@midscene/web 0.16.9-beta-20250508125518.0 → 0.16.9

package/dist/es/agent.js

@@ -336,7 +336,6 @@ import { sleep } from "@midscene/core/utils";
 import { UITarsModelVersion } from "@midscene/shared/env";
 import { uiTarsModelVersion } from "@midscene/shared/env";
 import { vlLocateMode } from "@midscene/shared/env";
-import { getElementInfosScriptContent } from "@midscene/shared/fs";
 import { imageInfo, resizeImgBase64 as resizeImgBase642 } from "@midscene/shared/img";
 import { getDebug as getDebug2 } from "@midscene/shared/logger";
 import { assert as assert4 } from "@midscene/shared/utils";
@@ -350,7 +349,6 @@ import { getAIConfigInBoolean } from "@midscene/shared/env";
 import { getRunningPkgInfo } from "@midscene/shared/fs";
 import { getDebug } from "@midscene/shared/logger";
 import { ifInBrowser as ifInBrowser2 } from "@midscene/shared/utils";
-import semver from "semver";
 
 // src/common/utils.ts
 import { uploadTestInfoToServer } from "@midscene/core/utils";
@@ -369,8 +367,7 @@ var WebElementInfo = class {
     locator,
     id,
     attributes,
-    indexId,
-    xpaths
+    indexId
   }) {
     this.content = content;
     this.rect = rect;
@@ -382,7 +379,6 @@ var WebElementInfo = class {
     this.id = id;
     this.attributes = attributes;
     this.indexId = indexId;
-    this.xpaths = xpaths;
   }
 };
 
@@ -448,20 +444,13 @@ function replaceIllegalPathCharsAndSpace(str) {
 // src/common/task-cache.ts
 var debug = getDebug("cache");
 var TaskCache = class {
-  constructor(page, opts) {
+  constructor(opts) {
     this.midscenePkgInfo = getRunningPkgInfo();
     this.cacheId = replaceIllegalPathCharsAndSpace(opts?.cacheId || "");
-    this.page = page;
     this.cache = this.readCacheFromFile() || {
-      pkgName: "",
-      pkgVersion: "",
-      cacheId: "",
       aiTasks: []
     };
     this.newCache = {
-      pkgName: this.midscenePkgInfo?.name || "",
-      pkgVersion: this.midscenePkgInfo?.version || "",
-      cacheId: this.cacheId,
       aiTasks: []
     };
   }
@@ -474,7 +463,7 @@ var TaskCache = class {
       tasks: newCacheGroup
     });
     return {
-      matchCache: async (pageContext, type, actionPrompt) => {
+      matchCache: (pageContext, type, actionPrompt) => {
         if (index === -1) {
           return false;
         }
@@ -512,7 +501,7 @@ var TaskCache = class {
       }
     };
   }
-  async matchCache(pageContext, type, userPrompt, cacheGroup) {
+  matchCache(pageContext, type, userPrompt, cacheGroup) {
     debug(
       "will read cache, type: %s, prompt: %s, cacheGroupLength: %s",
       type,
@@ -533,64 +522,21 @@ var TaskCache = class {
         taskRes?.response
       );
       if (taskRes?.type === "locate") {
-        const xpaths = taskRes.response?.elements?.[0]?.xpaths;
-        if (!xpaths || !xpaths.length) {
-          debug("no xpaths in cached response");
+        const id = taskRes.response?.elements[0].id;
+        if (!id) {
+          debug("no id in cached response");
           return false;
         }
-        for (const xpath of xpaths) {
-          if (this.page.pageType === "playwright") {
-            try {
-              const playwrightPage = this.page.underlyingPage;
-              const xpathLocator = playwrightPage.locator(`xpath=${xpath}`);
-              const xpathCount = await xpathLocator.count();
-              if (xpathCount === 1) {
-                debug(
-                  "cache hit, type: %s, prompt: %s, xpath: %s",
-                  type,
-                  userPrompt,
-                  xpath
-                );
-                const xpathElement = await xpathLocator.first();
-                await xpathElement.evaluate((element) => {
-                  element.setAttribute("data-midscene", "cache-hit");
-                  element.scrollIntoView();
-                });
-                return taskRes.response;
-              }
-            } catch (error) {
-              debug("playwright xpath locator error", error);
-            }
-          } else if (this.page.pageType === "puppeteer") {
-            try {
-              const puppeteerPage = this.page.underlyingPage;
-              const xpathElements = await puppeteerPage.$$(`xpath=${xpath}`);
-              if (xpathElements && xpathElements.length === 1) {
-                debug(
-                  "cache hit, type: %s, prompt: %s, xpath: %s",
-                  type,
-                  userPrompt,
-                  xpath
-                );
-                await xpathElements[0].evaluate((element) => {
-                  element.setAttribute("data-midscene", "cache-hit");
-                  element.scrollIntoView();
-                });
-                return taskRes.response;
-              }
-            } catch (error) {
-              debug("puppeteer xpath locator error", error);
-            }
-          } else {
-            debug("unknown page type, will not match cache", {
-              pageType: this.page.pageType
-            });
-          }
+        const foundInContext = pageContext.content.find(
+          (contentElement) => contentElement.id === id
+        );
+        if (!foundInContext) {
           debug("cannot match element with same id in current page", {
             element: taskRes.element
           });
           return false;
         }
+        return taskRes.response;
       }
       if (taskRes && taskRes.type === type && taskRes.prompt === userPrompt) {
         const contextEqual = this.pageContextEqual(
@@ -648,14 +594,11 @@ var TaskCache = class {
         const data = readFileSync(cacheFile, "utf8");
         const jsonData = JSON.parse(data);
         if (!this.midscenePkgInfo) {
-          debug("no midscene pkg info, will not read cache from file");
           return void 0;
         }
-        if (semver.lt(jsonData.pkgVersion, "0.17.0") || jsonData.pkgVersion.includes("beta")) {
-          console.warn(
-            `You are using an old version of Midscene cache file, and we cannot match any info from it. Starting from Midscene v0.17, we changed our strategy to use xpath for cache info, providing better performance. Please delete the existing cache and rebuild it. Sorry for the inconvenience.
-cache file: ${cacheFile}`
-          );
+        const jsonDataPkgVersion = jsonData.pkgVersion.split(".");
+        const midscenePkgInfoPkgVersion = this.midscenePkgInfo.version.split(".");
+        if (jsonDataPkgVersion[0] !== midscenePkgInfoPkgVersion[0] || jsonDataPkgVersion[1] !== midscenePkgInfoPkgVersion[1]) {
           return void 0;
         }
         debug("read cache from file: %s", cacheFile);
@@ -686,7 +629,15 @@ cache file: ${cacheFile}`
       writeLogFile({
         fileName: `${this.cacheId}`,
         fileExt: "json",
-        fileContent: stringifyDumpData(this.newCache, 2),
+        fileContent: stringifyDumpData(
+          {
+            pkgName: midscenePkgInfo.name,
+            pkgVersion: midscenePkgInfo.version,
+            cacheId: this.cacheId,
+            ...this.newCache
+          },
+          2
+        ),
         type: "cache"
       });
     }
@@ -777,7 +728,7 @@ var PageTaskExecutor = class {
     this.conversationHistory = [];
     this.page = page;
     this.insight = insight;
-    this.taskCache = new TaskCache(page, {
+    this.taskCache = new TaskCache({
       cacheId: opts?.cacheId
     });
     this.onTaskStartCallback = opts?.onTaskStart;
@@ -865,30 +816,16 @@ var PageTaskExecutor = class {
             };
             task.recorder = [recordItem];
             const cachePrompt = param.prompt;
-            const locateCache = await cacheGroup?.matchCache(
+            const locateCache = cacheGroup?.matchCache(
               pageContext,
               "locate",
               cachePrompt
             );
-            let elementInfo = null;
-            const oldId = locateCache?.elements?.[0]?.id;
-            let newId = null;
-            try {
-              if (oldId) {
-                const elementInfosScriptContent = getElementInfosScriptContent();
-                elementInfo = await this.page.evaluateJavaScript?.(
-                  `${elementInfosScriptContent}midscene_element_inspector.getElementInfoByNode(document.querySelector('[data-midscene="cache-hit"]'))`
-                );
-                newId = elementInfo?.id;
-                debug2("get new id by node", newId);
-              }
-            } catch (error) {
-              debug2("get element info by node error: ", error);
-            }
+            const idInCache = locateCache?.elements?.[0]?.id;
             let cacheHitFlag = false;
             let quickAnswerId = param?.id;
-            if (!quickAnswerId && newId) {
-              quickAnswerId = newId;
+            if (!quickAnswerId && idInCache) {
+              quickAnswerId = idInCache;
             }
             const quickAnswer = {
               id: quickAnswerId,
@@ -899,19 +836,10 @@ var PageTaskExecutor = class {
               quickAnswer
             });
             const aiCost = Date.now() - startTime;
-            if (element && element.id === quickAnswerId && newId) {
+            if (element && element.id === quickAnswerId && idInCache) {
               cacheHitFlag = true;
             }
             if (element) {
-              let xpaths;
-              try {
-                const elementInfosScriptContent = getElementInfosScriptContent();
-                xpaths = await this.page.evaluateJavaScript?.(
-                  `${elementInfosScriptContent}midscene_element_inspector.getXpathsById('${element.id}')`
-                );
-              } catch (error) {
-                debug2("getXpathsById error: ", error);
-              }
               cacheGroup?.saveCache({
                 type: "locate",
                 pageContext: {
@@ -922,8 +850,7 @@ var PageTaskExecutor = class {
                 response: {
                   elements: [
                     {
-                      id: newId || element.id,
-                      xpaths
+                      id: element.id
                     }
                   ]
                 },
@@ -1252,7 +1179,7 @@ var PageTaskExecutor = class {
         executorContext.task.recorder = [recordItem];
         executorContext.task.pageContext = pageContext;
         const cachePrompt = `${param.userInstruction} @ ${param.log || ""}`;
-        const planCache = await cacheGroup.matchCache(
+        const planCache = cacheGroup.matchCache(
           pageContext,
           "plan",
           cachePrompt
@@ -1416,7 +1343,7 @@ var PageTaskExecutor = class {
           ]
         });
         const startTime = Date.now();
-        const planCache = await cacheGroup.matchCache(
+        const planCache = cacheGroup.matchCache(
           pageContext,
           "ui-tars-plan",
           userInstruction