Jarvis-Brain 0.1.11.2__tar.gz → 0.1.11.10__tar.gz

{jarvis_brain-0.1.11.2 → jarvis_brain-0.1.11.10}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: Jarvis_Brain
-Version: 0.1.11.2
+Version: 0.1.11.10
 Summary: Jarvis brain mcp
 Requires-Python: >=3.10
 Requires-Dist: beautifulsoup4

{jarvis_brain-0.1.11.2 → jarvis_brain-0.1.11.10}/mcp_tools/dp_tools.py

@@ -7,6 +7,7 @@ import os
 import time
 from typing import Any
 
+import DrissionPage
 from fastmcp import FastMCP
 
 from tools.browser_manager import BrowserManager
@@ -162,13 +163,13 @@ def register_check_selector(mcp: FastMCP, browser_manager):
             attr_output = json.dumps(ele_attr_list, ensure_ascii=False)
         # 对attr_output逐个截断，截断的长度为：一轮最大token除以元素个数+3个点+两个引号和逗号
         return dp_mcp_message_pack(
-            f"已完成tab页:【{tab_id}】对：【{css_selector}】的检查",
+            f"已完成tab页:【{tab_id}】对：【{css_selector}】的检查，当前选中了 {len(target_eles)} 个元素",
             tab_id=tab_id,
             selector=css_selector,
             selector_ele_exist=exist_flag,
             page_size=page_size,
             offset=offset,
-            attr_output=attr_output
+            attr_output=attr_output,
         )
 
 
@@ -236,27 +237,33 @@ def register_assert_waf(mcp: FastMCP, browser_manager):
 
 
 def register_click_action(mcp: FastMCP, browser_manager):
-    @mcp.tool(name="click_action", description="尝试点击tab页中的元素，返回元素是否可以被点击，以及是否点击成功。")
-    async def click_action(browser_port: int, tab_id: str, css_selector: str) -> dict[str, Any]:
+    @mcp.tool(name="click_action",
+              description="尝试点击tab页中的元素，返回元素是否可以被点击，以及是否点击成功。"
+                          "其中target_element_index默认为0，当传入的Selector可以定位到多个元素时，需要传入target_element_index指定具体点击目标 ")
+    async def click_action(browser_port: int, tab_id: str, css_selector: str, target_element_index: int = 0) -> dict[
+        str, Any]:
         _browser = browser_manager.get_browser(browser_port)
         target_tab = _browser.get_tab(tab_id)
         css_selector = css_selector
         if "css:" not in css_selector:
             css_selector = "css:" + css_selector
         target_eles = target_tab.eles(css_selector)
-        click_success = False
-        element_clickable = False
-        if len(target_eles) == 1:
-            target_element = target_eles[0]
-            element_clickable = target_element.states.is_clickable
-            try:
-                target_element.click()
-                click_success = True
-            except Exception as e:
-                click_success = False
-            message = f"tab页:【{tab_id}】点击【{css_selector}】 {'成功' if click_success else '失败'} 了"
+        # click_success = False
+        # element_clickable = False
+        # if len(target_eles) == 1:
+        target_element = target_eles[target_element_index]
+        element_clickable = target_element.states.is_clickable
+        try:
+            target_element.click()
+            click_success = True
+        except Exception as e:
+            click_success = False
+        if target_element_index > 0:
+            message = f"tab页:【{tab_id}】点击【{css_selector}】【index={target_element_index}】的元素 {'成功' if click_success else '失败'} 了"
         else:
-            message = f"tab页:【{tab_id}】传入的css_selector找到了{len(target_eles)}个元素，请确保传入的css_selector可以找到唯一的一个元素"
+            message = f"tab页:【{tab_id}】点击【{css_selector}】 {'成功' if click_success else '失败'} 了"
+        # else:
+        #     message = f"tab页:【{tab_id}】传入的css_selector找到了{len(target_eles)}个元素，请确保传入的css_selector可以找到唯一的一个元素"
         return dp_mcp_message_pack(
             message=message,
             browser_port=browser_port,
@@ -308,16 +315,19 @@ def register_scroll_action(mcp: FastMCP, browser_manager):
 
 
 def register_get_screenshot(mcp: FastMCP, browser_manager):
-    @mcp.tool(name="get_tab_screenshot", description="尝试对传入tab页进行截图，并将截图压缩为1M大小png图片，会返回截图保存路径")
+    @mcp.tool(name="get_tab_screenshot",
+              description="尝试对传入tab页进行截图，并将截图压缩为1M大小png图片，会返回截图保存路径")
     async def get_tab_screenshot(browser_port: int, tab_id: str) -> dict[str, Any]:
         _browser = browser_manager.get_browser(browser_port)
         target_tab = _browser.get_tab(tab_id)
+        target_tab.wait.doc_loaded()
         if not os.path.exists(html_source_code_local_save_path):
             os.makedirs(html_source_code_local_save_path)
         timestamp = int(time.time() * 1000)
-        origin_png = target_tab.get_screenshot(as_bytes="png")
-        compress_png = compress_image_bytes(origin_png)
-        image_path = os.path.join(html_source_code_local_save_path, f"{browser_port}_{tab_id}_{timestamp}.png")
+        # time.sleep(1)
+        origin_png = target_tab.get_screenshot(as_bytes="jpg", full_page=True)
+        compress_png = compress_image_bytes(origin_png, 0.5)
+        image_path = os.path.join(html_source_code_local_save_path, f"{browser_port}_{tab_id}_{timestamp}.jpg")
         with open(image_path, "wb") as f:
             f.write(compress_png)
         return dp_mcp_message_pack(

{jarvis_brain-0.1.11.2 → jarvis_brain-0.1.11.10}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "Jarvis_Brain"  # 别人下载时用的名字，必须在 PyPI 上唯一
-version = "0.1.11.2"
+version = "0.1.11.10"
 description = "Jarvis brain mcp"
 dependencies = [
     "fastmcp",

{jarvis_brain-0.1.11.2 → jarvis_brain-0.1.11.10}/tools/browser_proxy.py

@@ -140,7 +140,10 @@ def check_data_packet(packet: DataPacket, client: DPProxyClient):
         data = packet.request.postData
     domain = urlparse(url).netloc
     body = packet.response.body
-    body_str = json.dumps(body, ensure_ascii=False, separators=(',', ':'))
+    if isinstance(body, dict):
+        body_str = json.dumps(body, ensure_ascii=False, separators=(',', ':'))
+    else:
+        body_str = str(body)
     body_str_list = [body_str[i:i + one_turn_max_token] for i in range(0, len(body_str), one_turn_max_token)]
     body_completed = True
     packet_filter = client.packet_filter
@@ -155,9 +158,9 @@ def check_data_packet(packet: DataPacket, client: DPProxyClient):
             continue
         if (index + 1) != len(body_str_list):
             body_completed = False
-        if packet.response:
+        try:
             response_headers = packet.response.headers
-        else:
+        except TypeError:
             response_headers = {}
         temp_dict = {
             "url": url,
@@ -171,59 +174,4 @@ def check_data_packet(packet: DataPacket, client: DPProxyClient):
         client.packet_queue.append(temp_dict)
 
 
-def check_data_packet(packet: DataPacket, client: DPProxyClient):
-    """
-    封装监听到的数据包，并将其存放在client的packet_queue中
-    :param packet:
-    :param client:
-    :return:
-    """
-    url = packet.url
-    method = packet.request.method
-    data = None
-    if packet.request.hasPostData:
-        data = packet.request.postData
-    domain = urlparse(url).netloc
-    body = packet.response.body
-    body_str = json.dumps(body, ensure_ascii=False, separators=(',', ':'))
-    body_str_list = [body_str[i:i + one_turn_max_token] for i in range(0, len(body_str), one_turn_max_token)]
-    body_completed = True
-    packet_filter = client.packet_filter
-    domain_filter = packet_filter.get("domain_filter", None)
-    method_filter = packet_filter.get("method_filter", ["GET", "POST"])
-    for index, body_str in enumerate(body_str_list):
-        # 如果给了domain_filter并且domain没有在domain_filter中时跳过该数据包
-        if domain_filter and domain not in domain_filter:
-            continue
-        # 如果method没有在method_filter中，则跳过该数据包
-        if method not in method_filter:
-            continue
-        if (index + 1) != len(body_str_list):
-            body_completed = False
-        temp_dict = {
-            "url": url,
-            "body_completed": body_completed,
-            "method": method,
-            "request_data": data,
-            "request_headers": dict(packet.request.headers),
-            "response_headers": dict(packet.response.headers),
-            "response_body_segment": body_str.replace("\\", ""),
-        }
-        client.packet_queue.append(temp_dict)
-
-
 client_manager = DPProxyClientManager()
-
-# if __name__ == '__main__':
-#     co = ChromiumOptions().set_user_agent(
-#         "Mozilla/5.0 (Linux; Android 6.0; Nexus 5 Build/MRA58N) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/143.0.0.0 Mobile Safari/537.36")
-#     tab = ChromiumPage(co).latest_tab
-#     client = DPProxyClient(tab, self_kill=False)
-#     # client = CaptchaClient(tab, self_kill=True)
-#     tab = client.get_driver(True)
-#     url = "https://api.toutiaoapi.com/feoffline/hotspot_and_local/html/hot_list/index.html?client_extra_params=%7B%22custom_log_pb%22%3A%22%7B%5C%22style_id%5C%22%3A%5C%2240030%5C%22%2C%5C%22entrance_hotspot%5C%22%3A%5C%22search%5C%22%2C%5C%22location%5C%22%3A%5C%22hot_board%5C%22%2C%5C%22category_name%5C%22%3A%5C%22hotboard_light%5C%22%7D%22%7D&count=50&log_pb=%7B%22style_id%22%3A%2240030%22%2C%22entrance_hotspot%22%3A%22search%22%2C%22location%22%3A%22hot_board%22%2C%22category_name%22%3A%22hotboard_light%22%7D&only_hot_list=1&tab_name=stream&enter_keyword=%23%E7%BE%8E%E5%9B%BD%E9%80%80%E5%87%BA66%E4%B8%AA%E5%9B%BD%E9%99%85%E7%BB%84%E7%BB%87%23"
-#     tab.get(url)
-#     for _ in range(5056):
-#         new_packet = client.pop_first_packet()
-#         print(new_packet, "23")
-#         time.sleep(1)

{jarvis_brain-0.1.11.2 → jarvis_brain-0.1.11.10}/tools/tools.py

@@ -11,84 +11,95 @@ from PIL import Image
 import io
 
 compress_html_js = """
-    function getSimplifiedDOM(node) {
-        // 1. 处理文本节点
-        if (node.nodeType === Node.TEXT_NODE) {
-            const text = node.textContent.trim();
-            // 限制文本长度，避免大段文章消耗 token，保留前100个字符通常足够定位
-            return text ? text.slice(0, 100) + (text.length > 100 ? '...' : '') : null;
-        }
-
-        // 2. 过滤无用标签
-        const ignoreTags = ['SCRIPT', 'STYLE', 'NOSCRIPT', 'IFRAME', 'SVG', 'LINK', 'META'];
-        if (ignoreTags.includes(node.tagName)) return null;
-        if (node.nodeType !== Node.ELEMENT_NODE) return null;
-
-        // 3. 过滤不可见元素
-        const style = window.getComputedStyle(node);
-        if (style.display === 'none' || style.visibility === 'hidden' || style.opacity === '0') return null;
-        // 过滤宽高太小的元素（往往是埋点空像素）
-        const rect = node.getBoundingClientRect();
-        if (rect.width === 0 || rect.height === 0) return null;
-
-        // --- 开始构建标签字符串 ---
-        const tagName = node.tagName.toLowerCase();
-        let tagStr = tagName;
-
-        // A. 基础标识符 (ID 和 Class)
-        if (node.id) tagStr += `#${node.id}`;
-        if (node.className && typeof node.className === 'string') {
-            // 过滤掉 Tailwind 等太长且无语义的 class，保留有意义的业务 class
-            // 这里简单处理，全部保留，让 LLM 自己判断
-            const classes = node.className.trim().split(/\s+/);
-            if (classes.length > 0) tagStr += `.${classes.join('.')}`;
-        }
-
-        // B. 关键属性白名单 (这是你指出问题的核心修复)
-        const props = [];
-
-        // 通用重要属性
-        if (node.getAttribute('role')) props.push(`role="${node.getAttribute('role')}"`);
-        if (node.getAttribute('aria-label')) props.push(`aria-label="${node.getAttribute('aria-label')}"`);
-        if (node.getAttribute('title')) props.push(`title="${node.getAttribute('title')}"`);
-
-        // 特定标签的特定属性
-        if (tagName === 'a') {
-            const href = node.getAttribute('href');
-            // 只保留有意义的链接，忽略 javascript:;
-            if (href && !href.startsWith('javascript')) props.push(`href="${href}"`);
-        } else if (tagName === 'input' || tagName === 'textarea' || tagName === 'select') {
-            if (node.getAttribute('type')) props.push(`type="${node.getAttribute('type')}"`);
-            if (node.getAttribute('name')) props.push(`name="${node.getAttribute('name')}"`);
-            if (node.getAttribute('placeholder')) props.push(`placeholder="${node.getAttribute('placeholder')}"`);
-            if (node.disabled) props.push('disabled');
-            if (node.checked) props.push('checked');
-        } else if (tagName === 'button') {
-            if (node.getAttribute('type')) props.push(`type="${node.getAttribute('type')}"`);
-        } else if (tagName === 'img') {
-            if (node.getAttribute('alt')) props.push(`alt="${node.getAttribute('alt')}"`);
-        }
-
-        if (props.length > 0) {
-            tagStr += ` ${props.join(' ')}`;
-        }
-
-        // 4. 递归子节点
-        const children = Array.from(node.childNodes)
-            .map(getSimplifiedDOM)
-            .filter(n => n !== null);
-
-        // 5. 组装输出
-        // 如果没有子节点，也没有ID/Class，也不是输入框/图片/链接，那这个标签可能只是布局用的 div，可以考虑跳过它直接返回子节点内容
-        // 但为了保持结构完整，我们暂时保留它
-        if (children.length === 0) {
-            // 自闭合标签或空标签
-            return `<${tagStr} />`;
-        }
-        return `<${tagStr}>${children.join('')}</${tagName}>`; // 结束标签只保留 tagName 节省 token
+function getSimplifiedDOM(node) {
+    // 1. 处理文本节点
+    if (node.nodeType === Node.TEXT_NODE) {
+        const text = node.textContent.trim();
+        return text ? text.slice(0, 100) + (text.length > 100 ? '...' : '') : null;
     }
 
-    return getSimplifiedDOM(document.body);
+    // 2. 过滤无用标签
+    const ignoreTags = ['SCRIPT', 'STYLE', 'NOSCRIPT', 'IFRAME', 'SVG', 'LINK', 'META'];
+    if (ignoreTags.includes(node.tagName)) return null;
+    if (node.nodeType !== Node.ELEMENT_NODE) return null;
+
+    // 3. 过滤不可见元素
+    // 【注意】这里声明了第一次 style
+    const style = window.getComputedStyle(node);
+    
+    if (style.display === 'none' || style.visibility === 'hidden' || style.opacity === '0') return null;
+    
+    // 过滤宽高太小的元素（往往是埋点空像素）
+    const rect = node.getBoundingClientRect();
+    
+    // 【修复点】删除了这里重复的 const style = ... 代码
+    // 直接使用上面已经定义好的 style 变量即可
+    
+    // 如果宽高为0，但溢出可见，说明可能有定位的子元素显示在外面
+    if ((rect.width === 0 || rect.height === 0) && style.overflow !== 'visible') return null;
+
+    // --- 开始构建标签字符串 ---
+    const tagName = node.tagName.toLowerCase();
+    let tagStr = tagName;
+
+    // A. 基础标识符 (ID 和 Class)
+    if (node.id) tagStr += `#${node.id}`;
+    if (node.className && typeof node.className === 'string') {
+        const classes = node.className.trim().split(/\s+/);
+        if (classes.length > 0) tagStr += `.${classes.join('.')}`;
+    }
+
+    // B. 关键属性白名单
+    const props = [];
+
+    // 通用重要属性
+    if (node.getAttribute('role')) props.push(`role="${node.getAttribute('role')}"`);
+    if (node.getAttribute('aria-label')) props.push(`aria-label="${node.getAttribute('aria-label')}"`);
+    if (node.getAttribute('title')) props.push(`title="${node.getAttribute('title')}"`);
+    // 建议增加这个，很多弹窗用这个属性
+    if (node.getAttribute('aria-modal')) props.push(`aria-modal="${node.getAttribute('aria-modal')}"`);
+
+    // 特定标签的特定属性
+    if (tagName === 'a') {
+        const href = node.getAttribute('href');
+        if (href && !href.startsWith('javascript')) props.push(`href="${href}"`);
+    } else if (tagName === 'input' || tagName === 'textarea' || tagName === 'select') {
+        if (node.getAttribute('type')) props.push(`type="${node.getAttribute('type')}"`);
+        if (node.getAttribute('name')) props.push(`name="${node.getAttribute('name')}"`);
+        if (node.getAttribute('placeholder')) props.push(`placeholder="${node.getAttribute('placeholder')}"`);
+        if (node.disabled) props.push('disabled');
+        if (node.checked) props.push('checked');
+    } else if (tagName === 'button') {
+        if (node.getAttribute('type')) props.push(`type="${node.getAttribute('type')}"`);
+    } else if (tagName === 'img') {
+        if (node.getAttribute('alt')) props.push(`alt="${node.getAttribute('alt')}"`);
+    } else if (tagName === 'dialog') {
+        // 保留 open 属性
+        if (node.open) props.push('open');
+    }
+
+    if (props.length > 0) {
+        tagStr += ` ${props.join(' ')}`;
+    }
+
+    // 4. 递归子节点 (包含 Shadow DOM 处理)
+    let childNodes = Array.from(node.childNodes);
+    if (node.shadowRoot) {
+        childNodes = [...childNodes, ...Array.from(node.shadowRoot.childNodes)];
+    }
+    
+    const children = childNodes
+        .map(getSimplifiedDOM)
+        .filter(n => n !== null);
+    
+    // 5. 组装输出
+    if (children.length === 0) {
+        return `<${tagStr} />`;
+    }
+    return `<${tagStr}>${children.join('')}</${tagName}>`; 
+}
+
+return getSimplifiedDOM(document.body);
 """