提示词
系统提示词
You are an AI agent designed to automate browser tasks. Your goal is to accomplish the ultimate task following the rules.
# Input Format
Task
Previous steps
Current URL
Open Tabs
Interactive Elements
[index]<type>text</type>
- index: Numeric identifier for interaction
- type: HTML element type (button, input, etc.)
- text: Element description
Example:
[33]<div>User form</div>
\t*[35]*<button aria-label='Submit form'>Submit</button>
- Only elements with numeric indexes in [] are interactive
- (stacked) indentation (with \t) is important and means that the element is a (html) child of the element above (with a lower index)
- Elements with \* are new elements that were added after the previous step (if url has not changed)
# Response Rules
1. RESPONSE FORMAT: You must ALWAYS respond with valid JSON in this exact format:
{"current_state": {"evaluation_previous_goal": "Success|Failed|Unknown - Analyze the current elements and the image to check if the previous goals/actions are successful like intended by the task. Mention if something unexpected happened. Shortly state why/why not",
"memory": "Description of what has been done and what you need to remember. Be very specific. Count here ALWAYS how many times you have done something and how many remain. E.g. 0 out of 10 websites analyzed. Continue with abc and xyz",
"next_goal": "What needs to be done with the next immediate action"},
"action":[{"one_action_name": {// action-specific parameter}}, // ... more actions in sequence]}
2. ACTIONS: You can specify multiple actions in the list to be executed in sequence. But always specify only one action name per item. Use maximum 10 actions per sequence.
Common action sequences:
- Form filling: [{"input_text": {"index": 1, "text": "username"}}, {"input_text": {"index": 2, "text": "password"}}, {"click_element": {"index": 3}}]
- Navigation and extraction: [{"go_to_url": {"url": "https://example.com"}}, {"extract_content": {"goal": "extract the names"}}]
- Actions are executed in the given order
- If the page changes after an action, the sequence is interrupted and you get the new state.
- Only provide the action sequence until an action which changes the page state significantly.
- Try to be efficient, e.g. fill forms at once, or chain actions where nothing changes on the page
- only use multiple actions if it makes sense.
3. ELEMENT INTERACTION:
- Only use indexes of the interactive elements
4. NAVIGATION & ERROR HANDLING:
- If no suitable elements exist, use other functions to complete the task
- If stuck, try alternative approaches - like going back to a previous page, new search, new tab etc.
- Handle popups/cookies by accepting or closing them
- Use scroll to find elements you are looking for
- If you want to research something, open a new tab instead of using the current tab
- If captcha pops up, try to solve it - else try a different approach
- If the page is not fully loaded, use wait action
5. TASK COMPLETION:
- Use the done action as the last action as soon as the ultimate task is complete
- Dont use "done" before you are done with everything the user asked you, except you reach the last step of max_steps.
- If you reach your last step, use the done action even if the task is not fully finished. Provide all the information you have gathered so far. If the ultimate task is completely finished set success to true. If not everything the user asked for is completed set success in done to false!
- If you have to do something repeatedly for example the task says for "each", or "for all", or "x times", count always inside "memory" how many times you have done it and how many remain. Don't stop until you have completed like the task asked you. Only call done after the last step.
- Don't hallucinate actions
- Make sure you include everything you found out for the ultimate task in the done text parameter. Do not just say you are done, but include the requested information of the task.
6. VISUAL CONTEXT:
- When an image is provided, use it to understand the page layout
- Bounding boxes with labels on their top right corner correspond to element indexes
7. Form filling:
- If you fill an input field and your action sequence is interrupted, most often something changed e.g. suggestions popped up under the field.
8. Long tasks:
- Keep track of the status and subresults in the memory.
- You are provided with procedural memory summaries that condense previous task history (every N steps). Use these summaries to maintain context about completed actions, current progress, and next steps. The summaries appear in chronological order and contain key information about navigation history, findings, errors encountered, and current state. Refer to these summaries to avoid repeating actions and to ensure consistent progress toward the task goal.
9. Extraction:
- If your task is to find information - call extract_content on the specific pages to get and store the information.
Your responses must be always JSON with the specified format.
user Your ultimate task is: """打开淘宝,找到最便宜的Mac Mini m4,给我链接,如果需要登录 请提示我登录""". If you achieved your ultimate task, stop everything and use the done action in the next step to complete the task. If not, continue as usual.
user Example output:
assistant null
{
"current_state": {
"evaluation_previous_goal": "Success – I successfully clicked on the 'Apple' link from the Google Search results page, which directed me to the Apple company homepage. This is a good start toward finding the best place to buy a new iPhone, as the Apple website often lists iPhones for sale.",
"memory": "I searched for 'iPhone retailers' on Google. From the Google Search results page, I used the ‘click_element_by_index’ tool to click on element at index [45] labeled 'Best Buy' but that did not navigate to a new page. I then clicked on element at index [82] labeled 'Apple', which redirected me to the Apple company homepage. Currently at step 3/15.",
"next_goal": "Looking at the reported structure of the current page, I can see the item '[127]<h3 iPhone/>' in the content. I believe this link will lead to more information and potentially prices for iPhones. I'll click on the link at index [127] using the ‘click_element_by_index’ tool and hope to see prices on the next page."
},
"action": [
{
"click_element_by_index": {
"index": 127
}
}
]
}
tool Browser started
user [Your task history memory starts here]
user
[Task history memory ends]
[Current state starts here]
The following is one-time information - if you need to remember it write it to memory:
Current url: about:blank
Available tabs:
[TabInfo(page_id=0, url='about:blank', title='', parent_page_id=None)]
Interactive elements from top layer of the current page inside the viewport:
empty page
Current step: 1/100 Current date and time: 2025-07-15 19:51
完整交互流程
{
"messages":
[
{
"content": "You are an AI agent designed to automate browser tasks. Your goal is to accomplish the ultimate task following the rules.\n\n# Input Format\n\nTask\nPrevious steps\nCurrent URL\nOpen Tabs\nInteractive Elements\n[index]<type>text</type>\n\n- index: Numeric identifier for interaction\n- type: HTML element type (button, input, etc.)\n- text: Element description\n Example:\n [33]<div>User form</div>\n \\t*[35]*<button aria-label='Submit form'>Submit</button>\n\n- Only elements with numeric indexes in [] are interactive\n- (stacked) indentation (with \\t) is important and means that the element is a (html) child of the element above (with a lower index)\n- Elements with \\* are new elements that were added after the previous step (if url has not changed)\n\n# Response Rules\n\n1. RESPONSE FORMAT: You must ALWAYS respond with valid JSON in this exact format:\n {\"current_state\": {\"evaluation_previous_goal\": \"Success|Failed|Unknown - Analyze the current elements and the image to check if the previous goals/actions are successful like intended by the task. Mention if something unexpected happened. Shortly state why/why not\",\n \"memory\": \"Description of what has been done and what you need to remember. Be very specific. Count here ALWAYS how many times you have done something and how many remain. E.g. 0 out of 10 websites analyzed. Continue with abc and xyz\",\n \"next_goal\": \"What needs to be done with the next immediate action\"},\n \"action\":[{\"one_action_name\": {// action-specific parameter}}, // ... more actions in sequence]}\n\n2. ACTIONS: You can specify multiple actions in the list to be executed in sequence. But always specify only one action name per item. Use maximum 10 actions per sequence.\nCommon action sequences:\n\n- Form filling: [{\"input_text\": {\"index\": 1, \"text\": \"username\"}}, {\"input_text\": {\"index\": 2, \"text\": \"password\"}}, {\"click_element\": {\"index\": 3}}]\n- Navigation and extraction: [{\"go_to_url\": {\"url\": \"https://example.com\"}}, {\"extract_content\": {\"goal\": \"extract the names\"}}]\n- Actions are executed in the given order\n- If the page changes after an action, the sequence is interrupted and you get the new state.\n- Only provide the action sequence until an action which changes the page state significantly.\n- Try to be efficient, e.g. fill forms at once, or chain actions where nothing changes on the page\n- only use multiple actions if it makes sense.\n\n3. ELEMENT INTERACTION:\n\n- Only use indexes of the interactive elements\n\n4. NAVIGATION & ERROR HANDLING:\n\n- If no suitable elements exist, use other functions to complete the task\n- If stuck, try alternative approaches - like going back to a previous page, new search, new tab etc.\n- Handle popups/cookies by accepting or closing them\n- Use scroll to find elements you are looking for\n- If you want to research something, open a new tab instead of using the current tab\n- If captcha pops up, try to solve it - else try a different approach\n- If the page is not fully loaded, use wait action\n\n5. TASK COMPLETION:\n\n- Use the done action as the last action as soon as the ultimate task is complete\n- Dont use \"done\" before you are done with everything the user asked you, except you reach the last step of max_steps.\n- If you reach your last step, use the done action even if the task is not fully finished. Provide all the information you have gathered so far. If the ultimate task is completely finished set success to true. If not everything the user asked for is completed set success in done to false!\n- If you have to do something repeatedly for example the task says for \"each\", or \"for all\", or \"x times\", count always inside \"memory\" how many times you have done it and how many remain. Don't stop until you have completed like the task asked you. Only call done after the last step.\n- Don't hallucinate actions\n- Make sure you include everything you found out for the ultimate task in the done text parameter. Do not just say you are done, but include the requested information of the task.\n\n6. VISUAL CONTEXT:\n\n- When an image is provided, use it to understand the page layout\n- Bounding boxes with labels on their top right corner correspond to element indexes\n\n7. Form filling:\n\n- If you fill an input field and your action sequence is interrupted, most often something changed e.g. suggestions popped up under the field.\n\n8. Long tasks:\n\n- Keep track of the status and subresults in the memory.\n- You are provided with procedural memory summaries that condense previous task history (every N steps). Use these summaries to maintain context about completed actions, current progress, and next steps. The summaries appear in chronological order and contain key information about navigation history, findings, errors encountered, and current state. Refer to these summaries to avoid repeating actions and to ensure consistent progress toward the task goal.\n\n9. Extraction:\n\n- If your task is to find information - call extract_content on the specific pages to get and store the information.\n Your responses must be always JSON with the specified format.\n",
"role": "system"
},
{
"content": "Your ultimate task is: \"\"\"打开淘宝,找到最便宜的Mac Mini m4,给我链接,如果需要登录 请提示我登录\"\"\". If you achieved your ultimate task, stop everything and use the done action in the next step to complete the task. If not, continue as usual.",
"role": "user"
},
{
"content": "Example output:",
"role": "user"
},
{
"content": null,
"role": "assistant",
"tool_calls":
[
{
"function":
{
"arguments": "{\"current_state\": {\"evaluation_previous_goal\": \"Success - I successfully clicked on the 'Apple' link from the Google Search results page, \\n\\t\\t\\t\\t\\t\\t\\twhich directed me to the 'Apple' company homepage. This is a good start toward finding \\n\\t\\t\\t\\t\\t\\t\\tthe best place to buy a new iPhone as the Apple website often list iPhones for sale.\", \"memory\": \"I searched for 'iPhone retailers' on Google. From the Google Search results page, \\n\\t\\t\\t\\t\\t\\t\\tI used the 'click_element_by_index' tool to click on element at index [45] labeled 'Best Buy' but calling \\n\\t\\t\\t\\t\\t\\t\\tthe tool did not direct me to a new page. I then used the 'click_element_by_index' tool to click \\n\\t\\t\\t\\t\\t\\t\\ton element at index [82] labeled 'Apple' which redirected me to the 'Apple' company homepage. \\n\\t\\t\\t\\t\\t\\t\\tCurrently at step 3/15.\", \"next_goal\": \"Looking at reported structure of the current page, I can see the item '[127]<h3 iPhone/>' \\n\\t\\t\\t\\t\\t\\t\\tin the content. I think this button will lead to more information and potentially prices \\n\\t\\t\\t\\t\\t\\t\\tfor iPhones. I'll click on the link at index [127] using the 'click_element_by_index' \\n\\t\\t\\t\\t\\t\\t\\ttool and hope to see prices on the next page.\"}, \"action\": [{\"click_element_by_index\": {\"index\": 127}}]}",
"name": "AgentOutput"
},
"id": "1",
"type": "function"
}
]
},
{
"content": "Browser started",
"role": "tool",
"tool_call_id": "1"
},
{
"content": "[Your task history memory starts here]",
"role": "user"
},
{
"content": null,
"role": "assistant",
"tool_calls":
[
{
"function":
{
"arguments": "{\"current_state\": {\"evaluation_previous_goal\": \"Unknown - The browser is currently on an empty page (about:blank). No previous actions have been taken yet.\", \"memory\": \"Starting the task to open Taobao and find the cheapest Mac Mini M4. Step 1 of 100.\", \"next_goal\": \"Navigate to Taobao's homepage to begin the search for the Mac Mini M4.\"}, \"action\": [{\"go_to_url\": {\"url\": \"https://www.taobao.com\"}}]}",
"name": "AgentOutput"
},
"id": "2",
"type": "function"
}
]
},
{
"content": "",
"role": "tool",
"tool_call_id": "2"
},
{
"content": "Action result: 🔗 Navigated to https://www.taobao.com",
"role": "user"
},
{
"content": null,
"role": "assistant",
"tool_calls":
[
{
"function":
{
"arguments": "{\"current_state\": {\"evaluation_previous_goal\": \"Success - Successfully navigated to Taobao's homepage.\", \"memory\": \"Navigated to Taobao's homepage. Step 1 of 100. Next, I need to search for 'Mac Mini M4' to find the cheapest option.\", \"next_goal\": \"Search for 'Mac Mini M4' on Taobao to find the cheapest option.\"}, \"action\": [{\"input_text\": {\"index\": 14, \"text\": \"Mac Mini M4\"}}, {\"click_element_by_index\": {\"index\": 13}}]}",
"name": "AgentOutput"
},
"id": "3",
"type": "function"
}
]
},
{
"content": "",
"role": "tool",
"tool_call_id": "3"
},
{
"content": "Action result: ⌨️ Input Mac Mini M4 into index 14",
"role": "user"
},
{
"content": "Action result: 🖱️ Clicked button with index 13: 搜索 - New tab opened - switching to it",
"role": "user"
},
{
"content": null,
"role": "assistant",
"tool_calls":
[
{
"function":
{
"arguments": "{\"current_state\": {\"evaluation_previous_goal\": \"Success - Successfully searched for 'Mac Mini M4' on Taobao and navigated to the search results page.\", \"memory\": \"Navigated to Taobao's homepage, searched for 'Mac Mini M4', and opened the search results page. Now, I need to find the cheapest Mac Mini M4 from the search results. Step 3 of 100.\", \"next_goal\": \"Extract the cheapest Mac Mini M4 from the search results and provide its link.\"}, \"action\": [{\"extract_content\": {\"goal\": \"Extract the cheapest Mac Mini M4 product details and link\", \"should_strip_link_urls\": false}}]}",
"name": "AgentOutput"
},
"id": "4",
"type": "function"
}
]
},
{
"content": "",
"role": "tool",
"tool_call_id": "4"
},
{
"content": "Action result: 📄 Extracted from page\n: ```json\n{\n \"status\": \"error\",\n \"message\": \"No product details found for Mac Mini M4 on the provided Taobao search page. The page appears to be a login or general search interface without specific product listings.\",\n \"details\": {\n \"page_content\": \"The page is primarily a search interface for Taobao with login prompts, navigation links, and general search functionalities. There are no visible product listings or details for Mac Mini M4 in the provided content.\",\n \"suggestions\": [\n \"Ensure you are logged in to Taobao to view complete search results.\",\n \"Perform a search specifically for 'Mac Mini M4' to see product listings.\",\n \"Check if the search results are being blocked by ad blockers or JavaScript restrictions.\"\n ]\n }\n}\n```\n",
"role": "user"
},
{
"content": null,
"role": "assistant",
"tool_calls":
[
{
"function":
{
"arguments": "{\"current_state\": {\"evaluation_previous_goal\": \"Failed - The search results page did not display any product listings for 'Mac Mini M4'. The page may require login or additional filters to show relevant results.\", \"memory\": \"Navigated to Taobao's homepage, searched for 'Mac Mini M4', and opened the search results page. However, the page did not display any product listings, possibly due to a login requirement. Step 4 of 100.\", \"next_goal\": \"Check if a login prompt is required and prompt the user to log in if necessary.\"}, \"action\": [{\"ask_for_assistant\": {\"query\": \"The Taobao search results page requires login to view product listings. Please log in to Taobao to proceed with finding the cheapest Mac Mini M4.\"}}]}",
"name": "AgentOutput"
},
"id": "5",
"type": "function"
}
]
},
{
"content": "",
"role": "tool",
"tool_call_id": "5"
},
{
"content": "Action result: AI ask: The Taobao search results page requires login to view product listings. Please log in to Taobao to proceed with finding the cheapest Mac Mini M4.. User response: Internal Error: Cannot display help request.",
"role": "user"
},
{
"content": null,
"role": "assistant",
"tool_calls":
[
{
"function":
{
"arguments": "{\"current_state\": {\"evaluation_previous_goal\": \"Failed - The Taobao search results page did not display any product listings for 'Mac Mini M4'. The page may require login or additional filters to show relevant results.\", \"memory\": \"Navigated to Taobao's homepage, searched for 'Mac Mini M4', and opened the search results page. However, the page did not display any product listings, possibly due to a login requirement. Step 5 of 100.\", \"next_goal\": \"Scroll down the search results page to check if product listings appear after scrolling.\"}, \"action\": [{\"scroll_down\": {}}]}",
"name": "AgentOutput"
},
"id": "6",
"type": "function"
}
]
},
{
"content": "",
"role": "tool",
"tool_call_id": "6"
},
{
"content": "Action result: 🔍 Scrolled down the page by one page",
"role": "user"
},
{
"content": null,
"role": "assistant",
"tool_calls":
[
{
"function":
{
"arguments": "{\"current_state\": {\"evaluation_previous_goal\": \"Success - Scrolled down the search results page and found multiple listings for Mac Mini M4.\", \"memory\": \"Navigated to Taobao's homepage, searched for 'Mac Mini M4', and opened the search results page. Scrolled down to view multiple product listings. Now, I need to identify the cheapest Mac Mini M4 and provide its link. Step 6 of 100.\", \"next_goal\": \"Identify the cheapest Mac Mini M4 from the search results and provide its link.\"}, \"action\": [{\"extract_content\": {\"goal\": \"Extract the details and link of the cheapest Mac Mini M4 from the search results\", \"should_strip_link_urls\": false}}]}",
"name": "AgentOutput"
},
"id": "7",
"type": "function"
}
]
},
{
"content": "",
"role": "tool",
"tool_call_id": "7"
},
{
"content": "Action result: 📄 Extracted from page\n: ```json\n{\n \"cheapest_mac_mini_m4\": {\n \"title\": \"Apple/苹果 2024款Mac Mini M4/M4 Pro芯片台式电脑主机未拆封\",\n \"price\": \"¥3388\",\n \"link\": \"//item.taobao.com/item.htm?id=851793447219\",\n \"specs\": [\n \"7 核图形处理器\",\n \"256GB\"\n ],\n \"location\": \"广东 广州\",\n \"seller\": \"壹佳靓机数码\",\n \"seller_link\": \"//store.taobao.com/shop/view_shop.htm?appUid=RAzN8HWT9fsZXRyNvJ86fpjQD8KK8TEvGhwE1qEYnDt7NESoocd\",\n \"shipping\": \"包邮\",\n \"sales\": \"5人付款\"\n }\n}\n```\n",
"role": "user"
},
{
"content": "\n[Task history memory ends]\n[Current state starts here]\nThe following is one-time information - if you need to remember it write it to memory:\nCurrent url: https://s.taobao.com/search?commend=all&ie=utf8&initiative_id=tbindexz_20170306&page=1&preLoadOrigin=https%3A%2F%2Fwww.taobao.com&q=Mac%20Mini%20M4&search_type=item&sourceId=tb.index&spm=a21bo.jianhua%2Fa.search_manual.0&ssid=s5-e&tab=all\nAvailable tabs:\n[TabInfo(page_id=0, url='https://www.taobao.com/', title='淘宝', parent_page_id=None), TabInfo(page_id=1, url='https://s.taobao.com/search?commend=all&ie=utf8&initiative_id=tbindexz_20170306&page=1&preLoadOrigin=https%3A%2F%2Fwww.taobao.com&q=Mac%20Mini%20M4&search_type=item&sourceId=tb.index&spm=a21bo.jianhua%2Fa.search_manual.0&ssid=s5-e&tab=all', title='Mac Mini M4_淘宝搜索', parent_page_id=None)]\nInteractive elements from top layer of the current page inside the viewport:\n... 1005 pixels above - scroll or extract content to see more ...\n[0]<a />\n[1]<button type='submit'>搜索 />\n[2]<input name='q' aria-label='请输入搜索文字' value='Mac Mini M4' />\n搜同款\n[3]<div />\n\t[4]<a >云飞侠小时达 />\n\t[5]<div />\n\t\t[6]<a title='点此可以直接和卖家交流选好的宝贝，或相互交流网购体验，还支持语音视频噢。'>旺旺在线 />\n[7]<div />\n\t[8]<a >优选数码通讯 />\n\t[9]<div />\n\t\t[10]<a title='点此可以直接和卖家交流选好的宝贝，或相互交流网购体验，还支持语音视频噢。'>旺旺在线 />\n[11]<div />\n\t[12]<a >回头客8万\nApple授权专营店 />\n\t[13]<div />\n\t\t[14]<a title='点此可以直接和卖家交流选好的宝贝，或相互交流网购体验，还支持语音视频噢。'>旺旺在线 />\n[15]<div />\n\t[16]<a >迷制数码 />\n\t[17]<div />\n\t\t[18]<a title='点此可以直接和卖家交流选好的宝贝，或相互交流网购体验，还支持语音视频噢。'>旺旺在线 />\n[19]<a >【教育版】Apple/苹果\nMac\nmini\nM4\n芯片 24GB+512GB 台式电脑主机\nmacOS\n512GB\n一级\n¥\n7499\n北京\n包邮 />\n\t[20]<a >回头客8万\nApple授权专营店 />\n\t[21]<div />\n\t\t[22]<a title='点此可以直接和卖家交流选好的宝贝，或相互交流网购体验，还支持语音视频噢。'>旺旺在线 />\n[23]<a >Apple/苹果\nMac\nmini\nM4\n系列芯片 一体机更快更高效正品\n7200转\n影音娱乐\nApple M4\n¥\n4999\n0人付款\n广东\n深圳\n包邮 />\n\t[24]<a >10年老店\nJohn Davison Rockefe />\n\t[25]<div />\n\t\t[26]<a title='点此可以直接和卖家交流选好的宝贝，或相互交流网购体验，还支持语音视频噢。'>旺旺在线 />\n[27]<a >Apple/苹果\nMac\nmini\n台式电脑主机国行原封2024新款\nM4\n芯片/\nM4\npro\nmacOS\n无光驱\n1kg\n¥\n3898\n7人付款\n重庆\n包邮 />\n\t[28]<a >7年老店\n重庆配之家科技 />\n\t[29]<div />\n\t\t[30]<a title='点此可以直接和卖家交流选好的宝贝，或相互交流网购体验，还支持语音视频噢。'>旺旺在线 />\n[31]<a >Apple/苹果 2024款\nMac\nMini\nM4\n/\nM4\nPro台式电脑迷你主机未拆封\n7 核图形处理器\nmacOS\n256GB\n¥\n3388\n5人付款\n广东\n广州\n包邮 />\n\t[32]<a >壹佳靓机数码 />\n\t[33]<div />\n\t\t[34]<a title='点此可以直接和卖家交流选好的宝贝，或相互交流网购体验，还支持语音视频噢。'>旺旺在线 />\n对本次搜索体验满意吗\n[35]<div >非常满意 />\n[36]<div >满意 />\n[37]<div >感觉一般 />\n[38]<div >不满意 />\n[39]<div >非常不满意 />\n[40]<div >不感兴趣，关闭 />\n[41]<a >【激活返50/政府补贴20%】 Apple/苹果\nMac\nmini\n台式电脑主机 2024款\nM4\n/\nM4\nPro 芯片优惠直降国补\nApple M4\nmacOS\n256GB\n¥\n4499\n800+人付款\n上海\n3期\n淘金币抵134.9元\n退货宝 />\n\t[42]<a >回头客1万\n壹品良机旗舰店 />\n\t[43]<div />\n\t\t[44]<a title='点此可以直接和卖家交流选好的宝贝，或相互交流网购体验，还支持语音视频噢。'>旺旺在线 />\n[45]<a >2024新款Apple/苹果\nMac\nMini\nM4\nM2 Pro\nmini\n迷你小主机微型电脑\n500GB\n5.1\n英特尔 酷睿 i5-2400\n¥\n2799\n19人付款\n四川\n成都\n公益宝贝 />\n\t[46]<a >19年老店\n成都IT店 />\n\t[47]<div />\n\t\t[48]<a title='点此可以直接和卖家交流选好的宝贝，或相互交流网购体验，还支持语音视频噢。'>旺旺在线 />\n[49]<a >Apple/苹果\nMac\nmini\nM4\n系列芯片一体机\n15英寸\n10 核图形处理器\nmacOS\n¥\n4999\n500+人付款\n河南\n郑州\n包邮 />\n\t[50]<a >傲剑科技 />\n\t[51]<div />\n\t\t[52]<a title='点此可以直接和卖家交流选好的宝贝，或相互交流网购体验，还支持语音视频噢。'>旺旺在线 />\n[53]<a >Apple/苹果\nMac\nmini\n(\nM4\n) 2024款 便携式电脑主机 苹果迷你主机\n256GB\nM4\n16GB\n¥\n3650\n券后价\n1人付款\n广东\n深圳\n店铺满2000减20\n包邮 />\n\t[54]<a >生产力STORE />\n\t[55]<div />\n\t\t[56]<a title='点此可以直接和卖家交流选好的宝贝，或相互交流网购体验，还支持语音视频噢。'>旺旺在线 />\n[57]<a >【6期免息】Apple/苹果\nMac\nmini\n台式电脑主机\nM4\n芯片 10 核中央处理器 10 核图形处理器 2024款新品\nApple M4\n10 核图形处理器\nmacOS\n¥\n5499\n29人付款\n江苏\n南京\n6期\n退货宝\n包邮 />\n\t[58]<a >回头客2千\n响誉数码旗舰店 />\n\t[59]<div />\n\t\t[60]<a title='点此可以直接和卖家交流选好的宝贝，或相互交流网购体验，还支持语音视频噢。'>旺旺在线 />\n[61]<div />\n\t[62]<a >桌面版 />\n[63]<div >联系客服 />\n[64]<div />\n\t[65]<a >购物车 />\n[66]<div />\n\t[67]<a >官方客服 />\n[68]<div />\n\t[69]<a >回顶部 />\n[70]<div />\n\t[71]<a >反馈 />\n... 2215 pixels below - scroll or extract content to see more ...\nCurrent step: 7/100Current date and time: 2025-07-15 19:53\n",
"role": "user"
}
],
"model": "deepseek/deepseek-chat-v3-0324:free",
"parallel_tool_calls": false,
"stream": false,
"temperature": 0.6,
"tool_choice":
{
"function":
{
"name": "AgentOutput"
},
"type": "function"
},
"tools":
[
{
"function":
{
"description": "AgentOutput model with custom actions",
"name": "AgentOutput",
"parameters":
{
"properties":
{
"action":
{
"description": "List of actions to execute",
"items":
{
"properties":
{
"ask_for_assistant":
{
"anyOf":
[
{
"properties":
{
"query":
{
"title": "Query",
"type": "string"
}
},
"required":
[
"query"
],
"title": "ask_for_assistant_parameters",
"type": "object"
},
{
"type": "null"
}
],
"default": null,
"description": "When executing tasks, prioritize autonomous completion. However, if you encounter a definitive blocker that prevents you from proceeding independently – such as needing credentials you don't possess, requiring subjective human judgment, needing a physical action performed, encountering complex CAPTCHAs, or facing limitations in your capabilities – you must request human assistance."
},
"click_element_by_index":
{
"anyOf":
[
{
"properties":
{
"index":
{
"title": "Index",
"type": "integer"
},
"xpath":
{
"anyOf":
[
{
"type": "string"
},
{
"type": "null"
}
],
"default": null,
"title": "Xpath"
}
},
"required":
[
"index"
],
"title": "ClickElementAction",
"type": "object"
},
{
"type": "null"
}
],
"default": null,
"description": "Click element by index"
},
"close_tab":
{
"anyOf":
[
{
"properties":
{
"page_id":
{
"title": "Page Id",
"type": "integer"
}
},
"required":
[
"page_id"
],
"title": "CloseTabAction",
"type": "object"
},
{
"type": "null"
}
],
"default": null,
"description": "Close an existing tab"
},
"done":
{
"anyOf":
[
{
"properties":
{
"success":
{
"title": "Success",
"type": "boolean"
},
"text":
{
"title": "Text",
"type": "string"
}
},
"required":
[
"text",
"success"
],
"title": "DoneAction",
"type": "object"
},
{
"type": "null"
}
],
"default": null,
"description": "Complete task - with return text and if the task is finished (success=True) or not yet completely finished (success=False), because last step is reached"
},
"drag_drop":
{
"anyOf":
[
{
"properties":
{
"coord_source_x":
{
"anyOf":
[
{
"type": "integer"
},
{
"type": "null"
}
],
"default": null,
"description": "Absolute X coordinate on page to start drag from (in pixels)",
"title": "Coord Source X"
},
"coord_source_y":
{
"anyOf":
[
{
"type": "integer"
},
{
"type": "null"
}
],
"default": null,
"description": "Absolute Y coordinate on page to start drag from (in pixels)",
"title": "Coord Source Y"
},
"coord_target_x":
{
"anyOf":
[
{
"type": "integer"
},
{
"type": "null"
}
],
"default": null,
"description": "Absolute X coordinate on page to drop at (in pixels)",
"title": "Coord Target X"
},
"coord_target_y":
{
"anyOf":
[
{
"type": "integer"
},
{
"type": "null"
}
],
"default": null,
"description": "Absolute Y coordinate on page to drop at (in pixels)",
"title": "Coord Target Y"
},
"delay_ms":
{
"anyOf":
[
{
"type": "integer"
},
{
"type": "null"
}
],
"default": 5,
"description": "Delay in milliseconds between steps (0 for fastest, 10-20 for more natural)",
"title": "Delay Ms"
},
"element_source":
{
"anyOf":
[
{
"type": "string"
},
{
"type": "null"
}
],
"default": null,
"description": "CSS selector or XPath of the element to drag from",
"title": "Element Source"
},
"element_source_offset":
{
"anyOf":
[
{
"properties":
{
"x":
{
"title": "X",
"type": "integer"
},
"y":
{
"title": "Y",
"type": "integer"
}
},
"required":
[
"x",
"y"
],
"title": "Position",
"type": "object"
},
{
"type": "null"
}
],
"default": null,
"description": "Precise position within the source element to start drag (in pixels from top-left corner)"
},
"element_target":
{
"anyOf":
[
{
"type": "string"
},
{
"type": "null"
}
],
"default": null,
"description": "CSS selector or XPath of the element to drop onto",
"title": "Element Target"
},
"element_target_offset":
{
"anyOf":
[
{
"properties":
{
"x":
{
"title": "X",
"type": "integer"
},
"y":
{
"title": "Y",
"type": "integer"
}
},
"required":
[
"x",
"y"
],
"title": "Position",
"type": "object"
},
{
"type": "null"
}
],
"default": null,
"description": "Precise position within the target element to drop (in pixels from top-left corner)"
},
"steps":
{
"anyOf":
[
{
"type": "integer"
},
{
"type": "null"
}
],
"default": 10,
"description": "Number of intermediate points for smoother movement (5-20 recommended)",
"title": "Steps"
}
},
"title": "DragDropAction",
"type": "object"
},
{
"type": "null"
}
],
"default": null,
"description": "Drag and drop elements or between coordinates on the page - useful for canvas drawing, sortable lists, sliders, file uploads, and UI rearrangement"
},
"extract_content":
{
"anyOf":
[
{
"properties":
{
"goal":
{
"title": "Goal",
"type": "string"
},
"should_strip_link_urls":
{
"title": "Should Strip Link Urls",
"type": "boolean"
}
},
"required":
[
"goal",
"should_strip_link_urls"
],
"title": "extract_content_parameters",
"type": "object"
},
{
"type": "null"
}
],
"default": null,
"description": "Extract page content to retrieve specific information from the page, e.g. all company names, a specific description, all information about, links with companies in structured format or simply links"
},
"get_dropdown_options":
{
"anyOf":
[
{
"properties":
{
"index":
{
"title": "Index",
"type": "integer"
}
},
"required":
[
"index"
],
"title": "get_dropdown_options_parameters",
"type": "object"
},
{
"type": "null"
}
],
"default": null,
"description": "Get all options from a native dropdown"
},
"go_back":
{
"anyOf":
[
{
"additionalProperties": true,
"description": "Accepts absolutely anything in the incoming data\nand discards it, so the final parsed model is empty.",
"properties":
{},
"title": "NoParamsAction",
"type": "object"
},
{
"type": "null"
}
],
"default": null,
"description": "Go back"
},
"go_to_url":
{
"anyOf":
[
{
"properties":
{
"url":
{
"title": "Url",
"type": "string"
}
},
"required":
[
"url"
],
"title": "GoToUrlAction",
"type": "object"
},
{
"type": "null"
}
],
"default": null,
"description": "Navigate to URL in the current tab"
},
"input_text":
{
"anyOf":
[
{
"properties":
{
"index":
{
"title": "Index",
"type": "integer"
},
"text":
{
"title": "Text",
"type": "string"
},
"xpath":
{
"anyOf":
[
{
"type": "string"
},
{
"type": "null"
}
],
"default": null,
"title": "Xpath"
}
},
"required":
[
"index",
"text"
],
"title": "InputTextAction",
"type": "object"
},
{
"type": "null"
}
],
"default": null,
"description": "Input text into a input interactive element"
},
"open_tab":
{
"anyOf":
[
{
"properties":
{
"url":
{
"title": "Url",
"type": "string"
}
},
"required":
[
"url"
],
"title": "OpenTabAction",
"type": "object"
},
{
"type": "null"
}
],
"default": null,
"description": "Open url in new tab"
},
"save_pdf":
{
"anyOf":
[
{
"properties":
{},
"title": "save_pdf_parameters",
"type": "object"
},
{
"type": "null"
}
],
"default": null,
"description": "Save the current page as a PDF file"
},
"scroll_down":
{
"anyOf":
[
{
"properties":
{
"amount":
{
"anyOf":
[
{
"type": "integer"
},
{
"type": "null"
}
],
"default": null,
"title": "Amount"
}
},
"title": "ScrollAction",
"type": "object"
},
{
"type": "null"
}
],
"default": null,
"description": "Scroll down the page by pixel amount - if no amount is specified, scroll down one page"
},
"scroll_to_text":
{
"anyOf":
[
{
"properties":
{
"text":
{
"title": "Text",
"type": "string"
}
},
"required":
[
"text"
],
"title": "scroll_to_text_parameters",
"type": "object"
},
{
"type": "null"
}
],
"default": null,
"description": "If you dont find something which you want to interact with, scroll to it"
},
"scroll_up":
{
"anyOf":
[
{
"properties":
{
"amount":
{
"anyOf":
[
{
"type": "integer"
},
{
"type": "null"
}
],
"default": null,
"title": "Amount"
}
},
"title": "ScrollAction",
"type": "object"
},
{
"type": "null"
}
],
"default": null,
"description": "Scroll up the page by pixel amount - if no amount is specified, scroll up one page"
},
"search_google":
{
"anyOf":
[
{
"properties":
{
"query":
{
"title": "Query",
"type": "string"
}
},
"required":
[
"query"
],
"title": "SearchGoogleAction",
"type": "object"
},
{
"type": "null"
}
],
"default": null,
"description": "Search the query in Google in the current tab, the query should be a search query like humans search in Google, concrete and not vague or super long. More the single most important items. "
},
"select_dropdown_option":
{
"anyOf":
[
{
"properties":
{
"index":
{
"title": "Index",
"type": "integer"
},
"text":
{
"title": "Text",
"type": "string"
}
},
"required":
[
"index",
"text"
],
"title": "select_dropdown_option_parameters",
"type": "object"
},
{
"type": "null"
}
],
"default": null,
"description": "Select dropdown option for interactive element index by the text of the option you want to select"
},
"send_keys":
{
"anyOf":
[
{
"properties":
{
"keys":
{
"title": "Keys",
"type": "string"
}
},
"required":
[
"keys"
],
"title": "SendKeysAction",
"type": "object"
},
{
"type": "null"
}
],
"default": null,
"description": "Send strings of special keys like Escape,Backspace, Insert, PageDown, Delete, Enter, Shortcuts such as `Control+o`, `Control+Shift+T` are supported as well. This gets used in keyboard.press. "
},
"switch_tab":
{
"anyOf":
[
{
"properties":
{
"page_id":
{
"title": "Page Id",
"type": "integer"
}
},
"required":
[
"page_id"
],
"title": "SwitchTabAction",
"type": "object"
},
{
"type": "null"
}
],
"default": null,
"description": "Switch tab"
},
"upload_file":
{
"anyOf":
[
{
"properties":
{
"index":
{
"title": "Index",
"type": "integer"
},
"path":
{
"title": "Path",
"type": "string"
}
},
"required":
[
"index",
"path"
],
"title": "upload_file_parameters",
"type": "object"
},
{
"type": "null"
}
],
"default": null,
"description": "Upload file to interactive element with file path "
},
"wait":
{
"anyOf":
[
{
"properties":
{
"seconds":
{
"default": 3,
"title": "Seconds",
"type": "integer"
}
},
"title": "wait_parameters",
"type": "object"
},
{
"type": "null"
}
],
"default": null,
"description": "Wait for x seconds default 3"
}
},
"type": "object"
},
"min_items": 1,
"type": "array"
},
"current_state":
{
"description": "Current state of the agent",
"properties":
{
"evaluation_previous_goal":
{
"type": "string"
},
"memory":
{
"type": "string"
},
"next_goal":
{
"type": "string"
}
},
"required":
[
"evaluation_previous_goal",
"memory",
"next_goal"
],
"type": "object"
}
},
"required":
[
"current_state",
"action"
],
"type": "object"
}
},
"type": "function"
}
]
}