV_KEYS: string[];
  * env keys for debug or basic run
  * can not be override by overrideAIConfig
  */
-export declare const BASIC_ENV_KEYS: readonly ["MIDSCENE_DEBUG_MODE", "MIDSCENE_DEBUG_AI_PROFILE", "MIDSCENE_DEBUG_AI_RESPONSE", "MIDSCENE_RUN_DIR"];
-export declare const BOOLEAN_ENV_KEYS: readonly ["MIDSCENE_CACHE", "MIDSCENE_LANGSMITH_DEBUG", "MIDSCENE_FORCE_DEEP_THINK", "MIDSCENE_MCP_USE_PUPPETEER_MODE", "MIDSCENE_MCP_ANDROID_MODE"];
+export declare const BASIC_ENV_KEYS: readonly ["MIDSCENE_DEBUG_MODE", "MIDSCENE_DEBUG_MODEL_PROFILE", "MIDSCENE_DEBUG_MODEL_RESPONSE", "MIDSCENE_RUN_DIR"];
+export declare const BOOLEAN_ENV_KEYS: readonly ["MIDSCENE_CACHE", "MIDSCENE_FORCE_DEEP_THINK", "MIDSCENE_MCP_USE_PUPPETEER_MODE", "MIDSCENE_MCP_ANDROID_MODE", "MIDSCENE_LANGSMITH_DEBUG", "MIDSCENE_LANGFUSE_DEBUG"];
 export declare const NUMBER_ENV_KEYS: readonly ["MIDSCENE_CACHE_MAX_FILENAME_LENGTH", "MIDSCENE_REPLANNING_CYCLE_LIMIT"];
-export declare const STRING_ENV_KEYS: readonly ["OPENAI_MAX_TOKENS", "MIDSCENE_ADB_PATH", "MIDSCENE_ADB_REMOTE_HOST", "MIDSCENE_ADB_REMOTE_PORT", "MIDSCENE_ANDROID_IME_STRATEGY", "MIDSCENE_IOS_DEVICE_UDID", "MIDSCENE_IOS_SIMULATOR_UDID", "MIDSCENE_REPORT_TAG_NAME", "MIDSCENE_PREFERRED_LANGUAGE", "MATCH_BY_POSITION", "MIDSCENE_MCP_CHROME_PATH", "DOCKER_CONTAINER"];
+export declare const STRING_ENV_KEYS: readonly ["MIDSCENE_MODEL_MAX_TOKENS", "OPENAI_MAX_TOKENS", "MIDSCENE_ADB_PATH", "MIDSCENE_ADB_REMOTE_HOST", "MIDSCENE_ADB_REMOTE_PORT", "MIDSCENE_ANDROID_IME_STRATEGY", "MIDSCENE_IOS_DEVICE_UDID", "MIDSCENE_IOS_SIMULATOR_UDID", "MIDSCENE_REPORT_TAG_NAME", "MIDSCENE_PREFERRED_LANGUAGE", "MATCH_BY_POSITION", "MIDSCENE_MCP_CHROME_PATH", "DOCKER_CONTAINER"];
 /**
  * Non model related env keys, used for globally controlling the behavior of midscene
  * Can not be override by agent.modelConfig but can be override by overrideAIConfig
  * Can be access at any time
  */
-export declare const GLOBAL_ENV_KEYS: readonly ["MIDSCENE_CACHE", "MIDSCENE_LANGSMITH_DEBUG", "MIDSCENE_FORCE_DEEP_THINK", "MIDSCENE_MCP_USE_PUPPETEER_MODE", "MIDSCENE_MCP_ANDROID_MODE", "MIDSCENE_CACHE_MAX_FILENAME_LENGTH", "MIDSCENE_REPLANNING_CYCLE_LIMIT", "OPENAI_MAX_TOKENS", "MIDSCENE_ADB_PATH", "MIDSCENE_ADB_REMOTE_HOST", "MIDSCENE_ADB_REMOTE_PORT", "MIDSCENE_ANDROID_IME_STRATEGY", "MIDSCENE_IOS_DEVICE_UDID", "MIDSCENE_IOS_SIMULATOR_UDID", "MIDSCENE_REPORT_TAG_NAME", "MIDSCENE_PREFERRED_LANGUAGE", "MATCH_BY_POSITION", "MIDSCENE_MCP_CHROME_PATH", "DOCKER_CONTAINER"];
+export declare const GLOBAL_ENV_KEYS: readonly ["MIDSCENE_CACHE", "MIDSCENE_FORCE_DEEP_THINK", "MIDSCENE_MCP_USE_PUPPETEER_MODE", "MIDSCENE_MCP_ANDROID_MODE", "MIDSCENE_LANGSMITH_DEBUG", "MIDSCENE_LANGFUSE_DEBUG", "MIDSCENE_CACHE_MAX_FILENAME_LENGTH", "MIDSCENE_REPLANNING_CYCLE_LIMIT", "MIDSCENE_MODEL_MAX_TOKENS", "OPENAI_MAX_TOKENS", "MIDSCENE_ADB_PATH", "MIDSCENE_ADB_REMOTE_HOST", "MIDSCENE_ADB_REMOTE_PORT", "MIDSCENE_ANDROID_IME_STRATEGY", "MIDSCENE_IOS_DEVICE_UDID", "MIDSCENE_IOS_SIMULATOR_UDID", "MIDSCENE_REPORT_TAG_NAME", "MIDSCENE_PREFERRED_LANGUAGE", "MATCH_BY_POSITION", "MIDSCENE_MCP_CHROME_PATH", "DOCKER_CONTAINER"];
 /**
  * Model related eve keys, used for declare which model to use.
  * Can be override by both agent.modelConfig and overrideAIConfig
  * Can only be access after agent.constructor
  */
-export declare const MODEL_ENV_KEYS: readonly ["MIDSCENE_MODEL_NAME", "MIDSCENE_OPENAI_INIT_CONFIG_JSON", "MIDSCENE_OPENAI_API_KEY", "MIDSCENE_OPENAI_BASE_URL", "MIDSCENE_OPENAI_USE_AZURE", "MIDSCENE_OPENAI_SOCKS_PROXY", "MIDSCENE_OPENAI_HTTP_PROXY", "MIDSCENE_USE_AZURE_OPENAI", "MIDSCENE_AZURE_OPENAI_SCOPE", "MIDSCENE_AZURE_OPENAI_INIT_CONFIG_JSON", "MIDSCENE_USE_ANTHROPIC_SDK", "MIDSCENE_USE_VLM_UI_TARS", "MIDSCENE_USE_QWEN_VL", "MIDSCENE_USE_QWEN3_VL", "MIDSCENE_USE_DOUBAO_VISION", "MIDSCENE_USE_GEMINI", "MIDSCENE_USE_VL_MODEL", "ANTHROPIC_API_KEY", "MIDSCENE_AZURE_OPENAI_ENDPOINT", "MIDSCENE_AZURE_OPENAI_KEY", "MIDSCENE_AZURE_OPENAI_API_VERSION", "MIDSCENE_AZURE_OPENAI_DEPLOYMENT", "MIDSCENE_VL_MODE", "OPENAI_API_KEY", "OPENAI_BASE_URL", "OPENAI_USE_AZURE", "ANTHROPIC_API_KEY", "AZURE_OPENAI_ENDPOINT", "AZURE_OPENAI_KEY", "AZURE_OPENAI_API_VERSION", "AZURE_OPENAI_DEPLOYMENT", "MIDSCENE_VQA_MODEL_NAME", "MIDSCENE_VQA_OPENAI_SOCKS_PROXY", "MIDSCENE_VQA_OPENAI_HTTP_PROXY", "MIDSCENE_VQA_OPENAI_BASE_URL", "MIDSCENE_VQA_OPENAI_API_KEY", "MIDSCENE_VQA_OPENAI_INIT_CONFIG_JSON", "MIDSCENE_VQA_OPENAI_USE_AZURE", "MIDSCENE_VQA_USE_AZURE_OPENAI", "MIDSCENE_VQA_AZURE_OPENAI_SCOPE", "MIDSCENE_VQA_AZURE_OPENAI_KEY", "MIDSCENE_VQA_AZURE_OPENAI_ENDPOINT", "MIDSCENE_VQA_AZURE_OPENAI_API_VERSION", "MIDSCENE_VQA_AZURE_OPENAI_DEPLOYMENT", "MIDSCENE_VQA_AZURE_OPENAI_INIT_CONFIG_JSON", "MIDSCENE_VQA_USE_ANTHROPIC_SDK", "MIDSCENE_VQA_ANTHROPIC_API_KEY", "MIDSCENE_VQA_VL_MODE", "MIDSCENE_PLANNING_MODEL_NAME", "MIDSCENE_PLANNING_OPENAI_SOCKS_PROXY", "MIDSCENE_PLANNING_OPENAI_HTTP_PROXY", "MIDSCENE_PLANNING_OPENAI_BASE_URL", "MIDSCENE_PLANNING_OPENAI_API_KEY", "MIDSCENE_PLANNING_OPENAI_INIT_CONFIG_JSON", "MIDSCENE_PLANNING_OPENAI_USE_AZURE", "MIDSCENE_PLANNING_USE_AZURE_OPENAI", "MIDSCENE_PLANNING_AZURE_OPENAI_SCOPE", "MIDSCENE_PLANNING_AZURE_OPENAI_KEY", "MIDSCENE_PLANNING_AZURE_OPENAI_ENDPOINT", "MIDSCENE_PLANNING_AZURE_OPENAI_API_VERSION", "MIDSCENE_PLANNING_AZURE_OPENAI_DEPLOYMENT", "MIDSCENE_PLANNING_AZURE_OPENAI_INIT_CONFIG_JSON", "MIDSCENE_PLANNING_USE_ANTHROPIC_SDK", "MIDSCENE_PLANNING_ANTHROPIC_API_KEY", "MIDSCENE_PLANNING_VL_MODE", "MIDSCENE_GROUNDING_MODEL_NAME", "MIDSCENE_GROUNDING_OPENAI_SOCKS_PROXY", "MIDSCENE_GROUNDING_OPENAI_HTTP_PROXY", "MIDSCENE_GROUNDING_OPENAI_BASE_URL", "MIDSCENE_GROUNDING_OPENAI_API_KEY", "MIDSCENE_GROUNDING_OPENAI_INIT_CONFIG_JSON", "MIDSCENE_GROUNDING_OPENAI_USE_AZURE", "MIDSCENE_GROUNDING_USE_AZURE_OPENAI", "MIDSCENE_GROUNDING_AZURE_OPENAI_SCOPE", "MIDSCENE_GROUNDING_AZURE_OPENAI_KEY", "MIDSCENE_GROUNDING_AZURE_OPENAI_ENDPOINT", "MIDSCENE_GROUNDING_AZURE_OPENAI_API_VERSION", "MIDSCENE_GROUNDING_AZURE_OPENAI_DEPLOYMENT", "MIDSCENE_GROUNDING_AZURE_OPENAI_INIT_CONFIG_JSON", "MIDSCENE_GROUNDING_USE_ANTHROPIC_SDK", "MIDSCENE_GROUNDING_ANTHROPIC_API_KEY", "MIDSCENE_GROUNDING_VL_MODE"];
-export declare const ALL_ENV_KEYS: readonly [...string[], "MIDSCENE_DEBUG_MODE", "MIDSCENE_DEBUG_AI_PROFILE", "MIDSCENE_DEBUG_AI_RESPONSE", "MIDSCENE_RUN_DIR", "MIDSCENE_CACHE", "MIDSCENE_LANGSMITH_DEBUG", "MIDSCENE_FORCE_DEEP_THINK", "MIDSCENE_MCP_USE_PUPPETEER_MODE", "MIDSCENE_MCP_ANDROID_MODE", "MIDSCENE_CACHE_MAX_FILENAME_LENGTH", "MIDSCENE_REPLANNING_CYCLE_LIMIT", "OPENAI_MAX_TOKENS", "MIDSCENE_ADB_PATH", "MIDSCENE_ADB_REMOTE_HOST", "MIDSCENE_ADB_REMOTE_PORT", "MIDSCENE_ANDROID_IME_STRATEGY", "MIDSCENE_IOS_DEVICE_UDID", "MIDSCENE_IOS_SIMULATOR_UDID", "MIDSCENE_REPORT_TAG_NAME", "MIDSCENE_PREFERRED_LANGUAGE", "MATCH_BY_POSITION", "MIDSCENE_MCP_CHROME_PATH", "DOCKER_CONTAINER", "MIDSCENE_MODEL_NAME", "MIDSCENE_OPENAI_INIT_CONFIG_JSON", "MIDSCENE_OPENAI_API_KEY", "MIDSCENE_OPENAI_BASE_URL", "MIDSCENE_OPENAI_USE_AZURE", "MIDSCENE_OPENAI_SOCKS_PROXY", "MIDSCENE_OPENAI_HTTP_PROXY", "MIDSCENE_USE_AZURE_OPENAI", "MIDSCENE_AZURE_OPENAI_SCOPE", "MIDSCENE_AZURE_OPENAI_INIT_CONFIG_JSON", "MIDSCENE_USE_ANTHROPIC_SDK", "MIDSCENE_USE_VLM_UI_TARS", "MIDSCENE_USE_QWEN_VL", "MIDSCENE_USE_QWEN3_VL", "MIDSCENE_USE_DOUBAO_VISION", "MIDSCENE_USE_GEMINI", "MIDSCENE_USE_VL_MODEL", "ANTHROPIC_API_KEY", "MIDSCENE_AZURE_OPENAI_ENDPOINT", "MIDSCENE_AZURE_OPENAI_KEY", "MIDSCENE_AZURE_OPENAI_API_VERSION", "MIDSCENE_AZURE_OPENAI_DEPLOYMENT", "MIDSCENE_VL_MODE", "OPENAI_API_KEY", "OPENAI_BASE_URL", "OPENAI_USE_AZURE", "ANTHROPIC_API_KEY", "AZURE_OPENAI_ENDPOINT", "AZURE_OPENAI_KEY", "AZURE_OPENAI_API_VERSION", "AZURE_OPENAI_DEPLOYMENT", "MIDSCENE_VQA_MODEL_NAME", "MIDSCENE_VQA_OPENAI_SOCKS_PROXY", "MIDSCENE_VQA_OPENAI_HTTP_PROXY", "MIDSCENE_VQA_OPENAI_BASE_URL", "MIDSCENE_VQA_OPENAI_API_KEY", "MIDSCENE_VQA_OPENAI_INIT_CONFIG_JSON", "MIDSCENE_VQA_OPENAI_USE_AZURE", "MIDSCENE_VQA_USE_AZURE_OPENAI", "MIDSCENE_VQA_AZURE_OPENAI_SCOPE", "MIDSCENE_VQA_AZURE_OPENAI_KEY", "MIDSCENE_VQA_AZURE_OPENAI_ENDPOINT", "MIDSCENE_VQA_AZURE_OPENAI_API_VERSION", "MIDSCENE_VQA_AZURE_OPENAI_DEPLOYMENT", "MIDSCENE_VQA_AZURE_OPENAI_INIT_CONFIG_JSON", "MIDSCENE_VQA_USE_ANTHROPIC_SDK", "MIDSCENE_VQA_ANTHROPIC_API_KEY", "MIDSCENE_VQA_VL_MODE", "MIDSCENE_PLANNING_MODEL_NAME", "MIDSCENE_PLANNING_OPENAI_SOCKS_PROXY", "MIDSCENE_PLANNING_OPENAI_HTTP_PROXY", "MIDSCENE_PLANNING_OPENAI_BASE_URL", "MIDSCENE_PLANNING_OPENAI_API_KEY", "MIDSCENE_PLANNING_OPENAI_INIT_CONFIG_JSON", "MIDSCENE_PLANNING_OPENAI_USE_AZURE", "MIDSCENE_PLANNING_USE_AZURE_OPENAI", "MIDSCENE_PLANNING_AZURE_OPENAI_SCOPE", "MIDSCENE_PLANNING_AZURE_OPENAI_KEY", "MIDSCENE_PLANNING_AZURE_OPENAI_ENDPOINT", "MIDSCENE_PLANNING_AZURE_OPENAI_API_VERSION", "MIDSCENE_PLANNING_AZURE_OPENAI_DEPLOYMENT", "MIDSCENE_PLANNING_AZURE_OPENAI_INIT_CONFIG_JSON", "MIDSCENE_PLANNING_USE_ANTHROPIC_SDK", "MIDSCENE_PLANNING_ANTHROPIC_API_KEY", "MIDSCENE_PLANNING_VL_MODE", "MIDSCENE_GROUNDING_MODEL_NAME", "MIDSCENE_GROUNDING_OPENAI_SOCKS_PROXY", "MIDSCENE_GROUNDING_OPENAI_HTTP_PROXY", "MIDSCENE_GROUNDING_OPENAI_BASE_URL", "MIDSCENE_GROUNDING_OPENAI_API_KEY", "MIDSCENE_GROUNDING_OPENAI_INIT_CONFIG_JSON", "MIDSCENE_GROUNDING_OPENAI_USE_AZURE", "MIDSCENE_GROUNDING_USE_AZURE_OPENAI", "MIDSCENE_GROUNDING_AZURE_OPENAI_SCOPE", "MIDSCENE_GROUNDING_AZURE_OPENAI_KEY", "MIDSCENE_GROUNDING_AZURE_OPENAI_ENDPOINT", "MIDSCENE_GROUNDING_AZURE_OPENAI_API_VERSION", "MIDSCENE_GROUNDING_AZURE_OPENAI_DEPLOYMENT", "MIDSCENE_GROUNDING_AZURE_OPENAI_INIT_CONFIG_JSON", "MIDSCENE_GROUNDING_USE_ANTHROPIC_SDK", "MIDSCENE_GROUNDING_ANTHROPIC_API_KEY", "MIDSCENE_GROUNDING_VL_MODE"];
+export declare const MODEL_ENV_KEYS: readonly ["MIDSCENE_MODEL_NAME", "MIDSCENE_MODEL_INIT_CONFIG_JSON", "MIDSCENE_MODEL_API_KEY", "MIDSCENE_MODEL_BASE_URL", "MIDSCENE_MODEL_SOCKS_PROXY", "MIDSCENE_MODEL_HTTP_PROXY", "MIDSCENE_MODEL_TIMEOUT", "MIDSCENE_USE_VLM_UI_TARS", "MIDSCENE_USE_QWEN_VL", "MIDSCENE_USE_QWEN3_VL", "MIDSCENE_USE_DOUBAO_VISION", "MIDSCENE_USE_GEMINI", "MIDSCENE_USE_VL_MODEL", "OPENAI_API_KEY", "OPENAI_BASE_URL", "MIDSCENE_OPENAI_INIT_CONFIG_JSON", "MIDSCENE_OPENAI_HTTP_PROXY", "MIDSCENE_OPENAI_SOCKS_PROXY", "MIDSCENE_INSIGHT_MODEL_NAME", "MIDSCENE_INSIGHT_MODEL_SOCKS_PROXY", "MIDSCENE_INSIGHT_MODEL_HTTP_PROXY", "MIDSCENE_INSIGHT_MODEL_BASE_URL", "MIDSCENE_INSIGHT_MODEL_API_KEY", "MIDSCENE_INSIGHT_MODEL_INIT_CONFIG_JSON", "MIDSCENE_INSIGHT_MODEL_TIMEOUT", "MIDSCENE_PLANNING_MODEL_NAME", "MIDSCENE_PLANNING_MODEL_SOCKS_PROXY", "MIDSCENE_PLANNING_MODEL_HTTP_PROXY", "MIDSCENE_PLANNING_MODEL_BASE_URL", "MIDSCENE_PLANNING_MODEL_API_KEY", "MIDSCENE_PLANNING_MODEL_INIT_CONFIG_JSON", "MIDSCENE_PLANNING_MODEL_TIMEOUT", "MIDSCENE_MODEL_FAMILY"];
+export declare const ALL_ENV_KEYS: readonly [...string[], "MIDSCENE_DEBUG_MODE", "MIDSCENE_DEBUG_MODEL_PROFILE", "MIDSCENE_DEBUG_MODEL_RESPONSE", "MIDSCENE_RUN_DIR", "MIDSCENE_CACHE", "MIDSCENE_FORCE_DEEP_THINK", "MIDSCENE_MCP_USE_PUPPETEER_MODE", "MIDSCENE_MCP_ANDROID_MODE", "MIDSCENE_LANGSMITH_DEBUG", "MIDSCENE_LANGFUSE_DEBUG", "MIDSCENE_CACHE_MAX_FILENAME_LENGTH", "MIDSCENE_REPLANNING_CYCLE_LIMIT", "MIDSCENE_MODEL_MAX_TOKENS", "OPENAI_MAX_TOKENS", "MIDSCENE_ADB_PATH", "MIDSCENE_ADB_REMOTE_HOST", "MIDSCENE_ADB_REMOTE_PORT", "MIDSCENE_ANDROID_IME_STRATEGY", "MIDSCENE_IOS_DEVICE_UDID", "MIDSCENE_IOS_SIMULATOR_UDID", "MIDSCENE_REPORT_TAG_NAME", "MIDSCENE_PREFERRED_LANGUAGE", "MATCH_BY_POSITION", "MIDSCENE_MCP_CHROME_PATH", "DOCKER_CONTAINER", "MIDSCENE_MODEL_NAME", "MIDSCENE_MODEL_INIT_CONFIG_JSON", "MIDSCENE_MODEL_API_KEY", "MIDSCENE_MODEL_BASE_URL", "MIDSCENE_MODEL_SOCKS_PROXY", "MIDSCENE_MODEL_HTTP_PROXY", "MIDSCENE_MODEL_TIMEOUT", "MIDSCENE_USE_VLM_UI_TARS", "MIDSCENE_USE_QWEN_VL", "MIDSCENE_USE_QWEN3_VL", "MIDSCENE_USE_DOUBAO_VISION", "MIDSCENE_USE_GEMINI", "MIDSCENE_USE_VL_MODEL", "OPENAI_API_KEY", "OPENAI_BASE_URL", "MIDSCENE_OPENAI_INIT_CONFIG_JSON", "MIDSCENE_OPENAI_HTTP_PROXY", "MIDSCENE_OPENAI_SOCKS_PROXY", "MIDSCENE_INSIGHT_MODEL_NAME", "MIDSCENE_INSIGHT_MODEL_SOCKS_PROXY", "MIDSCENE_INSIGHT_MODEL_HTTP_PROXY", "MIDSCENE_INSIGHT_MODEL_BASE_URL", "MIDSCENE_INSIGHT_MODEL_API_KEY", "MIDSCENE_INSIGHT_MODEL_INIT_CONFIG_JSON", "MIDSCENE_INSIGHT_MODEL_TIMEOUT", "MIDSCENE_PLANNING_MODEL_NAME", "MIDSCENE_PLANNING_MODEL_SOCKS_PROXY", "MIDSCENE_PLANNING_MODEL_HTTP_PROXY", "MIDSCENE_PLANNING_MODEL_BASE_URL", "MIDSCENE_PLANNING_MODEL_API_KEY", "MIDSCENE_PLANNING_MODEL_INIT_CONFIG_JSON", "MIDSCENE_PLANNING_MODEL_TIMEOUT", "MIDSCENE_MODEL_FAMILY"];
 export type TEnvKeys = (typeof ALL_ENV_KEYS)[number];
 export type TGlobalConfig = Record<TEnvKeys, string | undefined>;
-export type TVlModeValues = 'qwen-vl' | 'qwen3-vl' | 'doubao-vision' | 'gemini' | 'vlm-ui-tars' | 'vlm-ui-tars-doubao' | 'vlm-ui-tars-doubao-1.5';
-export type TVlModeTypes = 'qwen-vl' | 'qwen3-vl' | 'doubao-vision' | 'gemini' | 'vlm-ui-tars';
-export interface IModelConfigForVQA {
-    [MIDSCENE_VQA_MODEL_NAME]: string;
-    [MIDSCENE_VQA_OPENAI_SOCKS_PROXY]?: string;
-    [MIDSCENE_VQA_OPENAI_HTTP_PROXY]?: string;
-    [MIDSCENE_VQA_OPENAI_BASE_URL]?: string;
-    [MIDSCENE_VQA_OPENAI_API_KEY]?: string;
-    [MIDSCENE_VQA_OPENAI_INIT_CONFIG_JSON]?: string;
-    [MIDSCENE_VQA_OPENAI_USE_AZURE]?: string;
-    [MIDSCENE_VQA_USE_AZURE_OPENAI]?: string;
-    [MIDSCENE_VQA_AZURE_OPENAI_SCOPE]?: string;
-    [MIDSCENE_VQA_AZURE_OPENAI_KEY]?: string;
-    [MIDSCENE_VQA_AZURE_OPENAI_ENDPOINT]?: string;
-    [MIDSCENE_VQA_AZURE_OPENAI_API_VERSION]?: string;
-    [MIDSCENE_VQA_AZURE_OPENAI_DEPLOYMENT]?: string;
-    [MIDSCENE_VQA_AZURE_OPENAI_INIT_CONFIG_JSON]?: string;
-    [MIDSCENE_VQA_USE_ANTHROPIC_SDK]?: string;
-    [MIDSCENE_VQA_ANTHROPIC_API_KEY]?: string;
-    [MIDSCENE_VQA_VL_MODE]?: TVlModeValues;
+export type TVlModeValues = 'qwen2.5-vl' | 'qwen3-vl' | 'doubao-vision' | 'gemini' | 'vlm-ui-tars' | 'vlm-ui-tars-doubao' | 'vlm-ui-tars-doubao-1.5';
+export type TVlModeTypes = 'qwen2.5-vl' | 'qwen3-vl' | 'doubao-vision' | 'gemini' | 'vlm-ui-tars';
+export declare const VL_MODE_RAW_VALID_VALUES: TVlModeValues[];
+/**
+ * Model family values - unified model configuration approach
+ * Replaces the old MIDSCENE_USE_* environment variables
+ *
+ * Note: These values directly correspond to VL_MODE_RAW_VALID_VALUES
+ * - 'qwen2.5-vl' is Qwen 2.5
+ * - 'qwen3-vl' is Qwen 3
+ */
+export type TModelFamily = TVlModeValues;
+export declare const MODEL_FAMILY_VALUES: TVlModeValues[];
+export interface IModelConfigForInsight {
+    [MIDSCENE_INSIGHT_MODEL_NAME]: string;
+    [MIDSCENE_INSIGHT_MODEL_SOCKS_PROXY]?: string;
+    [MIDSCENE_INSIGHT_MODEL_HTTP_PROXY]?: string;
+    [MIDSCENE_INSIGHT_MODEL_BASE_URL]?: string;
+    [MIDSCENE_INSIGHT_MODEL_API_KEY]?: string;
+    [MIDSCENE_INSIGHT_MODEL_INIT_CONFIG_JSON]?: string;
+    [MIDSCENE_INSIGHT_MODEL_TIMEOUT]?: string;
 }
 export interface IModelConfigForPlanning {
     [MIDSCENE_PLANNING_MODEL_NAME]: string;
-    [MIDSCENE_PLANNING_OPENAI_SOCKS_PROXY]?: string;
-    [MIDSCENE_PLANNING_OPENAI_HTTP_PROXY]?: string;
-    [MIDSCENE_PLANNING_OPENAI_BASE_URL]?: string;
-    [MIDSCENE_PLANNING_OPENAI_API_KEY]?: string;
-    [MIDSCENE_PLANNING_OPENAI_INIT_CONFIG_JSON]?: string;
-    [MIDSCENE_PLANNING_OPENAI_USE_AZURE]?: string;
-    [MIDSCENE_PLANNING_USE_AZURE_OPENAI]?: string;
-    [MIDSCENE_PLANNING_AZURE_OPENAI_SCOPE]?: string;
-    [MIDSCENE_PLANNING_AZURE_OPENAI_KEY]?: string;
-    [MIDSCENE_PLANNING_AZURE_OPENAI_ENDPOINT]?: string;
-    [MIDSCENE_PLANNING_AZURE_OPENAI_API_VERSION]?: string;
-    [MIDSCENE_PLANNING_AZURE_OPENAI_DEPLOYMENT]?: string;
-    [MIDSCENE_PLANNING_AZURE_OPENAI_INIT_CONFIG_JSON]?: string;
-    [MIDSCENE_PLANNING_USE_ANTHROPIC_SDK]?: string;
-    [MIDSCENE_PLANNING_ANTHROPIC_API_KEY]?: string;
-    [MIDSCENE_PLANNING_VL_MODE]?: TVlModeValues;
-}
-export interface IModeConfigForGrounding {
-    [MIDSCENE_GROUNDING_MODEL_NAME]: string;
-    [MIDSCENE_GROUNDING_OPENAI_SOCKS_PROXY]?: string;
-    [MIDSCENE_GROUNDING_OPENAI_HTTP_PROXY]?: string;
-    [MIDSCENE_GROUNDING_OPENAI_BASE_URL]?: string;
-    [MIDSCENE_GROUNDING_OPENAI_API_KEY]?: string;
-    [MIDSCENE_GROUNDING_OPENAI_INIT_CONFIG_JSON]?: string;
-    [MIDSCENE_GROUNDING_OPENAI_USE_AZURE]?: string;
-    [MIDSCENE_GROUNDING_USE_AZURE_OPENAI]?: string;
-    [MIDSCENE_GROUNDING_AZURE_OPENAI_SCOPE]?: string;
-    [MIDSCENE_GROUNDING_AZURE_OPENAI_KEY]?: string;
-    [MIDSCENE_GROUNDING_AZURE_OPENAI_ENDPOINT]?: string;
-    [MIDSCENE_GROUNDING_AZURE_OPENAI_API_VERSION]?: string;
-    [MIDSCENE_GROUNDING_AZURE_OPENAI_DEPLOYMENT]?: string;
-    [MIDSCENE_GROUNDING_AZURE_OPENAI_INIT_CONFIG_JSON]?: string;
-    [MIDSCENE_GROUNDING_USE_ANTHROPIC_SDK]?: string;
-    [MIDSCENE_GROUNDING_ANTHROPIC_API_KEY]?: string;
-    [MIDSCENE_GROUNDING_VL_MODE]?: TVlModeValues;
+    [MIDSCENE_PLANNING_MODEL_SOCKS_PROXY]?: string;
+    [MIDSCENE_PLANNING_MODEL_HTTP_PROXY]?: string;
+    [MIDSCENE_PLANNING_MODEL_BASE_URL]?: string;
+    [MIDSCENE_PLANNING_MODEL_API_KEY]?: string;
+    [MIDSCENE_PLANNING_MODEL_INIT_CONFIG_JSON]?: string;
+    [MIDSCENE_PLANNING_MODEL_TIMEOUT]?: string;
 }
+/**
+ * Model configuration for Planning intent.
+ *
+ * IMPORTANT: Planning MUST use a vision language model (VL mode).
+ * DOM-based planning is not supported.
+ *
+ * Required: MIDSCENE_MODEL_FAMILY must be set to one of:
+ *   - 'qwen2.5-vl'
+ *   - 'qwen3-vl'
+ *   - 'gemini'
+ *   - 'doubao-vision'
+ *   - 'vlm-ui-tars'
+ *   - 'vlm-ui-tars-doubao'
+ *   - 'vlm-ui-tars-doubao-1.5'
+ */
 export interface IModelConfigForDefault {
     [MIDSCENE_MODEL_NAME]: string;
-    [MIDSCENE_OPENAI_SOCKS_PROXY]?: string;
-    [MIDSCENE_OPENAI_HTTP_PROXY]?: string;
-    [MIDSCENE_OPENAI_BASE_URL]?: string;
-    [MIDSCENE_OPENAI_API_KEY]?: string;
-    [MIDSCENE_OPENAI_INIT_CONFIG_JSON]?: string;
-    [MIDSCENE_OPENAI_USE_AZURE]?: string;
-    [MIDSCENE_USE_AZURE_OPENAI]?: string;
-    [MIDSCENE_AZURE_OPENAI_SCOPE]?: string;
-    [MIDSCENE_AZURE_OPENAI_KEY]?: string;
-    [MIDSCENE_AZURE_OPENAI_ENDPOINT]?: string;
-    [MIDSCENE_AZURE_OPENAI_API_VERSION]?: string;
-    [MIDSCENE_AZURE_OPENAI_DEPLOYMENT]?: string;
-    [MIDSCENE_AZURE_OPENAI_INIT_CONFIG_JSON]?: string;
-    [MIDSCENE_USE_ANTHROPIC_SDK]?: string;
-    [MIDSCENE_ANTHROPIC_API_KEY]?: string;
-    [MIDSCENE_VL_MODE]?: TVlModeValues;
+    [MIDSCENE_MODEL_SOCKS_PROXY]?: string;
+    [MIDSCENE_MODEL_HTTP_PROXY]?: string;
+    [MIDSCENE_MODEL_BASE_URL]?: string;
+    [MIDSCENE_MODEL_API_KEY]?: string;
+    [MIDSCENE_MODEL_INIT_CONFIG_JSON]?: string;
+    [MIDSCENE_MODEL_FAMILY]?: TVlModeValues;
 }
 export interface IModelConfigForDefaultLegacy {
     [MIDSCENE_MODEL_NAME]: string;
@@ -217,35 +164,51 @@ export interface IModelConfigForDefaultLegacy {
     [OPENAI_BASE_URL]?: string;
     [OPENAI_API_KEY]?: string;
     [MIDSCENE_OPENAI_INIT_CONFIG_JSON]?: string;
-    [OPENAI_USE_AZURE]?: string;
-    [MIDSCENE_USE_AZURE_OPENAI]?: string;
-    [MIDSCENE_AZURE_OPENAI_SCOPE]?: string;
-    [AZURE_OPENAI_KEY]?: string;
-    [AZURE_OPENAI_ENDPOINT]?: string;
-    [AZURE_OPENAI_API_VERSION]?: string;
-    [AZURE_OPENAI_DEPLOYMENT]?: string;
-    [MIDSCENE_AZURE_OPENAI_INIT_CONFIG_JSON]?: string;
-    [MIDSCENE_USE_ANTHROPIC_SDK]?: string;
-    [ANTHROPIC_API_KEY]?: string;
-    [MIDSCENE_VL_MODE]?: TVlModeValues;
 }
 /**
- * - VQA: Visual Question Answering
- * - grounding：short for Visual Grounding
+ * - insight: Visual Question Answering and Visual Grounding (unified)
  * - planning: planning
- * - default: all except VQA、grounding、planning
+ * - default: all except insight、planning
  */
-export type TIntent = 'VQA' | 'planning' | 'grounding' | 'default';
-export type TModelConfigFn = (options: {
-    intent: TIntent;
-}) => IModelConfigForVQA | IModelConfigForPlanning | IModeConfigForGrounding | IModelConfigForDefault;
+export type TIntent = 'insight' | 'planning' | 'default';
+/**
+ * Env-style model configuration map supplied directly to the agent.
+ * Numbers are allowed so callers can pass numeric env values (e.g. limits) without casting.
+ */
+export type TModelConfig = Record<string, string | number>;
 export declare enum UITarsModelVersion {
     V1_0 = "1.0",
     V1_5 = "1.5",
     DOUBAO_1_5_15B = "doubao-1.5-15B",
     DOUBAO_1_5_20B = "doubao-1.5-20B"
 }
-export declare const VL_MODE_RAW_VALID_VALUES: TVlModeValues[];
+/**
+ * Callback to create custom OpenAI client instance
+ * @param config - Resolved model configuration including apiKey, baseURL, modelName, intent, etc.
+ * @returns OpenAI client instance (can be wrapped with langsmith, langfuse, etc.)
+ *
+ * Note: Wrapper functions like langsmith's wrapOpenAI() return the same OpenAI instance
+ * with enhanced behavior, so the return type remains compatible with OpenAI.
+ *
+ * Note: The return type is `any` in the shared package to avoid requiring openai as a dependency.
+ * The actual implementation should return an OpenAI instance.
+ *
+ * @example
+ * ```typescript
+ * import OpenAI from 'openai';
+ * import { wrapOpenAI } from 'langsmith/wrappers';
+ *
+ * createOpenAIClient: async (openai, opts) => {
+ *   // Wrap with langsmith for planning tasks
+ *   if (opts.baseURL?.includes('planning')) {
+ *     return wrapOpenAI(openai, { metadata: { task: 'planning' } });
+ *   }
+ *
+ *   return openai;
+ * }
+ * ```
+ */
+export type CreateOpenAIClientFn = (openAIInstance: any, options: Record<string, unknown>) => Promise<any>;
 export interface IModelConfig {
     /**
      * proxy
@@ -263,32 +226,30 @@ export interface IModelConfig {
     openaiApiKey?: string;
     openaiExtraConfig?: Record<string, unknown>;
     /**
-     * Azure
-     */
-    openaiUseAzureDeprecated?: boolean;
-    useAzureOpenai?: boolean;
-    azureOpenaiScope?: string;
-    azureOpenaiKey?: string;
-    azureOpenaiEndpoint?: string;
-    azureOpenaiApiVersion?: string;
-    azureOpenaiDeployment?: string;
-    azureExtraConfig?: Record<string, unknown>;
-    /**
-     * Anthropic
+     * Timeout for API calls in milliseconds.
+     * If not set, uses OpenAI SDK default (10 minutes).
      */
-    useAnthropicSdk?: boolean;
-    anthropicApiKey?: string;
+    timeout?: number;
     /**
-     * - vlModeRaw: exists only in non-legacy logic. value can be 'doubao-vision', 'gemini', 'qwen-vl', 'vlm-ui-tars', 'vlm-ui-tars-doubao', 'vlm-ui-tars-doubao-1.5'
-     * - vlMode: based on the results of the vlModoRaw classification，value can be 'doubao-vision', 'gemini', 'qwen-vl', 'vlm-ui-tars'
+     * - vlModeRaw: exists only in non-legacy logic. value can be 'doubao-vision', 'gemini', 'qwen2.5-vl', 'vlm-ui-tars', 'vlm-ui-tars-doubao', 'vlm-ui-tars-doubao-1.5'
+     * - vlMode: based on the results of the vlModoRaw classification，value can be 'doubao-vision', 'gemini', 'qwen2.5-vl', 'vlm-ui-tars'
      */
     vlModeRaw?: string;
     vlMode?: TVlModeTypes;
     uiTarsModelVersion?: UITarsModelVersion;
     modelDescription: string;
     /**
-     * for debug
+     * original intent from the config
      */
     intent: TIntent;
-    from: 'modelConfig' | 'env' | 'legacy-env';
+    /**
+     * Custom OpenAI client factory function
+     *
+     * If provided, this function will be called to create OpenAI client instances
+     * for each AI call, allowing you to:
+     * - Wrap clients with observability tools (langsmith, langfuse)
+     * - Use custom OpenAI-compatible clients
+     * - Apply different configurations based on intent
+     */
+    createOpenAIClient?: CreateOpenAIClientFn;
 }

package/dist/types/extractor/dom-util.d.ts

@@ -1,4 +1,4 @@
-import { NodeType } from '../constants';
+import type { LocateResultElement } from '../types';
 export declare function isFormElement(node: globalThis.Node): boolean;
 export declare function isButtonElement(node: globalThis.Node): node is globalThis.HTMLButtonElement;
 export declare function isAElement(node: globalThis.Node): node is globalThis.HTMLButtonElement;
@@ -10,17 +10,4 @@ export declare function isContainerElement(node: globalThis.Node): node is globa
 export declare function generateElementByPosition(position: {
     x: number;
     y: number;
-}): {
-    id: string;
-    attributes: {
-        nodeType: NodeType;
-    };
-    rect: {
-        left: number;
-        top: number;
-        width: number;
-        height: number;
-    };
-    content: string;
-    center: number[];
-};
+}, description: string): LocateResultElement;

package/dist/types/extractor/index.d.ts

@@ -27,7 +27,6 @@ export { descriptionOfTree, traverseTree, treeToList, truncateText, trimAttribut
 export { extractTextWithPosition as webExtractTextWithPosition } from './web-extractor';
 export { extractTreeNode as webExtractNodeTree } from './web-extractor';
 export { extractTreeNodeAsString as webExtractNodeTreeAsString } from './web-extractor';
-export { setNodeHashCacheListOnWindow, getNodeFromCacheList } from './util';
-export { getXpathsById, getXpathsByPoint, getNodeInfoByXpath, getElementInfoByXpath, getElementXpath, } from './locator';
+export { getXpathsByPoint, getNodeInfoByXpath, getElementInfoByXpath, getElementXpath, } from './locator';
 export { generateElementByPosition } from './dom-util';
 export { isNotContainerElement } from './dom-util';

package/dist/types/extractor/locator.d.ts

@@ -1,7 +1,6 @@
 import type { ElementInfo } from '.';
 import type { Point } from '../types';
 export declare const getElementXpath: (element: Node, isOrderSensitive?: boolean, isLeafElement?: boolean) => string;
-export declare function getXpathsById(id: string): string[] | null;
 export declare function getXpathsByPoint(point: Point, isOrderSensitive: boolean): string[] | null;
 export declare function getNodeInfoByXpath(xpath: string): Node | null;
 export declare function getElementInfoByXpath(xpath: string): ElementInfo | null;

package/dist/types/extractor/tree.d.ts

@@ -1,9 +1,6 @@
 import type { BaseElement, ElementTreeNode } from '../types';
 export declare function truncateText(text: string | number | object | undefined, maxLength?: number): string;
-export declare function trimAttributes(attributes: Record<string, any>, truncateTextLength?: number): {
-    [key: string]: string;
-    nodeType: import("./constants").NodeType;
-};
+export declare function trimAttributes(attributes: Record<string, any>, truncateTextLength?: number): Record<string, string>;
 export declare function descriptionOfTree<ElementType extends BaseElement = BaseElement>(tree: ElementTreeNode<ElementType>, truncateTextLength?: number, filterNonTextContent?: boolean, visibleOnly?: boolean): string;
 export declare function treeToList<T extends BaseElement>(tree: ElementTreeNode<T>): T[];
 export declare function traverseTree<T extends BaseElement, ReturnNodeType extends BaseElement>(tree: ElementTreeNode<T>, onNode: (node: T) => ReturnNodeType): ElementTreeNode<ReturnNodeType>;

package/dist/types/extractor/util.d.ts

@@ -33,9 +33,6 @@ export declare function elementRect(el: globalThis.HTMLElement | globalThis.Node
 export declare function validTextNodeContent(node: globalThis.Node): string | false;
 export declare function getNodeAttributes(node: globalThis.HTMLElement | globalThis.Node, currentWindow: typeof globalThis.window): Record<string, string>;
 export declare function midsceneGenerateHash(node: globalThis.Node | null, content: string, rect: Rect): string;
-export declare function setNodeHashCacheListOnWindow(): void;
-export declare function setNodeToCacheList(node: globalThis.Node, id: string): void;
-export declare function getNodeFromCacheList(id: string): any;
 export declare function generateId(numberId: number): string;
 export declare function setGenerateHashOnWindow(): void;
 export declare function setMidsceneVisibleRectOnWindow(): void;

package/dist/types/index.d.ts

@@ -1,3 +1,4 @@
 export { createCopyStaticPlugin, createPlaygroundCopyPlugin, } from './build/copy-static';
+export { commonIgnoreWarnings } from './build/rspack-config';
 declare const _default: {};
 export default _default;

package/dist/types/mcp/base-server.d.ts

@@ -0,0 +1,77 @@
+import type { ParseArgsConfig } from 'node:util';
+import { McpServer } from '@modelcontextprotocol/sdk/server/mcp.js';
+import type { IMidsceneTools } from './types';
+export interface BaseMCPServerConfig {
+    name: string;
+    version: string;
+    description: string;
+}
+export interface HttpLaunchOptions {
+    port: number;
+    host?: string;
+}
+/**
+ * CLI argument configuration for MCP servers
+ */
+export declare const CLI_ARGS_CONFIG: ParseArgsConfig['options'];
+export interface CLIArgs {
+    mode?: string;
+    port?: string;
+    host?: string;
+}
+/**
+ * Launch an MCP server based on CLI arguments
+ * Shared helper to reduce duplication across platform CLI entry points
+ */
+export declare function launchMCPServer(server: BaseMCPServer, args: CLIArgs): Promise<void>;
+/**
+ * Base MCP Server class with programmatic launch() API
+ * Each platform extends this to provide their own tools manager
+ */
+export declare abstract class BaseMCPServer {
+    protected mcpServer: McpServer;
+    protected toolsManager?: IMidsceneTools;
+    protected config: BaseMCPServerConfig;
+    constructor(config: BaseMCPServerConfig);
+    /**
+     * Platform-specific: create tools manager instance
+     */
+    protected abstract createToolsManager(): IMidsceneTools;
+    /**
+     * Initialize tools manager and attach to MCP server
+     */
+    private initializeToolsManager;
+    /**
+     * Perform cleanup on shutdown
+     */
+    private performCleanup;
+    /**
+     * Initialize and launch the MCP server with stdio transport
+     */
+    launch(): Promise<void>;
+    /**
+     * Launch MCP server with HTTP transport
+     * Supports stateful sessions for web applications and service integration
+     */
+    launchHttp(options: HttpLaunchOptions): Promise<void>;
+    /**
+     * Create a new HTTP session with transport
+     */
+    private createHttpSession;
+    /**
+     * Start periodic session cleanup for inactive sessions
+     */
+    private startSessionCleanup;
+    /**
+     * Setup shutdown handlers for HTTP server
+     */
+    private setupHttpShutdownHandlers;
+    /**
+     * Get the underlying MCP server instance
+     */
+    getServer(): McpServer;
+    /**
+     * Get the tools manager instance
+     */
+    getToolsManager(): IMidsceneTools | undefined;
+}