@midscene/web 0.16.4 → 0.16.6

package/dist/es/ui-utils.js.map

@@ -1 +1 @@
-{"version":3,"mappings":";AAWO,SAAS,QAAQ,MAAqB;AAC3C,SAAO,KAAK,UAAU,GAAG,KAAK,IAAI,MAAM,KAAK,WAAW,EAAE,KAAK,KAAK;AACtE;AAEO,SAAS,eACd,OAC0C;AAE1C,QAAM,OAAO,MAAM,QAAQ,KAAK,IAAI,QAAQ,CAAC,KAAK;AAIlD,SAAO,KAAK,OAAO,CAAC,KAA+C,MAAM;AACvE,UAAM,cAAc,KAAK,SAAS,MAAM,KAAK,KAAK,SAAS,SAAS;AACpE,QAAI,gBAAgB,MAAM,OAAO,MAAM,MAAM;AAC3C,aAAO,IAAI,OAAO,CAAC,EAAE,KAAK,GAAG,SAAS,YAAY,CAAC,CAAC;AAAA,IACtD;AACA,QAAI,gBAAgB,MAAM,OAAO,MAAM,MAAM;AAC3C,aAAO,IAAI,OAAO,CAAC,EAAE,KAAK,GAAG,SAAS,OAAO,CAAC,CAAC;AAAA,IACjD;AACA,QAAI,gBAAgB,MAAM,OAAO,MAAM,MAAM;AAC3C,aAAO,IAAI,OAAO,CAAC,EAAE,KAAK,GAAG,SAAS,QAAQ,CAAC,CAAC;AAAA,IAClD;AACA,WAAO,IAAI,OAAO,CAAC,EAAE,KAAK,EAAE,CAAC,CAAC;AAAA,EAChC,GAAG,CAAC,CAAC;AACP;AAEO,SAAS,eAAe,QAA8B;AAC3D,MAAI,CAAC,QAAQ;AACX,WAAO;AAAA,EACT;AAEA,MAAI,OAAO,WAAW,UAAU;AAC9B,WAAO;AAAA,EACT;AAEA,SAAO,OAAO;AAChB;AAEO,SAAS,eAAe,aAAyC;AACtE,MAAI,CAAC,aAAa;AAChB,WAAO;AAAA,EACT;AACA,SAAO,GAAG,YAAY,aAAa,MAAM,KAAK,YAAY,cAAc,MAAM,KAAK,YAAY,YAAY,kBAAkB;AAC/H;AAEO,SAAS,aACd,MAUA,QACA;AACA,MAAI,QAAQ;AACV,WAAO,GAAG,IAAI,MAAM,MAAM;AAAA,EAC5B;AACA,SAAO;AACT;AAEO,SAAS,SAAS,MAAqB;AAC5C,MAAI;AACJ,MAAI,KAAK,SAAS,YAAY;AAC5B,YAAS,MAAgC,OAAO;AAAA,EAClD;AAEA,MAAI,KAAK,SAAS,WAAW;AAC3B,YACG,MAAqC,OAAO,UAC5C,MAAqC,OAAO,MAC5C,MAAoC,OAAO,cAC3C,MAAwC,OAAO;AAAA,EACpD;AAEA,MAAI,KAAK,SAAS,UAAU;AAC1B,UAAM,SAAU,MAA8B;AAC9C,UAAM,YAAY,SAAS,eAAe,MAAM,IAAI;AAEpD,YAAQ,KAAK,WAAW;AACxB,QAAI,OAAQ,MAA8B,OAAO,WAAW,UAAU;AACpE,cAAQ,GAAI,MAA8B,OAAO,MAAM;AAAA,IACzD,WACE,OAAQ,MAA8B,OAAO,eAAe,UAC5D;AACA,cAAQ,eAAgB,MAA8B,KAAK;AAAA,IAC7D,WACE,OAAQ,MAA8B,OAAO,UAAU,aACvD;AACA,cAAS,MAA8B,OAAO;AAAA,IAChD;AAEA,QAAI,WAAW;AACb,UAAI,OAAO;AACT,gBAAQ,GAAG,SAAS,MAAM,KAAK;AAAA,MACjC,OAAO;AACL,gBAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAEA,MAAI,OAAO,UAAU;AAAa,WAAO;AACzC,SAAO,OAAO,UAAU,WACpB,QACA,KAAK,UAAU,OAAO,QAAW,CAAC;AACxC;AAEO,IAAM,wBAAwB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA","names":[],"ignoreList":[],"sources":["../../src/common/ui-utils.ts"],"sourcesContent":["import type {\n  DetailedLocateParam,\n  ExecutionTask,\n  ExecutionTaskAction,\n  ExecutionTaskInsightAssertion,\n  ExecutionTaskInsightLocate,\n  ExecutionTaskInsightQuery,\n  ExecutionTaskPlanning,\n  PlanningActionParamScroll,\n} from '@midscene/core';\n\nexport function typeStr(task: ExecutionTask) {\n  return task.subType ? `${task.type} / ${task.subType || ''}` : task.type;\n}\n\nexport function getKeyCommands(\n  value: string | string[],\n): Array<{ key: string; command?: string }> {\n  // Ensure value is an array of keys\n  const keys = Array.isArray(value) ? value : [value];\n\n  // Process each key to attach a corresponding command if needed, based on the presence of 'Meta' or 'Control' in the keys array.\n  // ref: https://github.com/puppeteer/puppeteer/pull/9357/files#diff-32cf475237b000f980eb214a0a823e45a902bddb7d2426d677cae96397aa0ae4R94\n  return keys.reduce((acc: Array<{ key: string; command?: string }>, k) => {\n    const includeMeta = keys.includes('Meta') || keys.includes('Control');\n    if (includeMeta && (k === 'a' || k === 'A')) {\n      return acc.concat([{ key: k, command: 'SelectAll' }]);\n    }\n    if (includeMeta && (k === 'c' || k === 'C')) {\n      return acc.concat([{ key: k, command: 'Copy' }]);\n    }\n    if (includeMeta && (k === 'v' || k === 'V')) {\n      return acc.concat([{ key: k, command: 'Paste' }]);\n    }\n    return acc.concat([{ key: k }]);\n  }, []);\n}\n\nexport function locateParamStr(locate?: DetailedLocateParam) {\n  if (!locate) {\n    return '';\n  }\n\n  if (typeof locate === 'string') {\n    return locate;\n  }\n\n  return locate.prompt;\n}\n\nexport function scrollParamStr(scrollParam?: PlanningActionParamScroll) {\n  if (!scrollParam) {\n    return '';\n  }\n  return `${scrollParam.direction || 'down'}, ${scrollParam.scrollType || 'once'}, ${scrollParam.distance || 'distance-not-set'}`;\n}\n\nexport function taskTitleStr(\n  type:\n    | 'Tap'\n    | 'Hover'\n    | 'Input'\n    | 'KeyboardPress'\n    | 'Scroll'\n    | 'Action'\n    | 'Query'\n    | 'Assert'\n    | 'WaitFor',\n  prompt: string,\n) {\n  if (prompt) {\n    return `${type} - ${prompt}`;\n  }\n  return type;\n}\n\nexport function paramStr(task: ExecutionTask) {\n  let value: string | undefined | object;\n  if (task.type === 'Planning') {\n    value = (task as ExecutionTaskPlanning)?.param?.userInstruction;\n  }\n\n  if (task.type === 'Insight') {\n    value =\n      (task as ExecutionTaskInsightLocate)?.param?.prompt ||\n      (task as ExecutionTaskInsightLocate)?.param?.id ||\n      (task as ExecutionTaskInsightQuery)?.param?.dataDemand ||\n      (task as ExecutionTaskInsightAssertion)?.param?.assertion;\n  }\n\n  if (task.type === 'Action') {\n    const locate = (task as ExecutionTaskAction)?.locate;\n    const locateStr = locate ? locateParamStr(locate) : '';\n\n    value = task.thought || '';\n    if (typeof (task as ExecutionTaskAction)?.param?.timeMs === 'number') {\n      value = `${(task as ExecutionTaskAction)?.param?.timeMs}ms`;\n    } else if (\n      typeof (task as ExecutionTaskAction)?.param?.scrollType === 'string'\n    ) {\n      value = scrollParamStr((task as ExecutionTaskAction)?.param);\n    } else if (\n      typeof (task as ExecutionTaskAction)?.param?.value !== 'undefined'\n    ) {\n      value = (task as ExecutionTaskAction)?.param?.value;\n    }\n\n    if (locateStr) {\n      if (value) {\n        value = `${locateStr} - ${value}`;\n      } else {\n        value = locateStr;\n      }\n    }\n  }\n\n  if (typeof value === 'undefined') return '';\n  return typeof value === 'string'\n    ? value\n    : JSON.stringify(value, undefined, 2);\n}\n\nexport const limitOpenNewTabScript = `\nif (!window.__MIDSCENE_NEW_TAB_INTERCEPTOR_INITIALIZED__) {\n  window.__MIDSCENE_NEW_TAB_INTERCEPTOR_INITIALIZED__ = true;\n\n  // Intercept the window.open method (only once)\n  window.open = function(url) {\n    console.log('Blocked window.open:', url);\n    window.location.href = url;\n    return null;\n  };\n\n  // Block all a tag clicks with target=\"_blank\" (only once)\n  document.addEventListener('click', function(e) {\n    const target = e.target.closest('a');\n    if (target && target.target === '_blank') {\n      e.preventDefault();\n      console.log('Blocked new tab:', target.href);\n      window.location.href = target.href;\n      target.removeAttribute('target');\n    }\n  }, true);\n}\n`;\n"]}
+{"version":3,"mappings":";AAWO,SAAS,QAAQ,MAAqB;AAC3C,SAAO,KAAK,UAAU,GAAG,KAAK,IAAI,MAAM,KAAK,WAAW,EAAE,KAAK,KAAK;AACtE;AAEO,SAAS,eACd,OAC0C;AAE1C,QAAM,OAAO,MAAM,QAAQ,KAAK,IAAI,QAAQ,CAAC,KAAK;AAIlD,SAAO,KAAK,OAAO,CAAC,KAA+C,MAAM;AACvE,UAAM,cAAc,KAAK,SAAS,MAAM,KAAK,KAAK,SAAS,SAAS;AACpE,QAAI,gBAAgB,MAAM,OAAO,MAAM,MAAM;AAC3C,aAAO,IAAI,OAAO,CAAC,EAAE,KAAK,GAAG,SAAS,YAAY,CAAC,CAAC;AAAA,IACtD;AACA,QAAI,gBAAgB,MAAM,OAAO,MAAM,MAAM;AAC3C,aAAO,IAAI,OAAO,CAAC,EAAE,KAAK,GAAG,SAAS,OAAO,CAAC,CAAC;AAAA,IACjD;AACA,QAAI,gBAAgB,MAAM,OAAO,MAAM,MAAM;AAC3C,aAAO,IAAI,OAAO,CAAC,EAAE,KAAK,GAAG,SAAS,QAAQ,CAAC,CAAC;AAAA,IAClD;AACA,WAAO,IAAI,OAAO,CAAC,EAAE,KAAK,EAAE,CAAC,CAAC;AAAA,EAChC,GAAG,CAAC,CAAC;AACP;AAEO,SAAS,eAAe,QAA8B;AAC3D,MAAI,CAAC,QAAQ;AACX,WAAO;AAAA,EACT;AAEA,MAAI,OAAO,WAAW,UAAU;AAC9B,WAAO;AAAA,EACT;AAEA,SAAO,OAAO;AAChB;AAEO,SAAS,eAAe,aAAyC;AACtE,MAAI,CAAC,aAAa;AAChB,WAAO;AAAA,EACT;AACA,SAAO,GAAG,YAAY,aAAa,MAAM,KAAK,YAAY,cAAc,MAAM,KAAK,YAAY,YAAY,kBAAkB;AAC/H;AAEO,SAAS,aACd,MAcA,QACA;AACA,MAAI,QAAQ;AACV,WAAO,GAAG,IAAI,MAAM,MAAM;AAAA,EAC5B;AACA,SAAO;AACT;AAEO,SAAS,SAAS,MAAqB;AAC5C,MAAI;AACJ,MAAI,KAAK,SAAS,YAAY;AAC5B,YAAS,MAAgC,OAAO;AAAA,EAClD;AAEA,MAAI,KAAK,SAAS,WAAW;AAC3B,YACG,MAAqC,OAAO,UAC5C,MAAqC,OAAO,MAC5C,MAAoC,OAAO,cAC3C,MAAwC,OAAO;AAAA,EACpD;AAEA,MAAI,KAAK,SAAS,UAAU;AAC1B,UAAM,SAAU,MAA8B;AAC9C,UAAM,YAAY,SAAS,eAAe,MAAM,IAAI;AAEpD,YAAQ,KAAK,WAAW;AACxB,QAAI,OAAQ,MAA8B,OAAO,WAAW,UAAU;AACpE,cAAQ,GAAI,MAA8B,OAAO,MAAM;AAAA,IACzD,WACE,OAAQ,MAA8B,OAAO,eAAe,UAC5D;AACA,cAAQ,eAAgB,MAA8B,KAAK;AAAA,IAC7D,WACE,OAAQ,MAA8B,OAAO,UAAU,aACvD;AACA,cAAS,MAA8B,OAAO;AAAA,IAChD;AAEA,QAAI,WAAW;AACb,UAAI,OAAO;AACT,gBAAQ,GAAG,SAAS,MAAM,KAAK;AAAA,MACjC,OAAO;AACL,gBAAQ;AAAA,MACV;AAAA,IACF;AAAA,EACF;AAEA,MAAI,OAAO,UAAU;AAAa,WAAO;AACzC,SAAO,OAAO,UAAU,WACpB,QACA,KAAK,UAAU,OAAO,QAAW,CAAC;AACxC;AAEO,IAAM,wBAAwB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA","names":[],"ignoreList":[],"sources":["../../src/common/ui-utils.ts"],"sourcesContent":["import type {\n  DetailedLocateParam,\n  ExecutionTask,\n  ExecutionTaskAction,\n  ExecutionTaskInsightAssertion,\n  ExecutionTaskInsightLocate,\n  ExecutionTaskInsightQuery,\n  ExecutionTaskPlanning,\n  PlanningActionParamScroll,\n} from '@midscene/core';\n\nexport function typeStr(task: ExecutionTask) {\n  return task.subType ? `${task.type} / ${task.subType || ''}` : task.type;\n}\n\nexport function getKeyCommands(\n  value: string | string[],\n): Array<{ key: string; command?: string }> {\n  // Ensure value is an array of keys\n  const keys = Array.isArray(value) ? value : [value];\n\n  // Process each key to attach a corresponding command if needed, based on the presence of 'Meta' or 'Control' in the keys array.\n  // ref: https://github.com/puppeteer/puppeteer/pull/9357/files#diff-32cf475237b000f980eb214a0a823e45a902bddb7d2426d677cae96397aa0ae4R94\n  return keys.reduce((acc: Array<{ key: string; command?: string }>, k) => {\n    const includeMeta = keys.includes('Meta') || keys.includes('Control');\n    if (includeMeta && (k === 'a' || k === 'A')) {\n      return acc.concat([{ key: k, command: 'SelectAll' }]);\n    }\n    if (includeMeta && (k === 'c' || k === 'C')) {\n      return acc.concat([{ key: k, command: 'Copy' }]);\n    }\n    if (includeMeta && (k === 'v' || k === 'V')) {\n      return acc.concat([{ key: k, command: 'Paste' }]);\n    }\n    return acc.concat([{ key: k }]);\n  }, []);\n}\n\nexport function locateParamStr(locate?: DetailedLocateParam) {\n  if (!locate) {\n    return '';\n  }\n\n  if (typeof locate === 'string') {\n    return locate;\n  }\n\n  return locate.prompt;\n}\n\nexport function scrollParamStr(scrollParam?: PlanningActionParamScroll) {\n  if (!scrollParam) {\n    return '';\n  }\n  return `${scrollParam.direction || 'down'}, ${scrollParam.scrollType || 'once'}, ${scrollParam.distance || 'distance-not-set'}`;\n}\n\nexport function taskTitleStr(\n  type:\n    | 'Tap'\n    | 'Hover'\n    | 'Input'\n    | 'KeyboardPress'\n    | 'Scroll'\n    | 'Action'\n    | 'Query'\n    | 'Assert'\n    | 'WaitFor'\n    | 'Locate'\n    | 'Boolean'\n    | 'Number'\n    | 'String',\n  prompt: string,\n) {\n  if (prompt) {\n    return `${type} - ${prompt}`;\n  }\n  return type;\n}\n\nexport function paramStr(task: ExecutionTask) {\n  let value: string | undefined | object;\n  if (task.type === 'Planning') {\n    value = (task as ExecutionTaskPlanning)?.param?.userInstruction;\n  }\n\n  if (task.type === 'Insight') {\n    value =\n      (task as ExecutionTaskInsightLocate)?.param?.prompt ||\n      (task as ExecutionTaskInsightLocate)?.param?.id ||\n      (task as ExecutionTaskInsightQuery)?.param?.dataDemand ||\n      (task as ExecutionTaskInsightAssertion)?.param?.assertion;\n  }\n\n  if (task.type === 'Action') {\n    const locate = (task as ExecutionTaskAction)?.locate;\n    const locateStr = locate ? locateParamStr(locate) : '';\n\n    value = task.thought || '';\n    if (typeof (task as ExecutionTaskAction)?.param?.timeMs === 'number') {\n      value = `${(task as ExecutionTaskAction)?.param?.timeMs}ms`;\n    } else if (\n      typeof (task as ExecutionTaskAction)?.param?.scrollType === 'string'\n    ) {\n      value = scrollParamStr((task as ExecutionTaskAction)?.param);\n    } else if (\n      typeof (task as ExecutionTaskAction)?.param?.value !== 'undefined'\n    ) {\n      value = (task as ExecutionTaskAction)?.param?.value;\n    }\n\n    if (locateStr) {\n      if (value) {\n        value = `${locateStr} - ${value}`;\n      } else {\n        value = locateStr;\n      }\n    }\n  }\n\n  if (typeof value === 'undefined') return '';\n  return typeof value === 'string'\n    ? value\n    : JSON.stringify(value, undefined, 2);\n}\n\nexport const limitOpenNewTabScript = `\nif (!window.__MIDSCENE_NEW_TAB_INTERCEPTOR_INITIALIZED__) {\n  window.__MIDSCENE_NEW_TAB_INTERCEPTOR_INITIALIZED__ = true;\n\n  // Intercept the window.open method (only once)\n  window.open = function(url) {\n    console.log('Blocked window.open:', url);\n    window.location.href = url;\n    return null;\n  };\n\n  // Block all a tag clicks with target=\"_blank\" (only once)\n  document.addEventListener('click', function(e) {\n    const target = e.target.closest('a');\n    if (target && target.target === '_blank') {\n      e.preventDefault();\n      console.log('Blocked new tab:', target.href);\n      window.location.href = target.href;\n      target.removeAttribute('target');\n    }\n  }, true);\n}\n`;\n"]}

package/dist/lib/agent.js

@@ -1501,17 +1501,23 @@ var PageTaskExecutor = class {
       executor: taskExecutor
     };
   }
-  async query(demand) {
-    const description = typeof demand === "string" ? demand : JSON.stringify(demand);
-    const taskExecutor = new import_core.Executor(taskTitleStr("Query", description), {
-      onTaskStart: this.onTaskStartCallback
-    });
+  async createTypeQueryTask(type, demand) {
+    const taskExecutor = new import_core.Executor(
+      taskTitleStr(
+        type,
+        typeof demand === "string" ? demand : JSON.stringify(demand)
+      ),
+      {
+        onTaskStart: this.onTaskStartCallback
+      }
+    );
     const queryTask = {
       type: "Insight",
-      subType: "Query",
+      subType: type,
       locate: null,
       param: {
         dataDemand: demand
+        // for user param presentation in report right sidebar
       },
       executor: async (param) => {
         let insightDump;
@@ -1519,11 +1525,21 @@ var PageTaskExecutor = class {
           insightDump = dump;
         };
         this.insight.onceDumpUpdatedFn = dumpCollector;
-        const { data, usage } = await this.insight.extract(
-          param.dataDemand
-        );
+        const ifTypeRestricted = type !== "Query";
+        let demandInput = demand;
+        if (ifTypeRestricted) {
+          demandInput = {
+            result: `${type}, ${demand}`
+          };
+        }
+        const { data, usage } = await this.insight.extract(demandInput);
+        let outputResult = data;
+        if (ifTypeRestricted) {
+          (0, import_utils9.assert)(data?.result !== void 0, "No result in query data");
+          outputResult = data.result;
+        }
         return {
-          output: data,
+          output: outputResult,
           log: { dump: insightDump },
           usage
         };
@@ -1536,6 +1552,18 @@ var PageTaskExecutor = class {
       executor: taskExecutor
     };
   }
+  async query(demand) {
+    return this.createTypeQueryTask("Query", demand);
+  }
+  async boolean(prompt) {
+    return this.createTypeQueryTask("Boolean", prompt);
+  }
+  async number(prompt) {
+    return this.createTypeQueryTask("Number", prompt);
+  }
+  async string(prompt) {
+    return this.createTypeQueryTask("String", prompt);
+  }
   async assert(assertion) {
     const description = `assert: ${assertion}`;
     const taskExecutor = new import_core.Executor(taskTitleStr("Assert", description), {
@@ -1722,6 +1750,16 @@ function buildPlans(type, locateParam, param) {
     };
     returnPlans = [sleepPlan];
   }
+  if (type === "Locate") {
+    (0, import_utils10.assert)(locateParam, `missing locate info for action "${type}"`);
+    const locatePlan2 = {
+      type,
+      param: locateParam,
+      locate: locateParam,
+      thought: ""
+    };
+    returnPlans = [locatePlan2];
+  }
   if (returnPlans) {
     debug3("buildPlans", returnPlans);
     return returnPlans;
@@ -1921,6 +1959,35 @@ ${errorTask?.errorStack}`);
     this.afterTaskRunning(executor);
     return output;
   }
+  async aiBoolean(prompt) {
+    const { output, executor } = await this.taskExecutor.boolean(prompt);
+    this.afterTaskRunning(executor);
+    return output;
+  }
+  async aiNumber(prompt) {
+    const { output, executor } = await this.taskExecutor.number(prompt);
+    this.afterTaskRunning(executor);
+    return output;
+  }
+  async aiString(prompt) {
+    const { output, executor } = await this.taskExecutor.string(prompt);
+    this.afterTaskRunning(executor);
+    return output;
+  }
+  async aiLocate(prompt, opt) {
+    const detailedLocateParam = this.buildDetailedLocateParam(prompt, opt);
+    const plans = buildPlans("Locate", detailedLocateParam);
+    const { executor, output } = await this.taskExecutor.runPlans(
+      taskTitleStr("Locate", locateParamStr(detailedLocateParam)),
+      plans
+    );
+    this.afterTaskRunning(executor);
+    const { element } = output;
+    return {
+      rect: element?.rect,
+      center: element?.center
+    };
+  }
   async aiAssert(assertion, msg, opt) {
     const { output, executor } = await this.taskExecutor.assert(assertion);
     this.afterTaskRunning(executor, true);