@eko-ai/eko 2.0.9 → 2.1.0-alpha.2

package/dist/index.cjs.js

@@ -9,7 +9,7 @@ const config = {
     platform: "mac",
     maxReactNum: 200,
     maxTokens: 16000,
-    compressThreshold: 60,
+    compressThreshold: 80,
     largeTextLength: 5000,
     shortTextLength: 800,
 };
@@ -67,7 +67,7 @@ class Logger {
         };
         let formattedMessage = '';
         if (this.dateFormat) {
-            formattedMessage += `[${new Date().toISOString()}] `;
+            formattedMessage += `[${new Date().toLocaleString()}] `;
         }
         formattedMessage += `[${levelNames[level] || 'UNKNOWN'}] `;
         if (this.prefix) {
@@ -139,8 +139,238 @@ 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;
+    }, {});
+    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;
+}
+
 class Context {
     constructor(taskId, config, agents, chain) {
+        this.paused = false;
         this.taskId = taskId;
         this.config = config;
         this.agents = agents;
@@ -148,13 +378,21 @@ 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 +8681,6 @@ 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 +14745,7 @@ 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 +14754,7 @@ 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 +14762,10 @@ 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 +14797,7 @@ 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 +14806,7 @@ 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 +14814,10 @@ 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
@@ -17876,7 +17895,7 @@ class HumanInteractTool {
         this.name = TOOL_NAME$3;
         this.noPlan = true;
         this.description = `AI interacts with humans:
-confirm: Ask the user to confirm whether to execute an operation, especially when performing dangerous actions such as deleting system files.
+confirm: Ask the user to confirm whether to execute an operation, especially when performing dangerous actions such as deleting system files, users will choose Yes or No.
 input: Prompt the user to enter text; for example, when a task is ambiguous, the AI can choose to ask the user for details, and the user can respond by inputting.
 select: Allow the user to make a choice; in situations that require selection, the AI can ask the user to make a decision.
 request_help: Request assistance from the user; for instance, when an operation is blocked, the AI can ask the user for help, such as needing to log into a website or solve a CAPTCHA.`;
@@ -17972,7 +17991,7 @@ 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: {
@@ -17990,8 +18009,12 @@ class TaskNodeStatusTool {
                         type: "number",
                     },
                 },
+                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."
+                },
             },
-            required: ["doneIds", "todoIds"],
+            required: ["doneIds", "todoIds", "thought"],
         };
     }
     async execute(args, agentContext) {
@@ -18131,7 +18154,7 @@ 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}
@@ -18155,10 +18178,10 @@ The output language should follow the language corresponding to the user's task.
 const HUMAN_PROMPT = `
 * HUMAN INTERACT
 During the task execution process, you can use the \`${TOOL_NAME$3}\` tool to interact with humans, please call it in the following situations:
-- When performing dangerous operations such as deleting files, confirmation from humans is required
-- When encountering obstacles while accessing websites, such as requiring user login, you need to request human assistance
+- When performing dangerous operations such as deleting files, confirmation from humans is required.
+- When encountering obstacles while visiting a website, such as requiring user login or captcha, you need to request for manual assistance.
 - When requesting login, please only call the function when a login dialog box is clearly displayed.
-- Try not to use the \`${TOOL_NAME$3}\` tool
+- Try to minimize the use of \`${TOOL_NAME$3}\` tool.
 `;
 const VARIABLE_PROMPT = `
 * VARIABLE STORAGE
@@ -18234,7 +18257,7 @@ 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 +18311,7 @@ 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) {
@@ -18557,6 +18580,9 @@ class Agent {
         }
         return this.tools;
     }
+    addTool(tool) {
+        this.tools.push(tool);
+    }
     get Name() {
         return this.name;
     }
@@ -18600,7 +18626,7 @@ 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;
@@ -18759,7 +18785,7 @@ 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:
@@ -18916,12 +18942,12 @@ 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 = "";
@@ -18949,22 +18975,26 @@ 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 +19037,7 @@ 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 +19054,7 @@ 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 +19193,7 @@ 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) {
@@ -19200,6 +19230,16 @@ 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);
@@ -20064,6 +20104,16 @@ class BaseBrowserAgent extends Agent {
         }
         return null;
     }
+    toolUseNames(messages) {
+        let toolNames = [];
+        for (let i = 0; i < messages.length; i++) {
+            let message = messages[i];
+            if (message.role == "tool") {
+                toolNames.push(message.content[0].toolName);
+            }
+        }
+        return toolNames;
+    }
     async execute_mcp_script(agentContext, script) {
         return;
     }
@@ -20104,8 +20154,8 @@ function run_build_dom_tree() {
                 'name',
                 'role',
                 'class',
-                // 'href',
-                'tabindex',
+                'src',
+                'href',
                 'aria-label',
                 'placeholder',
                 'value',
@@ -20154,6 +20204,12 @@ 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}"`;
                             }
@@ -20712,15 +20768,13 @@ class BaseBrowserLabelsAgent extends BaseBrowserAgent {
         }, [index]);
         await sleep(200);
     }
-    async scroll_mouse_wheel(agentContext, amount) {
-        await this.execute_script(agentContext, (amount) => {
-            let viewportHeight = window.innerHeight ||
-                document.documentElement.clientHeight ||
-                document.body.clientHeight;
-            let y = Math.max(20, Math.min(viewportHeight / 10, 200));
-            window.scrollBy(0, y * amount);
-        }, [amount]);
+    async scroll_mouse_wheel(agentContext, amount, extract_page_content) {
+        await this.execute_script(agentContext, scroll_by, [{ amount }]);
         await sleep(200);
+        if (extract_page_content) {
+            let page_content = await this.extract_page_content(agentContext);
+            return "This is the latest page content:\n" + page_content;
+        }
     }
     async hover_to_element(agentContext, index) {
         await this.execute_script(agentContext, hover_to, [{ index }]);
@@ -20741,6 +20795,7 @@ class BaseBrowserLabelsAgent extends BaseBrowserAgent {
             for (let i = 0; i < 5; i++) {
                 await sleep(200);
                 await this.execute_script(agentContext, run_build_dom_tree, []);
+                await sleep(50);
                 element_result = (await this.execute_script(agentContext, () => {
                     return window.get_clickable_elements(true);
                 }, []));
@@ -20748,7 +20803,7 @@ class BaseBrowserLabelsAgent extends BaseBrowserAgent {
                     break;
                 }
             }
-            await sleep(50);
+            await sleep(100);
             let screenshot = await this.screenshot(agentContext);
             // agentContext.variables.set("selector_map", element_result.selector_map);
             let pseudoHtml = element_result.element_str;
@@ -20882,7 +20937,7 @@ class BaseBrowserLabelsAgent extends BaseBrowserAgent {
             },
             {
                 name: "scroll_mouse_wheel",
-                description: "Scroll the mouse wheel at current position, prioritize using extract_page_content, only scroll when you need to load more content",
+                description: "Scroll the mouse wheel at current position, only scroll when you need to load more content",
                 parameters: {
                     type: "object",
                     properties: {
@@ -20896,13 +20951,17 @@ class BaseBrowserLabelsAgent extends BaseBrowserAgent {
                             type: "string",
                             enum: ["up", "down"],
                         },
+                        extract_page_content: {
+                            type: "boolean",
+                            description: "After scrolling is completed, whether to extract the current latest page content",
+                        },
                     },
                     required: ["amount", "direction"],
                 },
                 execute: async (args, agentContext) => {
                     return await this.callInnerTool(async () => {
                         let amount = args.amount;
-                        await this.scroll_mouse_wheel(agentContext, args.direction == "up" ? -amount : amount);
+                        await this.scroll_mouse_wheel(agentContext, args.direction == "up" ? -amount : amount, args.extract_page_content == true);
                     });
                 },
             },
@@ -21024,6 +21083,7 @@ class BaseBrowserLabelsAgent extends BaseBrowserAgent {
         ];
     }
     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" &&
@@ -21042,14 +21102,55 @@ class BaseBrowserLabelsAgent extends BaseBrowserAgent {
                     },
                     {
                         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;
@@ -21188,6 +21289,50 @@ function select_option(params) {
         selectedText: option.text.trim(),
     };
 }
+function scroll_by(params) {
+    const amount = params.amount;
+    const documentElement = document.documentElement || document.body;
+    if (documentElement.scrollHeight > window.innerHeight * 1.2) {
+        const y = Math.max(20, Math.min((window.innerHeight || documentElement.clientHeight) / 10, 200));
+        window.scrollBy(0, y * amount);
+        return;
+    }
+    function findScrollableElements() {
+        const allElements = Array.from(document.querySelectorAll("*"));
+        return allElements.filter((el) => {
+            const style = window.getComputedStyle(el);
+            const overflowY = style.getPropertyValue("overflow-y");
+            return ((overflowY === "auto" || overflowY === "scroll") &&
+                el.scrollHeight > el.clientHeight);
+        });
+    }
+    function getVisibleArea(element) {
+        const rect = element.getBoundingClientRect();
+        const viewportHeight = window.innerHeight || documentElement.clientHeight;
+        const viewportWidth = window.innerWidth || documentElement.clientWidth;
+        const visibleLeft = Math.max(0, Math.min(rect.left, viewportWidth));
+        const visibleRight = Math.max(0, Math.min(rect.right, viewportWidth));
+        const visibleTop = Math.max(0, Math.min(rect.top, viewportHeight));
+        const visibleBottom = Math.max(0, Math.min(rect.bottom, viewportHeight));
+        const visibleWidth = visibleRight - visibleLeft;
+        const visibleHeight = visibleBottom - visibleTop;
+        return visibleWidth * visibleHeight;
+    }
+    const scrollableElements = findScrollableElements();
+    if (scrollableElements.length === 0) {
+        const y = Math.max(20, Math.min((window.innerHeight || documentElement.clientHeight) / 10, 200));
+        window.scrollBy(0, y * amount);
+        return false;
+    }
+    const sortedElements = scrollableElements.sort((a, b) => {
+        return getVisibleArea(b) - getVisibleArea(a);
+    });
+    const largestElement = sortedElements[0];
+    const viewportHeight = largestElement.clientHeight;
+    const y = Math.max(20, Math.min(viewportHeight / 10, 200));
+    largestElement.scrollBy(0, y * amount);
+    return true;
+}
 
 class BaseBrowserScreenAgent extends BaseBrowserAgent {
     constructor(llms, ext_tools, mcpClient) {