@sqaitech/shared 0.5.0

package/src/env/types.ts

@@ -0,0 +1,560 @@
+// config keys
+export const SQAI_OPENAI_INIT_CONFIG_JSON =
+  'SQAI_OPENAI_INIT_CONFIG_JSON';
+export const SQAI_MODEL_NAME = 'SQAI_MODEL_NAME';
+export const SQAI_LANGSMITH_DEBUG = 'SQAI_LANGSMITH_DEBUG';
+export const SQAI_DEBUG_AI_PROFILE = 'SQAI_DEBUG_AI_PROFILE';
+export const SQAI_DEBUG_AI_RESPONSE = 'SQAI_DEBUG_AI_RESPONSE';
+export const SQAI_DANGEROUSLY_PRINT_ALL_CONFIG =
+  'SQAI_DANGEROUSLY_PRINT_ALL_CONFIG';
+export const SQAI_DEBUG_MODE = 'SQAI_DEBUG_MODE';
+export const SQAI_MCP_USE_PUPPETEER_MODE =
+  'SQAI_MCP_USE_PUPPETEER_MODE';
+export const SQAI_MCP_CHROME_PATH = 'SQAI_MCP_CHROME_PATH';
+export const SQAI_MCP_ANDROID_MODE = 'SQAI_MCP_ANDROID_MODE';
+export const DOCKER_CONTAINER = 'DOCKER_CONTAINER';
+export const SQAI_FORCE_DEEP_THINK = 'SQAI_FORCE_DEEP_THINK';
+
+export const SQAI_OPENAI_SOCKS_PROXY = 'SQAI_OPENAI_SOCKS_PROXY';
+export const SQAI_OPENAI_HTTP_PROXY = 'SQAI_OPENAI_HTTP_PROXY';
+export const OPENAI_API_KEY = 'OPENAI_API_KEY';
+export const OPENAI_BASE_URL = 'OPENAI_BASE_URL';
+export const OPENAI_MAX_TOKENS = 'OPENAI_MAX_TOKENS';
+
+export const SQAI_ADB_PATH = 'SQAI_ADB_PATH';
+export const SQAI_ADB_REMOTE_HOST = 'SQAI_ADB_REMOTE_HOST';
+export const SQAI_ADB_REMOTE_PORT = 'SQAI_ADB_REMOTE_PORT';
+export const SQAI_ANDROID_IME_STRATEGY = 'SQAI_ANDROID_IME_STRATEGY';
+
+export const SQAI_IOS_DEVICE_UDID = 'SQAI_IOS_DEVICE_UDID';
+export const SQAI_IOS_SIMULATOR_UDID = 'SQAI_IOS_SIMULATOR_UDID';
+
+export const SQAI_CACHE = 'SQAI_CACHE';
+export const SQAI_USE_VLM_UI_TARS = 'SQAI_USE_VLM_UI_TARS';
+export const SQAI_USE_QWEN_VL = 'SQAI_USE_QWEN_VL';
+export const SQAI_USE_QWEN3_VL = 'SQAI_USE_QWEN3_VL';
+export const SQAI_USE_DOUBAO_VISION = 'SQAI_USE_DOUBAO_VISION';
+export const SQAI_USE_GEMINI = 'SQAI_USE_GEMINI';
+export const SQAI_USE_VL_MODEL = 'SQAI_USE_VL_MODEL';
+export const MATCH_BY_POSITION = 'MATCH_BY_POSITION';
+export const SQAI_API_TYPE = 'SQAI-API-TYPE';
+export const SQAI_REPORT_TAG_NAME = 'SQAI_REPORT_TAG_NAME';
+
+export const SQAI_REPLANNING_CYCLE_LIMIT =
+  'SQAI_REPLANNING_CYCLE_LIMIT';
+
+export const SQAI_PREFERRED_LANGUAGE = 'SQAI_PREFERRED_LANGUAGE';
+
+export const SQAI_USE_AZURE_OPENAI = 'SQAI_USE_AZURE_OPENAI';
+export const SQAI_AZURE_OPENAI_SCOPE = 'SQAI_AZURE_OPENAI_SCOPE';
+export const SQAI_AZURE_OPENAI_INIT_CONFIG_JSON =
+  'SQAI_AZURE_OPENAI_INIT_CONFIG_JSON';
+
+export const SQAI_CACHE_MAX_FILENAME_LENGTH =
+  'SQAI_CACHE_MAX_FILENAME_LENGTH';
+
+export const AZURE_OPENAI_ENDPOINT = 'AZURE_OPENAI_ENDPOINT';
+export const AZURE_OPENAI_KEY = 'AZURE_OPENAI_KEY';
+export const AZURE_OPENAI_API_VERSION = 'AZURE_OPENAI_API_VERSION';
+export const AZURE_OPENAI_DEPLOYMENT = 'AZURE_OPENAI_DEPLOYMENT';
+
+export const SQAI_USE_ANTHROPIC_SDK = 'SQAI_USE_ANTHROPIC_SDK';
+export const ANTHROPIC_API_KEY = 'ANTHROPIC_API_KEY';
+
+export const SQAI_RUN_DIR = 'SQAI_RUN_DIR';
+
+// default new
+export const SQAI_OPENAI_BASE_URL = 'SQAI_OPENAI_BASE_URL';
+export const SQAI_OPENAI_API_KEY = 'SQAI_OPENAI_API_KEY';
+export const SQAI_OPENAI_USE_AZURE = 'SQAI_OPENAI_USE_AZURE';
+export const SQAI_AZURE_OPENAI_KEY = 'SQAI_AZURE_OPENAI_KEY';
+export const SQAI_AZURE_OPENAI_ENDPOINT = 'SQAI_AZURE_OPENAI_ENDPOINT';
+export const SQAI_AZURE_OPENAI_API_VERSION =
+  'SQAI_AZURE_OPENAI_API_VERSION';
+export const SQAI_AZURE_OPENAI_DEPLOYMENT =
+  'SQAI_AZURE_OPENAI_DEPLOYMENT';
+export const SQAI_ANTHROPIC_API_KEY = 'SQAI_ANTHROPIC_API_KEY';
+export const SQAI_VL_MODE = 'SQAI_VL_MODE';
+
+// VQA
+export const SQAI_VQA_MODEL_NAME = 'SQAI_VQA_MODEL_NAME';
+export const SQAI_VQA_OPENAI_SOCKS_PROXY =
+  'SQAI_VQA_OPENAI_SOCKS_PROXY';
+export const SQAI_VQA_OPENAI_HTTP_PROXY = 'SQAI_VQA_OPENAI_HTTP_PROXY';
+export const SQAI_VQA_OPENAI_BASE_URL = 'SQAI_VQA_OPENAI_BASE_URL';
+export const SQAI_VQA_OPENAI_API_KEY = 'SQAI_VQA_OPENAI_API_KEY';
+export const SQAI_VQA_OPENAI_INIT_CONFIG_JSON =
+  'SQAI_VQA_OPENAI_INIT_CONFIG_JSON';
+export const SQAI_VQA_OPENAI_USE_AZURE = 'SQAI_VQA_OPENAI_USE_AZURE';
+export const SQAI_VQA_USE_AZURE_OPENAI = 'SQAI_VQA_USE_AZURE_OPENAI';
+export const SQAI_VQA_AZURE_OPENAI_SCOPE =
+  'SQAI_VQA_AZURE_OPENAI_SCOPE';
+export const SQAI_VQA_AZURE_OPENAI_KEY = 'SQAI_VQA_AZURE_OPENAI_KEY';
+export const SQAI_VQA_AZURE_OPENAI_ENDPOINT =
+  'SQAI_VQA_AZURE_OPENAI_ENDPOINT';
+export const SQAI_VQA_AZURE_OPENAI_API_VERSION =
+  'SQAI_VQA_AZURE_OPENAI_API_VERSION';
+export const SQAI_VQA_AZURE_OPENAI_DEPLOYMENT =
+  'SQAI_VQA_AZURE_OPENAI_DEPLOYMENT';
+export const SQAI_VQA_AZURE_OPENAI_INIT_CONFIG_JSON =
+  'SQAI_VQA_AZURE_OPENAI_INIT_CONFIG_JSON';
+export const SQAI_VQA_USE_ANTHROPIC_SDK = 'SQAI_VQA_USE_ANTHROPIC_SDK';
+export const SQAI_VQA_ANTHROPIC_API_KEY = 'SQAI_VQA_ANTHROPIC_API_KEY';
+export const SQAI_VQA_VL_MODE = 'SQAI_VQA_VL_MODE';
+
+// PLANNING
+export const SQAI_PLANNING_MODEL_NAME = 'SQAI_PLANNING_MODEL_NAME';
+export const SQAI_PLANNING_OPENAI_SOCKS_PROXY =
+  'SQAI_PLANNING_OPENAI_SOCKS_PROXY';
+export const SQAI_PLANNING_OPENAI_HTTP_PROXY =
+  'SQAI_PLANNING_OPENAI_HTTP_PROXY';
+export const SQAI_PLANNING_OPENAI_BASE_URL =
+  'SQAI_PLANNING_OPENAI_BASE_URL';
+export const SQAI_PLANNING_OPENAI_API_KEY =
+  'SQAI_PLANNING_OPENAI_API_KEY';
+export const SQAI_PLANNING_OPENAI_INIT_CONFIG_JSON =
+  'SQAI_PLANNING_OPENAI_INIT_CONFIG_JSON';
+export const SQAI_PLANNING_OPENAI_USE_AZURE =
+  'SQAI_PLANNING_OPENAI_USE_AZURE';
+export const SQAI_PLANNING_USE_AZURE_OPENAI =
+  'SQAI_PLANNING_USE_AZURE_OPENAI';
+export const SQAI_PLANNING_AZURE_OPENAI_SCOPE =
+  'SQAI_PLANNING_AZURE_OPENAI_SCOPE';
+export const SQAI_PLANNING_AZURE_OPENAI_KEY =
+  'SQAI_PLANNING_AZURE_OPENAI_KEY';
+export const SQAI_PLANNING_AZURE_OPENAI_ENDPOINT =
+  'SQAI_PLANNING_AZURE_OPENAI_ENDPOINT';
+export const SQAI_PLANNING_AZURE_OPENAI_API_VERSION =
+  'SQAI_PLANNING_AZURE_OPENAI_API_VERSION';
+export const SQAI_PLANNING_AZURE_OPENAI_DEPLOYMENT =
+  'SQAI_PLANNING_AZURE_OPENAI_DEPLOYMENT';
+export const SQAI_PLANNING_AZURE_OPENAI_INIT_CONFIG_JSON =
+  'SQAI_PLANNING_AZURE_OPENAI_INIT_CONFIG_JSON';
+export const SQAI_PLANNING_USE_ANTHROPIC_SDK =
+  'SQAI_PLANNING_USE_ANTHROPIC_SDK';
+export const SQAI_PLANNING_ANTHROPIC_API_KEY =
+  'SQAI_PLANNING_ANTHROPIC_API_KEY';
+export const SQAI_PLANNING_VL_MODE = 'SQAI_PLANNING_VL_MODE';
+
+// GROUNDING
+export const SQAI_GROUNDING_MODEL_NAME = 'SQAI_GROUNDING_MODEL_NAME';
+export const SQAI_GROUNDING_OPENAI_SOCKS_PROXY =
+  'SQAI_GROUNDING_OPENAI_SOCKS_PROXY';
+export const SQAI_GROUNDING_OPENAI_HTTP_PROXY =
+  'SQAI_GROUNDING_OPENAI_HTTP_PROXY';
+export const SQAI_GROUNDING_OPENAI_BASE_URL =
+  'SQAI_GROUNDING_OPENAI_BASE_URL';
+export const SQAI_GROUNDING_OPENAI_API_KEY =
+  'SQAI_GROUNDING_OPENAI_API_KEY';
+export const SQAI_GROUNDING_OPENAI_INIT_CONFIG_JSON =
+  'SQAI_GROUNDING_OPENAI_INIT_CONFIG_JSON';
+export const SQAI_GROUNDING_OPENAI_USE_AZURE =
+  'SQAI_GROUNDING_OPENAI_USE_AZURE';
+export const SQAI_GROUNDING_USE_AZURE_OPENAI =
+  'SQAI_GROUNDING_USE_AZURE_OPENAI';
+export const SQAI_GROUNDING_AZURE_OPENAI_SCOPE =
+  'SQAI_GROUNDING_AZURE_OPENAI_SCOPE';
+export const SQAI_GROUNDING_AZURE_OPENAI_KEY =
+  'SQAI_GROUNDING_AZURE_OPENAI_KEY';
+export const SQAI_GROUNDING_AZURE_OPENAI_ENDPOINT =
+  'SQAI_GROUNDING_AZURE_OPENAI_ENDPOINT';
+export const SQAI_GROUNDING_AZURE_OPENAI_API_VERSION =
+  'SQAI_GROUNDING_AZURE_OPENAI_API_VERSION';
+export const SQAI_GROUNDING_AZURE_OPENAI_DEPLOYMENT =
+  'SQAI_GROUNDING_AZURE_OPENAI_DEPLOYMENT';
+export const SQAI_GROUNDING_AZURE_OPENAI_INIT_CONFIG_JSON =
+  'SQAI_GROUNDING_AZURE_OPENAI_INIT_CONFIG_JSON';
+export const SQAI_GROUNDING_USE_ANTHROPIC_SDK =
+  'SQAI_GROUNDING_USE_ANTHROPIC_SDK';
+export const SQAI_GROUNDING_ANTHROPIC_API_KEY =
+  'SQAI_GROUNDING_ANTHROPIC_API_KEY';
+export const SQAI_GROUNDING_VL_MODE = 'SQAI_GROUNDING_VL_MODE';
+
+// @deprecated
+export const OPENAI_USE_AZURE = 'OPENAI_USE_AZURE';
+
+/**
+ * env keys declared but unused
+ */
+export const UNUSED_ENV_KEYS = [SQAI_DANGEROUSLY_PRINT_ALL_CONFIG];
+
+/**
+ * env keys for debug or basic run
+ * can not be override by overrideAIConfig
+ */
+export const BASIC_ENV_KEYS = [
+  SQAI_DEBUG_MODE,
+  SQAI_DEBUG_AI_PROFILE,
+  SQAI_DEBUG_AI_RESPONSE,
+  SQAI_RUN_DIR,
+] as const;
+
+export const BOOLEAN_ENV_KEYS = [
+  SQAI_CACHE,
+  SQAI_LANGSMITH_DEBUG,
+  SQAI_FORCE_DEEP_THINK,
+  SQAI_MCP_USE_PUPPETEER_MODE,
+  SQAI_MCP_ANDROID_MODE,
+] as const;
+
+export const NUMBER_ENV_KEYS = [
+  SQAI_CACHE_MAX_FILENAME_LENGTH,
+  SQAI_REPLANNING_CYCLE_LIMIT,
+] as const;
+
+export const STRING_ENV_KEYS = [
+  OPENAI_MAX_TOKENS,
+  SQAI_ADB_PATH,
+  SQAI_ADB_REMOTE_HOST,
+  SQAI_ADB_REMOTE_PORT,
+  SQAI_ANDROID_IME_STRATEGY,
+  SQAI_IOS_DEVICE_UDID,
+  SQAI_IOS_SIMULATOR_UDID,
+  SQAI_REPORT_TAG_NAME,
+  SQAI_PREFERRED_LANGUAGE,
+  MATCH_BY_POSITION,
+  SQAI_MCP_CHROME_PATH,
+  DOCKER_CONTAINER,
+] as const;
+
+/**
+ * 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 const GLOBAL_ENV_KEYS = [
+  ...BOOLEAN_ENV_KEYS,
+  ...NUMBER_ENV_KEYS,
+  ...STRING_ENV_KEYS,
+] as const;
+
+/**
+ * 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 const MODEL_ENV_KEYS = [
+  // model default
+  SQAI_MODEL_NAME,
+  SQAI_OPENAI_INIT_CONFIG_JSON,
+  SQAI_OPENAI_API_KEY,
+  SQAI_OPENAI_BASE_URL,
+  SQAI_OPENAI_USE_AZURE,
+  SQAI_OPENAI_SOCKS_PROXY,
+  SQAI_OPENAI_HTTP_PROXY,
+  SQAI_USE_AZURE_OPENAI,
+  SQAI_AZURE_OPENAI_SCOPE,
+  SQAI_AZURE_OPENAI_INIT_CONFIG_JSON,
+  SQAI_USE_ANTHROPIC_SDK,
+  SQAI_USE_VLM_UI_TARS,
+  SQAI_USE_QWEN_VL,
+  SQAI_USE_QWEN3_VL,
+  SQAI_USE_DOUBAO_VISION,
+  SQAI_USE_GEMINI,
+  SQAI_USE_VL_MODEL,
+  ANTHROPIC_API_KEY,
+  SQAI_AZURE_OPENAI_ENDPOINT,
+  SQAI_AZURE_OPENAI_KEY,
+  SQAI_AZURE_OPENAI_API_VERSION,
+  SQAI_AZURE_OPENAI_DEPLOYMENT,
+  SQAI_VL_MODE,
+  // model default legacy
+  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,
+  // VQA
+  SQAI_VQA_MODEL_NAME,
+  SQAI_VQA_OPENAI_SOCKS_PROXY,
+  SQAI_VQA_OPENAI_HTTP_PROXY,
+  SQAI_VQA_OPENAI_BASE_URL,
+  SQAI_VQA_OPENAI_API_KEY,
+  SQAI_VQA_OPENAI_INIT_CONFIG_JSON,
+  SQAI_VQA_OPENAI_USE_AZURE,
+  SQAI_VQA_USE_AZURE_OPENAI,
+  SQAI_VQA_AZURE_OPENAI_SCOPE,
+  SQAI_VQA_AZURE_OPENAI_KEY,
+  SQAI_VQA_AZURE_OPENAI_ENDPOINT,
+  SQAI_VQA_AZURE_OPENAI_API_VERSION,
+  SQAI_VQA_AZURE_OPENAI_DEPLOYMENT,
+  SQAI_VQA_AZURE_OPENAI_INIT_CONFIG_JSON,
+  SQAI_VQA_USE_ANTHROPIC_SDK,
+  SQAI_VQA_ANTHROPIC_API_KEY,
+  SQAI_VQA_VL_MODE,
+  // PLANNING
+  SQAI_PLANNING_MODEL_NAME,
+  SQAI_PLANNING_OPENAI_SOCKS_PROXY,
+  SQAI_PLANNING_OPENAI_HTTP_PROXY,
+  SQAI_PLANNING_OPENAI_BASE_URL,
+  SQAI_PLANNING_OPENAI_API_KEY,
+  SQAI_PLANNING_OPENAI_INIT_CONFIG_JSON,
+  SQAI_PLANNING_OPENAI_USE_AZURE,
+  SQAI_PLANNING_USE_AZURE_OPENAI,
+  SQAI_PLANNING_AZURE_OPENAI_SCOPE,
+  SQAI_PLANNING_AZURE_OPENAI_KEY,
+  SQAI_PLANNING_AZURE_OPENAI_ENDPOINT,
+  SQAI_PLANNING_AZURE_OPENAI_API_VERSION,
+  SQAI_PLANNING_AZURE_OPENAI_DEPLOYMENT,
+  SQAI_PLANNING_AZURE_OPENAI_INIT_CONFIG_JSON,
+  SQAI_PLANNING_USE_ANTHROPIC_SDK,
+  SQAI_PLANNING_ANTHROPIC_API_KEY,
+  SQAI_PLANNING_VL_MODE,
+  // GROUNDING
+  SQAI_GROUNDING_MODEL_NAME,
+  SQAI_GROUNDING_OPENAI_SOCKS_PROXY,
+  SQAI_GROUNDING_OPENAI_HTTP_PROXY,
+  SQAI_GROUNDING_OPENAI_BASE_URL,
+  SQAI_GROUNDING_OPENAI_API_KEY,
+  SQAI_GROUNDING_OPENAI_INIT_CONFIG_JSON,
+  SQAI_GROUNDING_OPENAI_USE_AZURE,
+  SQAI_GROUNDING_USE_AZURE_OPENAI,
+  SQAI_GROUNDING_AZURE_OPENAI_SCOPE,
+  SQAI_GROUNDING_AZURE_OPENAI_KEY,
+  SQAI_GROUNDING_AZURE_OPENAI_ENDPOINT,
+  SQAI_GROUNDING_AZURE_OPENAI_API_VERSION,
+  SQAI_GROUNDING_AZURE_OPENAI_DEPLOYMENT,
+  SQAI_GROUNDING_AZURE_OPENAI_INIT_CONFIG_JSON,
+  SQAI_GROUNDING_USE_ANTHROPIC_SDK,
+  SQAI_GROUNDING_ANTHROPIC_API_KEY,
+  SQAI_GROUNDING_VL_MODE,
+] as const;
+
+export const ALL_ENV_KEYS = [
+  ...UNUSED_ENV_KEYS,
+  ...BASIC_ENV_KEYS,
+  ...GLOBAL_ENV_KEYS,
+  ...MODEL_ENV_KEYS,
+] as const;
+
+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 {
+  // model name
+  [SQAI_VQA_MODEL_NAME]: string;
+  // proxy
+  [SQAI_VQA_OPENAI_SOCKS_PROXY]?: string;
+  [SQAI_VQA_OPENAI_HTTP_PROXY]?: string;
+  // OpenAI
+  [SQAI_VQA_OPENAI_BASE_URL]?: string;
+  [SQAI_VQA_OPENAI_API_KEY]?: string;
+  [SQAI_VQA_OPENAI_INIT_CONFIG_JSON]?: string;
+  // Azure
+  [SQAI_VQA_OPENAI_USE_AZURE]?: string;
+  [SQAI_VQA_USE_AZURE_OPENAI]?: string;
+  [SQAI_VQA_AZURE_OPENAI_SCOPE]?: string;
+  [SQAI_VQA_AZURE_OPENAI_KEY]?: string;
+  [SQAI_VQA_AZURE_OPENAI_ENDPOINT]?: string;
+  [SQAI_VQA_AZURE_OPENAI_API_VERSION]?: string;
+  [SQAI_VQA_AZURE_OPENAI_DEPLOYMENT]?: string;
+  [SQAI_VQA_AZURE_OPENAI_INIT_CONFIG_JSON]?: string;
+  // Anthropic
+  [SQAI_VQA_USE_ANTHROPIC_SDK]?: string;
+  [SQAI_VQA_ANTHROPIC_API_KEY]?: string;
+  // extra
+  [SQAI_VQA_VL_MODE]?: TVlModeValues;
+}
+
+export interface IModelConfigForPlanning {
+  // model name
+  [SQAI_PLANNING_MODEL_NAME]: string;
+  // proxy
+  [SQAI_PLANNING_OPENAI_SOCKS_PROXY]?: string;
+  [SQAI_PLANNING_OPENAI_HTTP_PROXY]?: string;
+  // OpenAI
+  [SQAI_PLANNING_OPENAI_BASE_URL]?: string;
+  [SQAI_PLANNING_OPENAI_API_KEY]?: string;
+  [SQAI_PLANNING_OPENAI_INIT_CONFIG_JSON]?: string;
+  // Azure
+  [SQAI_PLANNING_OPENAI_USE_AZURE]?: string;
+  [SQAI_PLANNING_USE_AZURE_OPENAI]?: string;
+  [SQAI_PLANNING_AZURE_OPENAI_SCOPE]?: string;
+  [SQAI_PLANNING_AZURE_OPENAI_KEY]?: string;
+  [SQAI_PLANNING_AZURE_OPENAI_ENDPOINT]?: string;
+  [SQAI_PLANNING_AZURE_OPENAI_API_VERSION]?: string;
+  [SQAI_PLANNING_AZURE_OPENAI_DEPLOYMENT]?: string;
+  [SQAI_PLANNING_AZURE_OPENAI_INIT_CONFIG_JSON]?: string;
+  // Anthropic
+  [SQAI_PLANNING_USE_ANTHROPIC_SDK]?: string;
+  [SQAI_PLANNING_ANTHROPIC_API_KEY]?: string;
+  // extra
+  [SQAI_PLANNING_VL_MODE]?: TVlModeValues;
+}
+
+export interface IModeConfigForGrounding {
+  // model name
+  [SQAI_GROUNDING_MODEL_NAME]: string;
+  // proxy
+  [SQAI_GROUNDING_OPENAI_SOCKS_PROXY]?: string;
+  [SQAI_GROUNDING_OPENAI_HTTP_PROXY]?: string;
+  // OpenAI
+  [SQAI_GROUNDING_OPENAI_BASE_URL]?: string;
+  [SQAI_GROUNDING_OPENAI_API_KEY]?: string;
+  [SQAI_GROUNDING_OPENAI_INIT_CONFIG_JSON]?: string;
+  // Azure
+  [SQAI_GROUNDING_OPENAI_USE_AZURE]?: string;
+  [SQAI_GROUNDING_USE_AZURE_OPENAI]?: string;
+  [SQAI_GROUNDING_AZURE_OPENAI_SCOPE]?: string;
+  [SQAI_GROUNDING_AZURE_OPENAI_KEY]?: string;
+  [SQAI_GROUNDING_AZURE_OPENAI_ENDPOINT]?: string;
+  [SQAI_GROUNDING_AZURE_OPENAI_API_VERSION]?: string;
+  [SQAI_GROUNDING_AZURE_OPENAI_DEPLOYMENT]?: string;
+  [SQAI_GROUNDING_AZURE_OPENAI_INIT_CONFIG_JSON]?: string;
+  // Anthropic
+  [SQAI_GROUNDING_USE_ANTHROPIC_SDK]?: string;
+  [SQAI_GROUNDING_ANTHROPIC_API_KEY]?: string;
+  // extra
+  [SQAI_GROUNDING_VL_MODE]?: TVlModeValues;
+}
+
+export interface IModelConfigForDefault {
+  // model name
+  [SQAI_MODEL_NAME]: string;
+  // proxy
+  [SQAI_OPENAI_SOCKS_PROXY]?: string;
+  [SQAI_OPENAI_HTTP_PROXY]?: string;
+  // OpenAI
+  [SQAI_OPENAI_BASE_URL]?: string;
+  [SQAI_OPENAI_API_KEY]?: string;
+  [SQAI_OPENAI_INIT_CONFIG_JSON]?: string;
+  // Azure
+  [SQAI_OPENAI_USE_AZURE]?: string;
+  [SQAI_USE_AZURE_OPENAI]?: string;
+  [SQAI_AZURE_OPENAI_SCOPE]?: string;
+  [SQAI_AZURE_OPENAI_KEY]?: string;
+  [SQAI_AZURE_OPENAI_ENDPOINT]?: string;
+  [SQAI_AZURE_OPENAI_API_VERSION]?: string;
+  [SQAI_AZURE_OPENAI_DEPLOYMENT]?: string;
+  [SQAI_AZURE_OPENAI_INIT_CONFIG_JSON]?: string;
+  // Anthropic
+  [SQAI_USE_ANTHROPIC_SDK]?: string;
+  [SQAI_ANTHROPIC_API_KEY]?: string;
+  // extra
+  [SQAI_VL_MODE]?: TVlModeValues;
+}
+
+export interface IModelConfigForDefaultLegacy {
+  // model name
+  [SQAI_MODEL_NAME]: string;
+  // proxy
+  [SQAI_OPENAI_SOCKS_PROXY]?: string;
+  [SQAI_OPENAI_HTTP_PROXY]?: string;
+  // OpenAI
+  [OPENAI_BASE_URL]?: string;
+  [OPENAI_API_KEY]?: string;
+  [SQAI_OPENAI_INIT_CONFIG_JSON]?: string;
+  // Azure
+  [OPENAI_USE_AZURE]?: string;
+  [SQAI_USE_AZURE_OPENAI]?: string;
+  [SQAI_AZURE_OPENAI_SCOPE]?: string;
+  [AZURE_OPENAI_KEY]?: string;
+  [AZURE_OPENAI_ENDPOINT]?: string;
+  [AZURE_OPENAI_API_VERSION]?: string;
+  [AZURE_OPENAI_DEPLOYMENT]?: string;
+  [SQAI_AZURE_OPENAI_INIT_CONFIG_JSON]?: string;
+  // Anthropic
+  [SQAI_USE_ANTHROPIC_SDK]?: string;
+  [ANTHROPIC_API_KEY]?: string;
+  // extra
+  [SQAI_VL_MODE]?: TVlModeValues;
+}
+
+/**
+ * - VQA: Visual Question Answering
+ * - grounding：short for Visual Grounding
+ * - planning: planning
+ * - default: all except VQA、grounding、planning
+ */
+export type TIntent = 'VQA' | 'planning' | 'grounding' | 'default';
+
+export type TModelConfigFn = (options: {
+  intent: TIntent;
+}) =>
+  | IModelConfigForVQA
+  | IModelConfigForPlanning
+  | IModeConfigForGrounding
+  | IModelConfigForDefault;
+
+export 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 const VL_MODE_RAW_VALID_VALUES: TVlModeValues[] = [
+  'doubao-vision',
+  'gemini',
+  'qwen-vl',
+  'qwen3-vl',
+  'vlm-ui-tars',
+  'vlm-ui-tars-doubao',
+  'vlm-ui-tars-doubao-1.5',
+];
+
+export interface IModelConfig {
+  /**
+   * proxy
+   */
+  socksProxy?: string;
+  httpProxy?: string;
+  /**
+   * model
+   */
+  modelName: string;
+  /**
+   * OpenAI
+   */
+  openaiBaseURL?: string;
+  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
+   */
+  useAnthropicSdk?: boolean;
+  anthropicApiKey?: string;
+  /**
+   * - 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?: string;
+  vlMode?: TVlModeTypes;
+  uiTarsModelVersion?: UITarsModelVersion;
+  modelDescription: string;
+  /**
+   * for debug
+   */
+  intent: TIntent;
+  from: 'modelConfig' | 'env' | 'legacy-env';
+}

package/src/env/utils.ts

@@ -0,0 +1,39 @@
+import { GlobalConfigManager } from './global-config-manager';
+import { ModelConfigManager } from './model-config-manager';
+import {
+  type GLOBAL_ENV_KEYS,
+  SQAI_PREFERRED_LANGUAGE,
+  type MODEL_ENV_KEYS,
+} from './types';
+
+export const globalModelConfigManager = new ModelConfigManager();
+
+export const globalConfigManager = new GlobalConfigManager();
+
+globalConfigManager.registerModelConfigManager(globalModelConfigManager);
+globalModelConfigManager.registerGlobalConfigManager(globalConfigManager);
+
+export const getPreferredLanguage = () => {
+  const prefer = globalConfigManager.getEnvConfigValue(
+    SQAI_PREFERRED_LANGUAGE,
+  );
+  if (prefer) {
+    return prefer;
+  }
+
+  const timeZone = Intl.DateTimeFormat().resolvedOptions().timeZone;
+  const isChina = timeZone === 'Asia/Shanghai';
+  return isChina ? 'Chinese' : 'English';
+};
+
+export const overrideAIConfig = (
+  newConfig: Partial<
+    Record<
+      (typeof GLOBAL_ENV_KEYS)[number] | (typeof MODEL_ENV_KEYS)[number],
+      string
+    >
+  >,
+  extendMode = false, // true: merge with global config, false: override global config
+) => {
+  globalConfigManager.overrideAIConfig(newConfig, extendMode);
+};

package/src/extractor/constants.ts

@@ -0,0 +1,5 @@
+export {
+  NodeType,
+  TEXT_MAX_SIZE,
+  TEXT_SIZE_THRESHOLD,
+} from '../constants/index';

package/src/extractor/debug.ts

@@ -0,0 +1,10 @@
+import { webExtractTextWithPosition } from '.';
+import {
+  setExtractTextWithPositionOnWindow,
+  setMidsceneVisibleRectOnWindow,
+} from './util';
+
+console.log(webExtractTextWithPosition(document.body, true));
+console.log(JSON.stringify(webExtractTextWithPosition(document.body, true)));
+setExtractTextWithPositionOnWindow();
+setMidsceneVisibleRectOnWindow();