opc-agent 4.1.0 → 4.1.2

package/src/cli/setup.ts

@@ -1,314 +1,314 @@
-import * as readline from 'readline';
-import * as fs from 'fs';
-import * as path from 'path';
-import * as http from 'http';
-import * as https from 'https';
-import * as os from 'os';
-
-// ── Colors ──────────────────────────────────────────────────────────────────
-const c = {
-  green: (s: string) => `\x1b[32m${s}\x1b[0m`,
-  red: (s: string) => `\x1b[31m${s}\x1b[0m`,
-  yellow: (s: string) => `\x1b[33m${s}\x1b[0m`,
-  blue: (s: string) => `\x1b[34m${s}\x1b[0m`,
-  cyan: (s: string) => `\x1b[36m${s}\x1b[0m`,
-  bold: (s: string) => `\x1b[1m${s}\x1b[0m`,
-  dim: (s: string) => `\x1b[2m${s}\x1b[0m`,
-  magenta: (s: string) => `\x1b[35m${s}\x1b[0m`,
-};
-
-// ── Types ───────────────────────────────────────────────────────────────────
-interface SetupConfig {
-  provider: 'ollama' | 'openai' | 'deepseek' | 'qwen' | 'anthropic';
-  model?: string;
-  embeddingModel?: string;
-  apiKey?: string;
-  baseUrl?: string;
-}
-
-interface AgentTemplate {
-  id: string;
-  icon: string;
-  name: string;
-  description: string;
-  defaultAgentName: string;
-}
-
-const TEMPLATES: AgentTemplate[] = [
-  { id: 'customer-service', icon: '🎧', name: '客服助手', description: '回答客户问题，自动学习产品知识', defaultAgentName: '我的客服助手' },
-  { id: 'content-writer', icon: '✍️', name: '写作助手', description: '帮你写文章、邮件、报告', defaultAgentName: '我的写作助手' },
-  { id: 'data-analyst', icon: '📊', name: '数据分析师', description: '分析数据，生成洞察', defaultAgentName: '我的数据分析师' },
-  { id: 'translator', icon: '🌐', name: '翻译助手', description: '多语言翻译，越用越准', defaultAgentName: '我的翻译助手' },
-  { id: 'general', icon: '🤖', name: '通用助手', description: '什么都能聊', defaultAgentName: '我的AI助手' },
-];
-
-const OPC_HOME = path.join(os.homedir(), '.opc');
-const CONFIG_PATH = path.join(OPC_HOME, 'config.json');
-
-// ── Readline helpers ────────────────────────────────────────────────────────
-export function createRL(input?: NodeJS.ReadableStream, output?: NodeJS.WritableStream): readline.Interface {
-  return readline.createInterface({
-    input: input ?? process.stdin,
-    output: output ?? process.stdout,
-  });
-}
-
-function ask(rl: readline.Interface, question: string): Promise<string> {
-  return new Promise((resolve) => rl.question(question, (a) => resolve(a.trim())));
-}
-
-// ── HTTP helpers ────────────────────────────────────────────────────────────
-function httpGet(url: string, timeout = 5000): Promise<{ status: number; body: string }> {
-  return new Promise((resolve, reject) => {
-    const mod = url.startsWith('https') ? https : http;
-    const req = mod.get(url, { timeout }, (res) => {
-      let body = '';
-      res.on('data', (chunk: Buffer) => { body += chunk; });
-      res.on('end', () => resolve({ status: res.statusCode ?? 0, body }));
-    });
-    req.on('error', reject);
-    req.on('timeout', () => { req.destroy(); reject(new Error('timeout')); });
-  });
-}
-
-// ── Step 1: Welcome ─────────────────────────────────────────────────────────
-function printWelcome(): void {
-  console.log('');
-  console.log(c.bold('  🎉 欢迎使用 OPC Agent！'));
-  console.log(c.dim('     让我们用 3 分钟配置你的第一个 AI Agent。'));
-  console.log('');
-}
-
-// ── Step 2: Model detection ─────────────────────────────────────────────────
-async function detectOllama(): Promise<{ running: boolean; models: string[] }> {
-  try {
-    const { body } = await httpGet('http://localhost:11434/api/tags');
-    const data = JSON.parse(body);
-    const models = (data.models || []).map((m: any) => m.name || m.model);
-    return { running: true, models };
-  } catch {
-    return { running: false, models: [] };
-  }
-}
-
-async function stepModel(rl: readline.Interface): Promise<SetupConfig> {
-  console.log(c.bold('📡 Step 1/4: 配置 AI 模型'));
-  console.log(c.dim('  正在检测本地 Ollama...'));
-  console.log('');
-
-  const ollama = await detectOllama();
-
-  if (ollama.running) {
-    console.log(c.green('  ✔ Ollama 运行中！已安装模型：'));
-    if (ollama.models.length === 0) {
-      console.log(c.dim('    （无模型）'));
-    } else {
-      for (const m of ollama.models) {
-        console.log(`    • ${m}`);
-      }
-    }
-    console.log('');
-
-    const hasChat = ollama.models.some((m) => m.startsWith('qwen2.5'));
-    const hasEmbed = ollama.models.some((m) => m.startsWith('nomic-embed-text'));
-
-    if (!hasChat) {
-      const dl = await ask(rl, `  推荐模型 ${c.cyan('qwen2.5:7b')} 未安装，是否下载？ [Y/n] `);
-      if (dl.toLowerCase() !== 'n') {
-        console.log(c.dim('  → 请在终端运行: ollama pull qwen2.5:7b'));
-      }
-    }
-
-    if (!hasEmbed) {
-      const dl = await ask(rl, `  推荐 Embedding 模型 ${c.cyan('nomic-embed-text')} 未安装，是否下载？ [Y/n] `);
-      if (dl.toLowerCase() !== 'n') {
-        console.log(c.dim('  → 请在终端运行: ollama pull nomic-embed-text'));
-      }
-    }
-
-    console.log('');
-    return {
-      provider: 'ollama',
-      model: hasChat ? ollama.models.find((m) => m.startsWith('qwen2.5'))! : 'qwen2.5:7b',
-      embeddingModel: hasEmbed ? 'nomic-embed-text' : 'nomic-embed-text',
-      baseUrl: 'http://localhost:11434',
-    };
-  }
-
-  // Ollama not running
-  console.log(c.yellow('  ⚠ 未检测到 Ollama'));
-  console.log('');
-  console.log('  请选择：');
-  console.log(`    ${c.cyan('A')} ) 安装 Ollama（推荐，免费本地运行）`);
-  console.log(`    ${c.cyan('B')} ) 使用云端 API`);
-  console.log('');
-
-  const choice = await ask(rl, '  你的选择 [A/B]: ');
-
-  if (choice.toUpperCase() !== 'B') {
-    console.log('');
-    console.log(c.bold('  📦 安装 Ollama：'));
-    console.log('');
-    if (process.platform === 'win32') {
-      console.log('    1. 访问 https://ollama.com/download');
-      console.log('    2. 下载 Windows 版并安装');
-      console.log('    3. 安装后运行: ollama pull qwen2.5:7b');
-    } else if (process.platform === 'darwin') {
-      console.log('    brew install ollama');
-      console.log('    ollama serve &');
-      console.log('    ollama pull qwen2.5:7b');
-    } else {
-      console.log('    curl -fsSL https://ollama.com/install.sh | sh');
-      console.log('    ollama serve &');
-      console.log('    ollama pull qwen2.5:7b');
-    }
-    console.log('');
-    console.log(c.dim('  安装完成后重新运行 opc setup'));
-    return { provider: 'ollama', model: 'qwen2.5:7b', embeddingModel: 'nomic-embed-text', baseUrl: 'http://localhost:11434' };
-  }
-
-  // Cloud API
-  console.log('');
-  console.log('  选择云端 Provider：');
-  console.log(`    ${c.cyan('1')} ) OpenAI`);
-  console.log(`    ${c.cyan('2')} ) DeepSeek`);
-  console.log(`    ${c.cyan('3')} ) 通义千问 (Qwen)`);
-  console.log(`    ${c.cyan('4')} ) Anthropic`);
-  console.log('');
-
-  const providerChoice = await ask(rl, '  选择 [1-4]: ');
-  const providers: Record<string, { provider: SetupConfig['provider']; baseUrl: string; model: string; testUrl: string }> = {
-    '1': { provider: 'openai', baseUrl: 'https://api.openai.com/v1', model: 'gpt-4o-mini', testUrl: 'https://api.openai.com/v1/models' },
-    '2': { provider: 'deepseek', baseUrl: 'https://api.deepseek.com/v1', model: 'deepseek-chat', testUrl: 'https://api.deepseek.com/v1/models' },
-    '3': { provider: 'qwen', baseUrl: 'https://dashscope.aliyuncs.com/compatible-mode/v1', model: 'qwen-plus', testUrl: 'https://dashscope.aliyuncs.com/compatible-mode/v1/models' },
-    '4': { provider: 'anthropic', baseUrl: 'https://api.anthropic.com/v1', model: 'claude-3-5-sonnet-20241022', testUrl: 'https://api.anthropic.com/v1/models' },
-  };
-
-  const p = providers[providerChoice] ?? providers['1'];
-  console.log('');
-  const apiKey = await ask(rl, `  请输入 ${c.bold(p.provider)} API Key: `);
-
-  if (apiKey) {
-    console.log(c.dim('  正在测试连接...'));
-    try {
-      await httpGet(p.testUrl);
-      console.log(c.green('  ✔ 连接成功！'));
-    } catch {
-      console.log(c.yellow('  ⚠ 无法验证连接，但配置已保存，稍后可测试'));
-    }
-  }
-
-  console.log('');
-  return { provider: p.provider, model: p.model, apiKey, baseUrl: p.baseUrl };
-}
-
-// ── Step 3: Choose template ─────────────────────────────────────────────────
-async function stepTemplate(rl: readline.Interface): Promise<AgentTemplate> {
-  console.log(c.bold('📋 Step 2/4: 选择 Agent 模板'));
-  console.log('');
-
-  for (let i = 0; i < TEMPLATES.length; i++) {
-    const t = TEMPLATES[i];
-    console.log(`    ${c.cyan(String(i + 1))} ) ${t.icon} ${c.bold(t.name)} — ${c.dim(t.description)}`);
-  }
-  console.log(`    ${c.cyan('6')} ) 📋 更多模板...（打开 Studio 浏览）`);
-  console.log('');
-
-  const choice = await ask(rl, '  选择模板 [1-6]: ');
-  const idx = parseInt(choice, 10) - 1;
-
-  if (choice === '6') {
-    console.log(c.dim('  → 运行 opc studio 浏览更多模板'));
-    return TEMPLATES[4]; // default to general
-  }
-
-  const template = TEMPLATES[idx] ?? TEMPLATES[4];
-  console.log(c.green(`  ✔ 已选择: ${template.icon} ${template.name}`));
-  console.log('');
-  return template;
-}
-
-// ── Step 4: Create agent ────────────────────────────────────────────────────
-async function stepCreateAgent(rl: readline.Interface, template: AgentTemplate, config: SetupConfig): Promise<{ name: string; id: string; dir: string }> {
-  console.log(c.bold('🤖 Step 3/4: 创建 Agent'));
-  console.log('');
-
-  const name = (await ask(rl, `  Agent 名称 [${template.defaultAgentName}]: `)) || template.defaultAgentName;
-  const description = await ask(rl, `  简短描述（可选）: `);
-
-  // Generate a slug id
-  const id = name
-    .replace(/[^\w\u4e00-\u9fff]/g, '-')
-    .replace(/-+/g, '-')
-    .replace(/^-|-$/g, '')
-    .toLowerCase() || `agent-${Date.now()}`;
-
-  const agentDir = path.join(OPC_HOME, 'agents', id);
-  fs.mkdirSync(agentDir, { recursive: true });
-
-  // Write agent config
-  const agentConfig = {
-    id,
-    name,
-    description: description || template.description,
-    template: template.id,
-    provider: config.provider,
-    model: config.model,
-    embeddingModel: config.embeddingModel,
-    apiKey: config.apiKey,
-    baseUrl: config.baseUrl,
-    createdAt: new Date().toISOString(),
-  };
-
-  fs.writeFileSync(path.join(agentDir, 'config.json'), JSON.stringify(agentConfig, null, 2));
-
-  // Initialize brain directory
-  const brainDir = path.join(agentDir, 'brain');
-  fs.mkdirSync(brainDir, { recursive: true });
-  fs.writeFileSync(path.join(brainDir, 'README.md'), `# ${name}\n\n将知识文档放在这里，Agent 会自动学习。\n`);
-
-  // Save global config
-  fs.mkdirSync(OPC_HOME, { recursive: true });
-  const globalConfig = fs.existsSync(CONFIG_PATH) ? JSON.parse(fs.readFileSync(CONFIG_PATH, 'utf-8')) : {};
-  globalConfig.defaultAgent = id;
-  globalConfig.provider = config.provider;
-  globalConfig.model = config.model;
-  globalConfig.baseUrl = config.baseUrl;
-  if (config.apiKey) globalConfig.apiKey = config.apiKey;
-  if (config.embeddingModel) globalConfig.embeddingModel = config.embeddingModel;
-  fs.writeFileSync(CONFIG_PATH, JSON.stringify(globalConfig, null, 2));
-
-  console.log('');
-  console.log(`  ${c.green('✔')} Agent 已创建: ${c.bold(name)}`);
-  console.log(`  ${c.dim('    目录: ' + agentDir)}`);
-  console.log('');
-
-  return { name, id, dir: agentDir };
-}
-
-// ── Step 5: Completion ──────────────────────────────────────────────────────
-function printCompletion(agentName: string): void {
-  console.log(c.bold(`  ✅ 你的 AI Agent「${agentName}」已创建！`));
-  console.log('');
-  console.log('  启动方式：');
-  console.log(`    ${c.cyan('opc studio')}    — 打开网页管理界面`);
-  console.log(`    ${c.cyan('opc chat')}      — 终端直接对话`);
-  console.log(`    ${c.cyan('opc start')}     — 后台运行`);
-  console.log('');
-  console.log(c.dim('  你的 Agent 会自动学习和进化，越用越聪明！🧬'));
-  console.log('');
-}
-
-// ── Main entry ──────────────────────────────────────────────────────────────
-export async function runSetup(input?: NodeJS.ReadableStream, output?: NodeJS.WritableStream): Promise<void> {
-  const rl = createRL(input, output);
-  try {
-    printWelcome();
-    const config = await stepModel(rl);
-    const template = await stepTemplate(rl);
-    const agent = await stepCreateAgent(rl, template, config);
-    printCompletion(agent.name);
-  } finally {
-    rl.close();
-  }
-}
+import * as readline from 'readline';
+import * as fs from 'fs';
+import * as path from 'path';
+import * as http from 'http';
+import * as https from 'https';
+import * as os from 'os';
+
+// ── Colors ──────────────────────────────────────────────────────────────────
+const c = {
+  green: (s: string) => `\x1b[32m${s}\x1b[0m`,
+  red: (s: string) => `\x1b[31m${s}\x1b[0m`,
+  yellow: (s: string) => `\x1b[33m${s}\x1b[0m`,
+  blue: (s: string) => `\x1b[34m${s}\x1b[0m`,
+  cyan: (s: string) => `\x1b[36m${s}\x1b[0m`,
+  bold: (s: string) => `\x1b[1m${s}\x1b[0m`,
+  dim: (s: string) => `\x1b[2m${s}\x1b[0m`,
+  magenta: (s: string) => `\x1b[35m${s}\x1b[0m`,
+};
+
+// ── Types ───────────────────────────────────────────────────────────────────
+interface SetupConfig {
+  provider: 'ollama' | 'openai' | 'deepseek' | 'qwen' | 'anthropic';
+  model?: string;
+  embeddingModel?: string;
+  apiKey?: string;
+  baseUrl?: string;
+}
+
+interface AgentTemplate {
+  id: string;
+  icon: string;
+  name: string;
+  description: string;
+  defaultAgentName: string;
+}
+
+const TEMPLATES: AgentTemplate[] = [
+  { id: 'customer-service', icon: '🎧', name: '客服助手', description: '回答客户问题，自动学习产品知识', defaultAgentName: '我的客服助手' },
+  { id: 'content-writer', icon: '✍️', name: '写作助手', description: '帮你写文章、邮件、报告', defaultAgentName: '我的写作助手' },
+  { id: 'data-analyst', icon: '📊', name: '数据分析师', description: '分析数据，生成洞察', defaultAgentName: '我的数据分析师' },
+  { id: 'translator', icon: '🌐', name: '翻译助手', description: '多语言翻译，越用越准', defaultAgentName: '我的翻译助手' },
+  { id: 'general', icon: '🤖', name: '通用助手', description: '什么都能聊', defaultAgentName: '我的AI助手' },
+];
+
+const OPC_HOME = path.join(os.homedir(), '.opc');
+const CONFIG_PATH = path.join(OPC_HOME, 'config.json');
+
+// ── Readline helpers ────────────────────────────────────────────────────────
+export function createRL(input?: NodeJS.ReadableStream, output?: NodeJS.WritableStream): readline.Interface {
+  return readline.createInterface({
+    input: input ?? process.stdin,
+    output: output ?? process.stdout,
+  });
+}
+
+function ask(rl: readline.Interface, question: string): Promise<string> {
+  return new Promise((resolve) => rl.question(question, (a) => resolve(a.trim())));
+}
+
+// ── HTTP helpers ────────────────────────────────────────────────────────────
+function httpGet(url: string, timeout = 5000): Promise<{ status: number; body: string }> {
+  return new Promise((resolve, reject) => {
+    const mod = url.startsWith('https') ? https : http;
+    const req = mod.get(url, { timeout }, (res) => {
+      let body = '';
+      res.on('data', (chunk: Buffer) => { body += chunk; });
+      res.on('end', () => resolve({ status: res.statusCode ?? 0, body }));
+    });
+    req.on('error', reject);
+    req.on('timeout', () => { req.destroy(); reject(new Error('timeout')); });
+  });
+}
+
+// ── Step 1: Welcome ─────────────────────────────────────────────────────────
+function printWelcome(): void {
+  console.log('');
+  console.log(c.bold('  🎉 欢迎使用 OPC Agent！'));
+  console.log(c.dim('     让我们用 3 分钟配置你的第一个 AI Agent。'));
+  console.log('');
+}
+
+// ── Step 2: Model detection ─────────────────────────────────────────────────
+async function detectOllama(): Promise<{ running: boolean; models: string[] }> {
+  try {
+    const { body } = await httpGet('http://localhost:11434/api/tags');
+    const data = JSON.parse(body);
+    const models = (data.models || []).map((m: any) => m.name || m.model);
+    return { running: true, models };
+  } catch {
+    return { running: false, models: [] };
+  }
+}
+
+async function stepModel(rl: readline.Interface): Promise<SetupConfig> {
+  console.log(c.bold('📡 Step 1/4: 配置 AI 模型'));
+  console.log(c.dim('  正在检测本地 Ollama...'));
+  console.log('');
+
+  const ollama = await detectOllama();
+
+  if (ollama.running) {
+    console.log(c.green('  ✔ Ollama 运行中！已安装模型：'));
+    if (ollama.models.length === 0) {
+      console.log(c.dim('    （无模型）'));
+    } else {
+      for (const m of ollama.models) {
+        console.log(`    • ${m}`);
+      }
+    }
+    console.log('');
+
+    const hasChat = ollama.models.some((m) => m.startsWith('qwen2.5'));
+    const hasEmbed = ollama.models.some((m) => m.startsWith('nomic-embed-text'));
+
+    if (!hasChat) {
+      const dl = await ask(rl, `  推荐模型 ${c.cyan('qwen2.5:7b')} 未安装，是否下载？ [Y/n] `);
+      if (dl.toLowerCase() !== 'n') {
+        console.log(c.dim('  → 请在终端运行: ollama pull qwen2.5:7b'));
+      }
+    }
+
+    if (!hasEmbed) {
+      const dl = await ask(rl, `  推荐 Embedding 模型 ${c.cyan('nomic-embed-text')} 未安装，是否下载？ [Y/n] `);
+      if (dl.toLowerCase() !== 'n') {
+        console.log(c.dim('  → 请在终端运行: ollama pull nomic-embed-text'));
+      }
+    }
+
+    console.log('');
+    return {
+      provider: 'ollama',
+      model: hasChat ? ollama.models.find((m) => m.startsWith('qwen2.5'))! : 'qwen2.5:7b',
+      embeddingModel: hasEmbed ? 'nomic-embed-text' : 'nomic-embed-text',
+      baseUrl: 'http://localhost:11434',
+    };
+  }
+
+  // Ollama not running
+  console.log(c.yellow('  ⚠ 未检测到 Ollama'));
+  console.log('');
+  console.log('  请选择：');
+  console.log(`    ${c.cyan('A')} ) 安装 Ollama（推荐，免费本地运行）`);
+  console.log(`    ${c.cyan('B')} ) 使用云端 API`);
+  console.log('');
+
+  const choice = await ask(rl, '  你的选择 [A/B]: ');
+
+  if (choice.toUpperCase() !== 'B') {
+    console.log('');
+    console.log(c.bold('  📦 安装 Ollama：'));
+    console.log('');
+    if (process.platform === 'win32') {
+      console.log('    1. 访问 https://ollama.com/download');
+      console.log('    2. 下载 Windows 版并安装');
+      console.log('    3. 安装后运行: ollama pull qwen2.5:7b');
+    } else if (process.platform === 'darwin') {
+      console.log('    brew install ollama');
+      console.log('    ollama serve &');
+      console.log('    ollama pull qwen2.5:7b');
+    } else {
+      console.log('    curl -fsSL https://ollama.com/install.sh | sh');
+      console.log('    ollama serve &');
+      console.log('    ollama pull qwen2.5:7b');
+    }
+    console.log('');
+    console.log(c.dim('  安装完成后重新运行 opc setup'));
+    return { provider: 'ollama', model: 'qwen2.5:7b', embeddingModel: 'nomic-embed-text', baseUrl: 'http://localhost:11434' };
+  }
+
+  // Cloud API
+  console.log('');
+  console.log('  选择云端 Provider：');
+  console.log(`    ${c.cyan('1')} ) OpenAI`);
+  console.log(`    ${c.cyan('2')} ) DeepSeek`);
+  console.log(`    ${c.cyan('3')} ) 通义千问 (Qwen)`);
+  console.log(`    ${c.cyan('4')} ) Anthropic`);
+  console.log('');
+
+  const providerChoice = await ask(rl, '  选择 [1-4]: ');
+  const providers: Record<string, { provider: SetupConfig['provider']; baseUrl: string; model: string; testUrl: string }> = {
+    '1': { provider: 'openai', baseUrl: 'https://api.openai.com/v1', model: 'gpt-4o-mini', testUrl: 'https://api.openai.com/v1/models' },
+    '2': { provider: 'deepseek', baseUrl: 'https://api.deepseek.com/v1', model: 'deepseek-chat', testUrl: 'https://api.deepseek.com/v1/models' },
+    '3': { provider: 'qwen', baseUrl: 'https://dashscope.aliyuncs.com/compatible-mode/v1', model: 'qwen-plus', testUrl: 'https://dashscope.aliyuncs.com/compatible-mode/v1/models' },
+    '4': { provider: 'anthropic', baseUrl: 'https://api.anthropic.com/v1', model: 'claude-3-5-sonnet-20241022', testUrl: 'https://api.anthropic.com/v1/models' },
+  };
+
+  const p = providers[providerChoice] ?? providers['1'];
+  console.log('');
+  const apiKey = await ask(rl, `  请输入 ${c.bold(p.provider)} API Key: `);
+
+  if (apiKey) {
+    console.log(c.dim('  正在测试连接...'));
+    try {
+      await httpGet(p.testUrl);
+      console.log(c.green('  ✔ 连接成功！'));
+    } catch {
+      console.log(c.yellow('  ⚠ 无法验证连接，但配置已保存，稍后可测试'));
+    }
+  }
+
+  console.log('');
+  return { provider: p.provider, model: p.model, apiKey, baseUrl: p.baseUrl };
+}
+
+// ── Step 3: Choose template ─────────────────────────────────────────────────
+async function stepTemplate(rl: readline.Interface): Promise<AgentTemplate> {
+  console.log(c.bold('📋 Step 2/4: 选择 Agent 模板'));
+  console.log('');
+
+  for (let i = 0; i < TEMPLATES.length; i++) {
+    const t = TEMPLATES[i];
+    console.log(`    ${c.cyan(String(i + 1))} ) ${t.icon} ${c.bold(t.name)} — ${c.dim(t.description)}`);
+  }
+  console.log(`    ${c.cyan('6')} ) 📋 更多模板...（打开 Studio 浏览）`);
+  console.log('');
+
+  const choice = await ask(rl, '  选择模板 [1-6]: ');
+  const idx = parseInt(choice, 10) - 1;
+
+  if (choice === '6') {
+    console.log(c.dim('  → 运行 opc studio 浏览更多模板'));
+    return TEMPLATES[4]; // default to general
+  }
+
+  const template = TEMPLATES[idx] ?? TEMPLATES[4];
+  console.log(c.green(`  ✔ 已选择: ${template.icon} ${template.name}`));
+  console.log('');
+  return template;
+}
+
+// ── Step 4: Create agent ────────────────────────────────────────────────────
+async function stepCreateAgent(rl: readline.Interface, template: AgentTemplate, config: SetupConfig): Promise<{ name: string; id: string; dir: string }> {
+  console.log(c.bold('🤖 Step 3/4: 创建 Agent'));
+  console.log('');
+
+  const name = (await ask(rl, `  Agent 名称 [${template.defaultAgentName}]: `)) || template.defaultAgentName;
+  const description = await ask(rl, `  简短描述（可选）: `);
+
+  // Generate a slug id
+  const id = name
+    .replace(/[^\w\u4e00-\u9fff]/g, '-')
+    .replace(/-+/g, '-')
+    .replace(/^-|-$/g, '')
+    .toLowerCase() || `agent-${Date.now()}`;
+
+  const agentDir = path.join(OPC_HOME, 'agents', id);
+  fs.mkdirSync(agentDir, { recursive: true });
+
+  // Write agent config
+  const agentConfig = {
+    id,
+    name,
+    description: description || template.description,
+    template: template.id,
+    provider: config.provider,
+    model: config.model,
+    embeddingModel: config.embeddingModel,
+    apiKey: config.apiKey,
+    baseUrl: config.baseUrl,
+    createdAt: new Date().toISOString(),
+  };
+
+  fs.writeFileSync(path.join(agentDir, 'config.json'), JSON.stringify(agentConfig, null, 2));
+
+  // Initialize brain directory
+  const brainDir = path.join(agentDir, 'brain');
+  fs.mkdirSync(brainDir, { recursive: true });
+  fs.writeFileSync(path.join(brainDir, 'README.md'), `# ${name}\n\n将知识文档放在这里，Agent 会自动学习。\n`);
+
+  // Save global config
+  fs.mkdirSync(OPC_HOME, { recursive: true });
+  const globalConfig = fs.existsSync(CONFIG_PATH) ? JSON.parse(fs.readFileSync(CONFIG_PATH, 'utf-8')) : {};
+  globalConfig.defaultAgent = id;
+  globalConfig.provider = config.provider;
+  globalConfig.model = config.model;
+  globalConfig.baseUrl = config.baseUrl;
+  if (config.apiKey) globalConfig.apiKey = config.apiKey;
+  if (config.embeddingModel) globalConfig.embeddingModel = config.embeddingModel;
+  fs.writeFileSync(CONFIG_PATH, JSON.stringify(globalConfig, null, 2));
+
+  console.log('');
+  console.log(`  ${c.green('✔')} Agent 已创建: ${c.bold(name)}`);
+  console.log(`  ${c.dim('    目录: ' + agentDir)}`);
+  console.log('');
+
+  return { name, id, dir: agentDir };
+}
+
+// ── Step 5: Completion ──────────────────────────────────────────────────────
+function printCompletion(agentName: string): void {
+  console.log(c.bold(`  ✅ 你的 AI Agent「${agentName}」已创建！`));
+  console.log('');
+  console.log('  启动方式：');
+  console.log(`    ${c.cyan('opc studio')}    — 打开网页管理界面`);
+  console.log(`    ${c.cyan('opc chat')}      — 终端直接对话`);
+  console.log(`    ${c.cyan('opc start')}     — 后台运行`);
+  console.log('');
+  console.log(c.dim('  你的 Agent 会自动学习和进化，越用越聪明！🧬'));
+  console.log('');
+}
+
+// ── Main entry ──────────────────────────────────────────────────────────────
+export async function runSetup(input?: NodeJS.ReadableStream, output?: NodeJS.WritableStream): Promise<void> {
+  const rl = createRL(input, output);
+  try {
+    printWelcome();
+    const config = await stepModel(rl);
+    const template = await stepTemplate(rl);
+    const agent = await stepCreateAgent(rl, template, config);
+    printCompletion(agent.name);
+  } finally {
+    rl.close();
+  }
+}