feat: CDP bridge batch support - cookies/tabs/cdp mixed commands, lazy attach, $N chain refs, file upload & screenshot verified

assets/tmwd_cdp_bridge/background.js

@@ -10,6 +10,10 @@ chrome.runtime.onMessage.addListener((msg, sender, sendResponse) => {
     handleCDP(msg, sender).then(sendResponse);
     return true;
   }
+  if (msg.action === 'batch') {
+    handleBatch(msg, sender).then(sendResponse);
+    return true;
+  }
   if (msg.action === 'tabs') {
     (async () => {
       try {
@@ -44,6 +48,38 @@ async function handleCookies(msg, sender) {
   }
 }
 
+async function handleBatch(msg, sender) {
+  const R = [];
+  let attached = null;
+  const resolve$N = (params) => JSON.parse(JSON.stringify(params || {}).replace(/"\$(\d+)\.([^"]+)"/g,
+    (_, i, path) => { let v = R[+i]; for (const k of path.split('.')) v = v[k]; return JSON.stringify(v); }));
+  try {
+    for (const c of msg.commands) {
+      if (c.cmd === 'cookies') {
+        R.push(await handleCookies(c, sender));
+      } else if (c.cmd === 'tabs') {
+        const tabs = await chrome.tabs.query({});
+        R.push({ ok: true, data: tabs.map(t => ({ id: t.id, url: t.url, title: t.title, active: t.active, windowId: t.windowId })) });
+      } else if (c.cmd === 'cdp') {
+        const tabId = c.tabId || msg.tabId || sender.tab?.id;
+        if (attached !== tabId) {
+          if (attached) { await chrome.debugger.detach({ tabId: attached }); attached = null; }
+          await chrome.debugger.attach({ tabId }, '1.3');
+          attached = tabId;
+        }
+        R.push(await chrome.debugger.sendCommand({ tabId }, c.method, resolve$N(c.params)));
+      } else {
+        R.push({ ok: false, error: 'unknown cmd: ' + c.cmd });
+      }
+    }
+    if (attached) await chrome.debugger.detach({ tabId: attached });
+    return { ok: true, results: R };
+  } catch (e) {
+    if (attached) try { await chrome.debugger.detach({ tabId: attached }); } catch (_) {}
+    return { ok: false, error: e.message, results: R };
+  }
+}
+
 async function handleCDP(msg, sender) {
   const tabId = msg.tabId || sender.tab?.id;
   if (!tabId) return { ok: false, error: 'no tabId' };

assets/tmwd_cdp_bridge/content.js

@@ -19,6 +19,8 @@ async function handle(el) {
       resp = await chrome.runtime.sendMessage({ action: 'cookies', url: req.url || location.href });
     } else if (cmd === 'cdp') {
       resp = await chrome.runtime.sendMessage({ action: 'cdp', method: req.method, params: req.params || {}, tabId: req.tabId });
+    } else if (cmd === 'batch') {
+      resp = await chrome.runtime.sendMessage({ action: 'batch', commands: req.commands, tabId: req.tabId });
     } else if (cmd === 'tabs') {
       resp = await chrome.runtime.sendMessage({ action: 'tabs', method: req.method, tabId: req.tabId });
     } else {

assets/tools_schema.json

@@ -44,7 +44,8 @@
     "description": "万能网页操控工具。通过执行 JavaScript 脚本实现对浏览器的完全控制（如点击、滚动、提取特定数据）。鼓励在有把握情况下（记忆中有selector/做法等）精准使用以减少web_scan调用。执行结果可选择保存到本地文件进行后续分析。",
     "parameters": {"type": "object", "properties": {
       "script": {"type": "string", "description": "要执行的 JavaScript 代码或JS文件路径。"},
-      "save_to_file": {"type": "string", "description": "可选。将 JS 执行结果（js_return）保存到的文件路径。该功能不支持 await 等异步结果。"}}, "required": ["script"]}
+      "save_to_file": {"type": "string", "description": "结果存文件，适合返回值较长时。不支持await。", "default": ""},
+      "no_monitor": {"type": "boolean", "description": "跳过页面变更监控，仅读取信息时用，省2-3秒。", "default": false}}, "required": ["script"]}
   }},
   {"type": "function", "function": {
     "name": "update_working_checkpoint",

ga.py

@@ -158,7 +158,7 @@ def log_memory_access(path):
     stats[fname] = {'count': stats.get(fname, {}).get('count', 0) + 1, 'last': datetime.now().strftime('%Y-%m-%d')}
     with open(stats_file, 'w', encoding='utf-8') as f: json.dump(stats, f, indent=2, ensure_ascii=False)
 
-def web_execute_js(script, switch_tab_id=None):
+def web_execute_js(script, switch_tab_id=None, no_monitor=False):
     """
     执行 JS 脚本来控制浏览器，并捕获结果和页面变化。
     script: 要执行的 JavaScript 代码字符串。
@@ -180,7 +180,7 @@ def web_execute_js(script, switch_tab_id=None):
         if len(driver.get_all_sessions()) == 0:
             return {"status": "error", "msg": "没有可用的浏览器标签页，请先打开一个浏览器标签页，且确认TMWebDriver浏览器tempermonkey插件已安装并启用。"}
         if switch_tab_id: driver.default_session_id = switch_tab_id
-        result = execute_js_rich(script, driver)
+        result = execute_js_rich(script, driver, no_monitor=no_monitor)
         return result
     except Exception as e:
         return {"status": "error", "msg": format_error(e)}
@@ -318,7 +318,8 @@ class GenericAgentHandler(BaseHandler):
             with open(abs_path, 'r', encoding='utf-8') as f: script = f.read()
         save_to_file = args.get("save_to_file", "")
         switch_tab_id = args.get("switch_tab_id") or args.get("tab_id")
-        result = web_execute_js(script, switch_tab_id=switch_tab_id)
+        no_monitor = args.get("no_monitor", False)
+        result = web_execute_js(script, switch_tab_id=switch_tab_id, no_monitor=no_monitor)
         if save_to_file and "js_return" in result:
             content = str(result["js_return"] or '')
             abs_path = self._get_abs_path(save_to_file)

memory/tmwebdriver_sop.md

@@ -8,13 +8,14 @@
 ## 限制(isTrusted)
 - JS dispatch的事件`isTrusted=false`，敏感操作(文件上传/部分按钮)会被浏览器拦截
 - ⭐**首选绕过：CDP桥**——CDP派发的Input事件是浏览器原生级别(isTrusted=true)，且无需前台，见下方CDP章节
-- 文件上传：JS无法填充`<input type=file>`，仍需ljqCtrl物理点击+Win32轮询文件对话框
+- 文件上传：JS无法填充`<input type=file>`
-  - 流程：SetForegroundWindow→ljqCtrl点上传按钮→FindWindow轮询对话框→输入路径→轮询关闭
+  - ⭐首选CDP batch：getDocument→querySelector→DOM.setFileInputFiles(无需前台/物理点击)
+  - 备选ljqCtrl物理点击：SetForegroundWindow→点上传按钮→FindWindow轮询对话框→输入路径→轮询关闭
 - 备选：元素→屏幕物理坐标(ljqCtrl/PostMessage点击前必算)：JS一次取rect+窗口信息，公式：
   - `physX = (screenX + rect中心x) * dpr`，`physY = (screenY + chromeH + rect中心y) * dpr`
   - chromeH = outerHeight - innerHeight，dpr = devicePixelRatio
   - 注意：screenX/Y也是CSS像素，所有值先加后统一乘dpr
-- 结论：读信息+普通操作用TMWebDriver；需isTrusted事件首选CDP桥；文件上传需配合ljqCtrl
+- 结论：读信息+普通操作用TMWebDriver；需isTrusted事件首选CDP桥；文件上传首选CDP三连(备选ljqCtrl)
 
 ## 导航
 - `web_scan` 仅读当前页不导航，切换网站用 `web_execute_js` + `location.href='url'`
@@ -54,8 +55,13 @@ el.id = '__ljq_ctrl'; el.style.display = 'none';
 el.textContent = JSON.stringify({cmd:'...', ...});
 document.body.appendChild(el);  // 响应写回el.textContent
 ```
-命令：`{cmd:'tabs'}` | `{cmd:'cookies'}` | `{cmd:'cdp', tabId:N, method:'...', params:{...}}`
+单命令：`{cmd:'tabs'}` | `{cmd:'cookies'}` | `{cmd:'cdp', tabId:N, method:'...', params:{...}}`
-- CDP可用任意方法(Input/Network/DOM/Page/Runtime/Emulation等)，每次attach→send→detach
+- ⭐batch混合：`{cmd:'batch', commands:[{cmd:'cookies'},{cmd:'tabs'},{cmd:'cdp',...},...]}`
+  - 返回`{ok:true, results:[...]}`，一次请求多命令，CDP懒attach复用session
+  - `$N.path`引用第N个结果字段(0-indexed)，如`"nodeId":"$2.root.nodeId"`
+  - 典型：文件上传三连 getDocument→querySelector(input[type=file])→setFileInputFiles
+  - ⚠tabId：CDP默认sender.tab.id(当前注入页)，跨tab需显式tabId或先batch内tabs查
+- CDP可用任意方法(Input/Network/DOM/Page/Runtime/Emulation等)，单条每次attach→send→detach
 - ⭐跨tab无需前台：指定tabId即可操作后台标签页
 - ⭐绕过isTrusted：CDP派发的Input事件是浏览器原生级别
 
@@ -66,7 +72,8 @@ document.body.appendChild(el);  // 响应写回el.textContent
   - 坑：多RenderWidgetHostHWND共存，必须按父窗口标题匹配再取子窗口
 
 ## 验证码/页面视觉截图
-- 优先：JS `canvas.toDataURL()` 直接拿base64（验证码是canvas/img时最干净，无需截屏）
+- ⭐首选CDP截图：`Page.captureScreenshot`(format:'png')→返回base64，无需前台/后台tab也行，全页高清
+- 验证码canvas/img：JS `canvas.toDataURL()` 直接拿base64最干净
 - 备选：`window.open(location.href,'_blank')` 前台开新标签→win32截图→完后close
   - GM_openInTab在web_execute_js不可用（非油猴上下文）
   - 浏览器无JS API切标签页，只能开新的来保证前台

simphtml.py

@@ -860,9 +860,11 @@ def get_html(driver, cutlist=False, maxchars=28000, instruction="", extra_js="")
     if len(ss) > maxchars: ss = ss[:maxchars] + ' ... [TRUNCATED]'
     return ss
 
-def execute_js_rich(script, driver):
+def execute_js_rich(script, driver, no_monitor=False):
+    last_html = None
+    if not no_monitor:
         try: last_html = get_html(driver, cutlist=False, extra_js=temp_monitor_js)
-    except: last_html = None
+        except: pass
     result = None;  error_msg = None;  reloaded = False; newTabs = []
     before_sids = set(driver.get_session_dict().keys())
     try:
@@ -889,6 +891,7 @@ def execute_js_rich(script, driver):
         rr['environment']['newTabs'] = newTabs
         rr['suggestion'] = "页面已刷新，以上新标签页在执行期间连接。"
     if error_msg: rr['error'] = error_msg
+    if no_monitor: return rr
     if not reloaded:
         try: rr['transients'] = get_temp_texts(driver)
         except: rr['transients'] = []