@godscene/shared 1.7.11

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

@@ -0,0 +1,339 @@
+export declare const MIDSCENE_MODEL_INIT_CONFIG_JSON = "MIDSCENE_MODEL_INIT_CONFIG_JSON";
+export declare const MIDSCENE_MODEL_EXTRA_BODY_JSON = "MIDSCENE_MODEL_EXTRA_BODY_JSON";
+export declare const MIDSCENE_MODEL_NAME = "MIDSCENE_MODEL_NAME";
+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";
+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_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";
+export declare const MIDSCENE_MODEL_TEMPERATURE = "MIDSCENE_MODEL_TEMPERATURE";
+export declare const MIDSCENE_MODEL_RETRY_COUNT = "MIDSCENE_MODEL_RETRY_COUNT";
+export declare const MIDSCENE_MODEL_RETRY_INTERVAL = "MIDSCENE_MODEL_RETRY_INTERVAL";
+export declare const MIDSCENE_MODEL_REASONING_EFFORT = "MIDSCENE_MODEL_REASONING_EFFORT";
+export declare const MIDSCENE_MODEL_REASONING_ENABLED = "MIDSCENE_MODEL_REASONING_ENABLED";
+export declare const MIDSCENE_MODEL_REASONING_BUDGET = "MIDSCENE_MODEL_REASONING_BUDGET";
+/**
+ * @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";
+export declare const MIDSCENE_ADB_REMOTE_PORT = "MIDSCENE_ADB_REMOTE_PORT";
+export declare const MIDSCENE_ANDROID_IME_STRATEGY = "MIDSCENE_ANDROID_IME_STRATEGY";
+export declare const MIDSCENE_IOS_DEVICE_UDID = "MIDSCENE_IOS_DEVICE_UDID";
+export declare const MIDSCENE_IOS_SIMULATOR_UDID = "MIDSCENE_IOS_SIMULATOR_UDID";
+export declare const MIDSCENE_IOS_DEVICE_CLASS_OVERRIDE = "MIDSCENE_IOS_DEVICE_CLASS_OVERRIDE";
+export declare const MIDSCENE_CACHE = "MIDSCENE_CACHE";
+export declare const MIDSCENE_USE_VLM_UI_TARS = "MIDSCENE_USE_VLM_UI_TARS";
+export declare const MIDSCENE_USE_QWEN_VL = "MIDSCENE_USE_QWEN_VL";
+export declare const MIDSCENE_USE_QWEN3_VL = "MIDSCENE_USE_QWEN3_VL";
+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_REPORT_TAG_NAME = "MIDSCENE_REPORT_TAG_NAME";
+export declare const MIDSCENE_REPORT_QUIET = "MIDSCENE_REPORT_QUIET";
+export declare const MIDSCENE_PREFERRED_LANGUAGE = "MIDSCENE_PREFERRED_LANGUAGE";
+export declare const MIDSCENE_CACHE_MAX_FILENAME_LENGTH = "MIDSCENE_CACHE_MAX_FILENAME_LENGTH";
+export declare const MIDSCENE_REPLANNING_CYCLE_LIMIT = "MIDSCENE_REPLANNING_CYCLE_LIMIT";
+export declare const MIDSCENE_RUN_DIR = "MIDSCENE_RUN_DIR";
+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_EXTRA_BODY_JSON = "MIDSCENE_INSIGHT_MODEL_EXTRA_BODY_JSON";
+export declare const MIDSCENE_INSIGHT_MODEL_TIMEOUT = "MIDSCENE_INSIGHT_MODEL_TIMEOUT";
+export declare const MIDSCENE_INSIGHT_MODEL_TEMPERATURE = "MIDSCENE_INSIGHT_MODEL_TEMPERATURE";
+export declare const MIDSCENE_INSIGHT_MODEL_RETRY_COUNT = "MIDSCENE_INSIGHT_MODEL_RETRY_COUNT";
+export declare const MIDSCENE_INSIGHT_MODEL_RETRY_INTERVAL = "MIDSCENE_INSIGHT_MODEL_RETRY_INTERVAL";
+export declare const MIDSCENE_INSIGHT_MODEL_FAMILY = "MIDSCENE_INSIGHT_MODEL_FAMILY";
+export declare const MIDSCENE_INSIGHT_MODEL_REASONING_EFFORT = "MIDSCENE_INSIGHT_MODEL_REASONING_EFFORT";
+export declare const MIDSCENE_INSIGHT_MODEL_REASONING_ENABLED = "MIDSCENE_INSIGHT_MODEL_REASONING_ENABLED";
+export declare const MIDSCENE_INSIGHT_MODEL_REASONING_BUDGET = "MIDSCENE_INSIGHT_MODEL_REASONING_BUDGET";
+export declare const YHT_DOMAIN = "YHT_DOMAIN";
+export declare const YHT_ACCESS_TOKEN = "YHT_ACCESS_TOKEN";
+export declare const YHT_MODEL = "YHT_MODEL";
+export declare const YHT_BASE_URL = "YHT_BASE_URL";
+export declare const YHT_CHAT_TYPE = "YHT_CHAT_TYPE";
+export declare const YHT_MODEL_CATEGORY = "YHT_MODEL_CATEGORY";
+export declare const YHT_STREAM = "YHT_STREAM";
+export declare const YHT_TEMPERATURE = "YHT_TEMPERATURE";
+export declare const YHT_TOP_P = "YHT_TOP_P";
+export declare const YHT_TRACE_ID = "YHT_TRACE_ID";
+export declare const MIDSCENE_PLANNING_MODEL_NAME = "MIDSCENE_PLANNING_MODEL_NAME";
+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_EXTRA_BODY_JSON = "MIDSCENE_PLANNING_MODEL_EXTRA_BODY_JSON";
+export declare const MIDSCENE_PLANNING_MODEL_TIMEOUT = "MIDSCENE_PLANNING_MODEL_TIMEOUT";
+export declare const MIDSCENE_PLANNING_MODEL_TEMPERATURE = "MIDSCENE_PLANNING_MODEL_TEMPERATURE";
+export declare const MIDSCENE_PLANNING_MODEL_RETRY_COUNT = "MIDSCENE_PLANNING_MODEL_RETRY_COUNT";
+export declare const MIDSCENE_PLANNING_MODEL_RETRY_INTERVAL = "MIDSCENE_PLANNING_MODEL_RETRY_INTERVAL";
+export declare const MIDSCENE_PLANNING_MODEL_FAMILY = "MIDSCENE_PLANNING_MODEL_FAMILY";
+export declare const MIDSCENE_PLANNING_MODEL_REASONING_EFFORT = "MIDSCENE_PLANNING_MODEL_REASONING_EFFORT";
+export declare const MIDSCENE_PLANNING_MODEL_REASONING_ENABLED = "MIDSCENE_PLANNING_MODEL_REASONING_ENABLED";
+export declare const MIDSCENE_PLANNING_MODEL_REASONING_BUDGET = "MIDSCENE_PLANNING_MODEL_REASONING_BUDGET";
+export declare const MIDSCENE_MODEL_FAMILY = "MIDSCENE_MODEL_FAMILY";
+/**
+ * env keys declared but unused
+ */
+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_MODEL_PROFILE", "MIDSCENE_DEBUG_MODEL_RESPONSE", "MIDSCENE_RUN_DIR"];
+export declare const BOOLEAN_ENV_KEYS: readonly ["MIDSCENE_CACHE", "MIDSCENE_MCP_USE_PUPPETEER_MODE", "MIDSCENE_MCP_ANDROID_MODE", "MIDSCENE_LANGSMITH_DEBUG", "MIDSCENE_LANGFUSE_DEBUG", "MIDSCENE_REPORT_QUIET"];
+export declare const NUMBER_ENV_KEYS: readonly ["MIDSCENE_MODEL_MAX_TOKENS", "MIDSCENE_CACHE_MAX_FILENAME_LENGTH", "MIDSCENE_REPLANNING_CYCLE_LIMIT"];
+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_MCP_USE_PUPPETEER_MODE", "MIDSCENE_MCP_ANDROID_MODE", "MIDSCENE_LANGSMITH_DEBUG", "MIDSCENE_LANGFUSE_DEBUG", "MIDSCENE_REPORT_QUIET", "MIDSCENE_MODEL_MAX_TOKENS", "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_MODEL_INIT_CONFIG_JSON", "MIDSCENE_MODEL_EXTRA_BODY_JSON", "MIDSCENE_MODEL_API_KEY", "MIDSCENE_MODEL_BASE_URL", "MIDSCENE_MODEL_SOCKS_PROXY", "MIDSCENE_MODEL_HTTP_PROXY", "MIDSCENE_MODEL_TIMEOUT", "MIDSCENE_MODEL_TEMPERATURE", "MIDSCENE_MODEL_RETRY_COUNT", "MIDSCENE_MODEL_RETRY_INTERVAL", "MIDSCENE_MODEL_REASONING_EFFORT", "MIDSCENE_MODEL_REASONING_ENABLED", "MIDSCENE_MODEL_REASONING_BUDGET", "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_EXTRA_BODY_JSON", "MIDSCENE_INSIGHT_MODEL_TIMEOUT", "MIDSCENE_INSIGHT_MODEL_TEMPERATURE", "MIDSCENE_INSIGHT_MODEL_RETRY_COUNT", "MIDSCENE_INSIGHT_MODEL_RETRY_INTERVAL", "MIDSCENE_INSIGHT_MODEL_FAMILY", "MIDSCENE_INSIGHT_MODEL_REASONING_EFFORT", "MIDSCENE_INSIGHT_MODEL_REASONING_ENABLED", "MIDSCENE_INSIGHT_MODEL_REASONING_BUDGET", "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_EXTRA_BODY_JSON", "MIDSCENE_PLANNING_MODEL_TIMEOUT", "MIDSCENE_PLANNING_MODEL_TEMPERATURE", "MIDSCENE_PLANNING_MODEL_RETRY_COUNT", "MIDSCENE_PLANNING_MODEL_RETRY_INTERVAL", "MIDSCENE_PLANNING_MODEL_FAMILY", "MIDSCENE_PLANNING_MODEL_REASONING_EFFORT", "MIDSCENE_PLANNING_MODEL_REASONING_ENABLED", "MIDSCENE_PLANNING_MODEL_REASONING_BUDGET", "MIDSCENE_MODEL_FAMILY", "YHT_DOMAIN", "YHT_ACCESS_TOKEN", "YHT_MODEL", "YHT_BASE_URL", "YHT_CHAT_TYPE", "YHT_MODEL_CATEGORY", "YHT_STREAM", "YHT_TEMPERATURE", "YHT_TOP_P"];
+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_MCP_USE_PUPPETEER_MODE", "MIDSCENE_MCP_ANDROID_MODE", "MIDSCENE_LANGSMITH_DEBUG", "MIDSCENE_LANGFUSE_DEBUG", "MIDSCENE_REPORT_QUIET", "MIDSCENE_MODEL_MAX_TOKENS", "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_EXTRA_BODY_JSON", "MIDSCENE_MODEL_API_KEY", "MIDSCENE_MODEL_BASE_URL", "MIDSCENE_MODEL_SOCKS_PROXY", "MIDSCENE_MODEL_HTTP_PROXY", "MIDSCENE_MODEL_TIMEOUT", "MIDSCENE_MODEL_TEMPERATURE", "MIDSCENE_MODEL_RETRY_COUNT", "MIDSCENE_MODEL_RETRY_INTERVAL", "MIDSCENE_MODEL_REASONING_EFFORT", "MIDSCENE_MODEL_REASONING_ENABLED", "MIDSCENE_MODEL_REASONING_BUDGET", "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_EXTRA_BODY_JSON", "MIDSCENE_INSIGHT_MODEL_TIMEOUT", "MIDSCENE_INSIGHT_MODEL_TEMPERATURE", "MIDSCENE_INSIGHT_MODEL_RETRY_COUNT", "MIDSCENE_INSIGHT_MODEL_RETRY_INTERVAL", "MIDSCENE_INSIGHT_MODEL_FAMILY", "MIDSCENE_INSIGHT_MODEL_REASONING_EFFORT", "MIDSCENE_INSIGHT_MODEL_REASONING_ENABLED", "MIDSCENE_INSIGHT_MODEL_REASONING_BUDGET", "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_EXTRA_BODY_JSON", "MIDSCENE_PLANNING_MODEL_TIMEOUT", "MIDSCENE_PLANNING_MODEL_TEMPERATURE", "MIDSCENE_PLANNING_MODEL_RETRY_COUNT", "MIDSCENE_PLANNING_MODEL_RETRY_INTERVAL", "MIDSCENE_PLANNING_MODEL_FAMILY", "MIDSCENE_PLANNING_MODEL_REASONING_EFFORT", "MIDSCENE_PLANNING_MODEL_REASONING_ENABLED", "MIDSCENE_PLANNING_MODEL_REASONING_BUDGET", "MIDSCENE_MODEL_FAMILY", "YHT_DOMAIN", "YHT_ACCESS_TOKEN", "YHT_MODEL", "YHT_BASE_URL", "YHT_CHAT_TYPE", "YHT_MODEL_CATEGORY", "YHT_STREAM", "YHT_TEMPERATURE", "YHT_TOP_P"];
+export type TEnvKeys = (typeof ALL_ENV_KEYS)[number];
+export type TGlobalConfig = Record<TEnvKeys, string | undefined>;
+/**
+ * valid Model family types
+ */
+export type TModelFamily = 'qwen2.5-vl' | 'qwen3-vl' | 'qwen3.5' | 'qwen3.6' | 'doubao-vision' | 'doubao-seed' | 'gemini' | 'vlm-ui-tars' | 'vlm-ui-tars-doubao' | 'vlm-ui-tars-doubao-1.5' | 'glm-v' | 'auto-glm' | 'auto-glm-multilingual' | 'gpt-5';
+export declare const MODEL_FAMILY_VALUES: TModelFamily[];
+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_EXTRA_BODY_JSON]?: string;
+    [MIDSCENE_INSIGHT_MODEL_TIMEOUT]?: string;
+    [MIDSCENE_INSIGHT_MODEL_TEMPERATURE]?: string;
+    [MIDSCENE_INSIGHT_MODEL_FAMILY]?: TModelFamily;
+    [YHT_DOMAIN]?: string;
+    [YHT_ACCESS_TOKEN]?: string;
+    [YHT_MODEL]?: string;
+    [YHT_BASE_URL]?: string;
+    [YHT_CHAT_TYPE]?: string;
+    [YHT_MODEL_CATEGORY]?: string;
+    [YHT_STREAM]?: string;
+    [YHT_TEMPERATURE]?: string;
+    [YHT_TOP_P]?: string;
+}
+export interface IModelConfigForPlanning {
+    [MIDSCENE_PLANNING_MODEL_NAME]: string;
+    [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_EXTRA_BODY_JSON]?: string;
+    [MIDSCENE_PLANNING_MODEL_TIMEOUT]?: string;
+    [MIDSCENE_PLANNING_MODEL_TEMPERATURE]?: string;
+    [MIDSCENE_PLANNING_MODEL_FAMILY]?: TModelFamily;
+}
+/**
+ * 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 「TModelFamily」
+ */
+export interface IModelConfigForDefault {
+    [MIDSCENE_MODEL_NAME]: string;
+    [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_EXTRA_BODY_JSON]?: string;
+    [MIDSCENE_MODEL_FAMILY]?: TModelFamily;
+    [MIDSCENE_MODEL_TEMPERATURE]?: string;
+    [MIDSCENE_MODEL_REASONING_EFFORT]?: string;
+    [MIDSCENE_MODEL_REASONING_ENABLED]?: string;
+    [MIDSCENE_MODEL_REASONING_BUDGET]?: string;
+}
+export interface IModelConfigForDefaultLegacy {
+    [MIDSCENE_MODEL_NAME]: string;
+    [MIDSCENE_OPENAI_SOCKS_PROXY]?: string;
+    [MIDSCENE_OPENAI_HTTP_PROXY]?: string;
+    [OPENAI_BASE_URL]?: string;
+    [OPENAI_API_KEY]?: string;
+    [MIDSCENE_OPENAI_INIT_CONFIG_JSON]?: string;
+}
+/**
+ * - insight: Visual Question Answering and Visual Grounding (unified)
+ * - planning: planning
+ * - default: all except insight、planning
+ */
+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"
+}
+/**
+ * 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
+     */
+    socksProxy?: string;
+    httpProxy?: string;
+    /**
+     * model
+     */
+    modelName: string;
+    /**
+     * OpenAI
+     */
+    openaiBaseURL?: string;
+    openaiApiKey?: string;
+    openaiExtraConfig?: Record<string, unknown>;
+    /**
+     * Extra body parameters merged into each chat completion request body.
+     * Unlike openaiExtraConfig (which configures the OpenAI client instance),
+     * this is spread directly into the completion.create() call body.
+     * Example: { "chat_template_kwargs": { "enable_thinking": true } }
+     */
+    extraBody?: Record<string, unknown>;
+    /**
+     * Timeout for API calls in milliseconds.
+     * If not set, uses OpenAI SDK default (10 minutes).
+     */
+    timeout?: number;
+    /**
+     * Temperature for model sampling.
+     */
+    temperature?: number;
+    /**
+     * Number of retries when AI call fails.
+     * Default is 1 (retry once after failure).
+     */
+    retryCount?: number;
+    /**
+     * Interval between retries in milliseconds.
+     * Default is 2000.
+     */
+    retryInterval?: number;
+    /**
+     * Reasoning effort level for the model.
+     * Passed through to model-family-specific parameters (e.g., reasoning_effort for doubao, reasoning.effort for gpt-5).
+     */
+    reasoningEffort?: string;
+    /**
+     * Enable/disable reasoning for the model.
+     * Passed through to model-family-specific parameters (e.g., enable_thinking for qwen, thinking.type for doubao/glm-v).
+     */
+    reasoningEnabled?: boolean;
+    /**
+     * Reasoning token budget for the model.
+     * Passed through to model-family-specific parameters (e.g., thinking_budget for qwen).
+     */
+    reasoningBudget?: number;
+    /**
+     * Model family - unified model configuration
+     * Maps directly to model families like 'qwen2.5-vl', 'qwen3-vl', 'doubao-vision', 'doubao-seed', etc.
+     */
+    modelFamily?: TModelFamily;
+    uiTarsModelVersion?: UITarsModelVersion;
+    modelDescription: string;
+    /**
+     * original intent from the config
+     */
+    intent: TIntent;
+    /**
+     * 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;
+    yht_domain?: string;
+    yht_access_token?: string;
+    yht_model?: string;
+    yht_base_url?: string;
+    yht_chat_type?: string;
+    yht_model_category?: string;
+    yht_stream?: string;
+    yht_temperature?: string;
+    yht_top_p?: string;
+    yht_trace_id?: string;
+}

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

@@ -0,0 +1,7 @@
+import { GlobalConfigManager } from './global-config-manager';
+import { ModelConfigManager } from './model-config-manager';
+import { type GLOBAL_ENV_KEYS, type MODEL_ENV_KEYS } from './types';
+export declare const globalModelConfigManager: ModelConfigManager;
+export declare const globalConfigManager: GlobalConfigManager;
+export declare const getPreferredLanguage: () => string;
+export declare const overrideAIConfig: (newConfig: Partial<Record<(typeof GLOBAL_ENV_KEYS)[number] | (typeof MODEL_ENV_KEYS)[number], string>>, extendMode?: boolean) => void;

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

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

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

@@ -0,0 +1,2 @@
+export declare const postWindowMessage: (targetWin: any, sourceWin: any, payload: any, targetOrigin?: any, timeout?: any) => Promise<any>;
+export declare const onWindowMessage: (win: any, fn: any) => () => any;

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

@@ -0,0 +1,69 @@
+/**
+ *
+ * 直到遇到fieldid或id为止
+ * @param {*} dom
+ * @param {*} onece
+ * @returns
+ */
+declare const getLocatorUntilFieldid: (dom: any, type?: string) => {
+    type: string;
+    value: any;
+};
+/**
+ * 全路径，适用于静态元素，当元素动态改变，而节点位置不变，可能不准确，比如百度的搜索结果，位置可能发生改变，指向一个固定的项
+ * 实现浏览器自带的Copy fullXPath方法
+ * 一直递归到 /html
+ * 忽略id
+ * 其他同上XPath
+ * @param {*} dom
+ * @returns string
+ */
+declare const getFullLocator: (dom: any, type?: string) => {
+    type: string;
+    value: any;
+};
+/**
+ * 对外的方法，依次调用上面的方法，生成可以的列表
+ * return []
+ * */
+declare const getLocators: (dom: any) => any[];
+/**
+ * 对外的方法，iframe在页面数量少，不需要全路径，要简洁
+ * @param {*} dom
+ */
+declare const getContainerLocatorData: (dom: any) => {
+    containerAllPaths: any[];
+    containerType: string;
+    containerPathType: any;
+    containerPathValue: any;
+    containerPathIndex: any;
+};
+/**
+ * 获取多个元素
+ * @param {*} locators
+ * @param {*} docOrHost
+ * @param {*} selected
+ * @returns
+ */
+declare const getManyElementByLocators: (docOrHost: Document | undefined, locators: any[], selected?: boolean) => unknown[];
+/**
+ * 获取单个元素
+ * @param {*} locators
+ * @param {*} docOrHost
+ * @returns
+ */
+declare const getOneElementByLocators: (docOrHost: any, locators: any[], selected?: boolean) => {} | null;
+/**
+ * 录制时先获取shadowRoot的容器树，再添加frame的容器树
+ * 获取shadowRoot的容器树
+ * @param {*} dom
+ * @returns
+ */
+declare const getShadowRootContainerStack: (dom: any) => {
+    containerAllPaths: any[];
+    containerType: string;
+    containerPathType: any;
+    containerPathValue: any;
+    containerPathIndex: any;
+}[];
+export { getLocators, getShadowRootContainerStack, getFullLocator, getLocatorUntilFieldid, getContainerLocatorData as getContainerPath, getManyElementByLocators as getElements, getOneElementByLocators as getUniqueElement, };

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

@@ -0,0 +1 @@
+export {};

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

@@ -0,0 +1,57 @@
+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;
+export declare function isSvgElement(node: globalThis.Node): node is globalThis.SVGSVGElement;
+export declare function isImgElement(node: globalThis.Node): node is globalThis.HTMLImageElement;
+export declare function isNotContainerElement(node: globalThis.Node): boolean;
+export declare function isTextElement(node: globalThis.Node): node is globalThis.HTMLTextAreaElement;
+export declare function isContainerElement(node: globalThis.Node): node is globalThis.HTMLElement;
+/**
+ * Generate a LocateResultElement from a point.
+ * This function creates an expanded rect around the given center point.
+ *
+ * Note: Center coordinates should be integers for pixel-aligned positioning.
+ * If decimal values are provided, they will be used as-is, which may result in
+ * non-pixel-aligned rect positions.
+ *
+ * The rect positioning behavior:
+ * - When edgeSize is even: center is at the top-left of the four center pixels
+ *   For example, with edgeSize=4 and center=[10, 10]:
+ *   □□□□
+ *   □■□□  (■ represents the center point at pixel 10)
+ *   □□□□
+ *   □□□□
+ *
+ * - When edgeSize is odd: center is at the exact middle pixel
+ *   For example, with edgeSize=5 and center=[10, 10]:
+ *   □□□□□
+ *   □□■□□  (■ represents the center point at pixel 10)
+ *   □□□□□
+ *
+ * @param center - Center point coordinates as [x, y] (should be integers)
+ * @param description - Description of the element
+ * @param edgeSize - Size to expand around the center point (default: 8)
+ * @returns A LocateResultElement with rect, center, and description
+ */
+export declare function generateElementByPoint(center: [number, number], description: string, edgeSize?: number): LocateResultElement;
+/**
+ * Generate a LocateResultElement from a rect.
+ * This function calculates the center point from the rect and preserves the
+ * original rect as the returned element boundary.
+ *
+ * Note: The rect uses inclusive coordinates where:
+ * - A rect from [left=10, top=10] with [width=1, height=1] covers exactly 1 pixel
+ * - The actual pixel range is [left, left+width) which means width pixels
+ *
+ * @param sourceRect - The source rect to generate element from (typically contains integer values)
+ * @param description - Description of the element
+ * @param edgeSize - Deprecated, retained for backward compatibility
+ * @returns A LocateResultElement with the original rect, center (always integers), and description
+ */
+export declare function generateElementByRect(sourceRect: {
+    left: number;
+    top: number;
+    width: number;
+    height: number;
+}, description: string, _edgeSize?: number): LocateResultElement;

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

@@ -0,0 +1,33 @@
+import type { NodeType } from '../constants/index';
+export interface ElementInfo {
+    id: string;
+    indexId: number;
+    nodeHashId: string;
+    xpaths?: string[];
+    attributes: {
+        nodeType: NodeType;
+        [key: string]: string;
+    };
+    nodeType: NodeType;
+    content: string;
+    rect: {
+        left: number;
+        top: number;
+        width: number;
+        height: number;
+    };
+    center: [number, number];
+    isVisible: boolean;
+}
+export interface ElementNode {
+    node: ElementInfo | null;
+    children: ElementNode[];
+}
+export { descriptionOfTree, traverseTree, treeToList, truncateText, trimAttributes, } from './tree';
+export { extractTextWithPosition as webExtractTextWithPosition } from './web-extractor';
+export { extractTreeNode as webExtractNodeTree } from './web-extractor';
+export { extractTreeNodeAsString as webExtractNodeTreeAsString } from './web-extractor';
+export { getElementFromPoint as webGetElementFromPoint } from './web-extractor';
+export { getXpathsByPoint, getXpathsById, getNodeInfoByXpath, getElementInfoByXpath, getElementXpath, } from './locator';
+export { generateElementByPoint, generateElementByRect } from './dom-util';
+export { isNotContainerElement } from './dom-util';

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

@@ -0,0 +1,9 @@
+import type { ElementInfo } from '.';
+import type { Point } from '../types';
+export declare const getElementXpath: (element: Node, isOrderSensitive?: boolean, isLeafElement?: boolean, limitToCurrentDocument?: boolean) => string;
+/** Retrieve XPath for a previously cached node by its hash ID.
+ *  Returns a local xpath within the node's own document (limitToCurrentDocument=true). */
+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

@@ -0,0 +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): 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

@@ -0,0 +1,47 @@
+import type { Rect } from '../types';
+export declare function setDebugMode(mode: boolean): void;
+export declare function getDebugMode(): boolean;
+export declare function logger(..._msg: any[]): void;
+export declare function isElementPartiallyInViewport(rect: ReturnType<typeof getRect>, currentWindow: typeof window, currentDocument: typeof document, visibleAreaRatio?: number): boolean;
+export declare function getPseudoElementContent(element: globalThis.Node, currentWindow: typeof globalThis.window): {
+    before: string;
+    after: string;
+};
+export declare function hasOverflowY(element: globalThis.HTMLElement, currentWindow: typeof globalThis.window): boolean;
+export interface ExtractedRect {
+    width: number;
+    height: number;
+    left: number;
+    top: number;
+    right: number;
+    bottom: number;
+    x: number;
+    y: number;
+    zoom: number;
+}
+export declare function overlappedRect(rect1: ExtractedRect, rect2: ExtractedRect): ExtractedRect | null;
+export declare function getRect(el: globalThis.HTMLElement | globalThis.Node, baseZoom: number, // base zoom
+currentWindow: typeof globalThis.window): ExtractedRect;
+export declare function elementRect(el: globalThis.HTMLElement | globalThis.Node | null, currentWindow: typeof globalThis.window, currentDocument: typeof globalThis.document, baseZoom?: number): {
+    left: number;
+    top: number;
+    width: number;
+    height: number;
+    zoom: number;
+    isVisible: boolean;
+} | false;
+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>;
+/**
+ * Reset the node hash cache. Call at the beginning of each extraction cycle
+ * to prevent stale DOM references from accumulating.
+ */
+export declare function setNodeHashCacheListOnWindow(): void;
+export declare function setNodeToCacheList(node: globalThis.Node, id: string): void;
+export declare function getNodeFromCacheList(id: string): globalThis.Node | undefined;
+export declare function midsceneGenerateHash(node: globalThis.Node | null, content: string, rect: Rect): string;
+export declare function generateId(numberId: number): string;
+export declare function setGenerateHashOnWindow(): void;
+export declare function setMidsceneVisibleRectOnWindow(): void;
+export declare function setExtractTextWithPositionOnWindow(): void;
+export declare function getTopDocument(): globalThis.HTMLElement;

package/dist/types/extractor/web-extractor.d.ts

@@ -0,0 +1,24 @@
+import type { WebElementInfo } from '../types';
+import type { Point } from '../types';
+export declare function collectElementInfo(node: Node, currentWindow: typeof window, currentDocument: typeof document, baseZoom?: number, basePoint?: Point, isContainer?: boolean): WebElementInfo | null | any;
+interface WebElementNode {
+    node: WebElementInfo | null;
+    children: WebElementNode[];
+}
+export declare function extractTextWithPosition(initNode: globalThis.Node, debugMode?: boolean): Promise<WebElementInfo[]>;
+export declare function extractTreeNodeAsString(initNode: globalThis.Node, visibleOnly?: boolean, debugMode?: boolean): Promise<string>;
+export declare function extractTreeNode(initNode: globalThis.Node, debugMode?: boolean, basePoint?: Point, baseZoom?: number): Promise<WebElementNode>;
+export declare function mergeElementAndChildrenRects(node: Node, currentWindow: typeof window, currentDocument: typeof document, baseZoom?: number): {
+    left: number;
+    top: number;
+    width: number;
+    height: number;
+    zoom: number;
+    isVisible: boolean;
+} | null;
+export declare const getElementFromPoint: (args: {
+    x: number;
+    y: number;
+    containerPaths: string[];
+}) => Promise<any>;
+export {};

package/dist/types/img/box-select.d.ts

@@ -0,0 +1,26 @@
+import type { BaseElement, Rect } from '../types';
+interface ElementForOverlay {
+    rect: Rect;
+    indexId?: number;
+}
+export declare const compositeElementInfoImg: (options: {
+    inputImgBase64: string;
+    elementsPositionInfo: Array<ElementForOverlay>;
+    size?: {
+        width: number;
+        height: number;
+    };
+    annotationPadding?: number;
+    borderThickness?: number;
+    prompt?: string;
+}) => Promise<string>;
+export declare const processImageElementInfo: (options: {
+    inputImgBase64: string;
+    elementsPositionInfo: Array<BaseElement>;
+    elementsPositionInfoWithoutText: Array<BaseElement>;
+}) => Promise<{
+    compositeElementInfoImgBase64: string;
+    compositeElementInfoImgWithoutTextBase64: string;
+}>;
+export declare function annotateRects(imgBase64: string, rects: Rect[], prompt?: string): Promise<string>;
+export {};