@eko-ai/eko 1.0.7 → 1.0.8

package/README.md

@@ -1,17 +1,31 @@
-# Eko
 
-[![License](https://img.shields.io/badge/license-MIT-blue.svg)](LICENSE) [![Build Status](https://img.shields.io/badge/build-passing-brightgreen.svg)](https://example.com/build-status) [![Version](https://img.shields.io/badge/version-1.0.5-yellow.svg)](https://eko.fellou.ai/docs/release/versions/)
 
-**Eko** is a revolutionary framework designed to empower developers and users alike to program their browser and operating system using natural language. With seamless integration of browser APIs, OS-level capabilities, and cutting-edge AI tools like Claude 3.5, Eko redefines how we interact with technology, making it intuitive, powerful, and accessible.
 
-## Key Features
+<h1 align="center">
+  <a href="https://github.com/FellouAI/eko" target="_blank">
+    <img src="https://github.com/user-attachments/assets/55dbdd6c-2b08-4e5f-a841-8fea7c2a0b92" alt="eko-logo" width="200" height="200">
+  </a>
+  <br>
+  <small>Eko - Build Production-ready Agentic Workflow with Natural Language</small>
+</h1>
 
-- **Natural Language Programming**: Transform human instructions into e- **Natural Language Programming**: Convert natural language task descriptions into executable workflows, making agent development more intuitive
-- **Two-Layer Execution Model**: Separate offline planning from online execution, making agent decisions more structured and explainable
-- **Comprehensive Tooling**: Rich built-in tools for browser automation, computer control, file operations, and web interactions
-- **Hybrid Drive System:** Combine LLM capabilities with developer control, enabling "human in the loop" and allowing interference at multiple levels of granularity
-- **Event-Driven Automation:** Trigger workflows based on browser or system events
-- **Environment Flexibility**: Work across different environments ( [Browser Extension](/docs/browseruse/browser-extension), [Web](/docs/browseruse/browser-web), [Node.js](/docs/computeruse/computer-node), [Next-Gen AI Browser Fellou](/docs/computeruse/computer-fellou) ) with consistent APIs
+
+
+[![License](https://img.shields.io/badge/license-MIT-blue.svg)](LICENSE) [![Build Status](https://img.shields.io/badge/build-passing-brightgreen.svg)](https://example.com/build-status) [![Version](https://img.shields.io/github/package-json/v/FellouAI/eko?color=yellow)](https://eko.fellou.ai/docs/release/versions/)
+
+Eko (pronounced like ‘echo’) is a production-ready JavaScript framework that enables developers to create reliable agents, **from simple commands to complex workflows**. It provides a unified interface for running agents in both **computer and browser environments**.
+
+# Framework Comparison
+
+| Feature                              | Eko   | Langchain  | Browser-use  | Dify.ai  | Coze   |
+|--------------------------------------|-------|------------|--------------|----------|--------|
+| **Supported Platform**               | **All platform**  | Server side  | Browser  | Web  | Web  |
+| **One sentence to multi-step workflow** | ✅    | ❌          | ✅            | ❌        | ❌      |
+| **Intervenability**                  | ✅    | ✅          | ❌            | ❌        | ❌      | 
+| **Development Efficiency**           | **High**  | Low      | Middle        | Middle    | Low    | 
+| **Task Complexity**           | High  | High      | Low        | Middle    | Middle    | Middle       |
+| **Open-source**                      | ✅    | ✅          | ✅            | ✅        | ❌      |
+| **Access to private web resources** | ✅ **(Coming soon)** | ❌          | ❌            | ❌        | ❌      |
 
 ## Quickstart
 
@@ -19,7 +33,7 @@
 npm install @eko-ai/eko
 ```
 
-> The following code is for reference only. For detailed usage, please refer to the [Eko Quickstart guide](https://eko.fellou.ai/docs/getting-started/quickstart/).
+> For detailed usage, please refer to the [Eko Quickstart guide](https://eko.fellou.ai/docs/getting-started/quickstart/).
 
 ```typescript
 import { Eko } from '@eko-ai/eko';
@@ -38,6 +52,46 @@ await eko.execute(sysWorkflow);
 
 ```
 
+## Demos
+
+**Propmt:** `Collect the latest NASDAQ data on Yahoo Finance, including price changes, market capitalization, trading volume of major stocks, analyze the data and generate visualization reports`.
+
+https://github.com/user-attachments/assets/4087b370-8eb8-4346-a549-c4ce4d1efec3
+
+Click [here](https://github.com/FellouAI/eko-demos/tree/main/browser-extension-stock) to get the source code.
+
+---
+
+**Propmt:** `Based on the README of FellouAI/eko on github, search for competitors, highlight the key contributions of Eko, write a blog post advertising Eko, and post it on Write.as.`
+
+https://github.com/user-attachments/assets/6feaea86-2fb9-4e5c-b510-479c2473d810
+
+Click [here](https://github.com/FellouAI/eko-demos/tree/main/browser-extension-blog) to get the source code.
+
+---
+
+**Propmt:** `Clean up all files in the current directory larger than 1MB`
+
+https://github.com/user-attachments/assets/ef7feb58-3ddd-4296-a1de-bb8b6c66e48b
+
+Click [here](https://eko.fellou.ai/docs/computeruse/computer-node/#example-file-cleanup-workflow) to Learn more.
+
+---
+
+**Propmt:** Automatic software testing
+```
+    Current login page automation test:
+    1. Correct account and password are: admin / 666666 
+    2. Please randomly combine usernames and passwords for testing to verify if login validation works properly, such as: username cannot be empty, password cannot be empty, incorrect username, incorrect password
+    3. Finally, try to login with the correct account and password to verify if login is successful
+    4. Generate test report and export
+```
+
+https://github.com/user-attachments/assets/7716300a-c51d-41f1-8d4f-e3f593c1b6d5
+
+
+Click [here](https://eko.fellou.ai/docs/browseruse/browser-web#example-login-automation-testing) to Learn more.
+
 ## Use Cases
 
 - Browser automation and web scraping
@@ -64,24 +118,13 @@ Eko can be used in multiple environments:
 - Browser Extension
 - Web Applications
 - Node.js Applications
-- [Fellou AI Browser](https://fellou.ai)
 
 ## Community and Support
 
 - Report issues on [GitHub Issues](https://github.com/FellouAI/eko/issues)
+- Join our [slack community discussions](https://join.slack.com/t/eko-ai/shared_invite/zt-2xhvkudv9-nHvD1g8Smp227sM51x_Meg)
 - Contribute tools and improvements
 - Share your use cases and feedback
-- Join our community discussions
-
-## Contributing
-
-We welcome contributions! See our [Contributing Guide](CONTRIBUTING.md) for details on:
-
-- Setting up the development environment
-- Code style guidelines
-- Submission process
-- Tool development
-- Use case optimization
 
 ## License
 

package/dist/extension/content/index.d.ts

@@ -13,3 +13,4 @@ declare function get_dropdown_options(request: any): {
     name: any;
 } | null;
 declare function select_dropdown_option(request: any): any;
+declare function request_user_help(task_id: string, failure_type: string, failure_message: string): void;

package/dist/extension/tools/browser.d.ts

@@ -10,7 +10,8 @@ export declare function right_click(tabId: number, coordinate?: [number, number]
 export declare function right_click_by(tabId: number, xpath?: string, highlightIndex?: number): Promise<any>;
 export declare function double_click(tabId: number, coordinate?: [number, number]): Promise<any>;
 export declare function double_click_by(tabId: number, xpath?: string, highlightIndex?: number): Promise<any>;
-export declare function screenshot(windowId: number): Promise<ScreenshotResult>;
+export declare function screenshot(windowId: number, compress?: boolean): Promise<ScreenshotResult>;
+export declare function compress_image(dataUrl: string, scale?: number, quality?: number): Promise<string>;
 export declare function scroll_to(tabId: number, coordinate: [number, number]): Promise<any>;
 export declare function scroll_to_by(tabId: number, xpath?: string, highlightIndex?: number): Promise<any>;
 export declare function get_dropdown_options(tabId: number, xpath?: string, highlightIndex?: number): Promise<any>;

package/dist/extension/tools/index.d.ts

@@ -7,4 +7,5 @@ import { OpenUrl } from './open_url';
 import { Screenshot } from './screenshot';
 import { TabManagement } from './tab_management';
 import { WebSearch } from './web_search';
-export { BrowserUse, ElementClick, ExportFile, ExtractContent, FindElementPosition, OpenUrl, Screenshot, TabManagement, WebSearch, };
+import { RequestLogin } from './request_login';
+export { BrowserUse, ElementClick, ExportFile, ExtractContent, FindElementPosition, OpenUrl, Screenshot, TabManagement, WebSearch, RequestLogin, };

package/dist/extension/tools/request_login.d.ts

@@ -0,0 +1,10 @@
+import { Tool, InputSchema, ExecutionContext } from '../../types/action.types';
+export declare class RequestLogin implements Tool<any, any> {
+    name: string;
+    description: string;
+    input_schema: InputSchema;
+    constructor();
+    execute(context: ExecutionContext, params: any): Promise<any>;
+    awaitLogin(tabId: number, task_id: string): Promise<boolean>;
+    isLoginIn(context: ExecutionContext): Promise<boolean>;
+}

package/dist/extension/utils.d.ts

@@ -5,6 +5,7 @@ export declare function getCurrentTabId(windowId?: number | undefined): Promise<
 export declare function open_new_tab(url: string, newWindow: boolean, windowId?: number): Promise<chrome.tabs.Tab>;
 export declare function executeScript(tabId: number, func: any, args: any[]): Promise<any>;
 export declare function waitForTabComplete(tabId: number, timeout?: number): Promise<chrome.tabs.Tab>;
+export declare function doesTabExists(tabId: number): Promise<unknown>;
 export declare function getPageSize(tabId?: number): Promise<[number, number]>;
 export declare function sleep(time: number): Promise<void>;
 export declare function injectScript(tabId: number, filename?: string): Promise<void>;

package/dist/extension.cjs.js

@@ -138,6 +138,19 @@ async function waitForTabComplete(tabId, timeout = 15000) {
         chrome.tabs.onUpdated.addListener(listener);
     });
 }
+async function doesTabExists(tabId) {
+    const tabExists = await new Promise((resolve) => {
+        chrome.tabs.get(tabId, (tab) => {
+            if (chrome.runtime.lastError) {
+                resolve(false);
+            }
+            else {
+                resolve(true);
+            }
+        });
+    });
+    return tabExists;
+}
 async function getPageSize(tabId) {
     if (!tabId) {
         tabId = await getCurrentTabId();
@@ -235,6 +248,7 @@ var utils = /*#__PURE__*/Object.freeze({
     __proto__: null,
     CountDownLatch: CountDownLatch,
     MsgEvent: MsgEvent,
+    doesTabExists: doesTabExists,
     executeScript: executeScript,
     getCurrentTabId: getCurrentTabId,
     getPageSize: getPageSize,
@@ -339,11 +353,21 @@ async function double_click_by(tabId, xpath, highlightIndex) {
         highlightIndex,
     });
 }
-async function screenshot(windowId) {
-    let dataUrl = await chrome.tabs.captureVisibleTab(windowId, {
-        format: 'jpeg', // jpeg / png
-        quality: 50, // 0-100
-    });
+async function screenshot(windowId, compress) {
+    let dataUrl;
+    if (compress) {
+        dataUrl = await chrome.tabs.captureVisibleTab(windowId, {
+            format: 'jpeg',
+            quality: 60, // 0-100
+        });
+        dataUrl = await compress_image(dataUrl, 0.7, 1);
+    }
+    else {
+        dataUrl = await chrome.tabs.captureVisibleTab(windowId, {
+            format: 'jpeg',
+            quality: 50,
+        });
+    }
     let data = dataUrl.substring(dataUrl.indexOf('base64,') + 7);
     return {
         image: {
@@ -353,6 +377,23 @@ async function screenshot(windowId) {
         },
     };
 }
+async function compress_image(dataUrl, scale = 0.8, quality = 0.8) {
+    const bitmap = await createImageBitmap(await (await fetch(dataUrl)).blob());
+    let width = bitmap.width * scale;
+    let height = bitmap.height * scale;
+    const canvas = new OffscreenCanvas(width, height);
+    const ctx = canvas.getContext('2d');
+    ctx.drawImage(bitmap, 0, 0, width, height);
+    const blob = await canvas.convertToBlob({
+        type: 'image/jpeg',
+        quality: quality,
+    });
+    return new Promise((resolve) => {
+        const reader = new FileReader();
+        reader.onloadend = () => resolve(reader.result);
+        reader.readAsDataURL(blob);
+    });
+}
 async function scroll_to(tabId, coordinate) {
     let from_coordinate = (await cursor_position(tabId)).coordinate;
     return await chrome.tabs.sendMessage(tabId, {
@@ -397,6 +438,7 @@ var browser = /*#__PURE__*/Object.freeze({
     __proto__: null,
     clear_input: clear_input,
     clear_input_by: clear_input_by,
+    compress_image: compress_image,
     cursor_position: cursor_position,
     double_click: double_click,
     double_click_by: double_click_by,
@@ -423,7 +465,6 @@ class BrowserUse {
         this.name = 'browser_use';
         this.description = `Use structured commands to interact with the browser, manipulating page elements through screenshots and webpage element extraction.
 * This is a browser GUI interface where you need to analyze webpages by taking screenshots and extracting page element structures, and specify action sequences to complete designated tasks.
-* Some operations may need time to process, so you might need to wait and continuously take screenshots and extract element structures to check the operation results.
 * Before any operation, you must first call the \`screenshot_extract_element\` command, which will return the browser page screenshot and structured element information, both specially processed.
 * ELEMENT INTERACTION:
    - Only use indexes that exist in the provided element list
@@ -433,17 +474,7 @@ class BrowserUse {
    - If no suitable elements exist, use other functions to complete the task
    - If stuck, try alternative approaches
    - Handle popups/cookies by accepting or closing them
-   - Use scroll to find elements you are looking for
-* Form filling:
-   - If you fill a input field and your action sequence is interrupted, most often a list with suggestions poped up under the field and you need to first select the right element from the suggestion list.
-* ACTION SEQUENCING:
-   - Actions are executed in the order they appear in the list 
-   - Each action should logically follow from the previous one
-   - If the page changes after an action, the sequence is interrupted and you get the new state.
-   - If content only disappears the sequence continues.
-   - Only provide the action sequence until you think the page will change.
-   - Try to be efficient, e.g. fill forms at once, or chain actions where nothing changes on the page like saving, extracting, checkboxes...
-   - only use multiple actions if it makes sense.`;
+   - Use scroll to find elements you are looking for`;
         this.input_schema = {
             type: 'object',
             properties: {
@@ -585,7 +616,7 @@ class BrowserUse {
                         return window.get_clickable_elements(true);
                     }, []);
                     context.selector_map = element_result.selector_map;
-                    let screenshot$1 = await screenshot(windowId);
+                    let screenshot$1 = await screenshot(windowId, true);
                     await executeScript(tabId, () => {
                         return window.remove_highlight();
                     }, []);
@@ -817,7 +848,7 @@ ${pseudoHtml}
 async function executeWithBrowserUse$1(context, task_prompt) {
     let tabId = await getTabId(context);
     let windowId = await getWindowId(context);
-    let screenshot_result = await screenshot(windowId);
+    let screenshot_result = await screenshot(windowId, false);
     let messages = [
         {
             role: 'user',
@@ -1065,7 +1096,7 @@ ${pseudoHtml}
 async function executeWithBrowserUse(context, task_prompt) {
     await getTabId(context);
     let windowId = await getWindowId(context);
-    let screenshot_result = await screenshot(windowId);
+    let screenshot_result = await screenshot(windowId, false);
     let messages = [
         {
             role: 'user',
@@ -1643,6 +1674,91 @@ async function doPageContent(taskId, detailLinkGroups, window) {
     return searchInfo;
 }
 
+class RequestLogin {
+    constructor() {
+        this.name = 'request_login';
+        this.description =
+            'Login to this website, assist with identity verification when manual intervention is needed, guide users through the login process, and wait for their confirmation of successful login.';
+        this.input_schema = {
+            type: 'object',
+            properties: {},
+        };
+    }
+    async execute(context, params) {
+        if (!params.force && await this.isLoginIn(context)) {
+            return true;
+        }
+        let tabId = await getTabId(context);
+        let task_id = 'login_required_' + tabId;
+        const request_user_help = async () => {
+            await chrome.tabs.sendMessage(tabId, {
+                type: 'request_user_help',
+                task_id,
+                failure_type: 'login_required',
+                failure_message: 'Access page require user authentication.',
+            });
+        };
+        const login_interval = setInterval(async () => {
+            try {
+                request_user_help();
+            }
+            catch (e) {
+                clearInterval(login_interval);
+            }
+        }, 2000);
+        try {
+            return await this.awaitLogin(tabId, task_id);
+        }
+        finally {
+            clearInterval(login_interval);
+        }
+    }
+    async awaitLogin(tabId, task_id) {
+        return new Promise((resolve) => {
+            const checkTabClosedInterval = setInterval(async () => {
+                const tabExists = await doesTabExists(tabId);
+                if (!tabExists) {
+                    clearInterval(checkTabClosedInterval);
+                    resolve(false);
+                    chrome.runtime.onMessage.removeListener(listener);
+                }
+            }, 1000);
+            const listener = (message) => {
+                if (message.type === 'issue_resolved' && message.task_id === task_id) {
+                    resolve(true);
+                    clearInterval(checkTabClosedInterval);
+                }
+            };
+            chrome.runtime.onMessage.addListener(listener);
+        });
+    }
+    async isLoginIn(context) {
+        let windowId = await getWindowId(context);
+        let screenshot_result = await screenshot(windowId, true);
+        let messages = [
+            {
+                role: 'user',
+                content: [
+                    {
+                        type: 'image',
+                        source: screenshot_result.image,
+                    },
+                    {
+                        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.',
+                    },
+                ],
+            },
+        ];
+        let response = await context.llmProvider.generateText(messages, { maxTokens: 256 });
+        let text = response.textContent;
+        if (!text) {
+            text = JSON.stringify(response.content);
+        }
+        return text.indexOf('LOGGED_IN') > -1;
+    }
+}
+
 var tools = /*#__PURE__*/Object.freeze({
     __proto__: null,
     BrowserUse: BrowserUse,
@@ -1651,6 +1767,7 @@ var tools = /*#__PURE__*/Object.freeze({
     ExtractContent: ExtractContent,
     FindElementPosition: FindElementPosition,
     OpenUrl: OpenUrl,
+    RequestLogin: RequestLogin,
     Screenshot: Screenshot,
     TabManagement: TabManagement,
     WebSearch: WebSearch

package/dist/extension.esm.js

@@ -136,6 +136,19 @@ async function waitForTabComplete(tabId, timeout = 15000) {
         chrome.tabs.onUpdated.addListener(listener);
     });
 }
+async function doesTabExists(tabId) {
+    const tabExists = await new Promise((resolve) => {
+        chrome.tabs.get(tabId, (tab) => {
+            if (chrome.runtime.lastError) {
+                resolve(false);
+            }
+            else {
+                resolve(true);
+            }
+        });
+    });
+    return tabExists;
+}
 async function getPageSize(tabId) {
     if (!tabId) {
         tabId = await getCurrentTabId();
@@ -233,6 +246,7 @@ var utils = /*#__PURE__*/Object.freeze({
     __proto__: null,
     CountDownLatch: CountDownLatch,
     MsgEvent: MsgEvent,
+    doesTabExists: doesTabExists,
     executeScript: executeScript,
     getCurrentTabId: getCurrentTabId,
     getPageSize: getPageSize,
@@ -337,11 +351,21 @@ async function double_click_by(tabId, xpath, highlightIndex) {
         highlightIndex,
     });
 }
-async function screenshot(windowId) {
-    let dataUrl = await chrome.tabs.captureVisibleTab(windowId, {
-        format: 'jpeg', // jpeg / png
-        quality: 50, // 0-100
-    });
+async function screenshot(windowId, compress) {
+    let dataUrl;
+    if (compress) {
+        dataUrl = await chrome.tabs.captureVisibleTab(windowId, {
+            format: 'jpeg',
+            quality: 60, // 0-100
+        });
+        dataUrl = await compress_image(dataUrl, 0.7, 1);
+    }
+    else {
+        dataUrl = await chrome.tabs.captureVisibleTab(windowId, {
+            format: 'jpeg',
+            quality: 50,
+        });
+    }
     let data = dataUrl.substring(dataUrl.indexOf('base64,') + 7);
     return {
         image: {
@@ -351,6 +375,23 @@ async function screenshot(windowId) {
         },
     };
 }
+async function compress_image(dataUrl, scale = 0.8, quality = 0.8) {
+    const bitmap = await createImageBitmap(await (await fetch(dataUrl)).blob());
+    let width = bitmap.width * scale;
+    let height = bitmap.height * scale;
+    const canvas = new OffscreenCanvas(width, height);
+    const ctx = canvas.getContext('2d');
+    ctx.drawImage(bitmap, 0, 0, width, height);
+    const blob = await canvas.convertToBlob({
+        type: 'image/jpeg',
+        quality: quality,
+    });
+    return new Promise((resolve) => {
+        const reader = new FileReader();
+        reader.onloadend = () => resolve(reader.result);
+        reader.readAsDataURL(blob);
+    });
+}
 async function scroll_to(tabId, coordinate) {
     let from_coordinate = (await cursor_position(tabId)).coordinate;
     return await chrome.tabs.sendMessage(tabId, {
@@ -395,6 +436,7 @@ var browser = /*#__PURE__*/Object.freeze({
     __proto__: null,
     clear_input: clear_input,
     clear_input_by: clear_input_by,
+    compress_image: compress_image,
     cursor_position: cursor_position,
     double_click: double_click,
     double_click_by: double_click_by,
@@ -421,7 +463,6 @@ class BrowserUse {
         this.name = 'browser_use';
         this.description = `Use structured commands to interact with the browser, manipulating page elements through screenshots and webpage element extraction.
 * This is a browser GUI interface where you need to analyze webpages by taking screenshots and extracting page element structures, and specify action sequences to complete designated tasks.
-* Some operations may need time to process, so you might need to wait and continuously take screenshots and extract element structures to check the operation results.
 * Before any operation, you must first call the \`screenshot_extract_element\` command, which will return the browser page screenshot and structured element information, both specially processed.
 * ELEMENT INTERACTION:
    - Only use indexes that exist in the provided element list
@@ -431,17 +472,7 @@ class BrowserUse {
    - If no suitable elements exist, use other functions to complete the task
    - If stuck, try alternative approaches
    - Handle popups/cookies by accepting or closing them
-   - Use scroll to find elements you are looking for
-* Form filling:
-   - If you fill a input field and your action sequence is interrupted, most often a list with suggestions poped up under the field and you need to first select the right element from the suggestion list.
-* ACTION SEQUENCING:
-   - Actions are executed in the order they appear in the list 
-   - Each action should logically follow from the previous one
-   - If the page changes after an action, the sequence is interrupted and you get the new state.
-   - If content only disappears the sequence continues.
-   - Only provide the action sequence until you think the page will change.
-   - Try to be efficient, e.g. fill forms at once, or chain actions where nothing changes on the page like saving, extracting, checkboxes...
-   - only use multiple actions if it makes sense.`;
+   - Use scroll to find elements you are looking for`;
         this.input_schema = {
             type: 'object',
             properties: {
@@ -583,7 +614,7 @@ class BrowserUse {
                         return window.get_clickable_elements(true);
                     }, []);
                     context.selector_map = element_result.selector_map;
-                    let screenshot$1 = await screenshot(windowId);
+                    let screenshot$1 = await screenshot(windowId, true);
                     await executeScript(tabId, () => {
                         return window.remove_highlight();
                     }, []);
@@ -815,7 +846,7 @@ ${pseudoHtml}
 async function executeWithBrowserUse$1(context, task_prompt) {
     let tabId = await getTabId(context);
     let windowId = await getWindowId(context);
-    let screenshot_result = await screenshot(windowId);
+    let screenshot_result = await screenshot(windowId, false);
     let messages = [
         {
             role: 'user',
@@ -1063,7 +1094,7 @@ ${pseudoHtml}
 async function executeWithBrowserUse(context, task_prompt) {
     await getTabId(context);
     let windowId = await getWindowId(context);
-    let screenshot_result = await screenshot(windowId);
+    let screenshot_result = await screenshot(windowId, false);
     let messages = [
         {
             role: 'user',
@@ -1641,6 +1672,91 @@ async function doPageContent(taskId, detailLinkGroups, window) {
     return searchInfo;
 }
 
+class RequestLogin {
+    constructor() {
+        this.name = 'request_login';
+        this.description =
+            'Login to this website, assist with identity verification when manual intervention is needed, guide users through the login process, and wait for their confirmation of successful login.';
+        this.input_schema = {
+            type: 'object',
+            properties: {},
+        };
+    }
+    async execute(context, params) {
+        if (!params.force && await this.isLoginIn(context)) {
+            return true;
+        }
+        let tabId = await getTabId(context);
+        let task_id = 'login_required_' + tabId;
+        const request_user_help = async () => {
+            await chrome.tabs.sendMessage(tabId, {
+                type: 'request_user_help',
+                task_id,
+                failure_type: 'login_required',
+                failure_message: 'Access page require user authentication.',
+            });
+        };
+        const login_interval = setInterval(async () => {
+            try {
+                request_user_help();
+            }
+            catch (e) {
+                clearInterval(login_interval);
+            }
+        }, 2000);
+        try {
+            return await this.awaitLogin(tabId, task_id);
+        }
+        finally {
+            clearInterval(login_interval);
+        }
+    }
+    async awaitLogin(tabId, task_id) {
+        return new Promise((resolve) => {
+            const checkTabClosedInterval = setInterval(async () => {
+                const tabExists = await doesTabExists(tabId);
+                if (!tabExists) {
+                    clearInterval(checkTabClosedInterval);
+                    resolve(false);
+                    chrome.runtime.onMessage.removeListener(listener);
+                }
+            }, 1000);
+            const listener = (message) => {
+                if (message.type === 'issue_resolved' && message.task_id === task_id) {
+                    resolve(true);
+                    clearInterval(checkTabClosedInterval);
+                }
+            };
+            chrome.runtime.onMessage.addListener(listener);
+        });
+    }
+    async isLoginIn(context) {
+        let windowId = await getWindowId(context);
+        let screenshot_result = await screenshot(windowId, true);
+        let messages = [
+            {
+                role: 'user',
+                content: [
+                    {
+                        type: 'image',
+                        source: screenshot_result.image,
+                    },
+                    {
+                        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.',
+                    },
+                ],
+            },
+        ];
+        let response = await context.llmProvider.generateText(messages, { maxTokens: 256 });
+        let text = response.textContent;
+        if (!text) {
+            text = JSON.stringify(response.content);
+        }
+        return text.indexOf('LOGGED_IN') > -1;
+    }
+}
+
 var tools = /*#__PURE__*/Object.freeze({
     __proto__: null,
     BrowserUse: BrowserUse,
@@ -1649,6 +1765,7 @@ var tools = /*#__PURE__*/Object.freeze({
     ExtractContent: ExtractContent,
     FindElementPosition: FindElementPosition,
     OpenUrl: OpenUrl,
+    RequestLogin: RequestLogin,
     Screenshot: Screenshot,
     TabManagement: TabManagement,
     WebSearch: WebSearch