@dpml/agent/dist/bin.mjs.map

Agent implementation for DPML

{"version":3,"sources":["../../../node_modules/.pnpm/tsup@8.4.0_@swc+core@1.11.24_postcss@8.5.3_typescript@5.8.2_yaml@2.7.1/node_modules/tsup/assets/esm_shims.js","../../../node_modules/.pnpm/dotenv@16.5.0/node_modules/dotenv/package.json","../../../node_modules/.pnpm/dotenv@16.5.0/node_modules/dotenv/lib/main.js","../src/index.ts","../src/config/index.ts","../src/config/schema.ts","../src/config/transformers.ts","../src/config/cli.ts","../src/api/agent.ts","../src/core/agentService.ts","../src/types/index.ts","../src/types/errors.ts","../src/core/AgentRunner.ts","../src/core/llm/llmFactory.ts","../src/core/llm/AnthropicClient.ts","../src/core/llm/OpenAIClient.ts","../src/core/session/InMemoryAgentSession.ts","../src/api/agentenv.ts","../src/core/agentenv/agentenvCore.ts","../src/core/agentenv/constants.ts","../src/api/index.ts","../src/bin.ts"],"sourcesContent":["// Shim globals in esm bundle\nimport { fileURLToPath } from 'url'\nimport path from 'path'\n\nconst getFilename = () => fileURLToPath(import.meta.url)\nconst getDirname = () => path.dirname(getFilename())\n\nexport const __dirname = /* @__PURE__ */ getDirname()\nexport const __filename = /* @__PURE__ */ getFilename()\n","{\n  \"name\": \"dotenv\",\n  \"version\": \"16.5.0\",\n  \"description\": \"Loads environment variables from .env file\",\n  \"main\": \"lib/main.js\",\n  \"types\": \"lib/main.d.ts\",\n  \"exports\": {\n    \".\": {\n      \"types\": \"./lib/main.d.ts\",\n      \"require\": \"./lib/main.js\",\n      \"default\": \"./lib/main.js\"\n    },\n    \"./config\": \"./config.js\",\n    \"./config.js\": \"./config.js\",\n    \"./lib/env-options\": \"./lib/env-options.js\",\n    \"./lib/env-options.js\": \"./lib/env-options.js\",\n    \"./lib/cli-options\": \"./lib/cli-options.js\",\n    \"./lib/cli-options.js\": \"./lib/cli-options.js\",\n    \"./package.json\": \"./package.json\"\n  },\n  \"scripts\": {\n    \"dts-check\": \"tsc --project tests/types/tsconfig.json\",\n    \"lint\": \"standard\",\n    \"pretest\": \"npm run lint && npm run dts-check\",\n    \"test\": \"tap run --allow-empty-coverage --disable-coverage --timeout=60000\",\n    \"test:coverage\": \"tap run --show-full-coverage --timeout=60000 --coverage-report=lcov\",\n    \"prerelease\": \"npm test\",\n    \"release\": \"standard-version\"\n  },\n  \"repository\": {\n    \"type\": \"git\",\n    \"url\": \"git://github.com/motdotla/dotenv.git\"\n  },\n  \"homepage\": \"https://github.com/motdotla/dotenv#readme\",\n  \"funding\": \"https://dotenvx.com\",\n  \"keywords\": [\n    \"dotenv\",\n    \"env\",\n    \".env\",\n    \"environment\",\n    \"variables\",\n    \"config\",\n    \"settings\"\n  ],\n  \"readmeFilename\": \"README.md\",\n  \"license\": \"BSD-2-Clause\",\n  \"devDependencies\": {\n    \"@types/node\": \"^18.11.3\",\n    \"decache\": \"^4.6.2\",\n    \"sinon\": \"^14.0.1\",\n    \"standard\": \"^17.0.0\",\n    \"standard-version\": \"^9.5.0\",\n    \"tap\": \"^19.2.0\",\n    \"typescript\": \"^4.8.4\"\n  },\n  \"engines\": {\n    \"node\": \">=12\"\n  },\n  \"browser\": {\n    \"fs\": false\n  }\n}\n","const fs = require('fs')\nconst path = require('path')\nconst os = require('os')\nconst crypto = require('crypto')\nconst packageJson = require('../package.json')\n\nconst version = packageJson.version\n\nconst LINE = /(?:^|^)\\s*(?:export\\s+)?([\\w.-]+)(?:\\s*=\\s*?|:\\s+?)(\\s*'(?:\\\\'|[^'])*'|\\s*\"(?:\\\\\"|[^\"])*\"|\\s*`(?:\\\\`|[^`])*`|[^#\\r\\n]+)?\\s*(?:#.*)?(?:$|$)/mg\n\n// Parse src into an Object\nfunction parse (src) {\n  const obj = {}\n\n  // Convert buffer to string\n  let lines = src.toString()\n\n  // Convert line breaks to same format\n  lines = lines.replace(/\\r\\n?/mg, '\\n')\n\n  let match\n  while ((match = LINE.exec(lines)) != null) {\n    const key = match[1]\n\n    // Default undefined or null to empty string\n    let value = (match[2] || '')\n\n    // Remove whitespace\n    value = value.trim()\n\n    // Check if double quoted\n    const maybeQuote = value[0]\n\n    // Remove surrounding quotes\n    value = value.replace(/^(['\"`])([\\s\\S]*)\\1$/mg, '$2')\n\n    // Expand newlines if double quoted\n    if (maybeQuote === '\"') {\n      value = value.replace(/\\\\n/g, '\\n')\n      value = value.replace(/\\\\r/g, '\\r')\n    }\n\n    // Add to object\n    obj[key] = value\n  }\n\n  return obj\n}\n\nfunction _parseVault (options) {\n  const vaultPath = _vaultPath(options)\n\n  // Parse .env.vault\n  const result = DotenvModule.configDotenv({ path: vaultPath })\n  if (!result.parsed) {\n    const err = new Error(`MISSING_DATA: Cannot parse ${vaultPath} for an unknown reason`)\n    err.code = 'MISSING_DATA'\n    throw err\n  }\n\n  // handle scenario for comma separated keys - for use with key rotation\n  // example: DOTENV_KEY=\"dotenv://:key_1234@dotenvx.com/vault/.env.vault?environment=prod,dotenv://:key_7890@dotenvx.com/vault/.env.vault?environment=prod\"\n  const keys = _dotenvKey(options).split(',')\n  const length = keys.length\n\n  let decrypted\n  for (let i = 0; i < length; i++) {\n    try {\n      // Get full key\n      const key = keys[i].trim()\n\n      // Get instructions for decrypt\n      const attrs = _instructions(result, key)\n\n      // Decrypt\n      decrypted = DotenvModule.decrypt(attrs.ciphertext, attrs.key)\n\n      break\n    } catch (error) {\n      // last key\n      if (i + 1 >= length) {\n        throw error\n      }\n      // try next key\n    }\n  }\n\n  // Parse decrypted .env string\n  return DotenvModule.parse(decrypted)\n}\n\nfunction _warn (message) {\n  console.log(`[dotenv@${version}][WARN] ${message}`)\n}\n\nfunction _debug (message) {\n  console.log(`[dotenv@${version}][DEBUG] ${message}`)\n}\n\nfunction _dotenvKey (options) {\n  // prioritize developer directly setting options.DOTENV_KEY\n  if (options && options.DOTENV_KEY && options.DOTENV_KEY.length > 0) {\n    return options.DOTENV_KEY\n  }\n\n  // secondary infra already contains a DOTENV_KEY environment variable\n  if (process.env.DOTENV_KEY && process.env.DOTENV_KEY.length > 0) {\n    return process.env.DOTENV_KEY\n  }\n\n  // fallback to empty string\n  return ''\n}\n\nfunction _instructions (result, dotenvKey) {\n  // Parse DOTENV_KEY. Format is a URI\n  let uri\n  try {\n    uri = new URL(dotenvKey)\n  } catch (error) {\n    if (error.code === 'ERR_INVALID_URL') {\n      const err = new Error('INVALID_DOTENV_KEY: Wrong format. Must be in valid uri format like dotenv://:key_1234@dotenvx.com/vault/.env.vault?environment=development')\n      err.code = 'INVALID_DOTENV_KEY'\n      throw err\n    }\n\n    throw error\n  }\n\n  // Get decrypt key\n  const key = uri.password\n  if (!key) {\n    const err = new Error('INVALID_DOTENV_KEY: Missing key part')\n    err.code = 'INVALID_DOTENV_KEY'\n    throw err\n  }\n\n  // Get environment\n  const environment = uri.searchParams.get('environment')\n  if (!environment) {\n    const err = new Error('INVALID_DOTENV_KEY: Missing environment part')\n    err.code = 'INVALID_DOTENV_KEY'\n    throw err\n  }\n\n  // Get ciphertext payload\n  const environmentKey = `DOTENV_VAULT_${environment.toUpperCase()}`\n  const ciphertext = result.parsed[environmentKey] // DOTENV_VAULT_PRODUCTION\n  if (!ciphertext) {\n    const err = new Error(`NOT_FOUND_DOTENV_ENVIRONMENT: Cannot locate environment ${environmentKey} in your .env.vault file.`)\n    err.code = 'NOT_FOUND_DOTENV_ENVIRONMENT'\n    throw err\n  }\n\n  return { ciphertext, key }\n}\n\nfunction _vaultPath (options) {\n  let possibleVaultPath = null\n\n  if (options && options.path && options.path.length > 0) {\n    if (Array.isArray(options.path)) {\n      for (const filepath of options.path) {\n        if (fs.existsSync(filepath)) {\n          possibleVaultPath = filepath.endsWith('.vault') ? filepath : `${filepath}.vault`\n        }\n      }\n    } else {\n      possibleVaultPath = options.path.endsWith('.vault') ? options.path : `${options.path}.vault`\n    }\n  } else {\n    possibleVaultPath = path.resolve(process.cwd(), '.env.vault')\n  }\n\n  if (fs.existsSync(possibleVaultPath)) {\n    return possibleVaultPath\n  }\n\n  return null\n}\n\nfunction _resolveHome (envPath) {\n  return envPath[0] === '~' ? path.join(os.homedir(), envPath.slice(1)) : envPath\n}\n\nfunction _configVault (options) {\n  const debug = Boolean(options && options.debug)\n  if (debug) {\n    _debug('Loading env from encrypted .env.vault')\n  }\n\n  const parsed = DotenvModule._parseVault(options)\n\n  let processEnv = process.env\n  if (options && options.processEnv != null) {\n    processEnv = options.processEnv\n  }\n\n  DotenvModule.populate(processEnv, parsed, options)\n\n  return { parsed }\n}\n\nfunction configDotenv (options) {\n  const dotenvPath = path.resolve(process.cwd(), '.env')\n  let encoding = 'utf8'\n  const debug = Boolean(options && options.debug)\n\n  if (options && options.encoding) {\n    encoding = options.encoding\n  } else {\n    if (debug) {\n      _debug('No encoding is specified. UTF-8 is used by default')\n    }\n  }\n\n  let optionPaths = [dotenvPath] // default, look for .env\n  if (options && options.path) {\n    if (!Array.isArray(options.path)) {\n      optionPaths = [_resolveHome(options.path)]\n    } else {\n      optionPaths = [] // reset default\n      for (const filepath of options.path) {\n        optionPaths.push(_resolveHome(filepath))\n      }\n    }\n  }\n\n  // Build the parsed data in a temporary object (because we need to return it).  Once we have the final\n  // parsed data, we will combine it with process.env (or options.processEnv if provided).\n  let lastError\n  const parsedAll = {}\n  for (const path of optionPaths) {\n    try {\n      // Specifying an encoding returns a string instead of a buffer\n      const parsed = DotenvModule.parse(fs.readFileSync(path, { encoding }))\n\n      DotenvModule.populate(parsedAll, parsed, options)\n    } catch (e) {\n      if (debug) {\n        _debug(`Failed to load ${path} ${e.message}`)\n      }\n      lastError = e\n    }\n  }\n\n  let processEnv = process.env\n  if (options && options.processEnv != null) {\n    processEnv = options.processEnv\n  }\n\n  DotenvModule.populate(processEnv, parsedAll, options)\n\n  if (lastError) {\n    return { parsed: parsedAll, error: lastError }\n  } else {\n    return { parsed: parsedAll }\n  }\n}\n\n// Populates process.env from .env file\nfunction config (options) {\n  // fallback to original dotenv if DOTENV_KEY is not set\n  if (_dotenvKey(options).length === 0) {\n    return DotenvModule.configDotenv(options)\n  }\n\n  const vaultPath = _vaultPath(options)\n\n  // dotenvKey exists but .env.vault file does not exist\n  if (!vaultPath) {\n    _warn(`You set DOTENV_KEY but you are missing a .env.vault file at ${vaultPath}. Did you forget to build it?`)\n\n    return DotenvModule.configDotenv(options)\n  }\n\n  return DotenvModule._configVault(options)\n}\n\nfunction decrypt (encrypted, keyStr) {\n  const key = Buffer.from(keyStr.slice(-64), 'hex')\n  let ciphertext = Buffer.from(encrypted, 'base64')\n\n  const nonce = ciphertext.subarray(0, 12)\n  const authTag = ciphertext.subarray(-16)\n  ciphertext = ciphertext.subarray(12, -16)\n\n  try {\n    const aesgcm = crypto.createDecipheriv('aes-256-gcm', key, nonce)\n    aesgcm.setAuthTag(authTag)\n    return `${aesgcm.update(ciphertext)}${aesgcm.final()}`\n  } catch (error) {\n    const isRange = error instanceof RangeError\n    const invalidKeyLength = error.message === 'Invalid key length'\n    const decryptionFailed = error.message === 'Unsupported state or unable to authenticate data'\n\n    if (isRange || invalidKeyLength) {\n      const err = new Error('INVALID_DOTENV_KEY: It must be 64 characters long (or more)')\n      err.code = 'INVALID_DOTENV_KEY'\n      throw err\n    } else if (decryptionFailed) {\n      const err = new Error('DECRYPTION_FAILED: Please check your DOTENV_KEY')\n      err.code = 'DECRYPTION_FAILED'\n      throw err\n    } else {\n      throw error\n    }\n  }\n}\n\n// Populate process.env with parsed values\nfunction populate (processEnv, parsed, options = {}) {\n  const debug = Boolean(options && options.debug)\n  const override = Boolean(options && options.override)\n\n  if (typeof parsed !== 'object') {\n    const err = new Error('OBJECT_REQUIRED: Please check the processEnv argument being passed to populate')\n    err.code = 'OBJECT_REQUIRED'\n    throw err\n  }\n\n  // Set process.env\n  for (const key of Object.keys(parsed)) {\n    if (Object.prototype.hasOwnProperty.call(processEnv, key)) {\n      if (override === true) {\n        processEnv[key] = parsed[key]\n      }\n\n      if (debug) {\n        if (override === true) {\n          _debug(`\"${key}\" is already defined and WAS overwritten`)\n        } else {\n          _debug(`\"${key}\" is already defined and was NOT overwritten`)\n        }\n      }\n    } else {\n      processEnv[key] = parsed[key]\n    }\n  }\n}\n\nconst DotenvModule = {\n  configDotenv,\n  _configVault,\n  _parseVault,\n  config,\n  decrypt,\n  parse,\n  populate\n}\n\nmodule.exports.configDotenv = DotenvModule.configDotenv\nmodule.exports._configVault = DotenvModule._configVault\nmodule.exports._parseVault = DotenvModule._parseVault\nmodule.exports.config = DotenvModule.config\nmodule.exports.decrypt = DotenvModule.decrypt\nmodule.exports.parse = DotenvModule.parse\nmodule.exports.populate = DotenvModule.populate\n\nmodule.exports = DotenvModule\n","/**\n * DPML Agent模块\n *\n * 用于创建和管理AI对话代理。提供一组简洁的API，\n * 使应用开发者能够通过DPML定义AI助手，并与大语言模型服务进行交互。\n *\n * @example\n * ```typescript\n * import { createAgent } from '@dpml/agent';\n *\n * const agent = createAgent({\n *   llm: {\n *     apiType: 'openai',\n *     apiKey: process.env.OPENAI_API_KEY,\n *     model: 'gpt-4-turbo'\n *   },\n *   prompt: '你是一个专业的JavaScript和TypeScript助手。'\n * });\n *\n * const response = await agent.chat('如何在TypeScript中实现单例模式？');\n * console.log(response);\n * ```\n *\n * 或者使用DPML声明式语法：\n * ```typescript\n * import { createAgent, compiler } from '@dpml/agent';\n *\n * // DPML配置\n * const dpmlContent = `\n * <agent>\n *   <llm api-type=\"openai\" api-key=\"@agentenv:OPENAI_API_KEY\" model=\"gpt-4-turbo\"></llm>\n *   <prompt>你是一个专业的JavaScript和TypeScript助手。</prompt>\n * </agent>\n * `;\n *\n * // 编译DPML为AgentConfig\n * const config = await compiler.compile(dpmlContent);\n *\n * // 创建Agent\n * const agent = createAgent(config);\n * ```\n */\n\nimport { createDomainDPML } from '@dpml/core';\n\nimport { schema, transformers, commandsConfig } from './config';\nimport type { AgentConfig } from './types';\n\n// 导出API\nexport * from './api';\n\n// 导出类型定义\nexport * from './types';\n\n/**\n * 创建Agent领域DPML实例\n *\n * 集成Schema、转换器和CLI配置，\n * 提供编译和命令行功能。\n */\nexport const agentDPML = createDomainDPML<AgentConfig>({\n  domain: 'agent',\n  description: 'AI Agent Domain',\n  schema,\n  transformers,\n  commands: commandsConfig,\n  options: {\n    strictMode: true,\n    errorHandling: 'throw'\n  }\n});\n\n/**\n * 导出编译器，便于直接使用\n */\nexport const compiler = agentDPML.compiler;\n","/**\n * 配置导出文件\n * 导出所有Agent的DPML配置组件\n */\n\n// 导出Schema\nexport { schema } from './schema';\n\n// 导出转换器\nexport { transformers, agentTransformer } from './transformers';\n\n// 导出CLI配置\nexport { commandsConfig } from './cli';\n","import type { DocumentSchema } from '@dpml/core';\n\n/**\n * Agent的DPML Schema定义\n *\n * 定义DPML文档的结构和约束规则，\n * 包括元素、属性和内容模型。\n */\nexport const schema: DocumentSchema = {\n  // 根元素定义\n  root: {\n    element: 'agent',\n    children: {\n      elements: [\n        { $ref: 'llm' },\n        { $ref: 'prompt' },\n        { $ref: 'experimental' }\n      ]\n    }\n  },\n  // 可复用类型定义\n  types: [\n    {\n      // LLM配置元素\n      element: 'llm',\n      attributes: [\n        {\n          name: 'api-type',\n          required: true\n        },\n        {\n          name: 'api-url'\n        },\n        {\n          name: 'api-key'\n        },\n        {\n          name: 'model',\n          required: true\n        }\n      ]\n    },\n    {\n      // 提示词元素\n      element: 'prompt',\n      content: {\n        type: 'text',\n        required: true\n      }\n    },\n    {\n      // 实验性功能元素\n      element: 'experimental',\n      children: {\n        elements: [\n          { $ref: 'tools' }\n        ]\n      }\n    },\n    {\n      // 工具集元素\n      element: 'tools',\n      children: {\n        elements: [\n          { $ref: 'tool' }\n        ]\n      }\n    },\n    {\n      // 工具元素\n      element: 'tool',\n      attributes: [\n        {\n          name: 'name',\n          required: true\n        },\n        {\n          name: 'description',\n          required: true\n        }\n      ]\n    }\n  ]\n};\n","import { createTransformerDefiner } from '@dpml/core';\nimport type { DPMLNode } from '@dpml/core';\n\nimport type { AgentConfig, LLMConfig } from '../types';\n\n// 创建转换器定义器\nconst definer = createTransformerDefiner();\n\n/**\n * Agent转换器\n *\n * 将DPML文档转换为AgentConfig对象\n */\nexport const agentTransformer = definer.defineStructuralMapper<unknown, AgentConfig>(\n  'agentTransformer',\n  [\n    {\n      // 将LLM元素转换为LLM配置\n      selector: \"agent > llm\",\n      targetPath: \"llm\",\n      transform: (value: unknown) => {\n        const node = value as DPMLNode;\n        const llmConfig: LLMConfig = {\n          apiType: node.attributes.get('api-type') || '',\n          apiUrl: node.attributes.get('api-url'),\n          apiKey: node.attributes.get('api-key'),\n          model: node.attributes.get('model') || ''\n        };\n\n        return llmConfig;\n      }\n    },\n    {\n      // 将prompt元素转换为提示词\n      selector: \"agent > prompt\",\n      targetPath: \"prompt\",\n      transform: (value: unknown) => {\n        const node = value as DPMLNode;\n\n        return node.content || '';\n      }\n    }\n  ]\n);\n\n// 导出所有转换器\nexport const transformers = [\n  agentTransformer\n];\n","import fs from 'fs/promises';\nimport path from 'path';\nimport readline from 'readline';\n\nimport type { DomainCommandsConfig } from '@dpml/core';\nimport dotenv from 'dotenv';\n\nimport { createAgent } from '../api/agent';\nimport { replaceEnvVars } from '../api/agentenv';\n\n/**\n * 加载环境变量\n */\nfunction loadEnvironmentVariables(options: any): void {\n  // 从env文件加载\n  if (options.envFile) {\n    const envPath = path.resolve(process.cwd(), options.envFile);\n\n    dotenv.config({ path: envPath });\n  }\n\n  // 从命令行选项加载\n  if (options.env) {\n    for (const envVar of options.env) {\n      const [key, value] = envVar.split('=');\n\n      if (key && value) {\n        process.env[key] = value;\n      }\n    }\n  }\n}\n\n/**\n * 处理常规聊天（非流式）\n */\nexport async function handleRegularChat(agent: any, input: string): Promise<void> {\n  try {\n    // 发送消息并获取响应\n    const response = await agent.chat(input);\n\n    console.log('\\n' + response + '\\n');\n  } catch (error) {\n    console.error('错误:', error instanceof Error ? error.message : String(error));\n  }\n}\n\n/**\n * 处理流式聊天\n */\nexport async function handleStreamChat(agent: any, input: string): Promise<void> {\n  try {\n    process.stdout.write('\\n'); // 输出开始新行\n\n    for await (const chunk of agent.chatStream(input)) {\n      process.stdout.write(chunk);\n    }\n\n    process.stdout.write('\\n\\n'); // 输出结束添加空行\n  } catch (error) {\n    console.error('\\n错误:', error instanceof Error ? error.message : String(error));\n  }\n}\n\n/**\n * 执行交互式聊天\n */\nasync function executeChat(actionContext: any, filePath: string, options: any): Promise<void> {\n  try {\n    // 加载环境变量\n    loadEnvironmentVariables(options);\n\n    console.log('\\nDPML Agent Chat');\n    console.log(`加载Agent配置: ${filePath}\\n`);\n\n    // 读取文件内容\n    const content = await fs.readFile(filePath, 'utf-8');\n\n    // 使用编译器解析DPML\n    const config = await actionContext.getCompiler().compile(content);\n\n    // 处理环境变量\n    const processedConfig = replaceEnvVars(config);\n\n    // 创建Agent实例\n    const agent = createAgent(processedConfig);\n\n    // 创建交互界面\n    const rl = readline.createInterface({\n      input: process.stdin,\n      output: process.stdout\n    });\n\n    console.log('你好，我是AI助手。有什么我可以帮你的？');\n\n    // 确定是否使用流式输出（默认启用）\n    const useStream = options.stream !== false;\n\n    // 交互式聊天循环\n    const askQuestion = () => {\n      rl.question('> ', async (input) => {\n        if (input.toLowerCase() === 'exit' || input.toLowerCase() === 'quit' || input.toLowerCase() === 'bye') {\n          console.log('会话结束。');\n          rl.close();\n\n          return;\n        }\n\n        // 根据选项选择处理方式\n        if (useStream) {\n          await handleStreamChat(agent, input);\n        } else {\n          await handleRegularChat(agent, input);\n        }\n\n        // 继续等待输入\n        askQuestion();\n      });\n    };\n\n    // 开始交互循环\n    askQuestion();\n  } catch (error) {\n    console.error('错误:', error instanceof Error ? error.message : String(error));\n    process.exit(1);\n  }\n}\n\n/**\n * Agent CLI命令配置\n */\nexport const commandsConfig: DomainCommandsConfig = {\n  // 包含标准validate命令\n  includeStandard: true,\n\n  // 自定义命令\n  actions: [\n    {\n      name: 'chat',\n      description: '启动与Agent的交互式聊天',\n      args: [\n        {\n          name: 'filePath',\n          description: 'Agent配置文件路径',\n          required: true\n        }\n      ],\n      options: [\n        {\n          flags: '-e, --env <KEY=VALUE...>',\n          description: '设置环境变量'\n        },\n        {\n          flags: '-f, --env-file <path>',\n          description: '指定环境变量文件路径'\n        },\n        {\n          flags: '-d, --debug',\n          description: '启用调试模式'\n        },\n        {\n          flags: '-s, --stream',\n          description: '启用流式输出模式（默认开启）',\n          defaultValue: true\n        }\n      ],\n      action: executeChat\n    }\n  ]\n};\n","import { createAgent as createAgentCore } from '../core/agentService';\nimport type { Agent, AgentConfig } from '../types';\n\n/**\n * 创建Agent实例\n *\n * 基于提供的配置创建一个符合Agent接口的实例，用于与LLM交互。\n * 使用闭包模式封装内部状态，提供简洁的交互接口。\n *\n * @param config Agent配置信息\n * @returns 符合Agent接口的实例\n */\nexport function createAgent(config: AgentConfig): Agent {\n  // 委托给agentService创建Agent实例\n  return createAgentCore(config);\n}\n","import type { Agent, AgentConfig, ChatInput, ChatOutput, Content } from '../types';\nimport { AgentError, AgentErrorType } from '../types';\n\nimport { AgentRunner } from './AgentRunner';\nimport { createClient } from './llm/llmFactory';\nimport { InMemoryAgentSession } from './session/InMemoryAgentSession';\n\n/**\n * 创建符合Agent接口的实例\n *\n * @param config Agent配置\n * @returns Agent实例\n */\nexport function createAgent(config: AgentConfig): Agent {\n  // 创建LLM客户端\n  const llmClient = createClient(config.llm);\n\n  // 创建会话管理器\n  const session = new InMemoryAgentSession();\n\n  // 创建AgentRunner实例\n  const runner = new AgentRunner(config, llmClient, session);\n\n  // 返回符合Agent接口的对象，使用闭包模式\n  return {\n    chat: (input: string | ChatInput) => handleChat(runner, input),\n    chatStream: (input: string | ChatInput) => handleChatStream(runner, input)\n  };\n}\n\n/**\n * 处理聊天请求\n */\nasync function handleChat(runner: AgentRunner, input: string | ChatInput): Promise<string> {\n  try {\n    // 标准化输入为ChatInput\n    const chatInput = normalizeChatInput(input);\n\n    // 发送消息并获取响应\n    const response = await runner.sendMessage(chatInput, false);\n\n    // 确保响应不是异步迭代器\n    if (Symbol.asyncIterator in response) {\n      throw new AgentError(\n        '意外收到流式响应',\n        AgentErrorType.UNKNOWN,\n        'UNEXPECTED_STREAM_RESPONSE'\n      );\n    }\n\n    // 提取回复中的文本内容\n    return extractTextFromContent(response.content);\n  } catch (error: unknown) {\n    // 已经是AgentError则直接抛出\n    if (error instanceof AgentError) {\n      throw error;\n    }\n\n    // 否则包装为AgentError\n    throw new AgentError(\n      `聊天请求处理失败: ${error instanceof Error ? error.message : String(error)}`,\n      AgentErrorType.UNKNOWN,\n      'CHAT_PROCESSING_ERROR',\n      error instanceof Error ? error : new Error(String(error))\n    );\n  }\n}\n\n/**\n * 处理流式聊天请求\n */\nasync function* handleChatStream(runner: AgentRunner, input: string | ChatInput): AsyncIterable<string> {\n  // 标准化输入为ChatInput\n  const chatInput = normalizeChatInput(input);\n\n  try {\n    // 发送消息并获取流式响应\n    const responseStream = await runner.sendMessage(chatInput, true);\n\n    // 确保响应是异步迭代器\n    if (!(Symbol.asyncIterator in responseStream)) {\n      // 如果收到单一响应而非流，也将其作为单个块返回\n      yield extractTextFromContent((responseStream as ChatOutput).content);\n\n      return;\n    }\n\n    // 处理流式响应\n    for await (const chunk of responseStream as AsyncIterable<ChatOutput>) {\n      // 提取每个块中的文本内容\n      const textContent = extractTextFromContent(chunk.content);\n\n      if (textContent) {\n        yield textContent;\n      }\n    }\n  } catch (error: unknown) {\n    // 已经是AgentError则直接抛出\n    if (error instanceof AgentError) {\n      throw error;\n    }\n\n    // 否则包装为AgentError\n    throw new AgentError(\n      `流式聊天请求处理失败: ${error instanceof Error ? error.message : String(error)}`,\n      AgentErrorType.UNKNOWN,\n      'STREAM_PROCESSING_ERROR',\n      error instanceof Error ? error : new Error(String(error))\n    );\n  }\n}\n\n/**\n * 将字符串或ChatInput标准化为ChatInput\n */\nfunction normalizeChatInput(input: string | ChatInput): ChatInput {\n  if (typeof input === 'string') {\n    return {\n      content: {\n        type: 'text',\n        value: input\n      }\n    };\n  }\n\n  return input;\n}\n\n/**\n * 从内容中提取文本\n */\nfunction extractTextFromContent(content: Content): string {\n  if (Array.isArray(content)) {\n    // 寻找数组中的第一个文本内容\n    const textItem = content.find(item => item.type === 'text');\n\n    return textItem ? textItem.value as string : '';\n  } else if (content.type === 'text') {\n    // 直接返回文本内容\n    return content.value as string;\n  }\n\n  return '';\n}\n","/**\n * Agent模块类型定义\n *\n * 统一导出所有公共类型\n */\nexport type { Agent } from './Agent';\nexport type { AgentConfig } from './AgentConfig';\nexport type { LLMConfig } from './LLMConfig';\nexport type { Content, ContentItem, ContentType } from './Content';\nexport type { ChatInput, ChatOutput } from './Chat';\nexport { AgentError, AgentErrorType } from './errors';\n","/**\n * Agent错误类型\n */\nexport enum AgentErrorType {\n  /**\n   * 配置错误\n   */\n  CONFIG = 'CONFIG',\n\n  /**\n   * LLM服务调用错误\n   */\n  LLM_SERVICE = 'LLM_SERVICE',\n\n  /**\n   * 内容处理错误\n   */\n  CONTENT = 'CONTENT',\n\n  /**\n   * 会话错误\n   */\n  SESSION = 'SESSION',\n\n  /**\n   * 未知错误\n   */\n  UNKNOWN = 'UNKNOWN'\n}\n\n/**\n * Agent错误\n *\n * 统一的错误类，用于处理Agent模块中的所有错误。\n */\nexport class AgentError extends Error {\n  /**\n   * 错误类型\n   */\n  readonly type: AgentErrorType;\n\n  /**\n   * 错误码\n   */\n  readonly code: string;\n\n  /**\n   * 原始错误\n   */\n  readonly cause?: Error;\n\n  /**\n   * 创建Agent错误\n   *\n   * @param message 错误消息\n   * @param type 错误类型\n   * @param code 错误码\n   * @param cause 原始错误\n   */\n  constructor(\n    message: string,\n    type: AgentErrorType = AgentErrorType.UNKNOWN,\n    code: string = 'AGENT_ERROR',\n    cause?: Error\n  ) {\n    super(message);\n    this.name = 'AgentError';\n    this.type = type;\n    this.code = code;\n    this.cause = cause;\n  }\n}\n","import type { AgentConfig, ChatInput, ChatOutput } from '../types';\n\nimport type { LLMClient } from './llm/LLMClient';\nimport type { AgentSession } from './session/AgentSession';\nimport type { Message } from './types';\n\n/**\n * Agent运行器\n *\n * 负责处理消息发送和接收的核心类。\n */\nexport class AgentRunner {\n  private config: AgentConfig;\n  private llmClient: LLMClient;\n  private session: AgentSession;\n\n  constructor(config: AgentConfig, llmClient: LLMClient, session: AgentSession) {\n    this.config = config;\n    this.llmClient = llmClient;\n    this.session = session;\n  }\n\n  /**\n   * 发送消息并获取响应\n   *\n   * @param input 输入内容\n   * @param stream 是否使用流式响应\n   * @returns 响应内容或流式响应迭代器\n   */\n  public sendMessage(input: ChatInput, stream: boolean): Promise<ChatOutput | AsyncIterable<ChatOutput>> {\n    // 创建用户消息\n    const userMessage: Message = {\n      role: 'user',\n      content: input.content\n    };\n\n    // 添加到会话历史\n    this.session.addMessage(userMessage);\n\n    // 准备发送给LLM的消息列表\n    const messages = this.prepareMessages();\n\n    // 发送消息给LLM客户端\n    return this.llmClient.sendMessages(messages, stream);\n  }\n\n  /**\n   * 准备发送给LLM的消息列表\n   *\n   * @returns 按照正确顺序排列的消息列表\n   */\n  private prepareMessages(): Message[] {\n    const messages: Message[] = [];\n\n    // 添加系统提示（总是在第一位）\n    if (this.config.prompt) {\n      messages.push({\n        role: 'system',\n        content: {\n          type: 'text',\n          value: this.config.prompt\n        }\n      });\n    }\n\n    // 添加所有历史消息\n    // 在测试中历史消息包括当前用户消息，所以不需要专门分离处理\n    const historyMessages = this.session.getMessages();\n\n    if (historyMessages.length > 0) {\n      messages.push(...historyMessages);\n    }\n\n    return messages;\n  }\n}\n","import { AgentError, AgentErrorType } from '../../types';\nimport type { LLMConfig } from '../../types';\n\nimport { AnthropicClient } from './AnthropicClient';\nimport type { LLMClient } from './LLMClient';\nimport { OpenAIClient } from './OpenAIClient';\n\n/**\n * 验证LLM配置的基本有效性\n * @param config LLM配置\n */\nfunction validateConfig(config: LLMConfig): void {\n  // 验证模型名称\n  if (!config.model) {\n    throw new AgentError(\n      '缺少必要的模型名称参数',\n      AgentErrorType.CONFIG,\n      'MISSING_MODEL_NAME'\n    );\n  }\n\n  // 验证OpenAI配置\n  if (config.apiType.toLowerCase() === 'openai' && !config.apiKey) {\n    throw new AgentError(\n      'OpenAI API密钥未提供',\n      AgentErrorType.CONFIG,\n      'MISSING_API_KEY'\n    );\n  }\n\n  // 验证Anthropic配置\n  if (config.apiType.toLowerCase() === 'anthropic' && !config.apiKey) {\n    throw new AgentError(\n      'Anthropic API密钥未提供',\n      AgentErrorType.CONFIG,\n      'MISSING_API_KEY'\n    );\n  }\n}\n\n/**\n * 创建LLM客户端实例\n *\n * @param config LLM配置\n * @returns LLM客户端实例\n */\nexport function createClient(config: LLMConfig): LLMClient {\n  // 首先验证配置的基本有效性\n  validateConfig(config);\n\n  switch (config.apiType.toLowerCase()) {\n    case 'openai':\n      return new OpenAIClient(config);\n    case 'anthropic':\n      return new AnthropicClient(config);\n    default:\n      throw new AgentError(\n        `不支持的API类型: ${config.apiType}`,\n        AgentErrorType.CONFIG,\n        'UNSUPPORTED_LLM_TYPE'\n      );\n  }\n}\n","import type { ChatOutput } from '../../types/Chat';\nimport type { ContentItem } from '../../types/Content';\nimport { AgentError, AgentErrorType } from '../../types/errors';\nimport type { LLMConfig } from '../../types/LLMConfig';\nimport type { Message } from '../types';\n\nimport type { LLMClient } from './LLMClient';\n\n/**\n * Anthropic客户端实现\n */\nexport class AnthropicClient implements LLMClient {\n  private apiKey: string;\n  private apiUrl: string;\n  private model: string;\n\n  constructor(config: LLMConfig) {\n    this.apiKey = config.apiKey || '';\n    this.apiUrl = config.apiUrl || 'https://api.anthropic.com/v1';\n    this.model = config.model;\n\n    if (!this.apiKey) {\n      throw new AgentError(\n        'Anthropic API密钥未提供',\n        AgentErrorType.CONFIG,\n        'MISSING_API_KEY'\n      );\n    }\n  }\n\n  public async sendMessages(messages: Message[], stream: boolean): Promise<ChatOutput | AsyncIterable<ChatOutput>> {\n    try {\n      // 转换为Anthropic消息格式\n      const anthropicMessages = this.convertToAnthropicMessages(messages);\n\n      // 选择同步或流式API\n      if (stream) {\n        return this.streamMessages(anthropicMessages);\n      } else {\n        return this.sendMessagesSync(anthropicMessages);\n      }\n    } catch (error: unknown) {\n      // 将底层错误包装为AgentError\n      throw new AgentError(\n        `LLM服务调用失败: ${error instanceof Error ? error.message : String(error)}`,\n        AgentErrorType.LLM_SERVICE,\n        'LLM_API_ERROR',\n        error instanceof Error ? error : new Error(String(error))\n      );\n    }\n  }\n\n  private async sendMessagesSync(anthropicMessages: Record<string, unknown>): Promise<ChatOutput> {\n    // 准备请求参数\n    const requestBody = {\n      model: this.model,\n      messages: anthropicMessages.messages,\n      max_tokens: 1024,\n      stream: false\n    };\n\n    // 发送请求到Anthropic API\n    const response = await fetch(`${this.apiUrl}/messages`, {\n      method: 'POST',\n      headers: {\n        'Content-Type': 'application/json',\n        'X-API-Key': this.apiKey,\n        'anthropic-version': '2023-06-01'\n      },\n      body: JSON.stringify(requestBody)\n    });\n\n    // 检查响应状态\n    if (!response.ok) {\n      const errorData = await response.text();\n\n      throw new Error(`Anthropic API请求失败: ${response.status} ${response.statusText} - ${errorData}`);\n    }\n\n    // 解析响应\n    const data = await response.json();\n    const content = data.content?.[0]?.text;\n\n    if (!content) {\n      throw new Error('Anthropic响应格式无效，没有找到内容');\n    }\n\n    // 返回标准化的输出\n    return {\n      content: {\n        type: 'text',\n        value: content\n      }\n    };\n  }\n\n  private async *streamMessages(anthropicMessages: Record<string, unknown>): AsyncIterable<ChatOutput> {\n    // 准备请求参数\n    const requestBody = {\n      model: this.model,\n      messages: anthropicMessages.messages,\n      max_tokens: 1024,\n      stream: true\n    };\n\n    // 发送请求到Anthropic API\n    const response = await fetch(`${this.apiUrl}/messages`, {\n      method: 'POST',\n      headers: {\n        'Content-Type': 'application/json',\n        'X-API-Key': this.apiKey,\n        'anthropic-version': '2023-06-01'\n      },\n      body: JSON.stringify(requestBody)\n    });\n\n    // 检查响应状态\n    if (!response.ok) {\n      const errorData = await response.text();\n\n      throw new Error(`Anthropic API流式请求失败: ${response.status} ${response.statusText} - ${errorData}`);\n    }\n\n    if (!response.body) {\n      throw new Error('Anthropic API流式响应没有响应体');\n    }\n\n    // 创建流式读取器\n    const reader = response.body.getReader();\n    const decoder = new TextDecoder();\n    let buffer = '';\n\n    try {\n      while (true) {\n        const { done, value } = await reader.read();\n\n        if (done) break;\n\n        // 解码新接收的数据\n        buffer += decoder.decode(value, { stream: true });\n\n        // 处理接收到的事件\n        const lines = buffer.split('\\n');\n\n        buffer = lines.pop() || '';\n\n        for (const line of lines) {\n          if (line.startsWith('data: ') && line !== 'data: [DONE]') {\n            const jsonData = line.substring(6); // 去掉 'data: ' 前缀\n\n            try {\n              const data = JSON.parse(jsonData);\n\n              if (data.type === 'content_block_delta' && data.delta?.text) {\n                yield {\n                  content: {\n                    type: 'text',\n                    value: data.delta.text\n                  }\n                };\n              }\n            } catch (e) {\n              console.error('解析SSE事件失败:', e);\n            }\n          }\n        }\n      }\n    } finally {\n      reader.releaseLock();\n    }\n  }\n\n  private convertToAnthropicMessages(messages: Message[]): Record<string, unknown> {\n    // Anthropic格式要求系统提示和用户消息分开处理\n    const systemMessage = messages.find(msg => msg.role === 'system');\n\n    // 过滤掉系统消息，只保留用户和助手消息\n    const userAssistantMessages = messages\n      .filter(msg => msg.role !== 'system')\n      .map(msg => ({\n        role: msg.role,\n        content: this.convertContent(msg.content)\n      }));\n\n    // 构建Anthropic请求格式\n    return {\n      system: systemMessage ? this.extractTextContent(systemMessage.content) : '',\n      messages: userAssistantMessages\n    };\n  }\n\n  private convertContent(content: ContentItem | ContentItem[]): unknown {\n    // Anthropic消息内容处理\n    if (Array.isArray(content)) {\n      // 暂时只支持文本内容，将数组中的文本合并\n      return content\n        .filter(item => item.type === 'text')\n        .map(item => item.value)\n        .join('\\n');\n    } else if (content.type === 'text') {\n      return content.value;\n    }\n\n    // 不支持的内容类型\n    return `[不支持的内容类型: ${Array.isArray(content) ? '多模态内容' : content.type}]`;\n  }\n\n  private extractTextContent(content: ContentItem | ContentItem[]): string {\n    if (Array.isArray(content)) {\n      return content\n        .filter(item => item.type === 'text')\n        .map(item => item.value as string)\n        .join('\\n');\n    } else if (content.type === 'text') {\n      return content.value as string;\n    }\n\n    return '';\n  }\n}\n","import type { ChatOutput } from '../../types/Chat';\nimport type { ContentItem } from '../../types/Content';\nimport { AgentError, AgentErrorType } from '../../types/errors';\nimport type { LLMConfig } from '../../types/LLMConfig';\nimport type { Message } from '../types';\n\nimport type { LLMClient } from './LLMClient';\n\n/**\n * OpenAI客户端实现\n */\nexport class OpenAIClient implements LLMClient {\n  private apiKey: string;\n  private apiUrl: string;\n  private model: string;\n\n  constructor(config: LLMConfig) {\n    this.apiKey = config.apiKey || '';\n    this.apiUrl = config.apiUrl || 'https://api.openai.com/v1';\n    this.model = config.model;\n\n    if (!this.apiKey) {\n      throw new AgentError(\n        'OpenAI API密钥未提供',\n        AgentErrorType.CONFIG,\n        'MISSING_API_KEY'\n      );\n    }\n  }\n\n  public async sendMessages(messages: Message[], stream: boolean): Promise<ChatOutput | AsyncIterable<ChatOutput>> {\n    try {\n      // 转换为OpenAI格式的消息\n      const openaiMessages = this.convertToOpenAIMessages(messages);\n\n      // 选择同步或流式API\n      if (stream) {\n        return this.streamMessages(openaiMessages);\n      } else {\n        return this.sendMessagesSync(openaiMessages);\n      }\n    } catch (error: unknown) {\n      // 将底层错误包装为AgentError\n      // 如果已经是AgentError类型，则直接抛出\n      if (error instanceof AgentError) {\n        throw error;\n      }\n\n      throw new AgentError(\n        `LLM服务调用失败: ${error instanceof Error ? error.message : String(error)}`,\n        AgentErrorType.LLM_SERVICE,\n        'LLM_API_ERROR',\n        error instanceof Error ? error : new Error(String(error))\n      );\n    }\n  }\n\n  private async sendMessagesSync(openaiMessages: Record<string, unknown>[]): Promise<ChatOutput> {\n    try {\n      // 准备请求参数\n      const requestBody = {\n        model: this.model,\n        messages: openaiMessages,\n        stream: false\n      };\n\n      // 发送请求到OpenAI API\n      const response = await fetch(`${this.apiUrl}/chat/completions`, {\n        method: 'POST',\n        headers: {\n          'Content-Type': 'application/json',\n          'Authorization': `Bearer ${this.apiKey}`\n        },\n        body: JSON.stringify(requestBody)\n      });\n\n      // 检查响应状态\n      if (!response.ok) {\n        const errorData = await response.text();\n\n        throw new AgentError(\n          `OpenAI API请求失败: ${response.status} ${response.statusText} - ${errorData}`,\n          AgentErrorType.LLM_SERVICE,\n          'LLM_API_ERROR'\n        );\n      }\n\n      // 解析响应\n      const data = await response.json();\n      const content = data.choices[0]?.message?.content;\n\n      if (!content) {\n        throw new AgentError(\n          'OpenAI响应格式无效，没有找到内容',\n          AgentErrorType.LLM_SERVICE,\n          'INVALID_RESPONSE_FORMAT'\n        );\n      }\n\n      // 返回标准化的输出\n      return {\n        content: {\n          type: 'text',\n          value: content\n        }\n      };\n    } catch (error) {\n      // 确保所有错误都被包装成AgentError\n      if (error instanceof AgentError) {\n        throw error;\n      }\n\n      throw new AgentError(\n        `LLM服务调用失败: ${error instanceof Error ? error.message : String(error)}`,\n        AgentErrorType.LLM_SERVICE,\n        'LLM_API_ERROR',\n        error instanceof Error ? error : new Error(String(error))\n      );\n    }\n  }\n\n  private async *streamMessages(openaiMessages: Record<string, unknown>[]): AsyncIterable<ChatOutput> {\n    try {\n      // 准备请求参数\n      const requestBody = {\n        model: this.model,\n        messages: openaiMessages,\n        stream: true\n      };\n\n      // 发送请求到OpenAI API\n      const response = await fetch(`${this.apiUrl}/chat/completions`, {\n        method: 'POST',\n        headers: {\n          'Content-Type': 'application/json',\n          'Authorization': `Bearer ${this.apiKey}`\n        },\n        body: JSON.stringify(requestBody)\n      });\n\n      // 检查响应状态\n      if (!response.ok) {\n        const errorData = await response.text();\n\n        throw new AgentError(\n          `OpenAI API流式请求失败: ${response.status} ${response.statusText} - ${errorData}`,\n          AgentErrorType.LLM_SERVICE,\n          'LLM_API_ERROR'\n        );\n      }\n\n      if (!response.body) {\n        throw new AgentError(\n          'OpenAI API流式响应没有响应体',\n          AgentErrorType.LLM_SERVICE,\n          'MISSING_RESPONSE_BODY'\n        );\n      }\n\n      // 创建流式读取器\n      const reader = response.body.getReader();\n      const decoder = new TextDecoder();\n      let buffer = '';\n\n      try {\n        while (true) {\n          const { done, value } = await reader.read();\n\n          if (done) break;\n\n          // 解码新接收的数据\n          buffer += decoder.decode(value, { stream: true });\n\n          // 处理接收到的事件\n          const lines = buffer.split('\\n');\n\n          buffer = lines.pop() || '';\n\n          for (const line of lines) {\n            if (line.startsWith('data: ') && line !== 'data: [DONE]') {\n              const jsonData = line.substring(6); // 去掉 'data: ' 前缀\n\n              try {\n                const data = JSON.parse(jsonData);\n                const content = data.choices[0]?.delta?.content;\n\n                if (content) {\n                  yield {\n                    content: {\n                      type: 'text',\n                      value: content\n                    }\n                  };\n                }\n              } catch (e) {\n                console.error('解析SSE事件失败:', e);\n                throw new AgentError(\n                  `解析SSE事件失败: ${e instanceof Error ? e.message : String(e)}`,\n                  AgentErrorType.LLM_SERVICE,\n                  'SSE_PARSING_ERROR',\n                  e instanceof Error ? e : new Error(String(e))\n                );\n              }\n            }\n          }\n        }\n      } finally {\n        reader.releaseLock();\n      }\n    } catch (error) {\n      // 确保所有错误都被包装成AgentError\n      if (error instanceof AgentError) {\n        throw error;\n      }\n\n      throw new AgentError(\n        `LLM服务调用失败: ${error instanceof Error ? error.message : String(error)}`,\n        AgentErrorType.LLM_SERVICE,\n        'LLM_API_ERROR',\n        error instanceof Error ? error : new Error(String(error))\n      );\n    }\n  }\n\n  private convertToOpenAIMessages(messages: Message[]): Record<string, unknown>[] {\n    // 转换消息格式为OpenAI API格式\n    return messages.map(msg => {\n      return {\n        role: msg.role,\n        content: this.convertContent(msg.content)\n      };\n    });\n  }\n\n  private convertContent(content: ContentItem | ContentItem[]): unknown {\n    // 处理多模态内容转换为OpenAI格式\n    if (Array.isArray(content)) {\n      // OpenAI格式支持内容数组\n      return content.map(item => this.convertContentItem(item));\n    }\n\n    return this.convertContentItem(content);\n  }\n\n  private convertContentItem(item: ContentItem): unknown {\n    switch (item.type) {\n      case 'text':\n        return item.value as string;\n      case 'image':\n        // 处理图像类型，转换为OpenAI的图像URL格式\n        return {\n          type: 'image_url',\n          image_url: {\n            url: `data:${item.mimeType || 'image/jpeg'};base64,${this.arrayBufferToBase64(item.value as Uint8Array)}`\n          }\n        };\n      case 'audio':\n      case 'video':\n      case 'file':\n      default:\n        // 暂不支持其他内容类型，转换为文本提示\n        return `[不支持的内容类型: ${item.type}]`;\n    }\n  }\n\n  private arrayBufferToBase64(buffer: Uint8Array): string {\n    // 将Uint8Array转换为Base64编码的字符串\n    if (typeof Buffer !== 'undefined') {\n      // Node.js环境\n      return Buffer.from(buffer).toString('base64');\n    } else {\n      // 浏览器环境\n      let binary = '';\n      const len = buffer.byteLength;\n\n      for (let i = 0; i < len; i++) {\n        binary += String.fromCharCode(buffer[i]);\n      }\n\n      // 使用btoa函数(浏览器内置)进行Base64编码\n      return btoa(binary);\n    }\n  }\n}\n","import type { Message } from '../types';\n\nimport type { AgentSession } from './AgentSession';\n\n/**\n * 内存会话实现\n */\nexport class InMemoryAgentSession implements AgentSession {\n  private messages: Message[] = [];\n  private capacity: number;\n\n  constructor(capacity: number = 100) {\n    this.capacity = capacity;\n  }\n\n  public addMessage(message: Message): void {\n    this.messages.push(message);\n\n    // 如果超出容量，删除最早的消息\n    if (this.messages.length > this.capacity) {\n      this.messages.shift();\n    }\n  }\n\n  public getMessages(): ReadonlyArray<Message> {\n    return [...this.messages];\n  }\n}\n","import { replaceEnvVars as replaceEnvVarsCore } from '../core/agentenv/agentenvCore';\n\n/**\n * 替换值中的环境变量引用\n *\n * 识别并替换字符串、数组或对象中的所有 @agentenv:ENV_NAME 格式的环境变量引用。\n * 支持任意嵌套的数据结构。\n *\n * @param value 包含环境变量引用的值（字符串、数组或对象）\n * @returns 替换后的值，保持原始类型\n * @example\n * // 替换字符串\n * const apiKey = replaceEnvVars('@agentenv:API_KEY');\n *\n * // 替换对象中的值\n * const config = replaceEnvVars({\n *   apiKey: '@agentenv:API_KEY',\n *   endpoint: '@agentenv:API_ENDPOINT'\n * });\n */\nexport function replaceEnvVars<T>(value: T): T {\n  return replaceEnvVarsCore(value);\n}\n","import { ENV_VAR_PATTERN } from './constants';\n\n/**\n * 替换值中的环境变量引用（Core实现）\n */\nexport function replaceEnvVars<T>(value: T): T {\n  // 处理null/undefined\n  if (value === null || value === undefined) {\n    return value;\n  }\n\n  // 处理字符串\n  if (typeof value === 'string') {\n    return replaceInString(value) as unknown as T;\n  }\n\n  // 处理数组\n  if (Array.isArray(value)) {\n    return value.map(item => replaceEnvVars(item)) as unknown as T;\n  }\n\n  // 处理对象\n  if (typeof value === 'object') {\n    const result: Record<string, any> = {};\n\n    for (const key in value as Record<string, any>) {\n      result[key] = replaceEnvVars((value as Record<string, any>)[key]);\n    }\n\n    return result as T;\n  }\n\n  // 其他类型直接返回\n  return value;\n}\n\n/**\n * 替换字符串中的环境变量引用\n */\nfunction replaceInString(value: string): string {\n  return value.replace(ENV_VAR_PATTERN, (_match, envName) => {\n    const envValue = process.env[envName];\n\n    if (envValue === undefined) {\n      console.warn(`警告: 环境变量 ${envName} 未定义`);\n\n      return _match; // 保留原始表达式\n    }\n\n    return envValue;\n  });\n}\n","/**\n * 环境变量引用的正则表达式模式\n * 匹配 @agentenv:ENV_NAME 格式\n */\nexport const ENV_VAR_PATTERN = /@agentenv:([A-Z0-9_]+)/g;\n\n/**\n * 环境变量引用前缀\n */\nexport const ENV_VAR_PREFIX = '@agentenv:';\n","/**\n * Agent模块的API层入口\n *\n * 导出所有公共API\n */\nexport * from './agent';\n","#!/usr/bin/env node\n/**\n * DPML Agent CLI工具\n *\n * 提供命令行接口，用于管理和使用通过DPML定义的Agent。\n * 基于@dpml/core的CLI基础设施实现。\n *\n * 使用示例:\n * ```bash\n * # 验证Agent配置\n * dpml agent validate agent-config.xml\n *\n * # 与Agent交互\n * dpml agent chat agent-config.xml\n *\n * # 使用环境变量\n * dpml agent chat agent-config.xml --env API_KEY=sk-xxxx\n * ```\n */\n\nimport { agentDPML } from './index';\n\n/**\n * CLI主函数\n */\nasync function main() {\n  try {\n    // 执行CLI命令\n    await agentDPML.cli.execute();\n  } catch (error) {\n    console.error('CLI执行出错:', error instanceof Error ? error.message : String(error));\n    process.exit(1);\n  }\n}\n\n// 运行主函数\nmain();\n"],"mappings":";;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;AAAA;;;;;;;ACAA;AAAA;AAAA;AAAA,MACE,MAAQ;AAAA,MACR,SAAW;AAAA,MACX,aAAe;AAAA,MACf,MAAQ;AAAA,MACR,OAAS;AAAA,MACT,SAAW;AAAA,QACT,KAAK;AAAA,UACH,OAAS;AAAA,UACT,SAAW;AAAA,UACX,SAAW;AAAA,QACb;AAAA,QACA,YAAY;AAAA,QACZ,eAAe;AAAA,QACf,qBAAqB;AAAA,QACrB,wBAAwB;AAAA,QACxB,qBAAqB;AAAA,QACrB,wBAAwB;AAAA,QACxB,kBAAkB;AAAA,MACpB;AAAA,MACA,SAAW;AAAA,QACT,aAAa;AAAA,QACb,MAAQ;AAAA,QACR,SAAW;AAAA,QACX,MAAQ;AAAA,QACR,iBAAiB;AAAA,QACjB,YAAc;AAAA,QACd,SAAW;AAAA,MACb;AAAA,MACA,YAAc;AAAA,QACZ,MAAQ;AAAA,QACR,KAAO;AAAA,MACT;AAAA,MACA,UAAY;AAAA,MACZ,SAAW;AAAA,MACX,UAAY;AAAA,QACV;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,QACA;AAAA,MACF;AAAA,MACA,gBAAkB;AAAA,MAClB,SAAW;AAAA,MACX,iBAAmB;AAAA,QACjB,eAAe;AAAA,QACf,SAAW;AAAA,QACX,OAAS;AAAA,QACT,UAAY;AAAA,QACZ,oBAAoB;AAAA,QACpB,KAAO;AAAA,QACP,YAAc;AAAA,MAChB;AAAA,MACA,SAAW;AAAA,QACT,MAAQ;AAAA,MACV;AAAA,MACA,SAAW;AAAA,QACT,IAAM;AAAA,MACR;AAAA,IACF;AAAA;AAAA;;;AC7DA;;;;QAAMA,MAAKC,UAAQ,IAAA;AACnB,QAAMC,QAAOD,UAAQ,MAAA;AACrB,QAAME,KAAKF,UAAQ,IAAA;AACnB,QAAMG,SAASH,UAAQ,QAAA;AACvB,QAAMI,cAAcJ;AAEpB,QAAMK,UAAUD,YAAYC;AAE5B,QAAMC,OAAO;AAGb,aAASC,MAAOC,KAAG;AACjB,YAAMC,MAAM,CAAC;AAGb,UAAIC,QAAQF,IAAIG,SAAQ;AAGxBD,cAAQA,MAAME,QAAQ,WAAW,IAAA;AAEjC,UAAIC;AACJ,cAAQA,QAAQP,KAAKQ,KAAKJ,KAAAA,MAAW,MAAM;AACzC,cAAMK,MAAMF,MAAM,CAAA;AAGlB,YAAIG,QAASH,MAAM,CAAA,KAAM;AAGzBG,gBAAQA,MAAMC,KAAI;AAGlB,cAAMC,aAAaF,MAAM,CAAA;AAGzBA,gBAAQA,MAAMJ,QAAQ,0BAA0B,IAAA;AAGhD,YAAIM,eAAe,KAAK;AACtBF,kBAAQA,MAAMJ,QAAQ,QAAQ,IAAA;AAC9BI,kBAAQA,MAAMJ,QAAQ,QAAQ,IAAA;QAChC;AAGAH,YAAIM,GAAAA,IAAOC;MACb;AAEA,aAAOP;IACT;AApCSF;AAsCT,aAASY,YAAaC,SAAO;AAC3B,YAAMC,YAAYC,WAAWF,OAAAA;AAG7B,YAAMG,SAASC,aAAaC,aAAa;QAAExB,MAAMoB;MAAU,CAAA;AAC3D,UAAI,CAACE,OAAOG,QAAQ;AAClB,cAAMC,MAAM,IAAIC,MAAM,8BAA8BP,SAAAA,wBAAiC;AACrFM,YAAIE,OAAO;AACX,cAAMF;MACR;AAIA,YAAMG,OAAOC,WAAWX,OAAAA,EAASY,MAAM,GAAA;AACvC,YAAMC,SAASH,KAAKG;AAEpB,UAAIC;AACJ,eAASC,IAAI,GAAGA,IAAIF,QAAQE,KAAK;AAC/B,YAAI;AAEF,gBAAMpB,MAAMe,KAAKK,CAAAA,EAAGlB,KAAI;AAGxB,gBAAMmB,QAAQC,cAAcd,QAAQR,GAAAA;AAGpCmB,sBAAYV,aAAac,QAAQF,MAAMG,YAAYH,MAAMrB,GAAG;AAE5D;QACF,SAASyB,OAAO;AAEd,cAAIL,IAAI,KAAKF,QAAQ;AACnB,kBAAMO;UACR;QAEF;MACF;AAGA,aAAOhB,aAAajB,MAAM2B,SAAAA;IAC5B;AAxCSf;AA0CT,aAASsB,MAAOC,SAAO;AACrBC,cAAQC,IAAI,WAAWvC,OAAAA