yiyan-browser-agent 1.11.0 → 1.11.2

package/README.md

@@ -82,55 +82,38 @@ npx playwright install-deps chromium  # 安装系统依赖
 
 ## 🚀 Quick Start
 
-### Windows
+### 第一步：登录文心一言（首次使用）
 
-**1. 首次运行 — 登录文心一言:**
+**Windows:**
 ```powershell
 yiyan-agent -i
 ```
-浏览器窗口打开后，登录你的百度账号，然后回到终端按 Enter。会话会保存 — 只需登录一次。
-
-**2. 给任务:**
-```powershell
-yiyan-agent "创建一个 Express REST API，带用户认证"
-```
-
-**3. 发送任务到已运行的服务器:**
-```powershell
-# 终端 1: 启动交互模式 (作为 HTTP 服务器)
+**Linux / macOS:**
+```bash
 yiyan-agent -i
-
-# 终端 2: 发送任务 (转发到服务器，不启动新浏览器)
-yiyan-agent "上海天气，20个字"
 ```
+浏览器窗口打开后，登录你的百度账号，然后回到终端按 Enter。会话会保存 — 只需登录一次。
 
-### Ubuntu / Linux
+### 第二步：通过 HTTP API 发送任务
 
-**1. 首次运行 — 登录文心一言:**
-```bash
-yiyan-agent -i
-```
+登录完成后，交互模式会自动启动 HTTP 服务（端口 9527），通过 API 发送任务：
 
-**2. 给任务:**
 ```bash
-yiyan-agent "build a REST API in Express with user authentication"
+curl -X POST http://localhost:9527/task \
+  -H "Content-Type: application/json" \
+  -d '{"task": "创建一个 Express REST API，带用户认证"}'
 ```
 
-**3. 从任意目录使用短别名 `ya`:**
-```bash
-cd ~/my-project
-ya "add input validation to all my API routes"
-```
+> 💡 更多用法见 [HTTP API](#-http-api) 章节。
 
 ---
 
 ## 💻 Usage
 
 ```
-yiyan-agent [OPTIONS] [TASK]
+yiyan-agent [OPTIONS]
 
-  -t, --task <task>    Task to run (or just type it as the last argument)
-  -i, --interactive    Keep browser open, run multiple tasks (starts HTTP server)
+  -i, --interactive    Start interactive mode with HTTP server on port 9527
   -d, --dir <path>     Set working directory (default: current directory)
   --debug              Print raw AI responses to the terminal
   --show-browser       Show browser window (non-interactive mode)
@@ -141,22 +124,22 @@ Aliases:
   ya                   Short form of yiyan-agent
 ```
 
-### Examples
+### 典型工作流
 
 ```bash
-# Single task — runs and exits
-yiyan-agent "create a Python script that scrapes Hacker News"
-
-# Interactive mode — keeps browser open, starts HTTP server on port 9527
+# 1. 启动交互模式（登录 + 启动 HTTP 服务）
 yiyan-agent -i
+# → Server listening on port 9527
 
-# Run on a specific project
-ya --dir ~/projects/my-app "refactor all callbacks to async/await"
+# 2. 通过 HTTP API 发送任务（另开终端）
+curl -X POST http://localhost:9527/task \
+  -H "Content-Type: application/json" \
+  -d '{"task": "创建一个 Python 爬虫"}'
 
-# Debug mode (shows what Yiyan is actually outputting)
-ya --debug "build a calculator"
+# 3. 查看任务状态
+curl http://localhost:9527/status
 
-# In interactive mode, type 'quit' or 'q' to exit:
+# 在交互模式中，输入 quit 或 q 退出：
 ❯ quit
 ```
 
@@ -193,9 +176,10 @@ When interactive mode (`-i`) is running, an HTTP server starts on port **9527**,
 yiyan-agent -i
 # → Server listening on port 9527
 
-# Terminal 2: Send task (forwarded to server, no new browser)
-yiyan-agent "北京天气，15个字"
-# → Found running server on port 9527, forwarding task...
+# Terminal 2: Send task via HTTP API
+curl -X POST http://localhost:9527/task \
+  -H "Content-Type: application/json" \
+  -d '{"task": "北京天气，15个字"}'
 ```
 
 ### HTTP POST API
@@ -220,8 +204,8 @@ yiyan-agent "北京天气，15个字"
 **Response:**
 ```json
 {
-  "question": "上海天气，20个字",
-  "answer": "上海今日晴，气温25°C...",
+  "question": "创建一个 Express REST API",
+  "answer": "好的，我来创建...",
   "duration": 5234,
   "status": "success"
 }
@@ -323,19 +307,21 @@ curl http://localhost:9527/task/abc123
 ### Windows CMD (curl)
 
 ```cmd
-curl -X POST http://localhost:9527/task -H "Content-Type: application/json" -d "{\"task\":\"上海天气，20个字\"}"
+curl -X POST http://localhost:9527/task -H "Content-Type: application/json" -d "{\"task\":\"创建一个 Express REST API\"}"
 ```
 
 ### PowerShell
 
 ```powershell
-Invoke-RestMethod -Uri "http://localhost:9527/task" -Method POST -Body '{"task":"上海天气"}' -ContentType "application/json"
+Invoke-RestMethod -Uri "http://localhost:9527/task" -Method POST -Body '{"task":"创建一个 Express REST API"}' -ContentType "application/json"
 ```
 
 ### Ubuntu / Linux (curl)
 
 ```bash
-curl -X POST http://localhost:9527/task -H "Content-Type: application/json" -d '{"task":"上海天气，20个字"}'
+curl -X POST http://localhost:9527/task \
+  -H "Content-Type: application/json" \
+  -d '{"task": "创建一个 Express REST API"}'
 ```
 
 ### From Other Programming Languages
@@ -343,11 +329,10 @@ curl -X POST http://localhost:9527/task -H "Content-Type: application/json" -d '
 **Python:**
 ```python
 import requests
-import json
 
 response = requests.post(
     'http://localhost:9527/task',
-    json={'task': '上海天气，20个字'}
+    json={'task': '创建一个 Express REST API'}
 )
 result = response.json()
 print(result)
@@ -357,7 +342,7 @@ print(result)
 ```javascript
 const http = require('http');
 
-const body = JSON.stringify({ task: '上海天气，20个字' });
+const body = JSON.stringify({ task: '创建一个 Express REST API' });
 
 const req = http.request({
   hostname: 'localhost',
@@ -495,13 +480,14 @@ Everything lives in `~/.yiyan-agent/` in your home directory:
 ### Agent responds but creates no files
 The browser DOM rendered the AI's response in a way the parser didn't catch. Run with `--debug` to see exactly what's being received:
 ```bash
-yiyan-agent --debug "build a calculator"
+yiyan-agent --debug -i
+# Then send a task via HTTP API and observe the raw output
 ```
 
 ### Agent stops responding / loops
 Yiyan's UI may have changed. Run the calibration tool — it inspects the live DOM and prints updated selectors:
 ```bash
-yiyan-agent --calibrate
+node src/calibrate.js
 ```
 
 ### Login session expired
@@ -605,7 +591,7 @@ Open an issue on GitHub with:
 - What you ran
 - What you expected
 - What actually happened
-- Output of `yiyan-agent --debug "your task"` if relevant
+- Debug output from `yiyan-agent --debug -i` if relevant
 
 ---
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "yiyan-browser-agent",
-  "version": "1.11.0",
+  "version": "1.11.2",
   "description": "AI coding agent powered by Yiyan (文心一言) via browser automation (chat.baidu.com) — no API key needed. Performance-optimized. Enhanced with comprehensive security.",
   "main": "src/index.js",
   "bin": {

package/src/browser.js

@@ -320,6 +320,7 @@ class YiyanBrowser {
     let lastThinkingText    = '';
     let stableCount         = 0;
     let lastStableTime      = Date.now();
+    let textChangedAt       = Date.now();  // 文本最后变化时间
     let dotCount            = 0;
     let hasCompletionMarker = false;
 
@@ -373,6 +374,7 @@ class YiyanBrowser {
         lastText = text;
         stableCount = 0;
         lastStableTime = Date.now();
+        textChangedAt = Date.now();
       }
 
       // ── 完成标记检测 (chat.baidu.com) ──
@@ -404,10 +406,12 @@ class YiyanBrowser {
       }
 
       // ── 完成判断 ──
-      // 条件1: 标记出现 + 两者都稳定 2 次
-      // 条件2(备用): 两者都稳定 3 次（即使没有完成标记）
+      // 条件1: 完成标记出现 + 稳定 2 次
+      // 条件2: 稳定 3 次（即使没有完成标记）
+      // 条件3(兜底): 文本长时间稳定不变（≥5秒），强制认为完成
       const condition1 = hasCompletionMarker && stableCount >= 2;
       const condition2 = stableCount >= 3;
+      const longStable = (Date.now() - textChangedAt) >= 5000 && text.length > 0;
 
       if (condition1 || condition2) {
         // 检查 _isGenerating 作为最终确认
@@ -420,6 +424,13 @@ class YiyanBrowser {
         // _isGenerating 说还在生成，重置计数继续等待
         stableCount = 0;
         lastStableTime = Date.now();
+      } else if (longStable) {
+        // 兜底：文本已经 5 秒没变化了，强制完成
+        logger.dim('Long stable fallback — text unchanged for 5s, forcing completion');
+        await this.page.waitForTimeout(300);
+        logger.clearLine();
+        logger.success('Response complete (stable timeout)');
+        break;
       }
 
       // Progress indicator
@@ -560,33 +571,13 @@ class YiyanBrowser {
 
   async _isGenerating() {
     return await this.page.evaluate(() => {
-      // ── 1. 检测 typing/generating 指示器 (chat.baidu.com) ──
-      const typingSelectors = [
-        '.cosd-markdown-content-typingall',
-        '.markdown-typing-all',
-        '[class*="typing"]',
-        '[class*="generating"]',
-        '[class*="loading-indicator"]',
-        'svg[class*="loading"]',
-        'svg[class*="spinner"]',
-        '[class*="blink"]',
-        '[class*="cursor-blink"]',
-        '[class*="pulsing"]',
-      ];
-      for (const sel of typingSelectors) {
-        const el = document.querySelector(sel);
-        if (el) {
-          const s = window.getComputedStyle(el);
-          if (s.display !== 'none' && s.visibility !== 'hidden') return true;
-        }
-      }
-
-      // ── 2. 检测停止按钮 ──
+      // ── 1. 检测停止按钮（最可靠的生成中信号）──
       const stopSelectors = [
         'button[aria-label*="Stop" i]',
         'button[aria-label*="停止"]',
         '[class*="stop-gen"]',
         '[class*="stopGen"]',
+        '[class*="stop-btn"]',
       ];
       for (const sel of stopSelectors) {
         const el = document.querySelector(sel);
@@ -596,35 +587,56 @@ class YiyanBrowser {
         }
       }
 
-      // ── 3. 检测完成标记 — 如果有回答但没有完成标记 → 还在生成 ──
+      // ── 2. 检测活跃的 CSS 动画（真正的 loading spinner，不是永久类名）──
+      // 只检测 svg 动画元素和明确的 loading 指示器
+      const activeAnimSelectors = [
+        'svg[class*="loading"]',
+        'svg[class*="spinner"]',
+        '[class*="loading-indicator"]',
+        '[class*="generating-indicator"]',
+      ];
+      for (const sel of activeAnimSelectors) {
+        const el = document.querySelector(sel);
+        if (el) {
+          const s = window.getComputedStyle(el);
+          if (s.display !== 'none' && s.visibility !== 'hidden') return true;
+        }
+      }
+
+      // ── 3. 如果有完成标记 → 肯定已完成 ──
       const completionMarkers = [
         '.cos-icon.cos-icon-copy',
         '.cos-icon-copy',
+        '.feedback-hover-show',
         '.cos-icon.cos-icon-share1',
-        '.cos-icon-feedback',
+        '.cos-icon.cos-icon-feedback',
         '[class*="copy-btn"]',
-        '[class*="copyBtn"]',
         '[aria-label*="Copy" i]',
         '[aria-label*="复制"]',
-        '[class*="regenerate"]',
-        '[class*="retry"]',
-        '[class*="action-btn"]',
-        '.feedback-hover-show',
       ];
-      let hasCompletionMarker = false;
       for (const sel of completionMarkers) {
-        if (document.querySelector(sel)) {
-          hasCompletionMarker = true;
-          break;
-        }
+        if (document.querySelector(sel)) return false;
       }
 
-      // 如果有响应内容但没有完成标记 → 还在生成
+      // ── 4. 检查是否有光标闪烁（生成中的光标）──
+      // 用 getComputedStyle 检测 animation，而不是类名
       const answerArea = document.querySelector(
-        '.ai-entry-block.ai-markdown, .answer-container, .cs-answer-container, .answer-box'
+        '.ai-entry-block.ai-markdown, .answer-container, .cs-answer-container'
       );
-      if (answerArea && answerArea.innerText && answerArea.innerText.length > 5) {
-        if (!hasCompletionMarker) {
+      if (answerArea) {
+        // 查找最后子元素是否有动画光标
+        const lastChild = answerArea.lastElementChild;
+        if (lastChild) {
+          const s = window.getComputedStyle(lastChild);
+          // 如果最后一个元素有活跃的动画 → 还在生成
+          if (s.animationName && s.animationName !== 'none' &&
+              s.display !== 'none' && s.visibility !== 'hidden') {
+            return true;
+          }
+        }
+
+        // 有回答内容但没有完成标记 → 可能还在生成
+        if (answerArea.innerText && answerArea.innerText.length > 5) {
           return true;
         }
       }