server.js•6.47 kB
import { FastMCP } from "fastmcp";
import { z } from "zod";
import axios from "axios";
import fs from "fs";
import path from "path";
import 'dotenv/config';
// DeepSeek API 配置
const DEEPSEEK_API_KEY = process.env.DEEPSEEK_API_KEY;
const DEEPSEEK_API_URL = "https://api.deepseek.com/v1/chat/completions";
// FastMCP 服务器
const mcpServer = new FastMCP({
name: "My Server",
version: "1.0.0",
});
// 原有的加法工具
mcpServer.addTool({
name: "add",
description: "Add two numbers",
parameters: z.object({
a: z.number(),
b: z.number(),
}),
execute: async (args) => {
return String(args.a + args.b);
},
});
// DeepSeek 聊天工具
mcpServer.addTool({
name: "deepseek_chat",
description: "使用Bach进行对话",
parameters: z.object({
message: z.string().describe("用户输入的消息"),
system_prompt: z.string().optional(), // 不再设置 default，后续逻辑动态赋值
model: z.string().optional().default("deepseek-chat").describe("使用的模型名称"),
temperature: z.number().optional().default(0.7).describe("生成文本的随机性，0-1之间"),
max_tokens: z.number().optional().default(1000).describe("最大生成token数"),
}),
execute: async (args) => {
try {
const messages = [];
let systemPrompt = args.system_prompt;
// 如果未传 system_prompt，则从 prompts.json 读取 frontend_architect 模板
if (!systemPrompt) {
const promptsPath = path.join(process.cwd(), 'prompts.json');
const promptsData = JSON.parse(fs.readFileSync(promptsPath, 'utf8'));
if (promptsData.prompts && promptsData.prompts.frontend_architect && promptsData.prompts.frontend_architect.content) {
systemPrompt = promptsData.prompts.frontend_architect.content;
}
}
// 添加系统提示词
if (systemPrompt) {
messages.push({
role: "system",
content: systemPrompt
});
}
// 添加用户消息
messages.push({
role: "user",
content: args.message
});
const response = await axios.post(
DEEPSEEK_API_URL,
{
model: args.model,
messages: messages,
temperature: args.temperature,
max_tokens: args.max_tokens,
},
{
headers: {
"Authorization": `Bearer ${DEEPSEEK_API_KEY}`,
"Content-Type": "application/json",
},
}
);
const aiResponse = response.data.choices[0].message.content;
return `DeepSeek AI 回复: ${aiResponse}`;
} catch (error) {
console.error("DeepSeek API 调用失败:", error.response?.data || error.message);
return `错误: ${error.response?.data?.error?.message || error.message}`;
}
},
});
// 获取提示词模板工具
mcpServer.addTool({
name: "get_prompts",
description: "获取可用的提示词模板列表",
parameters: z.object({
template: z.string().optional().describe("特定模板名称，如果不提供则返回所有模板"),
}),
execute: async (args) => {
try {
const promptsPath = path.join(process.cwd(), 'prompts.json');
const promptsData = JSON.parse(fs.readFileSync(promptsPath, 'utf8'));
if (args.template) {
if (promptsData.prompts[args.template]) {
return JSON.stringify({
template: args.template,
...promptsData.prompts[args.template]
}, null, 2);
} else {
return `错误: 未找到模板 "${args.template}"`;
}
} else {
const templates = Object.keys(promptsData.prompts).map(key => ({
key: key,
name: promptsData.prompts[key].name,
description: promptsData.prompts[key].description
}));
return `可用的提示词模板:\n${JSON.stringify(templates, null, 2)}`;
}
} catch (error) {
console.error("读取提示词模板失败:", error.message);
return `错误: ${error.message}`;
}
},
});
// 代码片段提问工具
mcpServer.addTool({
name: "code_selection_ask",
description: "基于选中代码片段和问题进行AI智能问答",
parameters: z.object({
code: z.string().describe("选中的代码片段"),
question: z.string().describe("用户的问题"),
system_prompt: z.string().optional(),
model: z.string().optional().default("deepseek-chat"),
temperature: z.number().optional().default(0.7),
max_tokens: z.number().optional().default(1000),
}),
execute: async (args) => {
// 拼接带代码上下文的提问
const message = `请基于以下代码片段回答问题：\n\n${args.code}\n\n问题：${args.question}`;
// 复用 deepseek_chat 的 system_prompt 逻辑
let systemPrompt = args.system_prompt;
if (!systemPrompt) {
const promptsPath = path.join(process.cwd(), 'prompts.json');
const promptsData = JSON.parse(fs.readFileSync(promptsPath, 'utf8'));
if (promptsData.prompts && promptsData.prompts.frontend_architect && promptsData.prompts.frontend_architect.content) {
systemPrompt = promptsData.prompts.frontend_architect.content;
}
}
const messages = [];
if (systemPrompt) {
messages.push({ role: "system", content: systemPrompt });
}
messages.push({ role: "user", content: message });
try {
const response = await axios.post(
DEEPSEEK_API_URL,
{
model: args.model,
messages: messages,
temperature: args.temperature,
max_tokens: args.max_tokens,
},
{
headers: {
"Authorization": `Bearer ${DEEPSEEK_API_KEY}`,
"Content-Type": "application/json",
},
}
);
const aiResponse = response.data.choices[0].message.content;
return `DeepSeek AI 回复: ${aiResponse}`;
} catch (error) {
console.error("DeepSeek API 调用失败:", error.response?.data || error.message);
return `错误: ${error.response?.data?.error?.message || error.message}`;
}
},
});
// 启动 FastMCP 服务器
console.log('🚀 启动 MCP 服务器...');
mcpServer.start({
transportType: "httpStream",
httpStream: {
port: 4001,
},
});
console.log('✅ MCP 服务器已启动在端口 4001');
console.log('📡 端点: http://localhost:4001/mcp');
console.log('🤖 DeepSeek AI 工具已集成，请设置 DEEPSEEK_API_KEY 环境变量');