yiyan-browser-agent 1.0.4 → 1.0.6

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "yiyan-browser-agent",
-  "version": "1.0.4",
+  "version": "1.0.6",
   "description": "AI coding agent powered by Yiyan (文心一言) via browser automation — no API key needed",
   "main": "src/index.js",
   "bin": {

package/src/browser.js

@@ -1,5 +1,5 @@
 // src/browser.js — Playwright controller for Yiyan (yiyan.baidu.com)
-// Originally designed for DeepSeek, adapted for Yiyan
+// Performance optimized: smart waits, paste input, MutationObserver
 'use strict';
 
 const { chromium } = require('playwright');
@@ -9,12 +9,9 @@ const logger       = require('./logger');
 
 // ─────────────────────────────────────────────────────────────────────────────
 //  Selector banks — ordered by likelihood, with fallbacks
-//  Supports both Yiyan (文心一言) and generic chat UI patterns
 // ─────────────────────────────────────────────────────────────────────────────
 
 const SEL = {
-  // Text input where the user types
-  // Yiyan uses contenteditable div with specific class names
   chatInput: [
     '.editable__T7WAW4uW',
     '[role="textbox"]',
@@ -30,8 +27,6 @@ const SEL = {
     '.input-area textarea',
   ],
 
-  // Button that submits the message
-  // Yiyan: 发送 button, often with icon
   sendButton: [
     'button[aria-label*="Send" i]',
     'button[aria-label*="发送"]',
@@ -45,8 +40,6 @@ const SEL = {
     '.send-btn',
   ],
 
-  // "Stop generating" button — visible while streaming
-  // Yiyan: 停止生成
   stopButton: [
     'button[aria-label*="Stop" i]',
     'button[aria-label*="停止"]',
@@ -57,8 +50,6 @@ const SEL = {
     '[class*="abort"]',
   ],
 
-  // "New chat" / "New conversation" button in sidebar
-  // Yiyan: 新对话
   newChat: [
     'button[aria-label*="New chat" i]',
     'button[aria-label*="新对话"]',
@@ -69,7 +60,14 @@ const SEL = {
     '[class*="newChat"]',
   ],
 
-  // The main chat messages container
+  // Response ready indicators
+  responseReady: [
+    '[class*="answer"]',
+    '[class*="response"]',
+    '[class*="markdown"]',
+    '.ds-markdown',
+  ],
+
   messageContainer: [
     '[class*="chat-content"]',
     '[class*="message-list"]',
@@ -79,7 +77,7 @@ const SEL = {
 };
 
 // ─────────────────────────────────────────────────────────────────────────────
-//  YiyanBrowser class
+//  YiyanBrowser class (optimized)
 // ─────────────────────────────────────────────────────────────────────────────
 
 class YiyanBrowser {
@@ -114,11 +112,9 @@ class YiyanBrowser {
       ignoreDefaultArgs: ['--enable-automation'],
     });
 
-    // Grab existing page or open a new one
     const pages   = this.context.pages();
     this.page     = pages.length > 0 ? pages[0] : await this.context.newPage();
 
-    // Mask automation signals
     await this.page.addInitScript(() => {
       Object.defineProperty(navigator, 'webdriver', { get: () => false });
     });
@@ -135,26 +131,29 @@ class YiyanBrowser {
     try { await this.context?.close(); } catch {}
   }
 
-  // ── Navigation ─────────────────────────────────────────────────────────────
+  // ── Navigation (optimized: smart wait) ─────────────────────────────────────
 
   async _navigate(url) {
     try {
-      await this.page.goto(url, { waitUntil: 'domcontentloaded', timeout: 30_000 });
-      await this.page.waitForTimeout(1_500);
+      // Wait for network idle, then extra wait for dynamic content
+      await this.page.goto(url, { waitUntil: 'networkidle', timeout: 30_000 });
+      // Give React/Vue SPA time to render
+      await this.page.waitForTimeout(800);
     } catch (err) {
       logger.warn(`Navigation warning: ${err.message}`);
+      // Fallback: longer wait if networkidle fails
+      await this.page.waitForTimeout(2000);
     }
   }
 
   async newChat() {
     try {
-      // Try clicking the "New Chat" button in the sidebar
       for (const sel of SEL.newChat) {
         try {
-          const el = await this.page.$(sel);
-          if (el && await el.isVisible()) {
+          const el = await this.page.waitForSelector(sel, { timeout: 2_000, state: 'visible' });
+          if (el) {
             await el.click();
-            await this.page.waitForTimeout(1_000);
+            await this.page.waitForTimeout(500);
             logger.dim('Started new chat session');
             return;
           }
@@ -162,15 +161,16 @@ class YiyanBrowser {
       }
     } catch {}
 
-    // Fallback: navigate to home which usually opens a fresh chat
     await this._navigate(config.YIYAN_URL);
     logger.dim('Navigated to Yiyan home (new chat)');
   }
 
-  // ── Login handling ─────────────────────────────────────────────────────────
+  // ── Login handling (optimized) ─────────────────────────────────────────────
 
   async _ensureLoggedIn() {
-    await this.page.waitForTimeout(2_000);
+    // Wait for page to be fully interactive
+    await this.page.waitForLoadState('domcontentloaded', { timeout: 5_000 }).catch(() => {});
+    await this.page.waitForTimeout(500);
 
     const needsLogin = await this.page.evaluate(() => {
       const url = window.location.href;
@@ -190,7 +190,7 @@ class YiyanBrowser {
     if (needsLogin) {
       this._printLoginBanner();
       await this._waitForEnter();
-      await this.page.waitForTimeout(2_000);
+      await this.page.waitForNavigation({ waitUntil: 'networkidle', timeout: 30_000 }).catch(() => {});
     }
   }
 
@@ -228,22 +228,51 @@ class YiyanBrowser {
     });
   }
 
-  // ── Sending Messages ───────────────────────────────────────────────────────
+  // ── Sending Messages (optimized: paste instead of type) ────────────────────
 
   async sendMessage(text) {
-    // Find input element
     const { el } = await this._findInput();
 
-    // Triple click to focus and select all
-    await el.click({ clickCount: 3, force: true });
-    await this.page.waitForTimeout(200);
+    // Focus the element
+    await el.focus();
+
+    // Clear existing content
+    await el.evaluate(e => {
+      e.textContent = '';
+      if (e.innerText) e.innerText = '';
+    });
+
+    // Use clipboard paste for instant input (much faster than typing)
+    await this.page.evaluate(async (text) => {
+      // Try modern clipboard API first
+      try {
+        await navigator.clipboard.writeText(text);
+        document.execCommand('paste');
+        return;
+      } catch {}
+
+      // Fallback: direct DOM manipulation with proper event dispatch
+      const activeEl = document.activeElement;
+      if (activeEl && activeEl.isContentEditable) {
+        activeEl.textContent = text;
+
+        // Dispatch input event so Yiyan recognizes the content
+        const inputEvent = new InputEvent('input', {
+          bubbles: true,
+          cancelable: true,
+          inputType: 'insertText',
+          data: text,
+        });
+        activeEl.dispatchEvent(inputEvent);
 
-    // Clear by pressing Delete
-    await this.page.keyboard.press('Delete');
-    await this.page.waitForTimeout(100);
+        // Also dispatch change event for good measure
+        const changeEvent = new Event('change', { bubbles: true });
+        activeEl.dispatchEvent(changeEvent);
+      }
+    }, text);
 
-    // Type text character by character (simulates real user input)
-    await this.page.keyboard.type(text, { delay: 50 });
+    // Small delay to let UI process the input
+    await this.page.waitForTimeout(50);
 
     // Press Enter to send
     await this.page.keyboard.press('Enter');
@@ -252,7 +281,8 @@ class YiyanBrowser {
   async _findInput() {
     for (const sel of SEL.chatInput) {
       try {
-        const el = await this.page.waitForSelector(sel, { timeout: 4_000, state: 'visible' });
+        // Increased timeout for initial page load
+        const el = await this.page.waitForSelector(sel, { timeout: 8_000, state: 'visible' });
         if (!el) continue;
         const tagName          = await el.evaluate(e => e.tagName.toLowerCase());
         const isContentEditable = await el.evaluate(e => e.isContentEditable);
@@ -267,74 +297,73 @@ class YiyanBrowser {
     );
   }
 
-  async _clickSendButton() {
-    for (const sel of SEL.sendButton) {
-      try {
-        const el = await this.page.$(sel);
-        if (el && await el.isVisible() && await el.isEnabled()) {
-          await el.click();
-          return true;
-        }
-      } catch {}
-    }
-    return false;
-  }
+  // ── Waiting for Response (optimized: MutationObserver) ────────────────────
 
-  // ── Waiting for Response ───────────────────────────────────────────────────
-
-  /**
-   * Wait until Yiyan finishes generating and return the response text.
-   *
-   * Algorithm:
-   *  1. Record how many assistant messages are on the page right now.
-   *  2. Wait until a new message appears (count goes up).
-   *  3. Poll the last message text every 500 ms.
-   *  4. When the text has not changed for STABLE_DELAY ms AND
-   *     no stop/loading indicator is visible → done.
-   */
   async waitForResponse() {
     const timeout     = config.RESPONSE_TIMEOUT;
     const stableDelay = config.STABLE_DELAY;
     const start       = Date.now();
 
-    // ── Phase 1: wait for a new message to appear ──────────────────────────
-    const initialCount = await this._getMessageCount();
-    let   appeared     = false;
+    // Use MutationObserver for efficient change detection
+    const responseId = await this.page.evaluate(() => {
+      let lastText    = '';
+      let stableStart = null;
+      let done        = false;
+
+      // Find response container
+      const container = document.querySelector('[class*="answer"], [class*="response"], [class*="markdown"], .ds-markdown, [class*="content"]');
+      if (!container) return null;
+
+      const observer = new MutationObserver(() => {
+        const currentText = container.textContent || '';
+
+        if (currentText !== lastText) {
+          lastText    = currentText;
+          stableStart = Date.now();
+        } else if (stableStart && Date.now() - stableStart >= 2500) {
+          // Text stable for 2.5s = done
+          done = true;
+        }
+      });
 
-    while (Date.now() - start < 12_000) {
-      const count = await this._getMessageCount();
-      if (count > initialCount) { appeared = true; break; }
-      await this.page.waitForTimeout(400);
-    }
+      observer.observe(container, { childList: true, subtree: true, characterData: true });
 
-    if (!appeared) logger.warn('Response may have been delayed — continuing to wait...');
+      // Store for cleanup
+      window.__yiyanObserver = observer;
+      window.__yiyanDone = () => done;
+      window.__yiyanText = () => lastText;
 
-    // ── Phase 2: wait for text to stabilise ───────────────────────────────
-    let lastText    = '';
-    let stableStart = null;
-    let dotCount    = 0;
+      return container.className || 'response-container';
+    });
 
+    // Wait for completion signal
+    let dotCount = 0;
     while (Date.now() - start < timeout) {
-      const text = await this._extractLastMessage();
-
-      if (text !== lastText) {
-        lastText    = text;
-        stableStart = null;
-      } else if (text.length > 0) {
-        if (!stableStart) stableStart = Date.now();
-        else if (Date.now() - stableStart >= stableDelay) {
-          if (!await this._isGenerating()) break;  // confirmed done
-          stableStart = null;                       // still generating, reset
-        }
+      const state = await this.page.evaluate(() => ({
+        done  : window.__yiyanDone ? window.__yiyanDone() : false,
+        text  : window.__yiyanText ? window.__yiyanText() : '',
+        length: (window.__yiyanText ? window.__yiyanText() : '').length,
+      }));
+
+      if (state.done && state.length > 10 && !await this._isGenerating()) {
+        break;
       }
 
       // Progress indicator
       dotCount = (dotCount + 1) % 4;
-      logger.thinking(`Receiving response${'.'.repeat(dotCount)}  (${text.length} chars)`);
-
-      await this.page.waitForTimeout(500);
+      logger.thinking(`Receiving response${'.'.repeat(dotCount)}  (${state.length} chars)`);
     }
 
+    // Cleanup observer
+    await this.page.evaluate(() => {
+      if (window.__yiyanObserver) {
+        window.__yiyanObserver.disconnect();
+        delete window.__yiyanObserver;
+        delete window.__yiyanDone;
+        delete window.__yiyanText;
+      }
+    });
+
     logger.clearLine();
 
     const final = await this._extractLastMessage();
@@ -343,16 +372,13 @@ class YiyanBrowser {
 
   // ── DOM Extraction ─────────────────────────────────────────────────────────
 
-  /** Count how many "response" blocks are visible */
   async _getMessageCount() {
     return await this.page.evaluate(() => {
       const candidates = [
-        // Yiyan specific selectors
         '[class*="answer"]',
         '[class*="response"]',
         '[class*="content"]',
         '[class*="markdown"]',
-        // Generic selectors
         '[class*="assistant"][class*="message"]',
         '[data-role="assistant"]',
         '[class*="markdown-content"]',
@@ -364,21 +390,14 @@ class YiyanBrowser {
         const els = document.querySelectorAll(sel);
         if (els.length > 0) return els.length;
       }
-      // Broad fallback - look for any text block that might contain response
       const textBlocks = Array.from(document.querySelectorAll('div, section, article'))
         .filter(el => el.innerText && el.innerText.length > 50);
       return textBlocks.length;
     });
   }
 
-  /** Extract the text of the last assistant message — including code blocks */
   async _extractLastMessage() {
     return await this.page.evaluate(() => {
-
-      // ── Helper: get all text including code blocks ────────────────────────
-      // Walks the DOM and reconstructs text, re-adding fence markers for code
-      // blocks so the parser can recognise tool_call fences even after the
-      // browser markdown renderer has converted them to <pre><code> elements.
       function getFullText(el) {
         if (!el) return '';
         let result = '';
@@ -391,8 +410,6 @@ class YiyanBrowser {
           if (node.nodeType !== Node.ELEMENT_NODE) return;
           const tag = node.tagName.toLowerCase();
 
-          // <pre> wraps a fenced code block — reconstruct the backtick fence
-          // so the parser can match the ```tool_call regex.
           if (tag === 'pre') {
             const codeEl = node.querySelector('code');
             if (codeEl) {
@@ -406,7 +423,6 @@ class YiyanBrowser {
             return;
           }
 
-          // Inline <code> — skip if inside a <pre> (already handled)
           if (tag === 'code') {
             const parentTag = node.parentElement && node.parentElement.tagName
               ? node.parentElement.tagName.toLowerCase() : '';
@@ -427,13 +443,10 @@ class YiyanBrowser {
         return result.trim();
       }
 
-      // ── Attempt 1: Specific assistant-message selectors ──────────────────
       const directSelectors = [
-        // Yiyan specific selectors (百度文心一言)
         '[class*="answer"]',
         '[class*="response"]',
         '[class*="message"][class*="content"]',
-        // Generic selectors
         '.ds-markdown',
         '[class*="assistant"] [class*="markdown"]',
         '[class*="assistant"] [class*="content"]',
@@ -454,7 +467,6 @@ class YiyanBrowser {
         } catch {}
       }
 
-      // ── Attempt 2: Any markdown/prose container ───────────────────────────
       try {
         const markdownEls = document.querySelectorAll(
           '[class*="markdown"], [class*="prose"], [class*="rendered"], [class*="content"]'
@@ -465,7 +477,6 @@ class YiyanBrowser {
         }
       } catch {}
 
-      // ── Attempt 3: Heuristic — large non-user text blocks ────────────────
       try {
         const allBlocks = Array.from(
           document.querySelectorAll('[class*="message"], [class*="chat-item"], [class*="turn"], [class*="answer"], [class*="content"]')
@@ -488,19 +499,15 @@ class YiyanBrowser {
         }
       } catch {}
 
-      // ── Attempt 4: Any large text block on page (last resort) ─────────────
       try {
         const allDivs = Array.from(document.querySelectorAll('div, section'));
         const textBlocks = allDivs.filter(el => {
           const text = el.innerText || '';
-          // Exclude input areas and user messages
           const cls = el.className || '';
           if (cls.includes('input') || cls.includes('editable') || cls.includes('user')) return false;
-          // Look for blocks with substantial text that might be AI response
           return text.length > 50 && !el.querySelector('textarea, [contenteditable]');
         });
         if (textBlocks.length > 0) {
-          // Sort by text length, prefer longer blocks (likely complete responses)
           textBlocks.sort((a, b) => (b.innerText || '').length - (a.innerText || '').length);
           return getFullText(textBlocks[0]);
         }
@@ -510,10 +517,8 @@ class YiyanBrowser {
     });
   }
 
-  /** True if Yiyan is still streaming / generating */
   async _isGenerating() {
     return await this.page.evaluate(() => {
-      // Check for stop button
       const stopSelectors = [
         'button[aria-label*="Stop" i]',
         '[class*="stop-gen"]',
@@ -528,7 +533,6 @@ class YiyanBrowser {
         }
       }
 
-      // Check for animated loading/typing indicators
       const loaderSelectors = [
         '[class*="typing"]',
         '[class*="loading"]',
@@ -557,23 +561,15 @@ class YiyanBrowser {
     if (!text) return '';
 
     return text
-      // Strip AI thinking blocks (DeepSeek R1 / Yiyan patterns)
       .replace(/<think>[\s\S]*?<\/think>\n?/gi, '')
-      // Strip "Thinking..." headers that sometimes prefix responses
       .replace(/^Thinking\.{0,3}\n[\s\S]*?\n\n/m, '')
-      // Strip copy-code button artifacts like "1CopyRunInsert"
       .replace(/^\d+(?:Copy|Run|Insert|Edit)\b.*$/gm, '')
-      // Collapse 3+ blank lines into 2
       .replace(/\n{3,}/g, '\n\n')
       .trim();
   }
 
   // ── Debug / Calibration Utilities ─────────────────────────────────────────
 
-  /**
-   * Dump useful DOM information to stdout.
-   * Called by `node src/calibrate.js` or `--debug` flag.
-   */
   async dumpDebugInfo() {
     const info = await this.page.evaluate(() => {
       const classFreq = {};
@@ -614,11 +610,10 @@ class YiyanBrowser {
     console.log('═'.repeat(60) + '\n');
   }
 
-  /** Take a screenshot (for debugging) */
   async screenshot(filePath = '/tmp/yiyan-agent-debug.png') {
     await this.page.screenshot({ path: filePath, fullPage: false });
     logger.info(`Screenshot saved: ${filePath}`);
   }
 }
 
-module.exports = YiyanBrowser;
+module.exports = YiyanBrowser;

package/src/config.js

@@ -12,10 +12,10 @@ const defaults = {
   SESSION_DIR    : path.join(os.homedir(), '.yiyan-agent', 'session'),
   HEADLESS       : false,
 
-  // Timing
+  // Timing (optimized for faster response detection)
   RESPONSE_TIMEOUT : 180_000,
-  STABLE_DELAY     : 2_500,
-  SEND_DELAY       : 400,
+  STABLE_DELAY     : 1_500,  // Reduced from 2500ms (MutationObserver is faster)
+  SEND_DELAY       : 100,    // Reduced from 400ms (paste is instant)
 
   // Agent
   MAX_ITERATIONS   : 40,