@eko-ai/eko 2.1.0 → 2.1.2

package/dist/index.cjs.js

@@ -7,11 +7,12 @@ var buffer = require('buffer');
 const config = {
     name: "Eko",
     platform: "mac",
-    maxReactNum: 200,
+    maxReactNum: 500,
     maxTokens: 16000,
     compressThreshold: 80,
     largeTextLength: 5000,
-    shortTextLength: 800,
+    fileTextMaxLength: 20000,
+    maxDialogueImgFileNum: 2,
 };
 
 var LogLevel;
@@ -67,7 +68,7 @@ class Logger {
         };
         let formattedMessage = '';
         if (this.dateFormat) {
-            formattedMessage += `[${new Date().toISOString()}] `;
+            formattedMessage += `[${new Date().toLocaleString()}] `;
         }
         formattedMessage += `[${levelNames[level] || 'UNKNOWN'}] `;
         if (this.prefix) {
@@ -139,8 +140,240 @@ class Logger {
 }
 const Log = new Logger();
 
+function sleep(time) {
+    return new Promise((resolve) => setTimeout(() => resolve(), time));
+}
+function uuidv4() {
+    return "xxxxxxxx-xxxx-4xxx-yxxx-xxxxxxxxxxxx".replace(/[xy]/g, function (c) {
+        const r = (Math.random() * 16) | 0;
+        const v = c === "x" ? r : (r & 0x3) | 0x8;
+        return v.toString(16);
+    });
+}
+function call_timeout(fun, timeout, error_callback) {
+    return new Promise(async (resolve, reject) => {
+        let timer = setTimeout(() => {
+            reject(new Error("Timeout"));
+            error_callback && error_callback("Timeout");
+        }, timeout);
+        try {
+            const result = await fun();
+            clearTimeout(timer);
+            resolve(result);
+        }
+        catch (e) {
+            clearTimeout(timer);
+            reject(e);
+            error_callback && error_callback(e + "");
+        }
+    });
+}
+function convertToolSchema(tool) {
+    if ("function" in tool) {
+        return {
+            type: "function",
+            name: tool.function.name,
+            description: tool.function.description,
+            parameters: tool.function.parameters,
+        };
+    }
+    else if ("input_schema" in tool) {
+        return {
+            type: "function",
+            name: tool.name,
+            description: tool.description,
+            parameters: tool.input_schema,
+        };
+    }
+    else if ("inputSchema" in tool) {
+        return {
+            type: "function",
+            name: tool.name,
+            description: tool.description,
+            parameters: tool.inputSchema,
+        };
+    }
+    else {
+        return {
+            type: "function",
+            name: tool.name,
+            description: tool.description,
+            parameters: tool.parameters,
+        };
+    }
+}
+function toImage(imageData) {
+    let image = null;
+    if (imageData.startsWith("http://") || imageData.startsWith("https://")) {
+        image = new URL(imageData);
+    }
+    else {
+        if (imageData.startsWith("data:image/")) {
+            imageData = imageData.substring(imageData.indexOf(",") + 1);
+        }
+        // @ts-ignore
+        if (typeof Buffer != "undefined") {
+            // @ts-ignore
+            const buffer = Buffer.from(imageData, "base64");
+            image = new Uint8Array(buffer);
+        }
+        else {
+            const binaryString = atob(imageData);
+            image = new Uint8Array(binaryString.length);
+            for (let i = 0; i < binaryString.length; i++) {
+                image[i] = binaryString.charCodeAt(i);
+            }
+        }
+    }
+    return image;
+}
+function mergeTools(tools1, tools2) {
+    let tools = [];
+    let toolMap2 = tools2.reduce((map, tool) => {
+        map[tool.name] = tool;
+        return map;
+    }, {});
+    let names = [];
+    for (let i = 0; i < tools1.length; i++) {
+        let tool1 = tools1[i];
+        let tool2 = toolMap2[tool1.name];
+        if (tool2) {
+            tools.push(tool2);
+            delete toolMap2[tool1.name];
+        }
+        else {
+            tools.push(tool1);
+        }
+    }
+    for (let i = 0; i < tools2.length; i++) {
+        let tool2 = tools2[i];
+        if (toolMap2[tool2.name] && names.indexOf(tool2.name) === -1) {
+            tools.push(tool2);
+            names.push(tool2.name);
+        }
+    }
+    return tools;
+}
+function mergeAgents(agents1, agents2) {
+    let tools = [];
+    let toolMap2 = agents2.reduce((map, tool) => {
+        map[tool.Name] = tool;
+        return map;
+    }, {});
+    for (let i = 0; i < agents1.length; i++) {
+        let tool1 = agents1[i];
+        let tool2 = toolMap2[tool1.Name];
+        if (tool2) {
+            tools.push(tool2);
+            delete toolMap2[tool1.Name];
+        }
+        else {
+            tools.push(tool1);
+        }
+    }
+    for (let i = 0; i < agents2.length; i++) {
+        let tool2 = agents2[i];
+        if (toolMap2[tool2.Name]) {
+            tools.push(tool2);
+        }
+    }
+    return tools;
+}
+function sub(str, maxLength, appendPoint = true) {
+    if (!str) {
+        return "";
+    }
+    if (str.length > maxLength) {
+        return str.substring(0, maxLength) + (appendPoint ? "..." : "");
+    }
+    return str;
+}
+function fixXmlTag(code) {
+    function fixDoubleChar(code) {
+        const stack = [];
+        for (let i = 0; i < code.length; i++) {
+            let s = code[i];
+            if (s === "<") {
+                stack.push(">");
+            }
+            else if (s === ">") {
+                stack.pop();
+            }
+            else if (s === '"') {
+                if (stack[stack.length - 1] === '"') {
+                    stack.pop();
+                }
+                else {
+                    stack.push('"');
+                }
+            }
+        }
+        const missingParts = [];
+        while (stack.length > 0) {
+            missingParts.push(stack.pop());
+        }
+        return code + missingParts.join("");
+    }
+    let eIdx = code.lastIndexOf(" ");
+    let endStr = eIdx > -1 ? code.substring(eIdx + 1) : "";
+    if (code.endsWith("=")) {
+        code += '""';
+    }
+    else if (endStr == "name" ||
+        endStr == "input" ||
+        endStr == "output" ||
+        endStr == "items" ||
+        endStr == "event" ||
+        endStr == "loop") {
+        let idx1 = code.lastIndexOf(">");
+        let idx2 = code.lastIndexOf("<");
+        if (idx1 < idx2 && code.lastIndexOf(" ") > idx2) {
+            code += '=""';
+        }
+    }
+    code = fixDoubleChar(code);
+    const stack = [];
+    function isSelfClosing(tag) {
+        return tag.endsWith("/>");
+    }
+    for (let i = 0; i < code.length; i++) {
+        let s = code[i];
+        if (s === "<") {
+            const isEndTag = code[i + 1] === "/";
+            let endIndex = code.indexOf(">", i);
+            let tagContent = code.slice(i, endIndex + 1);
+            if (isSelfClosing(tagContent)) ;
+            else if (isEndTag) {
+                stack.pop();
+            }
+            else {
+                stack.push(tagContent);
+            }
+            if (endIndex == -1) {
+                break;
+            }
+            i = endIndex;
+        }
+    }
+    const missingParts = [];
+    while (stack.length > 0) {
+        const top = stack.pop();
+        if (top.startsWith("<")) {
+            let arr = top.match(/<(\w+)/);
+            const tagName = arr[1];
+            missingParts.push(`</${tagName}>`);
+        }
+        else {
+            missingParts.push(top);
+        }
+    }
+    let completedCode = code + missingParts.join("");
+    return completedCode;
+}
+
 class Context {
     constructor(taskId, config, agents, chain) {
+        this.paused = false;
         this.taskId = taskId;
         this.config = config;
         this.agents = agents;
@@ -148,13 +381,21 @@ class Context {
         this.variables = new Map();
         this.controller = new AbortController();
     }
-    checkAborted() {
+    async checkAborted() {
         // this.controller.signal.throwIfAborted();
         if (this.controller.signal.aborted) {
             const error = new Error("Operation was interrupted");
             error.name = "AbortError";
             throw error;
         }
+        while (this.paused) {
+            await sleep(500);
+            if (this.controller.signal.aborted) {
+                const error = new Error("Operation was interrupted");
+                error.name = "AbortError";
+                throw error;
+            }
+        }
     }
 }
 class AgentContext {
@@ -8443,235 +8684,6 @@ createOpenAI({
   // strict for OpenAI API
 });
 
-function sleep(time) {
-    return new Promise((resolve) => setTimeout(() => resolve(), time));
-}
-function uuidv4() {
-    return "xxxxxxxx-xxxx-4xxx-yxxx-xxxxxxxxxxxx".replace(/[xy]/g, function (c) {
-        const r = (Math.random() * 16) | 0;
-        const v = c === "x" ? r : (r & 0x3) | 0x8;
-        return v.toString(16);
-    });
-}
-function call_timeout(fun, timeout, error_callback) {
-    return new Promise(async (resolve, reject) => {
-        let timer = setTimeout(() => {
-            reject(new Error("Timeout"));
-            error_callback && error_callback("Timeout");
-        }, timeout);
-        try {
-            const result = await fun();
-            clearTimeout(timer);
-            resolve(result);
-        }
-        catch (e) {
-            clearTimeout(timer);
-            reject(e);
-            error_callback && error_callback(e + "");
-        }
-    });
-}
-function convertToolSchema(tool) {
-    if ("function" in tool) {
-        return {
-            type: "function",
-            name: tool.function.name,
-            description: tool.function.description,
-            parameters: tool.function.parameters,
-        };
-    }
-    else if ("input_schema" in tool) {
-        return {
-            type: "function",
-            name: tool.name,
-            description: tool.description,
-            parameters: tool.input_schema,
-        };
-    }
-    else if ("inputSchema" in tool) {
-        return {
-            type: "function",
-            name: tool.name,
-            description: tool.description,
-            parameters: tool.inputSchema,
-        };
-    }
-    else {
-        return {
-            type: "function",
-            name: tool.name,
-            description: tool.description,
-            parameters: tool.parameters,
-        };
-    }
-}
-function toImage(imageData) {
-    let image = null;
-    if (imageData.startsWith("http://") || imageData.startsWith("https://")) {
-        image = new URL(imageData);
-    }
-    else {
-        if (imageData.startsWith("data:image/")) {
-            imageData = imageData.substring(imageData.indexOf(",") + 1);
-        }
-        // @ts-ignore
-        if (typeof Buffer != "undefined") {
-            // @ts-ignore
-            const buffer = Buffer.from(imageData, "base64");
-            image = new Uint8Array(buffer);
-        }
-        else {
-            const binaryString = atob(imageData);
-            image = new Uint8Array(binaryString.length);
-            for (let i = 0; i < binaryString.length; i++) {
-                image[i] = binaryString.charCodeAt(i);
-            }
-        }
-    }
-    return image;
-}
-function mergeTools(tools1, tools2) {
-    let tools = [];
-    let toolMap2 = tools2.reduce((map, tool) => {
-        map[tool.name] = tool;
-        return map;
-    }, {});
-    for (let i = 0; i < tools1.length; i++) {
-        let tool1 = tools1[i];
-        let tool2 = toolMap2[tool1.name];
-        if (tool2) {
-            tools.push(tool2);
-            delete toolMap2[tool1.name];
-        }
-        else {
-            tools.push(tool1);
-        }
-    }
-    for (let i = 0; i < tools2.length; i++) {
-        let tool2 = tools2[i];
-        if (toolMap2[tool2.name]) {
-            tools.push(tool2);
-        }
-    }
-    return tools;
-}
-function mergeAgents(agents1, agents2) {
-    let tools = [];
-    let toolMap2 = agents2.reduce((map, tool) => {
-        map[tool.Name] = tool;
-        return map;
-    }, {});
-    for (let i = 0; i < agents1.length; i++) {
-        let tool1 = agents1[i];
-        let tool2 = toolMap2[tool1.Name];
-        if (tool2) {
-            tools.push(tool2);
-            delete toolMap2[tool1.Name];
-        }
-        else {
-            tools.push(tool1);
-        }
-    }
-    for (let i = 0; i < agents2.length; i++) {
-        let tool2 = agents2[i];
-        if (toolMap2[tool2.Name]) {
-            tools.push(tool2);
-        }
-    }
-    return tools;
-}
-function sub(str, maxLength, appendPoint = true) {
-    if (!str) {
-        return "";
-    }
-    if (str.length > maxLength) {
-        return str.substring(0, maxLength) + (appendPoint ? "..." : "");
-    }
-    return str;
-}
-function fixXmlTag(code) {
-    function fixDoubleChar(code) {
-        const stack = [];
-        for (let i = 0; i < code.length; i++) {
-            let s = code[i];
-            if (s === "<") {
-                stack.push(">");
-            }
-            else if (s === ">") {
-                stack.pop();
-            }
-            else if (s === '"') {
-                if (stack[stack.length - 1] === '"') {
-                    stack.pop();
-                }
-                else {
-                    stack.push('"');
-                }
-            }
-        }
-        const missingParts = [];
-        while (stack.length > 0) {
-            missingParts.push(stack.pop());
-        }
-        return code + missingParts.join("");
-    }
-    let eIdx = code.lastIndexOf(" ");
-    let endStr = eIdx > -1 ? code.substring(eIdx + 1) : "";
-    if (code.endsWith("=")) {
-        code += '""';
-    }
-    else if (endStr == "name" ||
-        endStr == "input" ||
-        endStr == "output" ||
-        endStr == "items" ||
-        endStr == "event" ||
-        endStr == "loop") {
-        let idx1 = code.lastIndexOf(">");
-        let idx2 = code.lastIndexOf("<");
-        if (idx1 < idx2 && code.lastIndexOf(" ") > idx2) {
-            code += '=""';
-        }
-    }
-    code = fixDoubleChar(code);
-    const stack = [];
-    function isSelfClosing(tag) {
-        return tag.endsWith("/>");
-    }
-    for (let i = 0; i < code.length; i++) {
-        let s = code[i];
-        if (s === "<") {
-            const isEndTag = code[i + 1] === "/";
-            let endIndex = code.indexOf(">", i);
-            let tagContent = code.slice(i, endIndex + 1);
-            if (isSelfClosing(tagContent)) ;
-            else if (isEndTag) {
-                stack.pop();
-            }
-            else {
-                stack.push(tagContent);
-            }
-            if (endIndex == -1) {
-                break;
-            }
-            i = endIndex;
-        }
-    }
-    const missingParts = [];
-    while (stack.length > 0) {
-        const top = stack.pop();
-        if (top.startsWith("<")) {
-            let arr = top.match(/<(\w+)/);
-            const tagName = arr[1];
-            missingParts.push(`</${tagName}>`);
-        }
-        else {
-            missingParts.push(top);
-        }
-    }
-    let completedCode = code + missingParts.join("");
-    return completedCode;
-}
-
 // src/anthropic-provider.ts
 var anthropicErrorDataSchema = z.object({
   type: z.literal("error"),
@@ -14736,7 +14748,7 @@ class RetryLanguageModel {
                 toolChoice: request.toolChoice,
             },
             prompt: request.messages,
-            maxTokens: request.maxTokens || config.maxTokens,
+            maxTokens: request.maxTokens,
             temperature: request.temperature,
             topP: request.topP,
             topK: request.topK,
@@ -14745,6 +14757,7 @@ class RetryLanguageModel {
         });
     }
     async doGenerate(options) {
+        const maxTokens = options.maxTokens;
         const names = [...this.names, ...this.names];
         for (let i = 0; i < names.length; i++) {
             const name = names[i];
@@ -14752,6 +14765,10 @@ class RetryLanguageModel {
             if (!llm) {
                 continue;
             }
+            if (!maxTokens) {
+                options.maxTokens =
+                    this.llms[name].config?.maxTokens || config.maxTokens;
+            }
             try {
                 let result = await llm.doGenerate(options);
                 if (Log.isEnableDebug()) {
@@ -14783,7 +14800,7 @@ class RetryLanguageModel {
                 toolChoice: request.toolChoice,
             },
             prompt: request.messages,
-            maxTokens: request.maxTokens || config.maxTokens,
+            maxTokens: request.maxTokens,
             temperature: request.temperature,
             topP: request.topP,
             topK: request.topK,
@@ -14792,6 +14809,7 @@ class RetryLanguageModel {
         });
     }
     async doStream(options) {
+        const maxTokens = options.maxTokens;
         const names = [...this.names, ...this.names];
         for (let i = 0; i < names.length; i++) {
             const name = names[i];
@@ -14799,6 +14817,10 @@ class RetryLanguageModel {
             if (!llm) {
                 continue;
             }
+            if (!maxTokens) {
+                options.maxTokens =
+                    this.llms[name].config?.maxTokens || config.maxTokens;
+            }
             try {
                 const controller = new AbortController();
                 const signal = options.abortSignal
@@ -14855,10 +14877,7 @@ class RetryLanguageModel {
             return createOpenAI({
                 apiKey: llm.apiKey,
                 baseURL: llm.config?.baseURL,
-            }).languageModel(llm.model, {
-                // disable_parallel_tool_use
-                parallelToolCalls: llm.config?.parallelToolCalls || false,
-            });
+            }).languageModel(llm.model);
         }
         else if (llm.provider == "anthropic") {
             return createAnthropic({
@@ -17503,10 +17522,17 @@ function extractAgentXmlNode(agentXml, nodeId) {
     let nodesNode = doc.getElementsByTagName("nodes");
     if (nodesNode.length > 0) {
         let nodes = nodesNode[0].childNodes;
+        let _nodeId = 0;
         for (let i = 0; i < nodes.length; i++) {
             let node = nodes[i];
-            if (node.nodeType == 1 && node.getAttribute("id") == nodeId + "") {
-                return node;
+            if (node.nodeType == 1) {
+                if (node.getAttribute("id") == null || node.getAttribute("id") == "") {
+                    node.setAttribute("id", _nodeId + "");
+                }
+                _nodeId++;
+                if (node.getAttribute("id") == nodeId + "") {
+                    return node;
+                }
             }
         }
     }
@@ -17660,14 +17686,14 @@ async function compressAgentMessages(agentContext, rlm, messages, tools) {
     if (callback) {
         await callback.onMessage({
             taskId: agentContext.context.taskId,
-            agentName: toolCall.toolName,
+            agentName: agentContext.agent.Name,
             nodeId: agentContext.agentChain.agent.id,
             type: "tool_result",
             toolId: toolCall.toolCallId,
             toolName: toolCall.toolName,
             params: args,
             toolResult: toolResult,
-        });
+        }, agentContext);
     }
     // handle original messages
     let firstToolIndex = 3;
@@ -17686,6 +17712,7 @@ async function compressAgentMessages(agentContext, rlm, messages, tools) {
 function handleLargeContextMessages(messages) {
     let imageNum = 0;
     let fileNum = 0;
+    let maxNum = config.maxDialogueImgFileNum;
     let longTextTools = {};
     for (let i = messages.length - 1; i >= 0; i--) {
         let message = messages[i];
@@ -17693,7 +17720,7 @@ function handleLargeContextMessages(messages) {
             for (let j = 0; j < message.content.length; j++) {
                 let content = message.content[j];
                 if (content.type == "image") {
-                    if (++imageNum == 1) {
+                    if (++imageNum <= maxNum) {
                         break;
                     }
                     content = {
@@ -17703,7 +17730,7 @@ function handleLargeContextMessages(messages) {
                     message.content[j] = content;
                 }
                 else if (content.type == "file") {
-                    if (++fileNum == 1) {
+                    if (++fileNum <= maxNum) {
                         break;
                     }
                     content = {
@@ -17724,7 +17751,7 @@ function handleLargeContextMessages(messages) {
                 for (let r = 0; r < toolContent.length; r++) {
                     let _content = toolContent[r];
                     if (_content.type == "image") {
-                        if (++imageNum == 1) {
+                        if (++imageNum <= maxNum) {
                             break;
                         }
                         _content = {
@@ -17746,7 +17773,7 @@ function handleLargeContextMessages(messages) {
                         }
                         _content = {
                             type: "text",
-                            text: _content.text.substring(0, config.shortTextLength) + "...",
+                            text: _content.text.substring(0, config.largeTextLength) + "...",
                         };
                         toolContent[r] = _content;
                     }
@@ -17839,7 +17866,7 @@ const TOOL_NAME$4 = "foreach_task";
 class ForeachTaskTool {
     constructor() {
         this.name = TOOL_NAME$4;
-        this.description = `When executing the \`forEach\` node, please use it to complete the tasks corresponding to that forEach node, which will complete all tasks under the entire forEach node.`;
+        this.description = `When executing the \`forEach\` node, please use the current tool for counting to ensure tasks are executed sequentially, the tool needs to be called with each loop iteration.`;
         this.parameters = {
             type: "object",
             properties: {
@@ -17847,12 +17874,19 @@ class ForeachTaskTool {
                     type: "number",
                     description: "forEach node ID.",
                 },
+                progress: {
+                    type: "string",
+                    description: "Current execution progress.",
+                },
+                next_step: {
+                    type: "string",
+                    description: "Next task description.",
+                },
             },
-            required: ["nodeId"],
+            required: ["nodeId", "progress", "next_step"],
         };
     }
     async execute(args, agentContext) {
-        // 调用 forEach Agent 单独逻辑, 根据上下文判断并循环执行
         let nodeId = args.nodeId;
         let agentXml = agentContext.agentChain.agent.xml;
         let node = extractAgentXmlNode(agentXml, nodeId);
@@ -17863,10 +17897,27 @@ class ForeachTaskTool {
             throw new Error("Node ID is not a forEach node: " + nodeId);
         }
         let items = node.getAttribute("items");
+        let varValue = null;
+        let resultText = "Recorded";
         if (items && items != "list") {
-            agentContext.context.variables.get(items.trim());
+            varValue = agentContext.context.variables.get(items.trim());
+            if (varValue) {
+                let key = "foreach_" + nodeId;
+                let loop_count = agentContext.variables.get(key) || 0;
+                if (loop_count % 5 == 0) {
+                    resultText = `Variable information associated with the current loop task.\nvariable_name: ${items.trim()}\nvariable_value: ${varValue}`;
+                }
+                agentContext.variables.set(key, ++loop_count);
+            }
         }
-        return null;
+        return {
+            content: [
+                {
+                    type: "text",
+                    text: resultText,
+                },
+            ],
+        };
     }
 }
 
@@ -17938,6 +17989,11 @@ request_help: Request assistance from the user; for instance, when an operation
                     break;
                 case "request_help":
                     if (callback.onHumanHelp) {
+                        if (args.helpType == "request_login" &&
+                            (await this.checkIsLogined(agentContext))) {
+                            resultText = "Already logged in";
+                            break;
+                        }
                         let result = await callback.onHumanHelp(agentContext, (args.helpType || "request_assistance"), args.prompt);
                         resultText = `request_help result: ${result ? "Solved" : "Unresolved"}`;
                     }
@@ -17966,16 +18022,56 @@ request_help: Request assistance from the user; for instance, when an operation
             };
         }
     }
+    async checkIsLogined(agentContext) {
+        let screenshot = agentContext.agent["screenshot"];
+        if (!screenshot) {
+            return false;
+        }
+        try {
+            let imageResult = (await screenshot.call(agentContext.agent, agentContext));
+            let rlm = new RetryLanguageModel(agentContext.context.config.llms, agentContext.agent.Llms);
+            let image = toImage(imageResult.imageBase64);
+            let request = {
+                messages: [
+                    {
+                        role: "user",
+                        content: [
+                            {
+                                type: "image",
+                                image: image,
+                                mimeType: imageResult.imageType,
+                            },
+                            {
+                                type: "text",
+                                text: "Check if the current website is logged in. If not logged in, output `NOT_LOGIN`. If logged in, output `LOGGED_IN`. Output directly without explanation.",
+                            },
+                        ],
+                    },
+                ],
+                abortSignal: agentContext.context.controller.signal,
+            };
+            let result = await rlm.call(request);
+            return result.text && result.text.indexOf("LOGGED_IN") > -1;
+        }
+        catch (error) {
+            console.error("Error auto checking login status:", error);
+            return false;
+        }
+    }
 }
 
 const TOOL_NAME$2 = "task_node_status";
 class TaskNodeStatusTool {
     constructor() {
         this.name = TOOL_NAME$2;
-        this.description = `After completing each step of the task, you need to call this tool to update the status of the task node.`;
+        this.description = `After completing each step of the task, you need to call this tool to update the status of the task node, and think about the tasks to be processed and the next action plan.`;
         this.parameters = {
             type: "object",
             properties: {
+                thought: {
+                    type: "string",
+                    description: "Current thinking content, which can be analysis of the problem, assumptions, insights, reflections, or a summary of the previous, suggest the next action step to be taken, which should be specific, executable, and verifiable."
+                },
                 doneIds: {
                     type: "array",
                     description: "List of completed node IDs.",
@@ -17991,7 +18087,7 @@ class TaskNodeStatusTool {
                     },
                 },
             },
-            required: ["doneIds", "todoIds"],
+            required: ["thought", "doneIds", "todoIds"],
         };
     }
     async execute(args, agentContext) {
@@ -18099,21 +18195,156 @@ const TOOL_NAME = "watch_trigger";
 class WatchTriggerTool {
     constructor() {
         this.name = TOOL_NAME;
-        this.description = `When executing the \`watch\` node, please use it to complete the tasks corresponding to that watch node. It will complete all tasks under the entire watch node.`;
+        this.description = `When executing the \`watch\` node, please use it to monitor DOM element changes, it will block the listener until the element changes or times out.`;
         this.parameters = {
             type: "object",
             properties: {
                 nodeId: {
                     type: "number",
-                    description: "forEach node ID."
+                    description: "watch node ID.",
+                },
+                watch_area: {
+                    type: "array",
+                    description: "Element changes in monitoring area, eg: [x, y, width, height].",
+                    items: {
+                        type: "number",
+                    },
+                },
+                watch_index: {
+                    type: "array",
+                    description: "The index of elements to be monitoring multiple elements simultaneously.",
+                    items: {
+                        type: "number",
+                    },
+                },
+                frequency: {
+                    type: "number",
+                    description: "Check frequency, how many seconds between each check, default 1 seconds.",
+                    default: 1,
+                    minimum: 0.5,
+                    maximum: 30,
+                },
+                timeout: {
+                    type: "number",
+                    description: "Timeout in minute, default 5 minutes.",
+                    default: 5,
+                    minimum: 1,
+                    maximum: 30,
                 },
             },
             required: ["nodeId"],
         };
     }
     async execute(args, agentContext) {
-        // TODO Listen for changes to the DOM or file, and execute nodes
-        return null;
+        let nodeId = args.nodeId;
+        let agentXml = agentContext.agentChain.agent.xml;
+        let node = extractAgentXmlNode(agentXml, nodeId);
+        if (node == null) {
+            throw new Error("Node ID does not exist: " + nodeId);
+        }
+        if (node.tagName !== "watch") {
+            throw new Error("Node ID is not a watch node: " + nodeId);
+        }
+        let task_description = node.getElementsByTagName("description")[0]?.textContent || "";
+        if (!task_description) {
+            return {
+                content: [
+                    {
+                        type: "text",
+                        text: "The watch node does not have a description, skip.",
+                    },
+                ],
+            };
+        }
+        const screenshot = agentContext.agent["screenshot"];
+        const image1Result = (await screenshot.call(agentContext.agent, agentContext));
+        const image1 = toImage(image1Result.imageBase64);
+        const start = new Date().getTime();
+        const timeout = (args.timeout || 5) * 60000;
+        const frequency = Math.max(500, (args.frequency = args.frequency || 1) * 1000);
+        let rlm = new RetryLanguageModel(agentContext.context.config.llms, agentContext.agent.Llms);
+        while (new Date().getTime() - start < timeout) {
+            await agentContext.context.checkAborted();
+            await new Promise((resolve) => setTimeout(resolve, frequency));
+            const image2Result = (await screenshot.call(agentContext.agent, agentContext));
+            const image2 = toImage(image2Result.imageBase64);
+            const changeResult = await this.is_dom_change(agentContext, rlm, image1, image1Result.imageType, image2, image2Result.imageType, task_description);
+            if (changeResult.changed) {
+                return {
+                    content: [
+                        {
+                            type: "text",
+                            text: changeResult.changeInfo || "DOM change detected.",
+                        },
+                    ],
+                };
+            }
+        }
+        return {
+            content: [
+                {
+                    type: "text",
+                    text: "Timeout reached, no DOM changes detected.",
+                },
+            ],
+        };
+    }
+    async is_dom_change(agentContext, rlm, image1, image1Type, image2, image2Type, task_description) {
+        try {
+            let request = {
+                messages: [
+                    {
+                        role: "system",
+                        content: `You are a tool for detecting element changes. Given a task description, compare two images to determine whether the changes described in the task have occurred.
+If the changes have occurred, return an json with \`changed\` set to true and \`changeInfo\` containing a description of the changes. If no changes have occurred, return an object with \`changed\` set to false.
+
+## Example
+User: Monitor new messages in group chat
+### No changes detected
+Output:
+{
+  "changed": false
+}
+### Change detected
+Output:
+{
+  "changed": true,
+  "changeInfo": "New message received in the group chat. The message content is: 'Hello, how are you?'"
+}`,
+                    },
+                    {
+                        role: "user",
+                        content: [
+                            {
+                                type: "image",
+                                image: image1,
+                                mimeType: image1Type,
+                            },
+                            {
+                                type: "image",
+                                image: image2,
+                                mimeType: image2Type,
+                            },
+                            {
+                                type: "text",
+                                text: task_description,
+                            },
+                        ],
+                    },
+                ],
+                abortSignal: agentContext.context.controller.signal,
+            };
+            const result = await rlm.call(request);
+            let resultText = result.text || "{}";
+            resultText = resultText.substring(resultText.indexOf("{"), resultText.lastIndexOf("}") + 1);
+            return JSON.parse(resultText);
+        }
+        catch (error) {
+            Log.error("Error in is_dom_change:", error);
+        }
+        return {
+            changed: false,
+        };
     }
 }
 
@@ -18131,7 +18362,7 @@ class McpTool {
 
 const AGENT_SYSTEM_TEMPLATE = `
 You are {name}, an autonomous AI agent for {agent} agent.
-UTC datetime: {datetime}
+Current datetime: {datetime}
 
 # Task Description
 {description}
@@ -18175,7 +18406,7 @@ repetitive tasks, when executing to the forEach node, require the use of the \`$
 `;
 const WATCH_NODE = `
     <!-- monitor task node, the loop attribute specifies whether to listen in a loop or listen once -->
-    <watch event="dom or file" loop="true">
+    <watch event="dom" loop="true">
       <description>Monitor task description</description>
       <trigger>
         <node>Trigger step node</node>
@@ -18184,7 +18415,7 @@ const WATCH_NODE = `
     </watch>`;
 const WATCH_PROMPT = `
 * watch node
-monitor changes in webpage DOM or file content, when executing to the watch node, require the use of the \`${TOOL_NAME}\` tool.
+monitor changes in webpage DOM elements, when executing to the watch node, require the use of the \`${TOOL_NAME}\` tool.
 `;
 function getAgentSystemPrompt(agent, agentNode, context, tools, extSysPrompt) {
     let prompt = "";
@@ -18220,7 +18451,7 @@ function getAgentSystemPrompt(agent, agentNode, context, tools, extSysPrompt) {
     }
     if (context.chain.agents.length > 1) {
         prompt += "\n Main task: " + context.chain.taskPrompt;
-        prompt += "\n# Pre-task execution results";
+        prompt += "\n\n# Pre-task execution results";
         for (let i = 0; i < context.chain.agents.length; i++) {
             let agentChain = context.chain.agents[i];
             if (agentChain.agentResult) {
@@ -18234,7 +18465,7 @@ function getAgentSystemPrompt(agent, agentNode, context, tools, extSysPrompt) {
     return AGENT_SYSTEM_TEMPLATE.replace("{name}", config.name)
         .replace("{agent}", agent.Name)
         .replace("{description}", agent.Description)
-        .replace("{datetime}", new Date().toISOString())
+        .replace("{datetime}", new Date().toLocaleString())
         .replace("{prompt}", prompt)
         .replace("{nodePrompt}", nodePrompt)
         .trim();
@@ -18288,7 +18519,7 @@ class Agent {
         let rlm = new RetryLanguageModel(context.config.llms, this.llms);
         let agentTools = tools;
         while (loopNum < maxReactNum) {
-            context.checkAborted();
+            await context.checkAborted();
             if (mcpClient) {
                 let controlMcp = await this.controlMcpTools(agentContext, messages, loopNum);
                 if (controlMcp.mcpTools) {
@@ -18299,7 +18530,7 @@ class Agent {
                 }
             }
             await this.handleMessages(agentContext, messages, tools);
-            let results = await callLLM(agentContext, rlm, messages, this.convertTools(agentTools));
+            let results = await callLLM(agentContext, rlm, messages, this.convertTools(agentTools), false, undefined, false, this.callback);
             let finalResult = await this.handleCallResult(agentContext, messages, agentTools, results);
             if (finalResult) {
                 return finalResult;
@@ -18355,17 +18586,18 @@ class Agent {
                     throw e;
                 }
             }
-            if (context.config.callback) {
-                await context.config.callback.onMessage({
+            const callback = this.callback || context.config.callback;
+            if (callback) {
+                await callback.onMessage({
                     taskId: context.taskId,
-                    agentName: result.toolName,
+                    agentName: agentContext.agent.Name,
                     nodeId: agentContext.agentChain.agent.id,
                     type: "tool_result",
                     toolId: result.toolCallId,
                     toolName: result.toolName,
                     params: result.args || {},
                     toolResult: toolResult,
-                });
+                }, agentContext);
             }
             let llmToolResult = this.convertToolResult(result, toolResult, user_messages);
             toolResults.push(llmToolResult);
@@ -18394,6 +18626,14 @@ class Agent {
         if (hasVariable) {
             tools.push(new VariableStorageTool());
         }
+        let hasForeach = agentNodeXml.indexOf("</forEach>") > -1;
+        if (hasForeach) {
+            tools.push(new ForeachTaskTool());
+        }
+        let hasWatch = agentNodeXml.indexOf("</watch>") > -1;
+        if (hasWatch) {
+            tools.push(new WatchTriggerTool());
+        }
         let toolNames = this.tools.map((tool) => tool.name);
         return tools.filter((tool) => toolNames.indexOf(tool.name) == -1);
     }
@@ -18557,6 +18797,12 @@ class Agent {
         }
         return this.tools;
     }
+    addTool(tool) {
+        this.tools.push(tool);
+    }
+    get Llms() {
+        return this.llms;
+    }
     get Name() {
         return this.name;
     }
@@ -18573,14 +18819,15 @@ class Agent {
         return this.mcpClient;
     }
 }
-async function callLLM(agentContext, rlm, messages, tools, noCompress, toolChoice, retry) {
+async function callLLM(agentContext, rlm, messages, tools, noCompress, toolChoice, retry, callback) {
     if (messages.length >= config.compressThreshold && !noCompress) {
         await compressAgentMessages(agentContext, rlm, messages, tools);
     }
     let context = agentContext.context;
     let agentChain = agentContext.agentChain;
     let agentNode = agentChain.agent;
-    let streamCallback = context.config.callback || {
+    let streamCallback = callback ||
+        context.config.callback || {
         onMessage: async () => { },
     };
     let request = {
@@ -18600,7 +18847,7 @@ async function callLLM(agentContext, rlm, messages, tools, noCompress, toolChoic
     const reader = result.stream.getReader();
     try {
         while (true) {
-            context.checkAborted();
+            await context.checkAborted();
             const { done, value } = await reader.read();
             if (done) {
                 break;
@@ -18617,7 +18864,7 @@ async function callLLM(agentContext, rlm, messages, tools, noCompress, toolChoic
                         streamId,
                         streamDone: false,
                         text: streamText,
-                    });
+                    }, agentContext);
                     break;
                 }
                 case "reasoning": {
@@ -18630,7 +18877,7 @@ async function callLLM(agentContext, rlm, messages, tools, noCompress, toolChoic
                         streamId,
                         streamDone: false,
                         text: thinkText,
-                    });
+                    }, agentContext);
                     break;
                 }
                 case "tool-call-delta": {
@@ -18644,7 +18891,7 @@ async function callLLM(agentContext, rlm, messages, tools, noCompress, toolChoic
                             streamId,
                             streamDone: true,
                             text: streamText,
-                        });
+                        }, agentContext);
                     }
                     toolArgsText += chunk.argsTextDelta || "";
                     await streamCallback.onMessage({
@@ -18655,7 +18902,7 @@ async function callLLM(agentContext, rlm, messages, tools, noCompress, toolChoic
                         toolId: chunk.toolCallId,
                         toolName: chunk.toolName,
                         paramsText: toolArgsText,
-                    });
+                    }, agentContext);
                     break;
                 }
                 case "tool-call": {
@@ -18670,7 +18917,7 @@ async function callLLM(agentContext, rlm, messages, tools, noCompress, toolChoic
                         toolName: chunk.toolName,
                         params: args,
                     };
-                    await streamCallback.onMessage(message);
+                    await streamCallback.onMessage(message, agentContext);
                     toolParts.push({
                         type: "tool-call",
                         toolCallId: chunk.toolCallId,
@@ -18687,7 +18934,7 @@ async function callLLM(agentContext, rlm, messages, tools, noCompress, toolChoic
                         type: "file",
                         mimeType: chunk.mimeType,
                         data: chunk.data,
-                    });
+                    }, agentContext);
                     break;
                 }
                 case "error": {
@@ -18698,7 +18945,7 @@ async function callLLM(agentContext, rlm, messages, tools, noCompress, toolChoic
                         nodeId: agentNode.id,
                         type: "error",
                         error: chunk.error,
-                    });
+                    }, agentContext);
                     throw new Error("Plan Error");
                 }
                 case "finish": {
@@ -18712,7 +18959,7 @@ async function callLLM(agentContext, rlm, messages, tools, noCompress, toolChoic
                             streamId,
                             streamDone: true,
                             text: streamText,
-                        });
+                        }, agentContext);
                     }
                     await streamCallback.onMessage({
                         taskId: context.taskId,
@@ -18721,10 +18968,13 @@ async function callLLM(agentContext, rlm, messages, tools, noCompress, toolChoic
                         type: "finish",
                         finishReason: chunk.finishReason,
                         usage: chunk.usage,
-                    });
-                    if (chunk.finishReason === "length" && messages.length >= 10 && !noCompress && !retry) {
+                    }, agentContext);
+                    if (chunk.finishReason === "length" &&
+                        messages.length >= 10 &&
+                        !noCompress &&
+                        !retry) {
                         await compressAgentMessages(agentContext, rlm, messages, tools);
-                        return callLLM(agentContext, rlm, messages, tools, noCompress, toolChoice, true);
+                        return callLLM(agentContext, rlm, messages, tools, noCompress, toolChoice, true, streamCallback);
                     }
                     break;
                 }
@@ -18743,11 +18993,11 @@ async function callLLM(agentContext, rlm, messages, tools, noCompress, toolChoic
         : toolParts;
 }
 
-const AGENT_NAME$4 = "Chat";
+const AGENT_NAME$5 = "Chat";
 class BaseChatAgent extends Agent {
     constructor(llms, ext_tools, mcpClient) {
         super({
-            name: AGENT_NAME$4,
+            name: AGENT_NAME$5,
             description: "You are a helpful assistant.",
             tools: ext_tools || [],
             llms: llms,
@@ -18759,7 +19009,7 @@ class BaseChatAgent extends Agent {
 
 const PLAN_SYSTEM_TEMPLATE = `
 You are {name}, an autonomous AI Agent Planner.
-UTC datetime: {datetime}
+Current datetime: {datetime}
 
 ## Task Description
 Your task is to understand the user's requirements, dynamically plan the user's tasks based on the Agent list, and please follow the steps below:
@@ -18793,8 +19043,8 @@ Your task is to understand the user's requirements, dynamically plan the user's
         <forEach items="list or variable name">
           <node>forEach step node</node>
         </forEach>
-        <!-- When you need to monitor changes in webpage DOM or file content, you can use \`Watch\`, the loop attribute specifies whether to listen in a loop or listen once. -->
-        <watch event="dom or file" loop="true">
+        <!-- When you need to monitor changes in webpage DOM elements, you can use \`Watch\`, the loop attribute specifies whether to listen in a loop or listen once. -->
+        <watch event="dom" loop="true">
           <description>Monitor task description</description>
           <trigger>
             <node>Trigger step node</node>
@@ -18916,12 +19166,12 @@ Output result:
 ];
 const PLAN_USER_TEMPLATE = `
 User Platform: {platform}
-Task Description: {taskPrompt}
+Task Description: {task_prompt}
 `;
 const PLAN_USER_TASK_WEBSITE_TEMPLATE = `
 User Platform: {platform}
 Task Website: {task_website}
-Task Description: {taskPrompt}
+Task Description: {task_prompt}
 `;
 async function getPlanSystemPrompt(context) {
     let agents_prompt = "";
@@ -18940,7 +19190,7 @@ async function getPlanSystemPrompt(context) {
                 "\n</agent>\n\n";
     }
     let example_prompt = "";
-    let hasChatAgent = context.agents.filter((a) => a.Name == AGENT_NAME$4).length > 0;
+    let hasChatAgent = context.agents.filter((a) => a.Name == AGENT_NAME$5).length > 0;
     const example_list = hasChatAgent
         ? [PLAN_CHAT_EXAMPLE, ...PLAN_EXAMPLE_LIST]
         : [...PLAN_EXAMPLE_LIST];
@@ -18949,22 +19199,26 @@ async function getPlanSystemPrompt(context) {
     }
     return PLAN_SYSTEM_TEMPLATE.replace("{name}", config.name)
         .replace("{agents}", agents_prompt.trim())
-        .replace("{datetime}", new Date().toISOString())
+        .replace("{datetime}", new Date().toLocaleString())
         .replace("{example_prompt}", example_prompt)
         .trim();
 }
-function getPlanUserPrompt(taskPrompt, task_website) {
+function getPlanUserPrompt(task_prompt, task_website, ext_prompt) {
+    let prompt = "";
     if (task_website) {
-        return PLAN_USER_TASK_WEBSITE_TEMPLATE.replace("{taskPrompt}", taskPrompt)
+        prompt = PLAN_USER_TASK_WEBSITE_TEMPLATE.replace("{task_prompt}", task_prompt)
             .replace("{platform}", config.platform)
-            .replace("{task_website}", task_website)
-            .trim();
+            .replace("{task_website}", task_website);
     }
     else {
-        return PLAN_USER_TEMPLATE.replace("{taskPrompt}", taskPrompt)
-            .replace("{platform}", config.platform)
-            .trim();
+        prompt = PLAN_USER_TEMPLATE.replace("{task_prompt}", task_prompt)
+            .replace("{platform}", config.platform);
     }
+    prompt = prompt.trim();
+    if (ext_prompt) {
+        prompt += `\n${ext_prompt.trim()}`;
+    }
+    return prompt;
 }
 
 class Planner {
@@ -19007,7 +19261,7 @@ class Planner {
                     content: [
                         {
                             type: "text",
-                            text: getPlanUserPrompt(taskPrompt, this.context.variables.get("task_website")),
+                            text: getPlanUserPrompt(taskPrompt, this.context.variables.get("task_website"), this.context.variables.get("plan_ext_prompt")),
                         },
                     ],
                 },
@@ -19024,7 +19278,7 @@ class Planner {
         let streamText = "";
         try {
             while (true) {
-                this.context.checkAborted();
+                await this.context.checkAborted();
                 const { done, value } = await reader.read();
                 if (done) {
                     break;
@@ -19163,7 +19417,7 @@ class Eko {
         }, {});
         let results = [];
         for (let i = 0; i < workflow.agents.length; i++) {
-            context.checkAborted();
+            await context.checkAborted();
             let agentNode = workflow.agents[i];
             let agent = agentMap[agentNode.name];
             if (!agent) {
@@ -19173,6 +19427,9 @@ class Eko {
             context.chain.push(agentChain);
             agent.result = await agent.run(context, agentChain);
             results.push(agent.result);
+            if (agentNode.name === "Timer") {
+                break;
+            }
         }
         return {
             success: true,
@@ -19200,6 +19457,16 @@ class Eko {
             return false;
         }
     }
+    pauseTask(taskId, paused) {
+        let context = this.taskMap.get(taskId);
+        if (context) {
+            context.paused = paused;
+            return true;
+        }
+        else {
+            return false;
+        }
+    }
     addAgent(agent) {
         this.config.agents = this.config.agents || [];
         this.config.agents.push(agent);
@@ -19428,7 +19695,7 @@ function parseChunk(chunk) {
     return chunk_obj;
 }
 
-const AGENT_NAME$3 = "File";
+const AGENT_NAME$4 = "File";
 class BaseFileAgent extends Agent {
     constructor(work_path, llms, ext_tools, mcpClient, planDescription) {
         const _tools_ = [];
@@ -19436,7 +19703,7 @@ class BaseFileAgent extends Agent {
             ? `Your default working path is: ${work_path}`
             : "";
         super({
-            name: AGENT_NAME$3,
+            name: AGENT_NAME$4,
             description: `You are a file agent, handling file-related tasks such as creating, finding, reading, modifying files, etc.${prompt}`,
             tools: _tools_,
             llms: llms,
@@ -19450,6 +19717,37 @@ class BaseFileAgent extends Agent {
         }
         init_tools.forEach((tool) => _tools_.push(tool));
     }
+    async do_file_read(agentContext, path, write_variable) {
+        let file_context = await this.file_read(agentContext, path);
+        if (file_context && file_context.length > config.fileTextMaxLength) {
+            file_context = file_context.substring(0, config.fileTextMaxLength) + "...";
+        }
+        if (write_variable) {
+            agentContext.context.variables.set(write_variable, file_context);
+        }
+        return {
+            file_context: file_context,
+            write_variable: write_variable,
+        };
+    }
+    async do_file_write(agentContext, path, append, content, from_variable) {
+        if (content == null && from_variable == null) {
+            throw new Error(`content and from_variable cannot be both empty, cannot write to file ${path}`);
+        }
+        if (from_variable) {
+            let variable_value = agentContext.context.variables.get(from_variable) || "";
+            if (variable_value) {
+                content = variable_value;
+            }
+            if (!content) {
+                throw new Error(`Variable ${from_variable} is empty, cannot write to file ${path}`);
+            }
+        }
+        if (!content) {
+            throw new Error(`content is empty, cannot write to file ${path}`);
+        }
+        return await this.file_write(agentContext, path, content || "", append);
+    }
     buildInitTools() {
         return [
             {
@@ -19479,16 +19777,20 @@ class BaseFileAgent extends Agent {
                             type: "string",
                             description: "File path",
                         },
+                        write_variable: {
+                            type: "string",
+                            description: "Variable name, the content after reading is simultaneously written to the variable, facilitating direct loading from the variable in subsequent operations.",
+                        },
                     },
                     required: ["path"],
                 },
                 execute: async (args, agentContext) => {
-                    return await this.callInnerTool(() => this.file_read(agentContext, args.path));
+                    return await this.callInnerTool(() => this.do_file_read(agentContext, args.path, args.write_variable));
                 },
             },
             {
                 name: "file_write",
-                description: "Overwrite or append content to a file. Use for creating new files, appending content, or modifying existing files, only supports txt/md/csv or other text formats.",
+                description: "Overwrite or append content to a file. Use for creating new files, appending content, or modifying existing files, only supports txt/md/json/csv or other text formats.",
                 parameters: {
                     type: "object",
                     properties: {
@@ -19496,20 +19798,24 @@ class BaseFileAgent extends Agent {
                             type: "string",
                             description: "File path",
                         },
-                        content: {
-                            type: "string",
-                            description: "Text content",
-                        },
                         append: {
                             type: "boolean",
                             description: "(Optional) Whether to use append mode",
                             default: false,
                         },
+                        content: {
+                            type: "string",
+                            description: "Text content, write content directly to the file.",
+                        },
+                        from_variable: {
+                            type: "string",
+                            description: "Variable name, read content from the variable and write it.",
+                        },
                     },
-                    required: ["path", "content"],
+                    required: ["path"],
                 },
                 execute: async (args, agentContext) => {
-                    return await this.callInnerTool(() => this.file_write(agentContext, args.path, args.content, (args.append || false)));
+                    return await this.callInnerTool(() => this.do_file_write(agentContext, args.path, (args.append || false), args.content, args.from_variable));
                 },
             },
             {
@@ -19562,12 +19868,12 @@ class BaseFileAgent extends Agent {
     }
 }
 
-const AGENT_NAME$2 = "Shell";
+const AGENT_NAME$3 = "Shell";
 class BaseShellAgent extends Agent {
     constructor(llms, ext_tools, mcpClient, planDescription) {
         const _tools_ = [];
         super({
-            name: AGENT_NAME$2,
+            name: AGENT_NAME$3,
             description: `Run commands in a bash shell,
 * You must first call create_session to create a new session when using it for the first time.
 * Please execute delete commands with caution, and never perform dangerous operations like \`rm -rf /\`.
@@ -19644,7 +19950,45 @@ class BaseShellAgent extends Agent {
     }
 }
 
+const AGENT_NAME$2 = "Timer";
 class BaseTimerAgent extends Agent {
+    constructor(llms, ext_tools, mcpClient) {
+        super({
+            name: AGENT_NAME$2,
+            description: "You are a scheduled task scheduling agent.",
+            tools: ext_tools || [],
+            llms: llms,
+            mcpClient: mcpClient,
+        });
+        this.addTool(this.schedule_tool());
+    }
+    schedule_tool() {
+        return {
+            name: "task_schedule",
+            description: "Task scheduled trigger, the task is triggered at a scheduled time and will automatically create a scheduled task for execution.",
+            parameters: {
+                type: "object",
+                properties: {
+                    trigger_description: {
+                        type: "string",
+                        description: "Trigger time description.",
+                    },
+                    task_description: {
+                        type: "string",
+                        description: "Main task description, excluding trigger time.",
+                    },
+                    cron: {
+                        type: "string",
+                        description: "The cron expression of the trigger, for example, '0 9 * * *' indicates that it triggers at 9 a.m. every day.",
+                    },
+                },
+                required: ["cron"],
+            },
+            execute: async (args, agentContext) => {
+                return await this.callInnerTool(() => this.task_schedule(agentContext, args.trigger_description, args.task_description, args.cron));
+            },
+        };
+    }
 }
 
 const AGENT_NAME$1 = "Computer";
@@ -19873,7 +20217,7 @@ This is a computer GUI interface, observe the execution through screenshots, and
                             description: "Duration in millisecond",
                             default: 500,
                             minimum: 200,
-                            maximum: 2000,
+                            maximum: 10000,
                         },
                     },
                     required: ["duration"],
@@ -20066,6 +20410,9 @@ class BaseBrowserAgent extends Agent {
     }
     toolUseNames(messages) {
         let toolNames = [];
+        if (!messages) {
+            return toolNames;
+        }
         for (let i = 0; i < messages.length; i++) {
             let message = messages[i];
             if (message.role == "tool") {
@@ -20114,8 +20461,8 @@ function run_build_dom_tree() {
                 'name',
                 'role',
                 'class',
-                // 'href',
-                'tabindex',
+                'src',
+                'href',
                 'aria-label',
                 'placeholder',
                 'value',
@@ -20164,6 +20511,12 @@ function run_build_dom_tree() {
                                 let classList = value.split(" ").slice(0, 3);
                                 value = classList.join(" ");
                             }
+                            else if ((key == "src" || key == "href") && value && value.length > 200) {
+                                continue;
+                            }
+                            else if ((key == "src" || key == "href") && value && value.startsWith("/")) {
+                                value = window.location.origin + value;
+                            }
                             if (key && value) {
                                 attributes_str += ` ${key}="${value}"`;
                             }
@@ -20725,9 +21078,21 @@ class BaseBrowserLabelsAgent extends BaseBrowserAgent {
     async scroll_mouse_wheel(agentContext, amount, extract_page_content) {
         await this.execute_script(agentContext, scroll_by, [{ amount }]);
         await sleep(200);
+        if (!extract_page_content) {
+            const tools = this.toolUseNames(agentContext.agentChain.agentRequest?.messages);
+            let scroll_count = 0;
+            for (let i = tools.length - 1; i >= Math.max(tools.length - 8, 0); i--) {
+                if (tools[i] == "scroll_mouse_wheel") {
+                    scroll_count++;
+                }
+            }
+            if (scroll_count >= 3) {
+                extract_page_content = true;
+            }
+        }
         if (extract_page_content) {
             let page_content = await this.extract_page_content(agentContext);
-            return "This is the latest page content:\n" + page_content;
+            return "The current page content has been extracted, latest page content:\n" + page_content;
         }
     }
     async hover_to_element(agentContext, index) {
@@ -20836,7 +21201,7 @@ class BaseBrowserLabelsAgent extends BaseBrowserAgent {
                         },
                         enter: {
                             type: "boolean",
-                            description: "press the Enter key",
+                            description: "When text input is completed, press Enter (applicable to search boxes)",
                             default: false,
                         },
                     },
@@ -20872,23 +21237,30 @@ class BaseBrowserLabelsAgent extends BaseBrowserAgent {
                     return await this.callInnerTool(() => this.click_element(agentContext, args.index, (args.num_clicks || 1), (args.button || "left")));
                 },
             },
+            /*
             {
-                name: "scroll_to_element",
-                description: "Scroll to the element",
-                parameters: {
-                    type: "object",
-                    properties: {
-                        index: {
-                            type: "number",
-                            description: "The index of the element to input text into",
-                        },
-                    },
-                    required: ["index"],
-                },
-                execute: async (args, agentContext) => {
-                    return await this.callInnerTool(() => this.scroll_to_element(agentContext, args.index));
+              name: "scroll_to_element",
+              description: "Scroll to the element",
+              parameters: {
+                type: "object",
+                properties: {
+                  index: {
+                    type: "number",
+                    description: "The index of the element to input text into",
+                  },
                 },
+                required: ["index"],
+              },
+              execute: async (
+                args: Record<string, unknown>,
+                agentContext: AgentContext
+              ): Promise<ToolResult> => {
+                return await this.callInnerTool(() =>
+                  this.scroll_to_element(agentContext, args.index as number)
+                );
+              },
             },
+            */
             {
                 name: "scroll_mouse_wheel",
                 description: "Scroll the mouse wheel at current position, only scroll when you need to load more content",
@@ -20907,10 +21279,11 @@ class BaseBrowserLabelsAgent extends BaseBrowserAgent {
                         },
                         extract_page_content: {
                             type: "boolean",
+                            default: false,
                             description: "After scrolling is completed, whether to extract the current latest page content",
                         },
                     },
-                    required: ["amount", "direction"],
+                    required: ["amount", "direction", "extract_page_content"],
                 },
                 execute: async (args, agentContext) => {
                     return await this.callInnerTool(async () => {
@@ -21025,7 +21398,7 @@ class BaseBrowserLabelsAgent extends BaseBrowserAgent {
                             description: "Duration in millisecond",
                             default: 500,
                             minimum: 200,
-                            maximum: 2000,
+                            maximum: 10000,
                         },
                     },
                     required: ["duration"],
@@ -21036,33 +21409,94 @@ class BaseBrowserLabelsAgent extends BaseBrowserAgent {
             },
         ];
     }
+    async double_screenshots(agentContext, messages, tools) {
+        return true;
+    }
     async handleMessages(agentContext, messages, tools) {
+        const pseudoHtmlDescription = "This is the latest screenshot and page element information.\nindex and element:\n";
         let lastTool = this.lastToolResult(messages);
         if (lastTool &&
             lastTool.toolName !== "extract_page_content" &&
             lastTool.toolName !== "get_all_tabs" &&
             lastTool.toolName !== "variable_storage") {
             await sleep(300);
+            let image_contents = [];
+            if (await this.double_screenshots(agentContext, messages, tools)) {
+                let imageResult = await this.screenshot(agentContext);
+                let image = toImage(imageResult.imageBase64);
+                image_contents.push({
+                    type: "image",
+                    image: image,
+                    mimeType: imageResult.imageType,
+                });
+            }
             let result = await this.screenshot_and_html(agentContext);
             let image = toImage(result.imageBase64);
+            image_contents.push({
+                type: "image",
+                image: image,
+                mimeType: result.imageType,
+            });
             messages.push({
                 role: "user",
                 content: [
-                    {
-                        type: "image",
-                        image: image,
-                        mimeType: result.imageType,
-                    },
+                    ...image_contents,
                     {
                         type: "text",
-                        text: "This is the latest screenshot and page element information.\nindex and element:\n" +
-                            result.pseudoHtml,
+                        text: pseudoHtmlDescription + result.pseudoHtml,
                     },
                 ],
             });
         }
+        if (messages.length > 10) {
+            // compressed pseudoHtml
+            for (let i = 2; i < messages.length - 3; i++) {
+                let message = messages[i];
+                if (message.role == "user" && message.content.length == 2) {
+                    let content = message.content;
+                    for (let j = 0; j < content.length; j++) {
+                        let _content = content[j];
+                        if (_content.type == "text" &&
+                            _content.text.startsWith(pseudoHtmlDescription)) {
+                            _content.text = this.removePseudoHtmlAttr(_content.text, [
+                                "class",
+                                "src",
+                                "href",
+                            ]);
+                        }
+                    }
+                }
+            }
+        }
         super.handleMessages(agentContext, messages, tools);
     }
+    removePseudoHtmlAttr(pseudoHtml, remove_attrs) {
+        return pseudoHtml
+            .split("\n")
+            .map((line) => {
+            if (!line.startsWith("[") || line.indexOf("]:<") == -1) {
+                return line;
+            }
+            for (let i = 0; i < remove_attrs.length; i++) {
+                let sIdx = line.indexOf(remove_attrs[i] + '="');
+                if (sIdx == -1) {
+                    continue;
+                }
+                let eIdx = line.indexOf('"', sIdx + remove_attrs[i].length + 3);
+                if (eIdx == -1) {
+                    continue;
+                }
+                line =
+                    line.substring(0, sIdx) +
+                        line
+                            .substring(eIdx + 1)
+                            .trim()
+                            .replace('" >', '">');
+            }
+            return line;
+        })
+            .join("\n");
+    }
 }
 function typing(params) {
     let { index, text, enter } = params;
@@ -21094,6 +21528,19 @@ function typing(params) {
         }
     }
     input.focus && input.focus();
+    if (!text && enter) {
+        ["keydown", "keypress", "keyup"].forEach((eventType) => {
+            const event = new KeyboardEvent(eventType, {
+                key: "Enter",
+                code: "Enter",
+                keyCode: 13,
+                bubbles: true,
+                cancelable: true,
+            });
+            input.dispatchEvent(event);
+        });
+        return true;
+    }
     if (input.value == undefined) {
         input.textContent = text;
     }
@@ -21209,14 +21656,35 @@ function scroll_by(params) {
         window.scrollBy(0, y * amount);
         return;
     }
+    function findNodes(element = document, nodes = []) {
+        for (const node of Array.from(element.querySelectorAll("*"))) {
+            if (node.tagName === "IFRAME" && node.contentDocument) {
+                findNodes(node.contentDocument, nodes);
+            }
+            else {
+                nodes.push(node);
+            }
+        }
+        return nodes;
+    }
     function findScrollableElements() {
-        const allElements = Array.from(document.querySelectorAll("*"));
-        return allElements.filter((el) => {
+        const allElements = findNodes();
+        let elements = allElements.filter((el) => {
             const style = window.getComputedStyle(el);
             const overflowY = style.getPropertyValue("overflow-y");
             return ((overflowY === "auto" || overflowY === "scroll") &&
                 el.scrollHeight > el.clientHeight);
         });
+        if (elements.length == 0) {
+            elements = allElements.filter((el) => {
+                const style = window.getComputedStyle(el);
+                const overflowY = style.getPropertyValue("overflow-y");
+                return (overflowY === "auto" ||
+                    overflowY === "scroll" ||
+                    el.scrollHeight > el.clientHeight);
+            });
+        }
+        return elements;
     }
     function getVisibleArea(element) {
         const rect = element.getBoundingClientRect();
@@ -21243,6 +21711,12 @@ function scroll_by(params) {
     const viewportHeight = largestElement.clientHeight;
     const y = Math.max(20, Math.min(viewportHeight / 10, 200));
     largestElement.scrollBy(0, y * amount);
+    const maxHeightElement = sortedElements.sort((a, b) => b.getBoundingClientRect().height - a.getBoundingClientRect().height)[0];
+    if (maxHeightElement != largestElement) {
+        const viewportHeight = maxHeightElement.clientHeight;
+        const y = Math.max(20, Math.min(viewportHeight / 10, 200));
+        maxHeightElement.scrollBy(0, y * amount);
+    }
     return true;
 }
 
@@ -21510,7 +21984,7 @@ class BaseBrowserScreenAgent extends BaseBrowserAgent {
                             description: "Duration in millisecond",
                             default: 500,
                             minimum: 200,
-                            maximum: 2000,
+                            maximum: 10000,
                         },
                     },
                     required: ["duration"],