@bsbofmusic/cdper-doubao 1.0.0 → 1.0.1

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@bsbofmusic/cdper-doubao",
-  "version": "1.0.0",
+  "version": "1.0.1",
   "description": "Standalone Doubao CLI controlled through a user's real Chrome/Edge via CDP Bridge",
   "type": "commonjs",
   "main": "src/public-api.js",
@@ -39,7 +39,7 @@
     "node": ">=18.0.0"
   },
   "dependencies": {
-    "@bsbofmusic/cdper-core": "^1.2.4",
+    "@bsbofmusic/cdper-core": "^1.2.5",
     "yargs": "^17.7.2"
   }
-}
+}

package/src/doubao.js

@@ -43,7 +43,7 @@ function buildDoubaoUrlAction(queryText, mode = 'deep') {
       use_deep_think: '1',
     };
     action.options = {
-      deepThinkingActiveType: '3',
+      deepThinkingActiveType: '2',
       superTaskStatus: { switchValue: 0, taskMode: 0, safeToken: '' },
     };
   }
@@ -161,6 +161,7 @@ async function getDoubaoUiState(page) {
       return ['快速', 'Quick'].includes(meta.text) && meta.visible;
     });
     const expertButton = buttons.find((button) => /^(?:专家|Expert)($|\n)/.test(visibleButton(button).text) && visibleButton(button).visible);
+    const superModeButton = buttons.find((button) => /^(?:超能模式|Super|Deep)($|\n)/i.test(visibleButton(button).text) && visibleButton(button).visible);
     const menu = document.querySelector('[role="menu"]');
     const menuItems = menu
       ? Array.from(menu.querySelectorAll('[role="menuitem"]')).map((node) => ({
@@ -179,7 +180,7 @@ async function getDoubaoUiState(page) {
     });
     const loading = !!document.querySelector('[class*="loading"], [class*="spinner"], [class*="skeleton"]');
     const unavailable = bodyText.includes('该页面暂时不可用');
-    return { url: location.href, title: document.title || '', hasInput, hasQuickButton: !!quickButton, hasExpertButton: !!expertButton, menuVisible: !!menu, menuItems, loading, unavailable };
+    return { url: location.href, title: document.title || '', hasInput, hasQuickButton: !!quickButton, hasExpertButton: !!expertButton, hasSuperModeButton: !!superModeButton, menuVisible: !!menu, menuItems, loading, unavailable };
   });
 }
 
@@ -230,6 +231,17 @@ async function adoptExistingDoubaoPage(browser, fallbackPage) {
   return fallbackPage;
 }
 
+async function clickHandleAtCenter(page, handle, label) {
+  try {
+    const box = await handle.boundingBox();
+    if (box && box.width > 0 && box.height > 0) {
+      await page.mouse.click(box.x + box.width / 2, box.y + box.height / 2, { delay: 80 });
+      return true;
+    }
+  } catch (error) { log('WARN', '[doubao]', label + '坐标点击失败: ' + error.message); }
+  return false;
+}
+
 async function clickHandleWithoutCoordinates(handle, label) {
   try {
     await withLocalTimeout(handle.evaluate((element) => element.click()), label + ' DOM click');
@@ -270,6 +282,20 @@ async function clickButtonByExactText(page, text, label) {
 }
 
 async function clickMenuItemByPrefix(page, prefix, label) {
+  try {
+    const items = await page.$$('[role="menuitem"]');
+    for (const item of items) {
+      const meta = await item.evaluate((node) => {
+        const text = (node.innerText || '').trim();
+        const rect = node.getBoundingClientRect();
+        return { text, visible: rect.width > 0 && rect.height > 0 };
+      }).catch(() => null);
+      if (!meta || !meta.visible || !meta.text.startsWith(prefix)) continue;
+      if (await clickHandleAtCenter(page, item, label + '坐标')) return true;
+      if (await clickHandleWithoutCoordinates(item, label)) return true;
+    }
+  } catch (error) { log('WARN', '[doubao]', label + '坐标点击链路失败: ' + error.message); }
+
   try {
     const clicked = await page.evaluate((targetPrefix) => {
       const menu = document.querySelector('[role="menu"]');
@@ -352,12 +378,19 @@ async function getDoubaoDirectState(page) {
       return rect.width >= 80 && rect.height >= 20 && element.getAttribute('aria-hidden') !== 'true' && !element.disabled;
     });
     const inputText = String(input?.value || input?.innerText || input?.textContent || '').trim();
-    const stopButton = Array.from(document.querySelectorAll('button')).find((button) => {
+    const isVisible = (element) => {
+      if (!element) return false;
+      const rect = element.getBoundingClientRect();
+      const style = window.getComputedStyle(element);
+      return rect.width > 0 && rect.height > 0 && style.visibility !== 'hidden' && style.display !== 'none' && element.getAttribute('aria-hidden') !== 'true';
+    };
+    const stopButton = Array.from(document.querySelectorAll('button,[role="button"]')).find((button) => {
+      if (!isVisible(button) || button.disabled || button.getAttribute('aria-disabled') === 'true') return false;
       const text = button.innerText || '';
       const aria = button.getAttribute('aria-label') || '';
-      return /停止|stop/i.test(`${text} ${aria}`);
+      return /停止生成|停止回答|Stop generating|stop-button/i.test(`${text} ${aria}`) || !!button.querySelector('svg[icon="stop"]');
     });
-    const loading = document.querySelector('[class*="loading"], [class*="spinner"], [class*="skeleton"]');
+    const loading = Array.from(document.querySelectorAll('[class*="loading"], [class*="spinner"], [class*="skeleton"]')).find((element) => isVisible(element) && /loading|spinner|生成中|思考中|回答中/i.test(String(element.className || '') + ' ' + (element.innerText || '')));
     const replyContainers = Array.from(document.querySelectorAll('[class*="flow-markdown-body"], [class*="container-P2rR72"], [class*="mdbox-theme"], [class*="message"], [class*="response"], [class*="answer"], .prose'))
       .filter((element) => {
         const cls = String(element.className || '');
@@ -426,11 +459,9 @@ async function typeIntoDoubaoInput(page, inputHandle, text) {
   await inputHandle.focus();
   await humanPause(160, 420);
   const value = String(text || '');
-  if (shouldUseDirectTextInsert(value)) {
-    await insertTextViaCdp(page, value);
-    return;
-  }
-  await page.keyboard.type(value, { delay: computeTypingDelay(value) });
+  // Doubao composer is React-controlled; use CDP insertText for both short and
+  // multiline prompts to avoid stale keyboard/IME edge cases.
+  await insertTextViaCdp(page, value);
 }
 
 async function clickDoubaoSend(page) {
@@ -500,7 +531,10 @@ async function clickNewChat(page, options = {}) {
 }
 
 async function switchToExpert(page, sessionId) {
-  const isExpertButton = async () => (await getDoubaoUiState(page)).hasExpertButton;
+  const isExpertButton = async () => {
+    const state = await getDoubaoUiState(page);
+    return state.hasExpertButton;
+  };
 
   if (await isExpertButton()) { log('INFO', '[doubao]', '专家按钮已存在，模式已就绪'); return true; }
 
@@ -539,6 +573,9 @@ async function switchToExpert(page, sessionId) {
     };
 
     for (const candidate of candidates) {
+      if (await clickHandleAtCenter(page, candidate.handle, candidate.label + '坐标')) {
+        if (await waitForExpertMenu().catch(() => false)) return true;
+      }
       const clicked = await clickHandleWithoutCoordinates(candidate.handle, candidate.label);
       if (clicked && await waitForExpertMenu().catch(() => false)) return true;
       if (await activate(candidate.handle, 'ArrowDown')) return true;
@@ -657,17 +694,14 @@ async function queryDoubao(queryText, opts = {}) {
         }
         assertFollowupAnchorActive(page, session);
       } else if (conversationPlan.effective === 'fresh') {
-        // Fresh Doubao queries use the official URL action entrypoint. It
-        // submits the prompt server-side and enables deep thinking without
-        // brittle menu switching. The legacy composer path remains for
-        // anchored followups and rare manual-submit fallback.
+        // Fresh queries: open a clean doubao home and confirm no prior content.
+        // We intentionally avoid the url-action entrypoint because doubao
+        // cannot reliably open/fetch external URLs through it — the page says
+        // "无法打开外链".  Instead we open the composer, switch to expert mode
+        // (in ensure_mode), then type the full prompt including any URLs, which
+        // lets doubao's built-in link preview handle them natively.
+        inputReadyAfterOpen = await ensureFreshDoubaoHome(page, session.sessionId);
         initialReplySnapshot = emptyDoubaoSnapshot();
-        const direct = await openDoubaoUrlAction(page, queryText, urlActionMode, session.sessionId);
-        usedUrlAction = true;
-        urlActionNeedsManualSubmit = direct.needsManualSubmit;
-        inputReadyAfterOpen = Boolean(direct.state?.hasInput || direct.accepted);
-        promptSubmitted = direct.accepted;
-        if (!inputReadyAfterOpen) throw new Error('Doubao url-action did not prepare a prompt or conversation');
       } else {
         inputReadyAfterOpen = await openDoubaoHome(page, session.sessionId);
       }
@@ -696,14 +730,22 @@ async function queryDoubao(queryText, opts = {}) {
 
     throwIfAborted(abortSignal, 'Doubao query');
     expertModeConfirmed = await runState(flow, 'ensure_mode', async () => {
-      if (usedUrlAction) return false;
+      // If url-action already auto-submitted the prompt server-side, mode
+      // switching is too late for the current query — skip gracefully.
+      if (usedUrlAction && promptSubmitted && !urlActionNeedsManualSubmit) return false;
+      // For all other cases (including url-action that still needs manual
+      // submission), verify and switch to expert mode via the 快速→专家 menu.
       const confirmed = await switchToExpert(page, session.sessionId);
       await assertNoChallenge(page, session.sessionId);
       if (confirmed) await waitForDoubaoModeSettled(page, { timeoutMs: 6000, sessionId: session.sessionId }).catch(() => false);
       return confirmed;
     }, { timeoutMs: STATE_TIMEOUTS.ensure_mode });
     log('INFO', '[doubao]', { component: 'doubao', plugin: 'cdper-doubao', session_id: session.sessionId, step: 'flow', action: 'stage_complete', stage: 'ensure_mode', status: 'ok', ...buildDoubaoModeState(usedUrlAction, urlActionMode, expertModeConfirmed), ...(usedUrlAction ? {} : { expertModeConfirmed }) });
-    if (!usedUrlAction && !expertModeConfirmed) {
+    // Block sending only when expert mode could not be confirmed and the
+    // prompt hasn't already been submitted.  Already-submitted url-action
+    // prompts cannot be un-sent, so we let them through.
+    const modeBlockingBypass = usedUrlAction && promptSubmitted && !urlActionNeedsManualSubmit;
+    if (!modeBlockingBypass && !expertModeConfirmed) {
       log('WARN', '[doubao]', { component: 'doubao', plugin: 'cdper-doubao', session_id: session.sessionId, step: 'mode', action: 'expert_confirm', status: 'blocked', message: '未确认进入专家模式，停止发送以避免错误模式输出' });
       throw new Error('Doubao 专家模式未确认，需要人工接管');
     }
@@ -744,8 +786,11 @@ async function queryDoubao(queryText, opts = {}) {
         captureUserTurn: async (page, baseline) => {
           const baselineCount = Number(baseline?.candidateCount || 0);
           return page.evaluate((bc) => {
-            // Doubao: check if number of visible reply containers increased
-            const replyContainers = document.querySelectorAll('[class*="flow-markdown-body"], [class*="container-P2rR72"]');
+            // Broad selector set matching smart_wait's selectBestDoubaoReplyText
+            const replyContainers = document.querySelectorAll(
+              '[class*="flow-markdown-body"], [class*="container-P2rR72"],' +
+              '[class*="mdbox-theme"], [class*="response"], [class*="answer"], .prose'
+            );
             return replyContainers.length > bc;
           }, baselineCount);
         },
@@ -754,15 +799,24 @@ async function queryDoubao(queryText, opts = {}) {
             const textarea = document.querySelector('textarea') || document.querySelector('[contenteditable="true"]');
             if (!textarea) return false;
             const val = textarea.value || textarea.innerText || '';
-            return val.trim().length <= 5; // allow very short residual whitespace
+            return val.trim().length <= 5;
           });
         },
         captureGenerating: async (page) => {
           return page.evaluate(() => {
-            const stopBtn = Array.from(document.querySelectorAll('button')).find(
-              (b) => b.innerText?.includes('停止') || b.getAttribute('aria-label')?.includes('stop')
-            );
-            const loading = document.querySelector('[class*="loading"]') || document.querySelector('[class*="skeleton"]');
+            const isVisible = (el) => {
+              if (!el) return false;
+              const rect = el.getBoundingClientRect();
+              const style = window.getComputedStyle(el);
+              return rect.width > 0 && rect.height > 0 && style.visibility !== 'hidden' && style.display !== 'none' && el.getAttribute('aria-hidden') !== 'true';
+            };
+            const stopBtn = Array.from(document.querySelectorAll('button,[role="button"]')).find((button) => {
+              if (!isVisible(button) || button.disabled || button.getAttribute('aria-disabled') === 'true') return false;
+              const text = button.innerText || '';
+              const aria = button.getAttribute('aria-label') || '';
+              return /停止生成|停止回答|Stop generating|stop-button/i.test(`${text} ${aria}`) || !!button.querySelector('svg[icon="stop"]');
+            });
+            const loading = Array.from(document.querySelectorAll('[class*="loading"], [class*="skeleton"], [class*="spinner"]')).find((element) => isVisible(element) && /loading|spinner|生成中|思考中|回答中/i.test(String(element.className || '') + ' ' + (element.innerText || '')));
             return !!stopBtn || !!loading;
           });
         },