ideaco 1.1.5

package/src/core/agent/web-agent/backends/chatgpt/dom-scripts.js

@@ -0,0 +1,303 @@
+/**
+ * ChatGPT DOM 脚本 — 在浏览器页面中执行的 DOM 操作脚本
+ *
+ * 这些函数生成纯 JS 字符串，通过 webContents.executeJavaScript() 注入到 ChatGPT 页面中执行。
+ * 所有脚本必须是自包含的（不能引用外部变量），因为它们在渲染进程的沙盒中运行。
+ *
+ * 从 electron/main.cjs 中抽出，便于维护和扩展。
+ */
+import { getSelectors } from './config.js';
+
+/**
+ * 构建"发送消息"的 DOM 脚本
+ * 在 ChatGPT 页面中找到输入框，输入消息，然后点击发送按钮。
+ *
+ * @param {string} message - 要发送的消息
+ * @returns {string} 可执行的 JS 脚本字符串，返回 { ok: true } 或 { error: string }
+ */
+export function buildSendMessageScript(message) {
+  const escaped = JSON.stringify(message);
+  const inputSels = JSON.stringify(getSelectors('input'));
+  const sendSels = JSON.stringify(getSelectors('send'));
+  return `
+(async () => {
+  try {
+    // --- Step 1: 查找输入框 ---
+    const inputSelectors = ${inputSels};
+    let inputEl = null;
+    for (const sel of inputSelectors) {
+      const el = document.querySelector(sel);
+      if (el && el.offsetParent !== null) { inputEl = el; break; }
+    }
+    if (!inputEl) {
+      return { error: "Cannot find input element" };
+    }
+
+    // --- Step 2: 清除已有内容并输入消息 ---
+    inputEl.focus();
+    const msg = ${escaped};
+
+    if (inputEl.tagName === "TEXTAREA" || inputEl.tagName === "INPUT") {
+      const nativeSetter = Object.getOwnPropertyDescriptor(
+        window.HTMLTextAreaElement.prototype, 'value'
+      )?.set || Object.getOwnPropertyDescriptor(
+        window.HTMLInputElement.prototype, 'value'
+      )?.set;
+      if (nativeSetter) {
+        nativeSetter.call(inputEl, msg);
+      } else {
+        inputEl.value = msg;
+      }
+      inputEl.dispatchEvent(new Event("input", { bubbles: true }));
+    } else {
+      inputEl.innerHTML = "<p>" + msg.replace(/\\n/g, "</p><p>") + "</p>";
+      inputEl.dispatchEvent(new Event("input", { bubbles: true }));
+    }
+
+    await new Promise(r => setTimeout(r, 300));
+
+    // --- Step 3: 查找并点击发送按钮 ---
+    const sendSelectors = ${sendSels};
+    let sendBtn = null;
+    for (const sel of sendSelectors) {
+      const btn = document.querySelector(sel);
+      if (btn && !btn.disabled && btn.offsetParent !== null) { sendBtn = btn; break; }
+    }
+    if (!sendBtn) {
+      return { error: "Cannot find send button" };
+    }
+
+    sendBtn.click();
+    return { ok: true };
+  } catch (err) {
+    return { error: err.message || String(err) };
+  }
+})()
+`;
+}
+
+/**
+ * 构建"读取回复"的 DOM 脚本
+ * 轮询 ChatGPT 页面，读取最新的 assistant 回复内容。
+ *
+ * @returns {string} 可执行的 JS 脚本字符串，返回 { text, isStreaming, matchedSelector, matchedStopSelector, elementCount }
+ */
+export function buildReadResponseScript() {
+  const responseSels = JSON.stringify(getSelectors('response'));
+  const stopSels = JSON.stringify(getSelectors('stop'));
+  return `
+(() => {
+  const clean = (t) => t.replace(/[\\u200B-\\u200D\\uFEFF]/g, "").trim();
+
+  const selectors = ${responseSels};
+  let allEls = [];
+  let matchedSelector = null;
+  for (const sel of selectors) {
+    const els = document.querySelectorAll(sel);
+    if (els.length > 0) { allEls = Array.from(els); matchedSelector = sel; break; }
+  }
+
+  // 总是使用最后一个匹配的元素（最新的回复）
+  const last = allEls.length > 0 ? allEls[allEls.length - 1] : null;
+  let text = "";
+  if (last) {
+    const assistantInner = last.querySelector('[data-message-author-role="assistant"]');
+    const target = assistantInner || last;
+    const mdContainer = target.querySelector('[class*="markdown"]') || target.querySelector('.prose') || target;
+    text = clean(mdContainer.innerText || mdContainer.textContent || "");
+  }
+
+  // 动态回退 1：查找任何 assistant-like 内容
+  if (!text) {
+    const assistantEls = document.querySelectorAll('[data-message-author-role="assistant"]');
+    if (assistantEls.length > 0) {
+      const lastAssistant = assistantEls[assistantEls.length - 1];
+      const md = lastAssistant.querySelector('[class*="markdown"]') || lastAssistant.querySelector('.prose') || lastAssistant;
+      text = clean(md.innerText || md.textContent || "");
+      matchedSelector = '[data-message-author-role="assistant"] (dynamic fallback)';
+      allEls = Array.from(assistantEls);
+    }
+  }
+
+  // 动态回退 2：查找 article 元素
+  if (!text) {
+    const articles = document.querySelectorAll('article');
+    if (articles.length > 0) {
+      const lastArticle = articles[articles.length - 1];
+      const md = lastArticle.querySelector('[class*="markdown"]') || lastArticle.querySelector('.prose') || lastArticle;
+      const candidate = clean(md.innerText || md.textContent || "");
+      if (candidate.length > 5) {
+        text = candidate;
+        matchedSelector = 'article (dynamic fallback)';
+        allEls = Array.from(articles);
+      }
+    }
+  }
+
+  // 动态回退 3：暴力扫描 main 区域
+  if (!text) {
+    const main = document.querySelector('main') || document.querySelector('[role="main"]');
+    if (main) {
+      const walker = document.createTreeWalker(main, NodeFilter.SHOW_ELEMENT, {
+        acceptNode: (node) => {
+          if (node.closest('nav, header, footer, textarea, [contenteditable], input, button')) return NodeFilter.FILTER_REJECT;
+          const t = clean(node.innerText || '');
+          if (t.length > 30) return NodeFilter.FILTER_ACCEPT;
+          return NodeFilter.FILTER_SKIP;
+        }
+      });
+      let lastNode = null;
+      while (walker.nextNode()) { lastNode = walker.currentNode; }
+      if (lastNode) {
+        const candidate = clean(lastNode.innerText || '');
+        if (candidate.length > 30) {
+          text = candidate;
+          matchedSelector = 'TreeWalker brute-force (main)';
+          allEls = [lastNode];
+        }
+      }
+    }
+  }
+
+  // 停止按钮检测（判断是否还在流式输出）
+  const stopSelectors = ${stopSels};
+  let isStreaming = false;
+  let matchedStopSelector = null;
+  for (const sel of stopSelectors) {
+    const stopEl = document.querySelector(sel);
+    if (stopEl && stopEl.offsetParent !== null) { isStreaming = true; matchedStopSelector = sel; break; }
+  }
+
+  return { text, isStreaming, matchedSelector, matchedStopSelector, elementCount: allEls.length };
+})()
+`;
+}
+
+/**
+ * 构建"新建对话"的 DOM 脚本
+ * 点击 ChatGPT 的"新建对话"按钮，或直接导航到首页。
+ *
+ * @returns {string} 可执行的 JS 脚本字符串，返回 { ok: true }
+ */
+export function buildNewChatScript() {
+  const newChatSels = JSON.stringify(getSelectors('newChat'));
+  return `
+(() => {
+  const selectors = ${newChatSels};
+  for (const sel of selectors) {
+    const el = document.querySelector(sel);
+    if (el) { el.click(); return { ok: true }; }
+  }
+  window.location.href = 'https://chatgpt.com/';
+  return { ok: true, navigated: true };
+})()
+`;
+}
+
+/**
+ * 构建 DOM 诊断脚本（调试用）
+ * 当选择器匹配失败时，收集页面 DOM 结构信息帮助排查。
+ *
+ * @returns {string} 可执行的 JS 脚本字符串
+ */
+export function buildDiagnosticScript() {
+  return `
+(() => {
+  const url = location.href;
+  const title = document.title;
+  const bodyLen = document.body ? document.body.innerHTML.length : 0;
+  const hasDialog = !!document.querySelector('dialog, [role="dialog"], [role="alertdialog"]');
+  const hasCaptcha = !!document.querySelector('[class*="captcha"], [id*="captcha"], iframe[src*="captcha"]');
+  const hasLogin = !!document.querySelector('[data-testid="login-button"], [class*="login"], a[href*="/auth/login"]');
+  const probe = [];
+  const candidates = ['article', '[data-message-author-role]', '[class*="markdown"]', '.prose',
+    '[data-testid]', 'main', '[role="main"]', '[class*="conversation"]', '[class*="message"]',
+    '[class*="response"]', '[class*="chat"]', '[class*="turn"]', '[class*="agent"]'];
+  for (const sel of candidates) {
+    const count = document.querySelectorAll(sel).length;
+    if (count > 0) probe.push(sel + ':' + count);
+  }
+  const mainEl = document.querySelector('main') || document.querySelector('[role="main"]');
+  let mainChildren = '';
+  if (mainEl) {
+    mainChildren = Array.from(mainEl.children).slice(0, 5).map(c =>
+      c.tagName.toLowerCase() + (c.className ? '.' + String(c.className).split(' ').slice(0,3).join('.') : '') + (c.id ? '#' + c.id : '')
+    ).join(', ');
+  }
+  return { url, title, bodyLen, hasDialog, hasCaptcha, hasLogin, probe, mainChildren };
+})()
+`;
+}
+
+/**
+ * 构建最后手段的文本提取脚本
+ * 当所有选择器都失败时，尝试从页面 main 区域提取最长的文本块。
+ *
+ * @returns {string} 可执行的 JS 脚本字符串
+ */
+export function buildLastResortExtractionScript() {
+  return `
+(() => {
+  const clean = (t) => t.replace(/[\\u200B-\\u200D\\uFEFF]/g, "").trim();
+  const main = document.querySelector('main') || document.querySelector('[role="main"]');
+  if (!main) return '';
+  const allDivs = main.querySelectorAll('div, p, span, section');
+  let bestText = '';
+  let bestLen = 0;
+  for (const el of allDivs) {
+    if (el.closest('nav, header, footer, textarea, [contenteditable]')) continue;
+    const t = clean(el.innerText || '');
+    if (t.length > bestLen && t.length > 20) {
+      bestText = t;
+      bestLen = t.length;
+    }
+  }
+  return bestText;
+})()
+`;
+}
+
+/**
+ * 构建响应选择器泛化脚本（校准时使用）
+ * 用户点击一个回复气泡后，尝试找到能匹配所有类似气泡的通用选择器。
+ *
+ * @param {string} specificSelector - 用户点击的元素对应的具体选择器
+ * @returns {string} 可执行的 JS 脚本字符串，返回泛化后的选择器
+ */
+export function buildResponseSelectorGeneralizationScript(specificSelector) {
+  return `
+(function() {
+  const clicked = document.querySelector(${JSON.stringify(specificSelector)});
+  if (!clicked) return ${JSON.stringify(specificSelector)};
+
+  const directMatches = document.querySelectorAll(${JSON.stringify(specificSelector)});
+  if (directMatches.length > 1) return ${JSON.stringify(specificSelector)};
+
+  let el = clicked;
+  const maxDepth = 6;
+  for (let depth = 0; depth < maxDepth && el && el !== document.body; depth++) {
+    const role = el.getAttribute('data-message-author-role');
+    if (role === 'assistant') {
+      const sel = el.tagName.toLowerCase() + '[data-message-author-role="assistant"]';
+      if (document.querySelectorAll(sel).length >= 1) return sel;
+    }
+    const testId = el.getAttribute('data-testid');
+    if (testId && testId.includes('conversation') || testId && testId.includes('message')) {
+      const sel = el.tagName.toLowerCase() + '[data-testid="' + testId + '"]';
+      if (document.querySelectorAll(sel).length > 1) return sel;
+    }
+    if (el.classList.length > 0) {
+      for (const cls of el.classList) {
+        if (cls.startsWith('__') || cls.length < 3) continue;
+        const sel = el.tagName.toLowerCase() + '.' + cls;
+        const matches = document.querySelectorAll(sel);
+        if (matches.length > 1 && matches.length < 50) return sel;
+      }
+    }
+    el = el.parentElement;
+  }
+
+  return ${JSON.stringify(specificSelector)};
+})()
+`;
+}

package/src/core/agent/web-agent/backends/index.js

@@ -0,0 +1,91 @@
+/**
+ * Web Backends 模块入口
+ *
+ * 管理所有 web 后端（ChatGPT、Claude、DeepSeek 等）。
+ * 每个后端有自己的文件夹，包含配置（config.js）、DOM 脚本（dom-scripts.js）和客户端（client.js）。
+ *
+ * 目录结构:
+ * backends/
+ * ├── index.js           - 本文件，模块入口
+ * ├── base-backend.js    - Web 后端基类
+ * └── chatgpt/
+ *     ├── config.js      - ChatGPT 站点配置和选择器
+ *     ├── dom-scripts.js - ChatGPT DOM 交互脚本
+ *     └── client.js      - ChatGPTBackend 类
+ *
+ * 扩展新后端（如 Claude、DeepSeek）:
+ * 1. 新建 backends/claude/ 文件夹
+ * 2. 创建 config.js（站点 URL、选择器）、dom-scripts.js（DOM 脚本）、client.js（后端类）
+ * 3. 在此文件中注册新后端
+ */
+
+import { BaseWebBackend } from './base-backend.js';
+import { chatgptBackend } from './chatgpt/client.js';
+
+/**
+ * Web 后端注册表
+ * 通过 backendId 查找对应的后端实例。
+ */
+class WebBackendRegistry {
+  constructor() {
+    /** @type {Map<string, BaseWebBackend>} */
+    this._backends = new Map();
+  }
+
+  /**
+   * 注册一个后端
+   * @param {BaseWebBackend} backend
+   */
+  register(backend) {
+    this._backends.set(backend.id, backend);
+  }
+
+  /**
+   * 根据 backendId 获取后端
+   * @param {string} backendId - 如 'chatgpt', 'claude', 'deepseek'
+   * @returns {BaseWebBackend|null}
+   */
+  get(backendId) {
+    return this._backends.get(backendId) || null;
+  }
+
+  /**
+   * 根据 providerId 获取后端（兼容旧的 provider 命名，如 'web-chatgpt-xxx'）
+   * @param {string} providerId
+   * @returns {BaseWebBackend|null}
+   */
+  getByProviderId(providerId) {
+    if (!providerId) return null;
+    // 尝试精确匹配
+    for (const [id, backend] of this._backends) {
+      if (providerId === id || providerId.includes(id)) {
+        return backend;
+      }
+    }
+    return null;
+  }
+
+  /**
+   * 获取所有已注册的后端
+   * @returns {BaseWebBackend[]}
+   */
+  getAll() {
+    return Array.from(this._backends.values());
+  }
+
+  /**
+   * 获取所有后端 ID
+   * @returns {string[]}
+   */
+  getAllIds() {
+    return Array.from(this._backends.keys());
+  }
+}
+
+// 创建全局注册表单例并注册内置后端
+export const webBackendRegistry = new WebBackendRegistry();
+webBackendRegistry.register(chatgptBackend);
+
+// 导出
+export { BaseWebBackend } from './base-backend.js';
+export { ChatGPTBackend, chatgptBackend } from './chatgpt/client.js';

package/src/core/agent/web-agent/index.js

@@ -0,0 +1,278 @@
+import { BaseAgent } from '../base-agent.js';
+import { webClientRegistry } from './web-client.js';
+
+/**
+ * WebAgent — Communication engine powered by browser DOM automation.
+ *
+ * Opens a hidden Electron BrowserWindow, lets the user log in once, then controls
+ * the web chat interface (e.g. ChatGPT web, Claude web) via DOM scripting.
+ * No API key required — uses the user's existing subscription.
+ *
+ * Limitations vs LLMAgent:
+ * - No native tool calling (web UIs don't expose function calling)
+ * - Simulates tool calls by embedding tool definitions in the system prompt
+ * - Session may expire and need re-login
+ */
+export class WebAgent extends BaseAgent {
+  /**
+   * @param {object} config
+   * @param {object} config.provider - Web provider config
+   */
+  constructor(config) {
+    super();
+    this.provider = config.provider;
+    // Employee ID for per-employee session isolation (set by Employee layer)
+    this._employeeId = null;
+  }
+
+  /**
+   * Set the employee ID for session isolation.
+   * Called by the Employee layer to bind this agent to a specific employee.
+   * @param {string} employeeId
+   */
+  setEmployeeId(employeeId) {
+    this._employeeId = employeeId;
+  }
+
+  get agentType() {
+    return 'web';
+  }
+
+  isAvailable() {
+    // DOM mode: Electron manages cookies via Chromium session, no explicit cookie field needed
+    return !!(this.provider && this.provider.enabled);
+  }
+
+  canChat() {
+    return this.isAvailable();
+  }
+
+  async chat(messages, options = {}) {
+    if (!this.isAvailable()) {
+      throw new Error(`WebAgent provider "${this.provider?.name}" is not available`);
+    }
+    return await webClientRegistry.chat(this.provider.id, messages, {
+      ...options,
+      model: this.provider.webModel || this.provider.model,
+      sessionId: this._employeeId || options.sessionId || null,
+    });
+  }
+
+  /**
+   * Reset the ChatGPT conversation for this employee's session.
+   * Next chat will start a new conversation.
+   */
+  resetConversation() {
+    if (this.provider?.id) {
+      webClientRegistry.resetConversation(this.provider.id, this._employeeId || undefined);
+    }
+  }
+
+  /**
+   * Check if this employee's session needs a new conversation (too many messages).
+   * @param {number} [maxMessages]
+   * @returns {boolean}
+   */
+  needsNewSession(maxMessages) {
+    if (!this.provider?.id || !this._employeeId) return false;
+    return webClientRegistry.needsNewSession(this.provider.id, this._employeeId, maxMessages);
+  }
+
+  /**
+   * WebAgent's chatWithTools implementation.
+   * Since web APIs don't support native function calling, we simulate it
+   * by embedding tool definitions in the prompt and parsing structured output.
+   */
+  async chatWithTools(messages, toolExecutor, options = {}) {
+    if (!this.isAvailable()) {
+      throw new Error(`WebAgent provider "${this.provider?.name}" is not available`);
+    }
+
+    const maxIterations = options.maxIterations || 5;
+    const onToolCall = options.onToolCall || null;
+    const conversationMessages = [...messages];
+    const toolResults = [];
+
+    // Inject tool definitions into system message
+    const toolDefs = toolExecutor.definitions || [];
+    if (toolDefs.length > 0) {
+      const toolsPrompt = this._buildToolsPrompt(toolDefs);
+      // Prepend to first system message or add new one
+      const sysIdx = conversationMessages.findIndex(m => m.role === 'system');
+      if (sysIdx >= 0) {
+        conversationMessages[sysIdx] = {
+          ...conversationMessages[sysIdx],
+          content: conversationMessages[sysIdx].content + '\n\n' + toolsPrompt,
+        };
+      } else {
+        conversationMessages.unshift({ role: 'system', content: toolsPrompt });
+      }
+    }
+
+    for (let i = 0; i < maxIterations; i++) {
+      // Only force new conversation on the first iteration; subsequent iterations reuse
+      const iterOptions = i === 0 ? options : { ...options, newConversation: false };
+      const response = await this.chat(conversationMessages, iterOptions);
+
+      // Try to parse tool calls from the response
+      const parsedCalls = this._parseToolCalls(response.content);
+
+      if (!parsedCalls || parsedCalls.length === 0) {
+        // No tool calls found — this is the final response
+        return {
+          content: response.content,
+          toolResults,
+          messages: conversationMessages,
+          usage: response.usage,
+        };
+      }
+
+      // Process extracted tool calls
+      conversationMessages.push({ role: 'assistant', content: response.content });
+
+      const callResultTexts = [];
+      for (const call of parsedCalls) {
+        if (onToolCall) {
+          try { onToolCall({ tool: call.name, args: call.args, status: 'start' }); } catch {}
+        }
+
+        let result;
+        try {
+          result = await toolExecutor.execute(call.name, call.args);
+          toolResults.push({ tool: call.name, args: call.args, result, success: true });
+          if (onToolCall) {
+            try { onToolCall({ tool: call.name, args: call.args, status: 'done', success: true }); } catch {}
+          }
+        } catch (error) {
+          result = `Tool execution error: ${error.message}`;
+          toolResults.push({ tool: call.name, args: call.args, error: error.message, success: false });
+          if (onToolCall) {
+            try { onToolCall({ tool: call.name, args: call.args, status: 'error', error: error.message }); } catch {}
+          }
+        }
+
+        callResultTexts.push(
+          `[Tool Result: ${call.name}]\n${typeof result === 'string' ? result : JSON.stringify(result, null, 2)}`
+        );
+      }
+
+      // Feed tool results back as user message
+      conversationMessages.push({
+        role: 'user',
+        content: callResultTexts.join('\n\n'),
+      });
+    }
+
+    // Exceeded max iterations — one final call without tool prompt (reuse conversation)
+    const finalResponse = await this.chat(conversationMessages, { ...options, newConversation: false });
+    return {
+      content: finalResponse.content,
+      toolResults,
+      messages: conversationMessages,
+      usage: finalResponse.usage,
+    };
+  }
+
+  /**
+   * Build a prompt section that describes available tools.
+   */
+  _buildToolsPrompt(toolDefs) {
+    const toolDescriptions = toolDefs.map(t => {
+      const params = t.function?.parameters?.properties || {};
+      const required = t.function?.parameters?.required || [];
+      const paramList = Object.entries(params).map(([name, schema]) => {
+        const req = required.includes(name) ? ' (required)' : ' (optional)';
+        return `  - ${name}${req}: ${schema.description || schema.type || ''}`;
+      }).join('\n');
+      return `### ${t.function.name}\n${t.function.description || ''}\nParameters:\n${paramList}`;
+    }).join('\n\n');
+
+    return `## Available Tools
+
+You have access to the following tools. To use a tool, include a tool call block in your response using this exact format:
+
+\`\`\`tool_call
+{"name": "tool_name", "args": {"param1": "value1"}}
+\`\`\`
+
+You can make multiple tool calls in a single response. After receiving tool results, continue your work.
+
+${toolDescriptions}`;
+  }
+
+  /**
+   * Parse tool calls from LLM text response.
+   * Looks for ```tool_call blocks.
+   */
+  _parseToolCalls(content) {
+    if (!content) return null;
+    const calls = [];
+    const regex = /```tool_call\s*\n([\s\S]*?)```/g;
+    let match;
+    while ((match = regex.exec(content)) !== null) {
+      try {
+        const parsed = JSON.parse(match[1].trim());
+        if (parsed.name) {
+          calls.push({ name: parsed.name, args: parsed.args || {} });
+        }
+      } catch {
+        // Skip malformed tool calls
+      }
+    }
+    return calls.length > 0 ? calls : null;
+  }
+
+  getDisplayInfo() {
+    return {
+      name: this.provider?.name || 'Unknown Web Provider',
+      provider: this.provider?.provider || 'Web',
+      model: this.provider?.webModel || this.provider?.model || 'unknown',
+      type: 'web',
+      category: this.provider?.category || 'general',
+    };
+  }
+
+  getProviderDisplayInfo() {
+    return {
+      id: this.provider?.id,
+      name: this.provider?.name,
+      provider: this.provider?.provider,
+    };
+  }
+
+  switchProvider(newProvider) {
+    this.provider = newProvider;
+  }
+
+  getCostPerToken() {
+    return 0; // Web-based, no direct token cost
+  }
+
+  serializeAgent() {
+    return {
+      agentType: 'web',
+      provider: this.provider ? {
+        id: this.provider.id,
+        name: this.provider.name,
+        provider: this.provider.provider,
+        model: this.provider.model,
+        webModel: this.provider.webModel,
+        category: this.provider.category,
+        costPerToken: 0,
+        enabled: this.provider.enabled,
+        isWeb: true,
+      } : null,
+    };
+  }
+
+  /**
+   * Restore WebAgent from serialized data.
+   */
+  static deserialize(data, providerRegistry) {
+    let provider = data.provider;
+    if (providerRegistry && data.provider?.id) {
+      provider = providerRegistry.getById(data.provider.id) || data.provider;
+    }
+    return new WebAgent({ provider });
+  }
+}