p-api-agent 0.0.1

package/README.md

@@ -0,0 +1,309 @@
+# P-API-Agent
+
+一个基于 LLM 的智能 Agent 框架，支持工具函数调用和多轮对话。
+
+## 特性
+
+- 🤖 **智能对话**: 基于 LLM 的智能对话能力
+- 🔧 **工具调用**: 支持动态注册和调用工具函数
+- 📝 **类型安全**: 完整的 TypeScript 类型定义
+- 🛡️ **错误处理**: 完善的错误处理和日志系统
+- ⚙️ **可配置**: 灵活的配置选项
+- 📊 **日志系统**: 多级别日志输出
+
+## 安装
+
+```bash
+npm install p-api-agent
+```
+
+## 快速开始
+
+### 基础用法
+
+```typescript
+import { Agent, FunctionCall } from 'p-api-agent';
+
+// 创建 Agent 实例
+const agent = new Agent({
+    maxLoop: 20,      // 最大循环次数
+    debug: false      // 调试模式
+});
+
+// 注册 LLM 聊天能力
+agent.register_llm_text_ability(async (messages) => {
+    // 调用您的 LLM API
+    const response = await yourLLMAPI(messages);
+    return response;
+});
+
+// 注册工具函数（可选）
+const functionCall = new FunctionCall('/path/to/tools');
+agent.register_function_call(functionCall);
+
+// 开始对话
+const result = await agent.create_chat('你好，请帮我完成任务');
+console.log(result);
+```
+
+### 创建工具函数
+
+在工具目录下创建 `.js` 或 `.ts` 文件：
+
+```typescript
+// tools/get_weather.ts
+export function register() {
+    return {
+        name: 'get_weather',
+        description: '获取指定城市的天气信息',
+        input_schema: {
+            type: 'object',
+            properties: {
+                city: {
+                    type: 'string',
+                    description: '城市名称',
+                    examples: ['北京', '上海', '深圳']
+                }
+            },
+            required: ['city']
+        },
+        register_func: async (params: { city: string }) => {
+            // 实现您的工具逻辑
+            return {
+                city: params.city,
+                temperature: 25,
+                weather: '晴天'
+            };
+        }
+    };
+}
+```
+
+### 自定义提示词
+
+```typescript
+const agent = new Agent({
+    customPresetPrompt: `
+你是一个专业的助手...
+（自定义提示词内容）
+    `
+});
+```
+
+## API 文档
+
+### Agent 类
+
+#### 构造函数
+
+```typescript
+new Agent(options?: AgentOptions)
+```
+
+**选项：**
+- `maxLoop?: number` - 最大循环次数（默认：20）
+- `debug?: boolean` - 是否启用调试模式（默认：false）
+- `customPresetPrompt?: string` - 自定义预置提示词
+
+#### 方法
+
+##### register_llm_text_ability(func)
+
+注册 LLM 文字能力
+
+```typescript
+agent.register_llm_text_ability(async (userInput) => {
+    // 返回 LLM 响应
+    return llmResponse;
+});
+```
+
+##### register_function_call(functionCall)
+
+注册工具函数管理器
+
+```typescript
+const functionCall = new FunctionCall('/path/to/tools');
+agent.register_function_call(functionCall);
+```
+
+##### create_chat(userInput)
+
+创建聊天会话
+
+```typescript
+const result = await agent.create_chat('用户输入');
+// 或使用消息数组
+const result = await agent.create_chat([
+    {
+        role: 'user',
+        content: [{ type: 'text', text: '用户输入' }]
+    }
+]);
+```
+
+##### set_max_loop(maxLoop)
+
+设置最大循环次数
+
+```typescript
+agent.set_max_loop(30);
+```
+
+##### set_preset_prompt(prompt)
+
+设置自定义预置提示词
+
+```typescript
+agent.set_preset_prompt('自定义提示词内容');
+```
+
+### FunctionCall 类
+
+#### 构造函数
+
+```typescript
+new FunctionCall(toolPath: string)
+```
+
+**参数：**
+- `toolPath` - 工具函数所在目录的绝对路径
+
+#### 方法
+
+##### get_tools_list()
+
+获取所有工具函数列表
+
+```typescript
+const tools = functionCall.get_tools_list();
+// [{ name: 'tool1', description: '...' }, ...]
+```
+
+##### has_tool(name)
+
+检查工具函数是否存在
+
+```typescript
+const exists = functionCall.has_tool('get_weather');
+```
+
+##### exec_function(name, params)
+
+执行工具函数
+
+```typescript
+const result = await functionCall.exec_function('get_weather', { city: '北京' });
+```
+
+### Logger 工具
+
+```typescript
+import { Logger, LogLevel } from 'p-api-agent';
+
+// 设置日志级别
+Logger.setLevel(LogLevel.DEBUG);
+
+// 使用日志
+Logger.debug('调试信息');
+Logger.info('普通信息');
+Logger.warn('警告信息');
+Logger.error('错误信息');
+```
+
+### 错误类型
+
+```typescript
+import {
+    AgentError,
+    ToolNotFoundError,
+    ToolExecutionError,
+    ValidationError,
+    LLMError,
+    MaxLoopExceededError
+} from 'p-api-agent';
+
+try {
+    await agent.create_chat('...');
+} catch (error) {
+    if (error instanceof ToolNotFoundError) {
+        console.error('工具函数不存在');
+    } else if (error instanceof MaxLoopExceededError) {
+        console.error('超出最大循环次数');
+    }
+}
+```
+
+## 工作原理
+
+1. **用户输入** → Agent 接收用户请求
+2. **LLM 分析** → LLM 判断是否需要调用工具
+3. **工具调用** → 如需要，先获取工具文档，再执行工具
+4. **结果处理** → 整合工具结果，返回给用户
+5. **循环控制** → 最多循环 maxLoop 次，防止无限循环
+
+## LLM 响应格式
+
+LLM 需要返回 JSON 格式的指令：
+
+### 结束指令
+```json
+{
+    "command": "end",
+    "result": "最终返回给用户的结果"
+}
+```
+
+### 获取工具文档
+```json
+{
+    "command": "get_tool_doc",
+    "tool_name": "工具函数名称"
+}
+```
+
+### 调用工具
+```json
+{
+    "command": "use_tool",
+    "tool_name": "工具函数名称",
+    "params": { "参数名": "参数值" }
+}
+```
+
+## 最佳实践
+
+1. **错误处理**: 始终使用 try-catch 包裹 Agent 调用
+2. **日志记录**: 生产环境建议设置为 INFO 或 WARN 级别
+3. **循环次数**: 根据实际需求调整 maxLoop，避免过多循环
+4. **工具设计**: 工具函数应该功能单一、职责明确
+5. **参数验证**: 在工具函数中进行充分的参数验证
+
+## 开发
+
+```bash
+# 安装依赖
+npm install
+
+# 构建
+npm run build
+
+# 测试
+npm test
+```
+
+## 许可证
+
+ISC
+
+## 贡献
+
+欢迎提交 Issue 和 Pull Request！
+
+## 更新日志
+
+### v0.0.1
+- 初始版本发布
+- 支持基本的 Agent 功能
+- 工具函数调用
+- 完善的类型定义和错误处理

package/dist/index.cjs

@@ -0,0 +1,273 @@
+"use strict";
+var __create = Object.create;
+var __defProp = Object.defineProperty;
+var __getOwnPropDesc = Object.getOwnPropertyDescriptor;
+var __getOwnPropNames = Object.getOwnPropertyNames;
+var __getProtoOf = Object.getPrototypeOf;
+var __hasOwnProp = Object.prototype.hasOwnProperty;
+var __export = (target, all) => {
+  for (var name in all)
+    __defProp(target, name, { get: all[name], enumerable: true });
+};
+var __copyProps = (to, from, except, desc) => {
+  if (from && typeof from === "object" || typeof from === "function") {
+    for (let key of __getOwnPropNames(from))
+      if (!__hasOwnProp.call(to, key) && key !== except)
+        __defProp(to, key, { get: () => from[key], enumerable: !(desc = __getOwnPropDesc(from, key)) || desc.enumerable });
+  }
+  return to;
+};
+var __toESM = (mod, isNodeMode, target) => (target = mod != null ? __create(__getProtoOf(mod)) : {}, __copyProps(
+  // If the importer is in node compatibility mode or this is not an ESM
+  // file that has been converted to a CommonJS file using a Babel-
+  // compatible transform (i.e. "__esModule" has not been set), then set
+  // "default" to the CommonJS "module.exports" for node compatibility.
+  isNodeMode || !mod || !mod.__esModule ? __defProp(target, "default", { value: mod, enumerable: true }) : target,
+  mod
+));
+var __toCommonJS = (mod) => __copyProps(__defProp({}, "__esModule", { value: true }), mod);
+
+// src/index.ts
+var index_exports = {};
+__export(index_exports, {
+  Agent: () => Agent,
+  FunctionCall: () => FunctionCall,
+  LLM_Utils: () => LLM_Utils
+});
+module.exports = __toCommonJS(index_exports);
+
+// src/agent/index.ts
+var import_fs = __toESM(require("fs"), 1);
+
+// src/utils/llm_utils.ts
+var import_jsonrepair = require("jsonrepair");
+var _LLM_Utils = class {
+  /**
+   * 提取大模型返回的markdown
+   */
+  extract_markdown(content) {
+    const regex = /```markdown\n(.*?)\n```/s;
+    const match = content.match(regex);
+    if (!match) return content;
+    return match[1];
+  }
+  /**
+   * 提取大模型的json格式
+   */
+  extract_json(content) {
+    const regex = /```json\n(.*?)\n```/s;
+    const match = content.match(regex);
+    let json_content = match ? match[1] : content;
+    try {
+      return JSON.parse(json_content);
+    } catch (err) {
+      return JSON.parse((0, import_jsonrepair.jsonrepair)(json_content));
+    }
+  }
+  /**
+   * 提取mermaid格式
+   */
+  extract_mermaid(content) {
+    const regex = /```mermaid\n(.*?)\n```/s;
+    const match = content.match(regex);
+    if (!match) return content;
+    return match[1];
+  }
+  /**
+   * 提取svg格式
+   */
+  extract_svg(content) {
+    const regex = /```svg\n(.*?)\n```/s;
+    const match = content.match(regex);
+    if (!match) return content;
+    return match[1];
+  }
+  parse_json(content) {
+    if (typeof content === "object") {
+      return content;
+    }
+    try {
+      return JSON.parse(content);
+    } catch (err) {
+      return JSON.parse((0, import_jsonrepair.jsonrepair)(content));
+    }
+  }
+};
+var LLM_Utils = new _LLM_Utils();
+
+// src/agent/index.ts
+var import_path = __toESM(require("path"), 1);
+var Agent = class {
+  // 角色定义的提示词
+  constructor() {
+    this.function_call = null;
+    // 工具函数的说明文档, 供模型调用时参考
+    this.function_call_doc = "";
+    // 模型的聊天能力
+    this.llm_chat_func = null;
+    // 预置提示词
+    this.preset_prompt = this.get_preset_prompt();
+  }
+  /**
+   * 注册工具函数
+   */
+  register_function_call(function_call) {
+    const tools_list = function_call.get_tools_list();
+    this.function_call_doc = tools_list.map((val) => `\u5DE5\u5177\u51FD\u6570:${val.name}
+\u5DE5\u5177\u63CF\u8FF0:${val.description}`).join("\n\n");
+    this.function_call = function_call;
+    return true;
+  }
+  /**
+   * 注册LLM模型文字能力
+   */
+  register_llm_text_ability(func) {
+    this.llm_chat_func = func;
+  }
+  /**
+  * 注册LLM模型图片能力
+  */
+  register_llm_image_ability() {
+    return true;
+  }
+  /**
+   * 创建聊天
+   */
+  async create_chat(user_input) {
+    let dispose_user_input;
+    if (typeof user_input === "string") {
+      dispose_user_input = [
+        {
+          role: "user",
+          content: [
+            { type: "text", text: user_input }
+          ]
+        }
+      ];
+    } else {
+      dispose_user_input = JSON.parse(JSON.stringify(user_input));
+    }
+    if (this.function_call_doc) {
+      dispose_user_input.unshift({
+        role: "system",
+        content: [
+          { type: "text", text: `# \u5DE5\u5177\u51FD\u6570\u8BF4\u660E\u6587\u6863:
+${this.function_call_doc}` }
+        ]
+      });
+    }
+    dispose_user_input.unshift({
+      role: "system",
+      content: [
+        { type: "text", text: this.preset_prompt }
+      ]
+    });
+    let max_loop = 20;
+    let loop_count = 0;
+    while (loop_count < max_loop) {
+      const llm_res_str = await this.llm_chat_func(dispose_user_input);
+      const llm_res = LLM_Utils.extract_json(llm_res_str);
+      if (!llm_res.command) throw "\u5F02\u5E38\u6307\u4EE4, \u672A\u5305\u542Bcommand\u5B57\u6BB5";
+      if (llm_res.command === "end") {
+        return llm_res.result;
+      } else if (llm_res.command === "get_tool_doc") {
+        const tool_name = llm_res.tool_name;
+        const tool_doc = this.function_call.gen_tool_doc(tool_name);
+        dispose_user_input.push({
+          role: "user",
+          content: [
+            { type: "text", text: `\u5DE5\u5177\u51FD\u6570${tool_name}\u7684\u8BF4\u660E\u6587\u6863:
+${tool_doc}` }
+          ]
+        });
+      } else if (llm_res.command === "use_tool") {
+        const tool_name = llm_res.tool_name;
+        const tool_params = LLM_Utils.parse_json(llm_res.params);
+        const tool_result = await this.function_call.exec_function(tool_name, tool_params);
+        dispose_user_input.push({
+          role: "user",
+          content: [
+            { type: "text", text: `\u5DE5\u5177\u51FD\u6570${tool_name}\u7684\u8C03\u7528\u7ED3\u679C:
+${JSON.stringify(tool_result)}` }
+          ]
+        });
+      }
+      loop_count++;
+    }
+  }
+  /**
+   * agent预置提示词
+   */
+  get_preset_prompt() {
+    const prompt = import_fs.default.readFileSync(import_path.default.join(__dirname, "./preset_prompt.md"), "utf-8");
+    return prompt;
+  }
+};
+
+// src/function_call/index.ts
+var import_path2 = __toESM(require("path"), 1);
+var import_fs2 = __toESM(require("fs"), 1);
+var FunctionCall = class {
+  constructor(tool_path) {
+    this.tool_path = tool_path;
+    this.tools = {};
+    this.init();
+  }
+  init() {
+    for (const file_name of import_fs2.default.readdirSync(this.tool_path)) {
+      const route_path = import_path2.default.join(this.tool_path, file_name);
+      const mcp = require(route_path);
+      const info = mcp.register();
+      this.tools[info.name] = info;
+    }
+  }
+  /**
+   * 生成完整的工具函数说明文档
+   */
+  gen_tool_doc(name) {
+    const info = this.tools[name];
+    const schema = info.input_schema;
+    const schemaDesc = `\u7C7B\u578B: ${schema.type}`;
+    const params = Object.keys(schema.properties).map((key) => {
+      const item = schema.properties[key];
+      const required = schema.required?.includes(key) ? "\u5FC5\u586B" : "\u53EF\u9009";
+      const examples = item.examples && item.examples.length > 0 ? `\u793A\u4F8B\u503C: ${item.examples.join(", ")}` : "\u65E0\u793A\u4F8B";
+      return `  - ${key} (${item.type}, ${required})
+    \u8BF4\u660E: ${item.description}
+    ${examples}`;
+    }).join("\n");
+    return `\u5DE5\u5177\u51FD\u6570\u540D\u79F0: ${info.name}
+\u529F\u80FD\u63CF\u8FF0: ${info.description}
+\u8F93\u5165\u7ED3\u6784: ${schemaDesc}
+\u53C2\u6570\u5217\u8868:
+${params}`;
+  }
+  /**
+   * 获取工具函数列表
+   */
+  get_tools_list() {
+    return Object.keys(this.tools).map((key) => {
+      const info = this.tools[key];
+      return {
+        name: info.name,
+        description: info.description
+      };
+    });
+  }
+  /**
+   * 执行工具函数
+   */
+  async exec_function(name, params) {
+    const info = this.tools[name];
+    if (!info) {
+      throw `\u5DE5\u5177\u51FD\u6570${name}\u4E0D\u5B58\u5728`;
+    }
+    return await info.register_func(params);
+  }
+};
+// Annotate the CommonJS export names for ESM import in node:
+0 && (module.exports = {
+  Agent,
+  FunctionCall,
+  LLM_Utils
+});

package/dist/index.d.cts

@@ -0,0 +1,103 @@
+interface RegisterInfo {
+    name: string;
+    description: string;
+    input_schema: {
+        type: string;
+        properties: {
+            [key: string]: {
+                type: string;
+                description: string;
+                examples?: string[];
+            };
+        };
+        required?: string[];
+    };
+    register_func: (...args: any[]) => Promise<any>;
+}
+declare class FunctionCall {
+    tool_path: string;
+    private tools;
+    constructor(tool_path: string);
+    private init;
+    /**
+     * 生成完整的工具函数说明文档
+     */
+    gen_tool_doc(name: string): string;
+    /**
+     * 获取工具函数列表
+     */
+    get_tools_list(): {
+        name: string;
+        description: string;
+    }[];
+    /**
+     * 执行工具函数
+     */
+    exec_function(name: string, params: any): Promise<any>;
+}
+
+interface TextContent {
+    type: 'text';
+    text: string;
+}
+interface ImageContent {
+    type: 'image_url';
+    image_url: {
+        url: string;
+    };
+}
+type Content = TextContent | ImageContent;
+type UserChatInput = string | {
+    role: 'system' | 'user' | 'assistant';
+    content: Content[];
+}[];
+declare class Agent {
+    private function_call;
+    private function_call_doc;
+    private llm_chat_func;
+    preset_prompt: string;
+    constructor();
+    /**
+     * 注册工具函数
+     */
+    register_function_call(function_call: FunctionCall): boolean;
+    /**
+     * 注册LLM模型文字能力
+     */
+    register_llm_text_ability(func: (user_input: UserChatInput) => Promise<string>): void;
+    /**
+    * 注册LLM模型图片能力
+    */
+    register_llm_image_ability(): boolean;
+    /**
+     * 创建聊天
+     */
+    create_chat(user_input: UserChatInput): Promise<any>;
+    /**
+     * agent预置提示词
+     */
+    get_preset_prompt(): string;
+}
+
+declare class _LLM_Utils {
+    /**
+     * 提取大模型返回的markdown
+     */
+    extract_markdown(content: any): any;
+    /**
+     * 提取大模型的json格式
+     */
+    extract_json(content: any): any;
+    /**
+     * 提取mermaid格式
+     */
+    extract_mermaid(content: any): any;
+    /**
+     * 提取svg格式
+     */
+    extract_svg(content: any): any;
+    parse_json(content: string | object): any;
+}
+declare const LLM_Utils: _LLM_Utils;
+
+export { Agent, FunctionCall, LLM_Utils, type RegisterInfo };

package/dist/index.d.ts

@@ -0,0 +1,103 @@
+interface RegisterInfo {
+    name: string;
+    description: string;
+    input_schema: {
+        type: string;
+        properties: {
+            [key: string]: {
+                type: string;
+                description: string;
+                examples?: string[];
+            };
+        };
+        required?: string[];
+    };
+    register_func: (...args: any[]) => Promise<any>;
+}
+declare class FunctionCall {
+    tool_path: string;
+    private tools;
+    constructor(tool_path: string);
+    private init;
+    /**
+     * 生成完整的工具函数说明文档
+     */
+    gen_tool_doc(name: string): string;
+    /**
+     * 获取工具函数列表
+     */
+    get_tools_list(): {
+        name: string;
+        description: string;
+    }[];
+    /**
+     * 执行工具函数
+     */
+    exec_function(name: string, params: any): Promise<any>;
+}
+
+interface TextContent {
+    type: 'text';
+    text: string;
+}
+interface ImageContent {
+    type: 'image_url';
+    image_url: {
+        url: string;
+    };
+}
+type Content = TextContent | ImageContent;
+type UserChatInput = string | {
+    role: 'system' | 'user' | 'assistant';
+    content: Content[];
+}[];
+declare class Agent {
+    private function_call;
+    private function_call_doc;
+    private llm_chat_func;
+    preset_prompt: string;
+    constructor();
+    /**
+     * 注册工具函数
+     */
+    register_function_call(function_call: FunctionCall): boolean;
+    /**
+     * 注册LLM模型文字能力
+     */
+    register_llm_text_ability(func: (user_input: UserChatInput) => Promise<string>): void;
+    /**
+    * 注册LLM模型图片能力
+    */
+    register_llm_image_ability(): boolean;
+    /**
+     * 创建聊天
+     */
+    create_chat(user_input: UserChatInput): Promise<any>;
+    /**
+     * agent预置提示词
+     */
+    get_preset_prompt(): string;
+}
+
+declare class _LLM_Utils {
+    /**
+     * 提取大模型返回的markdown
+     */
+    extract_markdown(content: any): any;
+    /**
+     * 提取大模型的json格式
+     */
+    extract_json(content: any): any;
+    /**
+     * 提取mermaid格式
+     */
+    extract_mermaid(content: any): any;
+    /**
+     * 提取svg格式
+     */
+    extract_svg(content: any): any;
+    parse_json(content: string | object): any;
+}
+declare const LLM_Utils: _LLM_Utils;
+
+export { Agent, FunctionCall, LLM_Utils, type RegisterInfo };

package/dist/index.js

@@ -0,0 +1,241 @@
+var __require = /* @__PURE__ */ ((x) => typeof require !== "undefined" ? require : typeof Proxy !== "undefined" ? new Proxy(x, {
+  get: (a, b) => (typeof require !== "undefined" ? require : a)[b]
+}) : x)(function(x) {
+  if (typeof require !== "undefined") return require.apply(this, arguments);
+  throw Error('Dynamic require of "' + x + '" is not supported');
+});
+
+// src/agent/index.ts
+import fs from "fs";
+
+// src/utils/llm_utils.ts
+import { jsonrepair } from "jsonrepair";
+var _LLM_Utils = class {
+  /**
+   * 提取大模型返回的markdown
+   */
+  extract_markdown(content) {
+    const regex = /```markdown\n(.*?)\n```/s;
+    const match = content.match(regex);
+    if (!match) return content;
+    return match[1];
+  }
+  /**
+   * 提取大模型的json格式
+   */
+  extract_json(content) {
+    const regex = /```json\n(.*?)\n```/s;
+    const match = content.match(regex);
+    let json_content = match ? match[1] : content;
+    try {
+      return JSON.parse(json_content);
+    } catch (err) {
+      return JSON.parse(jsonrepair(json_content));
+    }
+  }
+  /**
+   * 提取mermaid格式
+   */
+  extract_mermaid(content) {
+    const regex = /```mermaid\n(.*?)\n```/s;
+    const match = content.match(regex);
+    if (!match) return content;
+    return match[1];
+  }
+  /**
+   * 提取svg格式
+   */
+  extract_svg(content) {
+    const regex = /```svg\n(.*?)\n```/s;
+    const match = content.match(regex);
+    if (!match) return content;
+    return match[1];
+  }
+  parse_json(content) {
+    if (typeof content === "object") {
+      return content;
+    }
+    try {
+      return JSON.parse(content);
+    } catch (err) {
+      return JSON.parse(jsonrepair(content));
+    }
+  }
+};
+var LLM_Utils = new _LLM_Utils();
+
+// src/agent/index.ts
+import path from "path";
+var Agent = class {
+  // 角色定义的提示词
+  constructor() {
+    this.function_call = null;
+    // 工具函数的说明文档, 供模型调用时参考
+    this.function_call_doc = "";
+    // 模型的聊天能力
+    this.llm_chat_func = null;
+    // 预置提示词
+    this.preset_prompt = this.get_preset_prompt();
+  }
+  /**
+   * 注册工具函数
+   */
+  register_function_call(function_call) {
+    const tools_list = function_call.get_tools_list();
+    this.function_call_doc = tools_list.map((val) => `\u5DE5\u5177\u51FD\u6570:${val.name}
+\u5DE5\u5177\u63CF\u8FF0:${val.description}`).join("\n\n");
+    this.function_call = function_call;
+    return true;
+  }
+  /**
+   * 注册LLM模型文字能力
+   */
+  register_llm_text_ability(func) {
+    this.llm_chat_func = func;
+  }
+  /**
+  * 注册LLM模型图片能力
+  */
+  register_llm_image_ability() {
+    return true;
+  }
+  /**
+   * 创建聊天
+   */
+  async create_chat(user_input) {
+    let dispose_user_input;
+    if (typeof user_input === "string") {
+      dispose_user_input = [
+        {
+          role: "user",
+          content: [
+            { type: "text", text: user_input }
+          ]
+        }
+      ];
+    } else {
+      dispose_user_input = JSON.parse(JSON.stringify(user_input));
+    }
+    if (this.function_call_doc) {
+      dispose_user_input.unshift({
+        role: "system",
+        content: [
+          { type: "text", text: `# \u5DE5\u5177\u51FD\u6570\u8BF4\u660E\u6587\u6863:
+${this.function_call_doc}` }
+        ]
+      });
+    }
+    dispose_user_input.unshift({
+      role: "system",
+      content: [
+        { type: "text", text: this.preset_prompt }
+      ]
+    });
+    let max_loop = 20;
+    let loop_count = 0;
+    while (loop_count < max_loop) {
+      const llm_res_str = await this.llm_chat_func(dispose_user_input);
+      const llm_res = LLM_Utils.extract_json(llm_res_str);
+      if (!llm_res.command) throw "\u5F02\u5E38\u6307\u4EE4, \u672A\u5305\u542Bcommand\u5B57\u6BB5";
+      if (llm_res.command === "end") {
+        return llm_res.result;
+      } else if (llm_res.command === "get_tool_doc") {
+        const tool_name = llm_res.tool_name;
+        const tool_doc = this.function_call.gen_tool_doc(tool_name);
+        dispose_user_input.push({
+          role: "user",
+          content: [
+            { type: "text", text: `\u5DE5\u5177\u51FD\u6570${tool_name}\u7684\u8BF4\u660E\u6587\u6863:
+${tool_doc}` }
+          ]
+        });
+      } else if (llm_res.command === "use_tool") {
+        const tool_name = llm_res.tool_name;
+        const tool_params = LLM_Utils.parse_json(llm_res.params);
+        const tool_result = await this.function_call.exec_function(tool_name, tool_params);
+        dispose_user_input.push({
+          role: "user",
+          content: [
+            { type: "text", text: `\u5DE5\u5177\u51FD\u6570${tool_name}\u7684\u8C03\u7528\u7ED3\u679C:
+${JSON.stringify(tool_result)}` }
+          ]
+        });
+      }
+      loop_count++;
+    }
+  }
+  /**
+   * agent预置提示词
+   */
+  get_preset_prompt() {
+    const prompt = fs.readFileSync(path.join(__dirname, "./preset_prompt.md"), "utf-8");
+    return prompt;
+  }
+};
+
+// src/function_call/index.ts
+import path2 from "path";
+import fs2 from "fs";
+var FunctionCall = class {
+  constructor(tool_path) {
+    this.tool_path = tool_path;
+    this.tools = {};
+    this.init();
+  }
+  init() {
+    for (const file_name of fs2.readdirSync(this.tool_path)) {
+      const route_path = path2.join(this.tool_path, file_name);
+      const mcp = __require(route_path);
+      const info = mcp.register();
+      this.tools[info.name] = info;
+    }
+  }
+  /**
+   * 生成完整的工具函数说明文档
+   */
+  gen_tool_doc(name) {
+    const info = this.tools[name];
+    const schema = info.input_schema;
+    const schemaDesc = `\u7C7B\u578B: ${schema.type}`;
+    const params = Object.keys(schema.properties).map((key) => {
+      const item = schema.properties[key];
+      const required = schema.required?.includes(key) ? "\u5FC5\u586B" : "\u53EF\u9009";
+      const examples = item.examples && item.examples.length > 0 ? `\u793A\u4F8B\u503C: ${item.examples.join(", ")}` : "\u65E0\u793A\u4F8B";
+      return `  - ${key} (${item.type}, ${required})
+    \u8BF4\u660E: ${item.description}
+    ${examples}`;
+    }).join("\n");
+    return `\u5DE5\u5177\u51FD\u6570\u540D\u79F0: ${info.name}
+\u529F\u80FD\u63CF\u8FF0: ${info.description}
+\u8F93\u5165\u7ED3\u6784: ${schemaDesc}
+\u53C2\u6570\u5217\u8868:
+${params}`;
+  }
+  /**
+   * 获取工具函数列表
+   */
+  get_tools_list() {
+    return Object.keys(this.tools).map((key) => {
+      const info = this.tools[key];
+      return {
+        name: info.name,
+        description: info.description
+      };
+    });
+  }
+  /**
+   * 执行工具函数
+   */
+  async exec_function(name, params) {
+    const info = this.tools[name];
+    if (!info) {
+      throw `\u5DE5\u5177\u51FD\u6570${name}\u4E0D\u5B58\u5728`;
+    }
+    return await info.register_func(params);
+  }
+};
+export {
+  Agent,
+  FunctionCall,
+  LLM_Utils
+};

package/package.json

@@ -0,0 +1,34 @@
+{
+  "type": "module",
+  "name": "p-api-agent",
+  "version": "0.0.1",
+  "main": "./dist/index.cjs",
+  "module": "./dist/index.mjs",
+  "files": [
+    "dist"
+  ],
+  "exports": {
+    ".": {
+      "import": "./dist/index.mjs",
+      "require": "./dist/index.cjs"
+    }
+  },
+  "types": "dist/index.d.ts",
+  "scripts": {
+    "build": "tsup"
+  },
+  "author": "",
+  "license": "ISC",
+  "description": "",
+  "devDependencies": {
+    "@types/lodash": "4.17.16",
+    "@types/node": "^25.3.5",
+    "javascript-obfuscator": "^4.1.1",
+    "tsup": "^8.4.0",
+    "typescript": "5.8.3"
+  },
+  "dependencies": {
+    "axios": "^1.13.6",
+    "jsonrepair": "^3.13.2"
+  }
+}