@midscene/core 0.2.1-beta-20240807115314.0 → 0.2.1

package/dist/es/index.js

@@ -1076,7 +1076,7 @@ function writeDumpFile(opts) {
         `${gitIgnoreContent}
 # Midscene.js dump files
 ${logDirName}/report
-${logDirName}/dump-logger
+${logDirName}/dump
 `,
         "utf-8"
       );
@@ -1090,6 +1090,19 @@ ${logDirName}/dump-logger
   }
   return filePath;
 }
+function replacerForPageObject(key, value) {
+  var _a, _b;
+  if (value && ((_a = value.constructor) == null ? void 0 : _a.name) === "Page") {
+    return "[Page object]";
+  }
+  if (value && ((_b = value.constructor) == null ? void 0 : _b.name) === "Browser") {
+    return "[Browser object]";
+  }
+  return value;
+}
+function stringifyDumpData(data, indents) {
+  return JSON.stringify(data, replacerForPageObject, indents);
+}
 
 // src/action/executor.ts
 var Executor = class {
@@ -4983,7 +4996,7 @@ function writeInsightDump(data, logId, dumpSubscriber) {
       logFileName = `${pid}_${baseData.logTime}-${Math.random()}`;
     }
   }
-  const dataString = JSON.stringify(finalData, null, 2);
+  const dataString = stringifyDumpData(finalData, 2);
   if (typeof logIdIndexMap[id] === "number") {
     logContent[logIdIndexMap[id]] = dataString;
   } else {
@@ -5329,11 +5342,14 @@ function systemPromptToTaskPlanning(query) {
     * param: { scrollType: 'ScrollUntilBottom', 'ScrollUntilTop', 'ScrollDown', 'ScrollUp' }
   * type: 'Error'
     * param: { message: string }, the error message
+  * type: 'Sleep'
+    * param: { timeMs: number }, wait for timeMs milliseconds 
   
   Here is an example of how to decompose a task.
-  When a user says 'Input "Weather in Shanghai" into the search bar, hit enter', by viewing the page screenshot and description, you my decompose this task into something like this:
+  When a user says 'Input "Weather in Shanghai" into the search bar, wait 1 second, hit enter', by viewing the page screenshot and description, you my decompose this task into something like this:
   * Find: 'The search bar'
   * Input: 'Weather in Shanghai'
+  * Sleep: 1000
   * KeyboardPress: 'Enter'
   
   Remember: 

package/dist/es/utils.js

@@ -64,7 +64,7 @@ function writeDumpFile(opts) {
         `${gitIgnoreContent}
 # Midscene.js dump files
 ${logDirName}/report
-${logDirName}/dump-logger
+${logDirName}/dump
 `,
         "utf-8"
       );
@@ -94,6 +94,19 @@ async function sleep(ms) {
   return new Promise((resolve) => setTimeout(resolve, ms));
 }
 var commonScreenshotParam = { type: "jpeg", quality: 75 };
+function replacerForPageObject(key, value) {
+  var _a, _b;
+  if (value && ((_a = value.constructor) == null ? void 0 : _a.name) === "Page") {
+    return "[Page object]";
+  }
+  if (value && ((_b = value.constructor) == null ? void 0 : _b.name) === "Browser") {
+    return "[Browser object]";
+  }
+  return value;
+}
+function stringifyDumpData(data, indents) {
+  return JSON.stringify(data, replacerForPageObject, indents);
+}
 export {
   commonScreenshotParam,
   getDumpDir,
@@ -104,7 +117,9 @@ export {
   groupedActionDumpFileExt,
   insightDumpFileExt,
   overlapped,
+  replacerForPageObject,
   setDumpDir,
   sleep,
+  stringifyDumpData,
   writeDumpFile
 };

package/dist/lib/index.js

@@ -1091,7 +1091,7 @@ function writeDumpFile(opts) {
         `${gitIgnoreContent}
 # Midscene.js dump files
 ${logDirName}/report
-${logDirName}/dump-logger
+${logDirName}/dump
 `,
         "utf-8"
       );
@@ -1105,6 +1105,19 @@ ${logDirName}/dump-logger
   }
   return filePath;
 }
+function replacerForPageObject(key, value) {
+  var _a, _b;
+  if (value && ((_a = value.constructor) == null ? void 0 : _a.name) === "Page") {
+    return "[Page object]";
+  }
+  if (value && ((_b = value.constructor) == null ? void 0 : _b.name) === "Browser") {
+    return "[Browser object]";
+  }
+  return value;
+}
+function stringifyDumpData(data, indents) {
+  return JSON.stringify(data, replacerForPageObject, indents);
+}
 
 // src/action/executor.ts
 var Executor = class {
@@ -4998,7 +5011,7 @@ function writeInsightDump(data, logId, dumpSubscriber) {
       logFileName = `${pid}_${baseData.logTime}-${Math.random()}`;
     }
   }
-  const dataString = JSON.stringify(finalData, null, 2);
+  const dataString = stringifyDumpData(finalData, 2);
   if (typeof logIdIndexMap[id] === "number") {
     logContent[logIdIndexMap[id]] = dataString;
   } else {
@@ -5344,11 +5357,14 @@ function systemPromptToTaskPlanning(query) {
     * param: { scrollType: 'ScrollUntilBottom', 'ScrollUntilTop', 'ScrollDown', 'ScrollUp' }
   * type: 'Error'
     * param: { message: string }, the error message
+  * type: 'Sleep'
+    * param: { timeMs: number }, wait for timeMs milliseconds 
   
   Here is an example of how to decompose a task.
-  When a user says 'Input "Weather in Shanghai" into the search bar, hit enter', by viewing the page screenshot and description, you my decompose this task into something like this:
+  When a user says 'Input "Weather in Shanghai" into the search bar, wait 1 second, hit enter', by viewing the page screenshot and description, you my decompose this task into something like this:
   * Find: 'The search bar'
   * Input: 'Weather in Shanghai'
+  * Sleep: 1000
   * KeyboardPress: 'Enter'
   
   Remember: 

package/dist/lib/utils.js

@@ -39,8 +39,10 @@ __export(utils_exports, {
   groupedActionDumpFileExt: () => groupedActionDumpFileExt,
   insightDumpFileExt: () => insightDumpFileExt,
   overlapped: () => overlapped,
+  replacerForPageObject: () => replacerForPageObject,
   setDumpDir: () => setDumpDir,
   sleep: () => sleep,
+  stringifyDumpData: () => stringifyDumpData,
   writeDumpFile: () => writeDumpFile
 });
 module.exports = __toCommonJS(utils_exports);
@@ -103,7 +105,7 @@ function writeDumpFile(opts) {
         `${gitIgnoreContent}
 # Midscene.js dump files
 ${logDirName}/report
-${logDirName}/dump-logger
+${logDirName}/dump
 `,
         "utf-8"
       );
@@ -133,6 +135,19 @@ async function sleep(ms) {
   return new Promise((resolve) => setTimeout(resolve, ms));
 }
 var commonScreenshotParam = { type: "jpeg", quality: 75 };
+function replacerForPageObject(key, value) {
+  var _a, _b;
+  if (value && ((_a = value.constructor) == null ? void 0 : _a.name) === "Page") {
+    return "[Page object]";
+  }
+  if (value && ((_b = value.constructor) == null ? void 0 : _b.name) === "Browser") {
+    return "[Browser object]";
+  }
+  return value;
+}
+function stringifyDumpData(data, indents) {
+  return JSON.stringify(data, replacerForPageObject, indents);
+}
 // Annotate the CommonJS export names for ESM import in node:
 0 && (module.exports = {
   commonScreenshotParam,
@@ -144,7 +159,9 @@ var commonScreenshotParam = { type: "jpeg", quality: 75 };
   groupedActionDumpFileExt,
   insightDumpFileExt,
   overlapped,
+  replacerForPageObject,
   setDumpDir,
   sleep,
+  stringifyDumpData,
   writeDumpFile
 });

package/dist/types/ai-model.d.ts

@@ -1,7 +1,7 @@
-import { c as callToGetJSONObject } from './util-3931e76e.js';
-export { d as describeUserPage } from './util-3931e76e.js';
+import { c as callToGetJSONObject } from './util-7b3c592c.js';
+export { d as describeUserPage } from './util-7b3c592c.js';
 export { ChatCompletionMessageParam } from 'openai/resources';
-import { B as BaseElement, U as UIContext, f as AIElementParseResponse, g as AISectionParseResponse } from './types-2c92867c.js';
+import { B as BaseElement, U as UIContext, f as AIElementParseResponse, g as AISectionParseResponse } from './types-3ef00c38.js';
 
 declare function systemPromptToFindElement(description: string, multi?: boolean): string;
 

package/dist/types/image.d.ts

@@ -1,5 +1,5 @@
 import { Buffer } from 'node:buffer';
-import { S as Size, R as Rect, i as UISection, U as UIContext, w as Color } from './types-2c92867c.js';
+import { S as Size, R as Rect, i as UISection, U as UIContext, x as Color } from './types-3ef00c38.js';
 import 'openai/resources';
 
 /**

package/dist/types/index.d.ts

@@ -1,6 +1,6 @@
-import { E as ExecutionTask, a as ExecutionTaskApply, b as ExecutionDump, B as BaseElement, U as UIContext, D as DumpSubscriber, I as InsightTaskInfo, c as InsightOptions, d as InsightAssertionResponse, P as PlanningAction } from './types-2c92867c.js';
-export { h as AIAssertionResponse, f as AIElementParseResponse, A as AIResponseFormat, g as AISectionParseResponse, x as BaseAgentParserOpt, k as BasicSectionQuery, C as CallAIFn, w as Color, m as DumpMeta, p as ElementById, j as EnsureObject, F as ExecutionRecorderItem, a1 as ExecutionTaskAction, a0 as ExecutionTaskActionApply, $ as ExecutionTaskInsightAssertion, _ as ExecutionTaskInsightAssertionApply, Z as ExecutionTaskInsightAssertionParam, N as ExecutionTaskInsightDumpLog, Q as ExecutionTaskInsightLocate, O as ExecutionTaskInsightLocateApply, M as ExecutionTaskInsightLocateOutput, K as ExecutionTaskInsightLocateParam, Y as ExecutionTaskInsightQuery, X as ExecutionTaskInsightQueryApply, W as ExecutionTaskInsightQueryOutput, V as ExecutionTaskInsightQueryParam, a3 as ExecutionTaskPlanning, a2 as ExecutionTaskPlanningApply, J as ExecutionTaskReturn, G as ExecutionTaskType, H as ExecutorContext, a4 as GroupedActionDump, n as InsightDump, l as InsightExtractParam, L as LiteUISection, o as PartialInsightDumpFromSDK, q as PlanningAIResponse, v as PlanningActionParamAssert, s as PlanningActionParamHover, t as PlanningActionParamInputOrKeyPress, u as PlanningActionParamScroll, r as PlanningActionParamTap, z as PlaywrightParserOpt, e as Point, y as PuppeteerParserOpt, R as Rect, S as Size, T as TaskCacheInfo, i as UISection } from './types-2c92867c.js';
-import { c as callToGetJSONObject, r as retrieveElement, a as retrieveSection } from './util-3931e76e.js';
+import { E as ExecutionTask, a as ExecutionTaskApply, b as ExecutionDump, B as BaseElement, U as UIContext, D as DumpSubscriber, I as InsightTaskInfo, c as InsightOptions, d as InsightAssertionResponse, P as PlanningAction } from './types-3ef00c38.js';
+export { h as AIAssertionResponse, f as AIElementParseResponse, A as AIResponseFormat, g as AISectionParseResponse, y as BaseAgentParserOpt, k as BasicSectionQuery, C as CallAIFn, x as Color, m as DumpMeta, p as ElementById, j as EnsureObject, G as ExecutionRecorderItem, a2 as ExecutionTaskAction, a1 as ExecutionTaskActionApply, a0 as ExecutionTaskInsightAssertion, $ as ExecutionTaskInsightAssertionApply, _ as ExecutionTaskInsightAssertionParam, O as ExecutionTaskInsightDumpLog, V as ExecutionTaskInsightLocate, Q as ExecutionTaskInsightLocateApply, N as ExecutionTaskInsightLocateOutput, M as ExecutionTaskInsightLocateParam, Z as ExecutionTaskInsightQuery, Y as ExecutionTaskInsightQueryApply, X as ExecutionTaskInsightQueryOutput, W as ExecutionTaskInsightQueryParam, a4 as ExecutionTaskPlanning, a3 as ExecutionTaskPlanningApply, K as ExecutionTaskReturn, H as ExecutionTaskType, J as ExecutorContext, a5 as GroupedActionDump, n as InsightDump, l as InsightExtractParam, L as LiteUISection, o as PartialInsightDumpFromSDK, q as PlanningAIResponse, v as PlanningActionParamAssert, s as PlanningActionParamHover, t as PlanningActionParamInputOrKeyPress, u as PlanningActionParamScroll, w as PlanningActionParamSleep, r as PlanningActionParamTap, F as PlaywrightParserOpt, e as Point, z as PuppeteerParserOpt, R as Rect, S as Size, T as TaskCacheInfo, i as UISection } from './types-3ef00c38.js';
+import { c as callToGetJSONObject, r as retrieveElement, a as retrieveSection } from './util-7b3c592c.js';
 export { setDumpDir } from './utils.js';
 import 'openai/resources';
 

package/dist/types/{types-2c92867c.d.ts → types-3ef00c38.d.ts}

@@ -124,7 +124,7 @@ type InsightAssertionResponse = AIAssertionResponse;
  */
 interface PlanningAction<ParamType = any> {
     thought?: string;
-    type: 'Locate' | 'Tap' | 'Hover' | 'Input' | 'KeyboardPress' | 'Scroll' | 'Error' | 'Assert';
+    type: 'Locate' | 'Tap' | 'Hover' | 'Input' | 'KeyboardPress' | 'Scroll' | 'Error' | 'Assert' | 'Sleep';
     param: ParamType;
 }
 interface PlanningAIResponse {
@@ -143,6 +143,9 @@ interface PlanningActionParamScroll {
 interface PlanningActionParamAssert {
     assertion: string;
 }
+interface PlanningActionParamSleep {
+    timeMs: number;
+}
 /**
  * misc
  */
@@ -234,4 +237,4 @@ interface GroupedActionDump {
     executions: ExecutionDump[];
 }
 
-export { type ExecutionTaskInsightAssertion as $, AIResponseFormat as A, BaseElement as B, type CallAIFn as C, type DumpSubscriber as D, type ExecutionTask as E, type ExecutionRecorderItem as F, type ExecutionTaskType as G, type ExecutorContext as H, type InsightTaskInfo as I, type ExecutionTaskReturn as J, type ExecutionTaskInsightLocateParam as K, type LiteUISection as L, type ExecutionTaskInsightLocateOutput as M, type ExecutionTaskInsightDumpLog as N, type ExecutionTaskInsightLocateApply as O, type PlanningAction as P, type ExecutionTaskInsightLocate as Q, type Rect as R, type Size as S, type TaskCacheInfo as T, UIContext as U, type ExecutionTaskInsightQueryParam as V, type ExecutionTaskInsightQueryOutput as W, type ExecutionTaskInsightQueryApply as X, type ExecutionTaskInsightQuery as Y, type ExecutionTaskInsightAssertionParam as Z, type ExecutionTaskInsightAssertionApply as _, type ExecutionTaskApply as a, type ExecutionTaskActionApply as a0, type ExecutionTaskAction as a1, type ExecutionTaskPlanningApply as a2, type ExecutionTaskPlanning as a3, type GroupedActionDump as a4, type ExecutionDump as b, type InsightOptions as c, type InsightAssertionResponse as d, type Point as e, type AIElementParseResponse as f, type AISectionParseResponse as g, type AIAssertionResponse as h, type UISection as i, type EnsureObject as j, type BasicSectionQuery as k, type InsightExtractParam as l, type DumpMeta as m, type InsightDump as n, type PartialInsightDumpFromSDK as o, type ElementById as p, type PlanningAIResponse as q, type PlanningActionParamTap as r, type PlanningActionParamHover as s, type PlanningActionParamInputOrKeyPress as t, type PlanningActionParamScroll as u, type PlanningActionParamAssert as v, type Color as w, type BaseAgentParserOpt as x, type PuppeteerParserOpt as y, type PlaywrightParserOpt as z };
+export { type ExecutionTaskInsightAssertionApply as $, AIResponseFormat as A, BaseElement as B, type CallAIFn as C, type DumpSubscriber as D, type ExecutionTask as E, type PlaywrightParserOpt as F, type ExecutionRecorderItem as G, type ExecutionTaskType as H, type InsightTaskInfo as I, type ExecutorContext as J, type ExecutionTaskReturn as K, type LiteUISection as L, type ExecutionTaskInsightLocateParam as M, type ExecutionTaskInsightLocateOutput as N, type ExecutionTaskInsightDumpLog as O, type PlanningAction as P, type ExecutionTaskInsightLocateApply as Q, type Rect as R, type Size as S, type TaskCacheInfo as T, UIContext as U, type ExecutionTaskInsightLocate as V, type ExecutionTaskInsightQueryParam as W, type ExecutionTaskInsightQueryOutput as X, type ExecutionTaskInsightQueryApply as Y, type ExecutionTaskInsightQuery as Z, type ExecutionTaskInsightAssertionParam as _, type ExecutionTaskApply as a, type ExecutionTaskInsightAssertion as a0, type ExecutionTaskActionApply as a1, type ExecutionTaskAction as a2, type ExecutionTaskPlanningApply as a3, type ExecutionTaskPlanning as a4, type GroupedActionDump as a5, type ExecutionDump as b, type InsightOptions as c, type InsightAssertionResponse as d, type Point as e, type AIElementParseResponse as f, type AISectionParseResponse as g, type AIAssertionResponse as h, type UISection as i, type EnsureObject as j, type BasicSectionQuery as k, type InsightExtractParam as l, type DumpMeta as m, type InsightDump as n, type PartialInsightDumpFromSDK as o, type ElementById as p, type PlanningAIResponse as q, type PlanningActionParamTap as r, type PlanningActionParamHover as s, type PlanningActionParamInputOrKeyPress as t, type PlanningActionParamScroll as u, type PlanningActionParamAssert as v, type PlanningActionParamSleep as w, type Color as x, type BaseAgentParserOpt as y, type PuppeteerParserOpt as z };

package/dist/types/{util-3931e76e.d.ts → util-7b3c592c.d.ts}

@@ -1,5 +1,5 @@
 import { ChatCompletionMessageParam } from 'openai/resources';
-import { B as BaseElement, U as UIContext } from './types-2c92867c.js';
+import { B as BaseElement, U as UIContext } from './types-3ef00c38.js';
 
 declare function callToGetJSONObject<T>(messages: ChatCompletionMessageParam[]): Promise<T>;
 

package/dist/types/utils.d.ts

@@ -1,4 +1,4 @@
-import { R as Rect } from './types-2c92867c.js';
+import { R as Rect } from './types-3ef00c38.js';
 import 'openai/resources';
 
 interface PkgInfo {
@@ -22,5 +22,7 @@ declare function getTmpFile(fileExtWithoutDot: string): string;
 declare function overlapped(container: Rect, target: Rect): boolean;
 declare function sleep(ms: number): Promise<unknown>;
 declare const commonScreenshotParam: any;
+declare function replacerForPageObject(key: string, value: any): any;
+declare function stringifyDumpData(data: any, indents?: number): string;
 
-export { commonScreenshotParam, getDumpDir, getDumpDirPath, getPkgInfo, getTmpDir, getTmpFile, groupedActionDumpFileExt, insightDumpFileExt, overlapped, setDumpDir, sleep, writeDumpFile };
+export { commonScreenshotParam, getDumpDir, getDumpDirPath, getPkgInfo, getTmpDir, getTmpFile, groupedActionDumpFileExt, insightDumpFileExt, overlapped, replacerForPageObject, setDumpDir, sleep, stringifyDumpData, writeDumpFile };

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@midscene/core",
   "description": "Hello, It's Midscene",
-  "version": "0.2.1-beta-20240807115314.0",
+  "version": "0.2.1",
   "jsnext:source": "./src/index.ts",
   "main": "./dist/lib/index.js",
   "module": "./dist/es/index.js",