Create index.js

index.js

@@ -0,0 +1,844 @@
+const http = require('http');
+const crypto = require('crypto');
+const fs = require('fs');
+const path = require('path');
+
+const ZAI_BASE = 'https://chat.z.ai';
+const PORT = process.env.PORT || 7860;
+
+// ================= 日志开关配置 =================
+const IS_DEBUG = process.env.DEBUG !== 'false' && process.env.LOG_LEVEL !== 'error';
+
+// 首字等待超时：10分钟；中途无流数据超时：10分钟
+const FIRST_TOKEN_TIMEOUT_MS = 600000; 
+const STREAM_IDLE_TIMEOUT_MS = 600000; 
+
+// 持久化路径配置
+const DATA_DIR = fs.existsSync('/data') ? '/data' : path.join(__dirname, 'data');
+const TOKENS_FILE = path.join(DATA_DIR, 'tokens.json');
+
+if (!fs.existsSync(DATA_DIR)) {
+  try { fs.mkdirSync(DATA_DIR, { recursive: true }); } catch (e) {}
+}
+
+function getTS() {
+  const d = new Date();
+  return d.toTimeString().split(' ')[0] + '.' + String(d.getMilliseconds()).padStart(3, '0');
+}
+
+function logInfo(...args) {
+  if (IS_DEBUG) console.log(`[${getTS()}]`, ...args);
+}
+function logWarn(...args) {
+  console.warn(`[${getTS()}]`, ...args);
+}
+function logError(...args) {
+  console.error(`[${getTS()}]`, ...args);
+}
+
+const COMMON_HEADERS = {
+  'accept': 'application/json, text/event-stream',
+  'accept-language': 'zh-CN,zh;q=0.9,en-US;q=0.8,en;q=0.7',
+  'content-type': 'application/json',
+  'origin': ZAI_BASE,
+  'referer': `${ZAI_BASE}/`,
+  'sec-ch-ua': '"Chromium";v="132", "Not A(Brand";v="24", "Google Chrome";v="132"',
+  'sec-ch-ua-mobile': '?0',
+  'sec-ch-ua-platform': '"Windows"',
+  'sec-fetch-dest': 'empty',
+  'sec-fetch-mode': 'cors',
+  'sec-fetch-site': 'same-origin',
+  'user-agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/132.0.0.0 Safari/537.36',
+  'x-region': 'overseas'
+};
+
+/* =========================================================
+   1. JWT Payload 解析与 Token 状态管理
+   ========================================================= */
+function parseTokenEmail(tokenStr) {
+  if (!tokenStr || typeof tokenStr !== 'string') return '';
+  try {
+    const parts = tokenStr.split('.');
+    if (parts.length < 2) return '';
+    let base64 = parts[1].replace(/-/g, '+').replace(/_/g, '/');
+    while (base64.length % 4 !== 0) base64 += '=';
+    const payloadJson = Buffer.from(base64, 'base64').toString('utf8');
+    const payload = JSON.parse(payloadJson);
+    return payload.email || payload.id || '';
+  } catch (e) {
+    return '';
+  }
+}
+
+let tokensPool = [];
+
+function formatToken(tokenStr) {
+  if (!tokenStr) return 'Guest/None';
+  const email = parseTokenEmail(tokenStr);
+  const emailTag = email ? ` [${email}]` : '';
+  if (tokenStr.length <= 16) return tokenStr + emailTag;
+  return `${tokenStr.slice(0, 8)}...${tokenStr.slice(-6)}${emailTag}`;
+}
+
+function loadTokens() {
+  if (fs.existsSync(TOKENS_FILE)) {
+    try {
+      const content = fs.readFileSync(TOKENS_FILE, 'utf8');
+      tokensPool = JSON.parse(content);
+      tokensPool.forEach(item => { item.email = parseTokenEmail(item.token); });
+      logInfo(`[+] Loaded ${tokensPool.length} token(s) from persistent storage (${TOKENS_FILE}).`);
+      return;
+    } catch (e) {
+      logError(`[-] Failed to parse ${TOKENS_FILE}, fallback to ENV:`, e.message);
+    }
+  }
+
+  const envVal = process.env.TOKENS || '';
+  let rawList = [];
+  try {
+    const parsed = JSON.parse(envVal);
+    if (Array.isArray(parsed)) rawList = parsed;
+  } catch (e) {
+    rawList = envVal.split(',').map(t => t.trim()).filter(Boolean);
+  }
+
+  tokensPool = rawList.map((t, idx) => {
+    const str = typeof t === 'string' ? t.trim() : '';
+    return {
+      id: `tok_${Date.now()}_${idx}_${Math.random().toString(36).slice(2, 6)}`,
+      token: str,
+      email: parseTokenEmail(str),
+      enabled: true,
+      errorCount: 0,
+      lastError: '',
+      inputTokens: 0,
+      outputTokens: 0,
+      totalRequests: 0,
+      lastUsed: null
+    };
+  }).filter(item => item.token);
+
+  saveTokens();
+}
+
+function saveTokens() {
+  try {
+    fs.writeFileSync(TOKENS_FILE, JSON.stringify(tokensPool, null, 2), 'utf8');
+  } catch (e) {
+    logError(`[-] Failed to save tokens to ${TOKENS_FILE}:`, e.message);
+  }
+}
+
+function getActiveToken() {
+  const enabledTokens = tokensPool.filter(t => t.enabled);
+  return enabledTokens.length > 0 ? enabledTokens[0].token : null;
+}
+
+function demoteToken(badToken, reason = 'Rate-limited/Exhausted') {
+  if (!badToken) return;
+  const idx = tokensPool.findIndex(t => t.token === badToken);
+  if (idx !== -1) {
+    const item = tokensPool[idx];
+    item.errorCount = (item.errorCount || 0) + 1;
+    item.lastError = `[${getTS()}] ${reason}`;
+    tokensPool.splice(idx, 1);
+    tokensPool.push(item);
+    saveTokens();
+    logWarn(`[!] Token 轮换 -> ${formatToken(badToken)} 已移至队尾。原因: ${reason}`);
+  }
+}
+
+function recordTokenUsage(tokenStr, inputChars = 0, outputChars = 0) {
+  if (!tokenStr) return;
+  const item = tokensPool.find(t => t.token === tokenStr);
+  if (item) {
+    item.totalRequests = (item.totalRequests || 0) + 1;
+    item.inputTokens = (item.inputTokens || 0) + Math.ceil(inputChars / 4);
+    item.outputTokens = (item.outputTokens || 0) + Math.ceil(outputChars / 4);
+    item.lastUsed = new Date().toISOString();
+    saveTokens();
+  }
+}
+
+loadTokens();
+
+/* =========================================================
+   2. 模型映射与配置
+   ========================================================= */
+const modelAliasMap = new Map([
+  ['glm-5.3-flash', 'x-preview-l'],
+  ['glm-5.3-flash-preview', 'x-preview-l'],
+  ['glm-5.3', 'glm-5.3'],
+  ['glm-5.2', 'glm-5.2'],
+  ['glm-5-turbo', 'GLM-5-Turbo'],
+  ['glm-5v-turbo', 'GLM-5v-Turbo'],
+  ['glm-4.7', 'glm-4.7'],
+  ['glm-4.5-air', '0727-106B-API'],
+  ['glm-4.5', '0727-360B-API']
+]);
+
+function parseModelConfig(rawModel) {
+  const wantSearch = rawModel.includes('-search');
+  let cleanModel = rawModel.replace(/-search/g, '');
+
+  let enableThinking = true;
+  let reasoningEffort = 'max';
+
+  if (cleanModel.includes('-thinking-off') || cleanModel.includes('-no-thinking')) {
+    enableThinking = false;
+    reasoningEffort = null;
+    cleanModel = cleanModel.replace('-thinking-off', '').replace('-no-thinking', '');
+  } else if (cleanModel.includes('-thinking-low')) {
+    enableThinking = true;
+    reasoningEffort = 'low';
+    cleanModel = cleanModel.replace('-thinking-low', '');
+  } else if (cleanModel.includes('-thinking-high')) {
+    enableThinking = true;
+    reasoningEffort = 'high';
+    cleanModel = cleanModel.replace('-thinking-high', '');
+  } else if (cleanModel.includes('-thinking-max')) {
+    enableThinking = true;
+    reasoningEffort = 'max';
+    cleanModel = cleanModel.replace('-thinking-max', '');
+  }
+
+  const key = cleanModel.toLowerCase();
+  const realBaseModel = modelAliasMap.get(key) || cleanModel;
+
+  return {
+    requestedModel: rawModel,
+    baseModel: realBaseModel,
+    webSearch: wantSearch,
+    enableThinking,
+    reasoningEffort
+  };
+}
+
+/* =========================================================
+   3. 原生 HTTP 接口实现 (彻底替代 Puppeteer)
+   ========================================================= */
+async function deleteChatSession(chatId, token) {
+  if (!chatId || !token) return;
+  try {
+    await fetch(`${ZAI_BASE}/api/v1/chats/${chatId}`, {
+      method: 'DELETE',
+      headers: { ...COMMON_HEADERS, 'Cookie': `token=${token}`, 'Authorization': `Bearer ${token}` }
+    });
+    logInfo(`[+] Cleaned Session ID: ${chatId}`);
+  } catch (e) {}
+}
+
+async function stopTask(messageId, token) {
+  if (!messageId || !token) return;
+  try {
+    await fetch(`${ZAI_BASE}/api/tasks/stop/${messageId}`, {
+      method: 'POST',
+      headers: { ...COMMON_HEADERS, 'Cookie': `token=${token}`, 'Authorization': `Bearer ${token}` },
+      body: JSON.stringify({})
+    });
+    logInfo(`[+] Stopped Task ID: ${messageId}`);
+  } catch (e) {}
+}
+
+// 创建新会话拿到 chatId 和初始 messageId
+async function createNewChatSession(prompt, token, modelOptions) {
+  const msgId = crypto.randomUUID();
+  const payload = {
+    chat: {
+      auto_web_search: modelOptions.webSearch,
+      enable_thinking: modelOptions.enableThinking,
+      models: [modelOptions.baseModel],
+      history: {
+        currentId: msgId,
+        messages: {
+          [msgId]: {
+            id: msgId,
+            role: 'user',
+            content: prompt
+          }
+        }
+      }
+    }
+  };
+  if (modelOptions.enableThinking && modelOptions.reasoningEffort) {
+    payload.chat.reasoning_effort = modelOptions.reasoningEffort;
+  }
+
+  const res = await fetch(`${ZAI_BASE}/api/v1/chats/new`, {
+    method: 'POST',
+    headers: { ...COMMON_HEADERS, 'Cookie': `token=${token}`, 'Authorization': `Bearer ${token}` },
+    body: JSON.stringify(payload)
+  });
+
+  if (!res.ok) {
+    const errText = await res.text();
+    throw new Error(`UPSTREAM_HTTP_ERROR_${res.status}: ${errText.slice(0, 300)}`);
+  }
+
+  const data = await res.json();
+  const chatId = data?.id || data?.chat_id || data?.chat?.id;
+  return { chatId, msgId };
+}
+
+// 直连 /api/v2/chat/completions 获取流式 SSE
+async function sendCompletionsHttp(prompt, token, modelOptions, chatId, msgId, abortSignal) {
+  const payload = {
+    models: [modelOptions.baseModel],
+    model: modelOptions.baseModel,
+    auto_web_search: modelOptions.webSearch,
+    enable_thinking: modelOptions.enableThinking,
+    chat_id: chatId,
+    features: {
+      auto_web_search: modelOptions.webSearch,
+      web_search: modelOptions.webSearch,
+      enable_thinking: modelOptions.enableThinking
+    }
+  };
+  if (modelOptions.enableThinking && modelOptions.reasoningEffort) {
+    payload.reasoning_effort = modelOptions.reasoningEffort;
+    payload.features.reasoning_effort = modelOptions.reasoningEffort;
+  }
+
+  const res = await fetch(`${ZAI_BASE}/api/v2/chat/completions`, {
+    method: 'POST',
+    headers: { ...COMMON_HEADERS, 'Cookie': `token=${token}`, 'Authorization': `Bearer ${token}` },
+    body: JSON.stringify(payload),
+    signal: abortSignal
+  });
+
+  if (!res.ok) {
+    const errText = await res.text();
+    throw new Error(`UPSTREAM_HTTP_ERROR_${res.status}: ${errText.slice(0, 300)}`);
+  }
+
+  return res.body; // 原生 ReadableStream
+}
+
+/* =========================================================
+   4. 队列调度管理
+   ========================================================= */
+let isProcessing = false;
+let lastTaskTime = Date.now();
+let currentProcessingContext = null;
+const requestQueue = [];
+
+function queueChatRequest(prompt, targetModel, modelOptions, streamCallback, taskContext) {
+  return new Promise((resolve, reject) => {
+    requestQueue.push({ prompt, targetModel, modelOptions, streamCallback, taskContext, resolve, reject });
+    processNextInQueue();
+  });
+}
+
+async function processNextInQueue() {
+  if (isProcessing || requestQueue.length === 0) return;
+  isProcessing = true;
+  lastTaskTime = Date.now();
+
+  const task = requestQueue.shift();
+  const { prompt, targetModel, modelOptions, streamCallback, taskContext, resolve, reject } = task;
+
+  currentProcessingContext = taskContext;
+
+  if (taskContext.isAborted) {
+    isProcessing = false;
+    currentProcessingContext = null;
+    processNextInQueue();
+    return;
+  }
+
+  // 基础防风控间隔
+  const randomDelay = 1500 + Math.floor(Math.random() * 2000);
+  await new Promise(r => setTimeout(r, randomDelay));
+
+  const MAX_ATTEMPTS = 3;
+  let attempt = 0;
+  let executionSuccess = false;
+  let lastError = null;
+
+  while (attempt < MAX_ATTEMPTS && !executionSuccess && !taskContext.isAborted) {
+    attempt++;
+    const activeToken = getActiveToken();
+    taskContext.activeToken = activeToken;
+
+    if (!activeToken) {
+      lastError = new Error('No enabled token available in pool.');
+      break;
+    }
+
+    const abortController = new AbortController();
+    taskContext.abortController = abortController;
+
+    try {
+      logInfo(`[>] HTTP Task (Attempt ${attempt}/${MAX_ATTEMPTS}) Model="${targetModel}" Token="${formatToken(activeToken)}"`);
+
+      // 1. 创建 Chat 会话
+      const { chatId, msgId } = await createNewChatSession(prompt, activeToken, modelOptions);
+      taskContext.createdChatId = chatId;
+      taskContext.capturedMessageId = msgId;
+
+      if (taskContext.isAborted) break;
+
+      // 2. 直连 HTTP SSE 流
+      const stream = await sendCompletionsHttp(prompt, activeToken, modelOptions, chatId, msgId, abortController.signal);
+      
+      // 3. 读取 Node.js ReadableStream / WebStream
+      const reader = stream.getReader();
+      const decoder = new TextDecoder('utf-8');
+
+      while (true) {
+        const { done, value } = await reader.read();
+        if (done) {
+          if (streamCallback) streamCallback('[DONE]');
+          break;
+        }
+        const chunk = decoder.decode(value, { stream: true });
+        if (streamCallback) streamCallback(chunk);
+      }
+
+      executionSuccess = true;
+      resolve();
+    } catch (err) {
+      lastError = err;
+      logWarn(`[-] Attempt ${attempt} failed: ${err.message}`);
+
+      const errStr = String(err.message);
+      const is405 = errStr.includes('405');
+      const isWaf = is405 || errStr.includes('403') || errStr.includes('401');
+
+      if (isWaf) {
+        demoteToken(activeToken, `WAF_HTTP_STATUS_${errStr.slice(0, 30)}`);
+      }
+
+      if (attempt < MAX_ATTEMPTS && !taskContext.isAborted) {
+        const wait = is405 ? 15 : 2;
+        await new Promise(r => setTimeout(r, wait * 1000));
+      }
+    } finally {
+      if (taskContext.createdChatId && activeToken) {
+        deleteChatSession(taskContext.createdChatId, activeToken).catch(() => {});
+      }
+    }
+  }
+
+  if (!executionSuccess && !taskContext.isAborted) {
+    if (lastError && String(lastError.message).includes('405')) {
+      logWarn(`[!] 3 次尝试均遭遇 405，触发兜底输出 "123"`);
+      taskContext.fallbackTo123 = true;
+      resolve();
+    } else {
+      reject(lastError);
+    }
+  }
+
+  currentProcessingContext = null;
+  isProcessing = false;
+  processNextInQueue();
+}
+
+/* =========================================================
+   5. OpenAI 格式兼容处理与路由 (handleChat)
+   ========================================================= */
+function makeChunk(id, model, content, reasoningContent, finishReason) {
+  const chunk = {
+    id, object: 'chat.completion.chunk', created: Math.floor(Date.now() / 1000), model,
+    choices: [{ index: 0, delta: {}, finish_reason: finishReason || null }],
+  };
+  if (content !== undefined) chunk.choices[0].delta.content = content;
+  if (reasoningContent !== undefined) chunk.choices[0].delta.reasoning_content = reasoningContent;
+  return chunk;
+}
+
+function makeCompletion(id, model, content, reasoningContent) {
+  const msg = { role: 'assistant', content };
+  if (reasoningContent) msg.reasoning_content = reasoningContent;
+  return {
+    id, object: 'chat.completion', created: Math.floor(Date.now() / 1000), model,
+    choices: [{ index: 0, message: msg, finish_reason: 'stop' }],
+    usage: { prompt_tokens: 0, completion_tokens: 0, total_tokens: 0 },
+  };
+}
+
+function parseBody(req) {
+  return new Promise((resolve) => {
+    let body = '';
+    req.on('data', c => body += c);
+    req.on('end', () => { try { resolve(JSON.parse(body || '{}')); } catch (e) { resolve({}); } });
+  });
+}
+
+function buildPromptFromMessages(messages) {
+  if (!Array.isArray(messages) || messages.length === 0) return '';
+  let systemText = '';
+  const conversationLines = [];
+
+  for (const m of messages) {
+    let contentStr = typeof m.content === 'string' ? m.content : JSON.stringify(m.content || '');
+    if (m.role === 'system') systemText += (systemText ? '\n' : '') + contentStr;
+    else conversationLines.push(`${m.role === 'user' ? 'User' : 'Assistant'}: ${contentStr}`);
+  }
+
+  if (!systemText && conversationLines.length === 1 && messages[0].role === 'user') return messages[0].content;
+  return (systemText ? `[System Instruction]\n${systemText}\n\n` : '') + conversationLines.join('\n\n');
+}
+
+async function handleChat(req, res) {
+  let isFinished = false;
+  const taskContext = { 
+    activeToken: null, 
+    capturedMessageId: null, 
+    createdChatId: null, 
+    abortController: null,
+    isAborted: false, 
+    fallbackTo123: false 
+  };
+
+  try {
+    const body = await parseBody(req);
+    const model = body.model || 'glm-5.3';
+    const modelOptions = parseModelConfig(model);
+    const messages = body.messages || [];
+    const wantStream = body.stream ?? false;
+    const chatId = 'chatcmpl-' + crypto.randomUUID().replace(/-/g, '').slice(0, 24);
+
+    const prompt = buildPromptFromMessages(messages);
+
+    if (!getActiveToken()) {
+      res.writeHead(400, { 'Content-Type': 'application/json', 'Access-Control-Allow-Origin': '*' });
+      return res.end(JSON.stringify({ error: { message: "No enabled token available in pool." } }));
+    }
+
+    let fullContent = '';
+    let fullReasoning = '';
+    let sseBuffer = '';
+    let contentStreamBuffer = '';
+    const STOP_WORD_REGEX = /(\n?\s*(?:User|用户|Human|HumanMessage)\s*[:：])/i;
+
+    if (wantStream) {
+      res.writeHead(200, {
+        'Content-Type': 'text/event-stream',
+        'Cache-Control': 'no-cache',
+        'Connection': 'keep-alive',
+        'X-Accel-Buffering': 'no',
+        'Access-Control-Allow-Origin': '*',
+      });
+      res.write(`data: ${JSON.stringify(makeChunk(chatId, model, '', undefined, null))}\n\n`);
+    }
+
+    const markFinished = (reason = 'normal') => {
+      if (!isFinished) {
+        isFinished = true;
+        if (wantStream && !res.writableEnded && contentStreamBuffer) {
+          res.write(`data: ${JSON.stringify(makeChunk(chatId, model, contentStreamBuffer, undefined, null))}\n\n`);
+          contentStreamBuffer = '';
+        }
+        if (wantStream && !res.writableEnded) {
+          res.write(`data: ${JSON.stringify(makeChunk(chatId, model, undefined, undefined, 'stop'))}\n\n`);
+          res.write('data: [DONE]\n\n');
+          res.end();
+        }
+        if (taskContext.capturedMessageId && taskContext.activeToken) {
+          stopTask(taskContext.capturedMessageId, taskContext.activeToken).catch(() => {});
+        }
+      }
+    };
+
+    req.on('close', () => {
+      if (!isFinished) {
+        taskContext.isAborted = true;
+        if (taskContext.abortController) taskContext.abortController.abort();
+        markFinished('client_disconnected');
+      }
+    });
+
+    const processLine = (line) => {
+      if (!line.startsWith('data: ')) return;
+      const raw = line.slice(6).trim();
+      if (!raw || raw === '[DONE]') return;
+
+      let parsed;
+      try { parsed = JSON.parse(raw); } catch { return; }
+
+      const d = (parsed.type === 'chat:completion' && parsed.data) ? parsed.data : parsed;
+      if (!d || typeof d !== 'object') return;
+
+      if (d.phase === 'done' || d.done === true) {
+        markFinished('phase_done');
+        return;
+      }
+
+      const delta = d.delta_content || d.content || d.reasoning_content || d.thinking_content || d.edit_content || d.text || '';
+      const phase = (d.phase || '').toLowerCase();
+      const isThinking = ['thinking', 'thought', 'think', 'reasoning'].includes(phase) || !!d.reasoning_content || !!d.thinking_content;
+
+      if (delta) {
+        if (isThinking) {
+          if (modelOptions.enableThinking !== false) {
+            fullReasoning += delta;
+            if (wantStream && !res.writableEnded) {
+              res.write(`data: ${JSON.stringify(makeChunk(chatId, model, undefined, delta, null))}\n\n`);
+            }
+          }
+        } else {
+          contentStreamBuffer += delta;
+          const match = contentStreamBuffer.match(STOP_WORD_REGEX);
+          if (match) {
+            const validPart = contentStreamBuffer.slice(0, match.index);
+            fullContent += validPart;
+            if (wantStream && !res.writableEnded && validPart) {
+              res.write(`data: ${JSON.stringify(makeChunk(chatId, model, validPart, undefined, null))}\n\n`);
+            }
+            contentStreamBuffer = '';
+            markFinished('hallucination_user_detected');
+            return;
+          }
+
+          const SAFE_WINDOW = 16;
+          if (contentStreamBuffer.length > SAFE_WINDOW) {
+            const emitLen = contentStreamBuffer.length - SAFE_WINDOW;
+            const toEmit = contentStreamBuffer.slice(0, emitLen);
+            contentStreamBuffer = contentStreamBuffer.slice(emitLen);
+            fullContent += toEmit;
+            if (wantStream && !res.writableEnded) {
+              res.write(`data: ${JSON.stringify(makeChunk(chatId, model, toEmit, undefined, null))}\n\n`);
+            }
+          }
+        }
+      }
+    };
+
+    const handleSseChunk = (rawChunk) => {
+      if (isFinished || taskContext.isAborted) return;
+      if (rawChunk === '[DONE]') {
+        if (sseBuffer.trim()) { processLine(sseBuffer.trim()); sseBuffer = ''; }
+        markFinished('sse_done_signal');
+        return;
+      }
+      sseBuffer += rawChunk;
+      const lines = sseBuffer.split(/\r?\n/);
+      sseBuffer = lines.pop() || '';
+      for (const line of lines) {
+        if (isFinished) break;
+        processLine(line);
+      }
+    };
+
+    await queueChatRequest(prompt, model, modelOptions, handleSseChunk, taskContext);
+
+    if (taskContext.fallbackTo123) {
+      if (wantStream && !res.writableEnded) {
+        res.write(`data: ${JSON.stringify(makeChunk(chatId, model, "123", undefined, null))}\n\n`);
+        res.write(`data: ${JSON.stringify(makeChunk(chatId, model, undefined, undefined, 'stop'))}\n\n`);
+        res.write('data: [DONE]\n\n');
+        return res.end();
+      } else if (!wantStream && !res.writableEnded) {
+        res.writeHead(200, { 'Content-Type': 'application/json', 'Access-Control-Allow-Origin': '*' });
+        return res.end(JSON.stringify(makeCompletion(chatId, model, "123", undefined)));
+      }
+    }
+
+    if (!wantStream && !res.writableEnded && !taskContext.isAborted) {
+      if (contentStreamBuffer) fullContent += contentStreamBuffer;
+      const match = fullContent.match(STOP_WORD_REGEX);
+      if (match) fullContent = fullContent.slice(0, match.index).trimEnd();
+
+      res.writeHead(200, { 'Content-Type': 'application/json', 'Access-Control-Allow-Origin': '*' });
+      res.end(JSON.stringify(makeCompletion(chatId, model, fullContent, modelOptions.enableThinking !== false ? (fullReasoning || undefined) : undefined)));
+    }
+
+    if (taskContext.activeToken) {
+      recordTokenUsage(taskContext.activeToken, prompt.length, fullContent.length + fullReasoning.length);
+    }
+
+  } catch (e) {
+    if (!res.headersSent) {
+      res.writeHead(500, { 'Content-Type': 'application/json', 'Access-Control-Allow-Origin': '*' });
+      res.end(JSON.stringify({ error: { message: e.message } }));
+    } else if (!res.writableEnded) {
+      res.write(`data: ${JSON.stringify({ error: { message: e.message } })}\n\n`);
+      res.end();
+    }
+  }
+}
+
+/* =========================================================
+   6. Dashboard 页面与轻量 Server 启动
+   ========================================================= */
+const ADMIN_HTML = `
+<!DOCTYPE html>
+<html lang="zh-CN">
+<head>
+  <meta charset="UTF-8">
+  <title>Token Dashboard (Pure HTTP Engine)</title>
+  <style>
+    body { font-family: -apple-system, sans-serif; background: #f0f2f5; margin: 0; padding: 20px; color: #1f2937; }
+    .container { max-width: 1200px; margin: 0 auto; }
+    .card { background: #fff; border-radius: 8px; padding: 20px; box-shadow: 0 1px 3px rgba(0,0,0,0.1); margin-bottom: 20px; }
+    .stats-grid { display: grid; grid-template-columns: repeat(auto-fit, minmax(200px, 1fr)); gap: 15px; margin-bottom: 20px; }
+    .stat-card { background: #fff; padding: 15px; border-radius: 8px; box-shadow: 0 1px 3px rgba(0,0,0,0.1); }
+    .stat-card .num { font-size: 22px; font-weight: bold; color: #2563eb; margin-top: 5px; }
+    textarea { width: 100%; height: 90px; border: 1px solid #d1d5db; border-radius: 6px; padding: 10px; font-family: monospace; box-sizing: border-box; }
+    button { background: #2563eb; color: white; border: none; padding: 8px 14px; border-radius: 6px; cursor: pointer; margin-right: 5px; }
+    button.danger { background: #ef4444; }
+    button.secondary { background: #6b7280; }
+    table { width: 100%; border-collapse: collapse; margin-top: 10px; }
+    th, td { text-align: left; padding: 12px; border-bottom: 1px solid #e5e7eb; font-size: 14px; }
+    .badge { padding: 4px 8px; border-radius: 12px; font-size: 12px; font-weight: 600; }
+    .bg-green { background: #d1fae5; color: #065f46; }
+    .bg-red { background: #fee2e2; color: #991b1b; }
+    .tok-text { font-family: monospace; font-size: 13px; }
+  </style>
+</head>
+<body>
+  <div class="container">
+    <h2>Token 状态仪表盘 (轻量无头 HTTP 引擎)</h2>
+    <div class="stats-grid">
+      <div class="stat-card"><div>总 Token 数</div><div class="num" id="statTotal">0</div></div>
+      <div class="stat-card"><div>已启用 Token</div><div class="num" id="statEnabled" style="color:#10b981">0</div></div>
+      <div class="stat-card"><div>累计总调用次数</div><div class="num" id="statRequests" style="color:#8b5cf6">0</div></div>
+    </div>
+    <div class="card">
+      <h3>批量添加 Token</h3>
+      <textarea id="newTokenInput" placeholder="请输入 Token，每行一个..."></textarea>
+      <div style="margin-top:10px"><button onclick="addTokens()">批量添加</button></div>
+    </div>
+    <div class="card">
+      <h3>Token 列表</h3>
+      <table>
+        <thead>
+          <tr><th>账号/邮箱</th><th>Token</th><th>状态</th><th>调用次数</th><th>错误数</th><th>最后错误</th><th>操作</th></tr>
+        </thead>
+        <tbody id="tokenTable"></tbody>
+      </table>
+    </div>
+  </div>
+  <script>
+    function formatTok(s) { return !s || s.length <= 16 ? s : s.slice(0, 8) + '...' + s.slice(-6); }
+    async function loadData() {
+      const res = await fetch('/admin/api/tokens');
+      const tokens = await res.json();
+      document.getElementById('statTotal').innerText = tokens.length;
+      document.getElementById('statEnabled').innerText = tokens.filter(t => t.enabled).length;
+      document.getElementById('statRequests').innerText = tokens.reduce((a, b) => a + (b.totalRequests || 0), 0);
+      const tbody = document.getElementById('tokenTable');
+      tbody.innerHTML = '';
+      tokens.forEach(t => {
+        const tr = document.createElement('tr');
+        tr.innerHTML = \`
+          <td>\${t.email || '-'}</td>
+          <td class="tok-text">\${formatTok(t.token)}</td>
+          <td><span class="badge \${t.enabled ? 'bg-green' : 'bg-red'}">\${t.enabled ? '已启用' : '已禁用'}</span></td>
+          <td>\${t.totalRequests || 0}</td>
+          <td>\${t.errorCount || 0}</td>
+          <td><small style="color:#ef4444">\${t.lastError || '-'}</small></td>
+          <td>
+            <button class="\${t.enabled ? 'secondary' : ''}" onclick="toggleToken('\${t.id}', \${!t.enabled})">\${t.enabled ? '禁用' : '启用'}</button>
+            <button class="danger" onclick="deleteToken('\${t.id}')">删除</button>
+          </td>
+        \`;
+        tbody.appendChild(tr);
+      });
+    }
+    async function addTokens() {
+      const text = document.getElementById('newTokenInput').value;
+      if (!text.trim()) return;
+      await fetch('/admin/api/tokens/add', { method: 'POST', headers: {'Content-Type': 'application/json'}, body: JSON.stringify({ rawText: text }) });
+      document.getElementById('newTokenInput').value = '';
+      loadData();
+    }
+    async function toggleToken(id, enabled) {
+      await fetch('/admin/api/tokens/toggle', { method: 'POST', headers: {'Content-Type': 'application/json'}, body: JSON.stringify({ id, enabled }) });
+      loadData();
+    }
+    async function deleteToken(id) {
+      if (!confirm('确定删除？')) return;
+      await fetch('/admin/api/tokens/delete', { method: 'POST', headers: {'Content-Type': 'application/json'}, body: JSON.stringify({ id }) });
+      loadData();
+    }
+    loadData();
+  </script>
+</body>
+</html>
+`;
+
+const server = http.createServer(async (req, res) => {
+  const urlObj = new URL(req.url, `http://${req.headers.host}`);
+  let reqPath = urlObj.pathname.replace(/\/$/, '') || '/';
+
+  if (req.method === 'OPTIONS') {
+    res.writeHead(200, {
+      'Access-Control-Allow-Origin': '*',
+      'Access-Control-Allow-Methods': 'POST, GET, OPTIONS',
+      'Access-Control-Allow-Headers': 'Content-Type, Authorization',
+    });
+    return res.end();
+  }
+
+  if (reqPath === '/' || reqPath === '/admin') {
+    res.writeHead(200, { 'Content-Type': 'text/html; charset=utf-8' });
+    return res.end(ADMIN_HTML);
+  }
+
+  if (reqPath === '/admin/api/tokens' && req.method === 'GET') {
+    res.writeHead(200, { 'Content-Type': 'application/json' });
+    return res.end(JSON.stringify(tokensPool));
+  }
+
+  if (reqPath === '/admin/api/tokens/add' && req.method === 'POST') {
+    const body = await parseBody(req);
+    const raw = body.rawText || '';
+    const list = raw.split(/[\n,]/).map(s => s.trim()).filter(Boolean);
+    list.forEach((t, i) => {
+      if (!tokensPool.some(x => x.token === t)) {
+        tokensPool.push({
+          id: `tok_${Date.now()}_${i}_${Math.random().toString(36).slice(2, 6)}`,
+          token: t,
+          email: parseTokenEmail(t),
+          enabled: true,
+          errorCount: 0,
+          lastError: '',
+          inputTokens: 0,
+          outputTokens: 0,
+          totalRequests: 0,
+          lastUsed: null
+        });
+      }
+    });
+    saveTokens();
+    res.writeHead(200, { 'Content-Type': 'application/json' });
+    return res.end(JSON.stringify({ success: true }));
+  }
+
+  if (reqPath === '/admin/api/tokens/toggle' && req.method === 'POST') {
+    const body = await parseBody(req);
+    const item = tokensPool.find(t => t.id === body.id);
+    if (item) { item.enabled = !!body.enabled; saveTokens(); }
+    res.writeHead(200, { 'Content-Type': 'application/json' });
+    return res.end(JSON.stringify({ success: true }));
+  }
+
+  if (reqPath === '/admin/api/tokens/delete' && req.method === 'POST') {
+    const body = await parseBody(req);
+    tokensPool = tokensPool.filter(t => t.id !== body.id);
+    saveTokens();
+    res.writeHead(200, { 'Content-Type': 'application/json' });
+    return res.end(JSON.stringify({ success: true }));
+  }
+
+  if (reqPath === '/v1/models' || reqPath === '/models') {
+    const models = Array.from(modelAliasMap.keys()).map(m => ({ id: m, object: 'model', created: 1677610602, owned_by: 'zhipu' }));
+    res.writeHead(200, { 'Content-Type': 'application/json', 'Access-Control-Allow-Origin': '*' });
+    return res.end(JSON.stringify({ object: 'list', data: models }));
+  }
+
+  if ((reqPath === '/v1/chat/completions' || reqPath === '/chat/completions') && req.method === 'POST') {
+    return handleChat(req, res);
+  }
+
+  res.writeHead(404, { 'Content-Type': 'application/json' });
+  res.end(JSON.stringify({ error: 'Not found' }));
+});
+
+server.listen(PORT, () => {
+  console.log(`\n> Pure HTTP Engine listening on port ${PORT}`);
+  console.log(`> UI Dashboard: http://localhost:${PORT}/\n`);
+});