Mastra：TypeScript 原生 Agent 框架

import { Mastra } from "@mastra/core";
import { Agent } from "@mastra/core/agent";
import { openai } from "@ai-sdk/openai";

// 创建 Agent
const assistant = new Agent({
  name: "Assistant",
  instructions: "你是一个有帮助的助手，用中文回答。",
  model: openai("gpt-4o-mini"),
});

// 注册到 Mastra 实例
const mastra = new Mastra({
  agents: { assistant },
});

// 运行
const result = await mastra.getAgent("assistant").generate(
  "介绍一下 Mastra 框架"
);
console.log(result.text);

import { createTool } from "@mastra/core/tools";
import { z } from "zod";

const getWeatherTool = createTool({
  id: "get-weather",
  description: "获取指定城市的当前天气",
  inputSchema: z.object({
    city: z.string().describe("城市名称，例如：北京、上海"),
  }),
  outputSchema: z.object({
    temperature: z.number(),
    condition: z.string(),
  }),
  execute: async ({ context }) => {
    const { city } = context;
    // 实际调用天气 API
    const data: Record<string, { temperature: number; condition: string }> = {
      北京: { temperature: 25, condition: "晴天" },
      上海: { temperature: 22, condition: "多云" },
    };
    return data[city] ?? { temperature: 0, condition: "暂无数据" };
  },
});

// 挂载到 Agent
const weatherAgent = new Agent({
  name: "WeatherAgent",
  instructions: "你可以查询天气，根据天气给出出行建议。",
  model: openai("gpt-4o-mini"),
  tools: { getWeatherTool },
});

import { Workflow, Step } from "@mastra/core/workflows";
import { z } from "zod";

// 定义步骤
const fetchDataStep = new Step({
  id: "fetch-data",
  outputSchema: z.object({ data: z.string() }),
  execute: async ({ context }) => {
    // context.triggerData 是工作流入参
    const topic = context.triggerData.topic as string;
    return { data: `关于 ${topic} 的原始数据...` };
  },
});

const analyzeStep = new Step({
  id: "analyze",
  outputSchema: z.object({ summary: z.string() }),
  execute: async ({ context }) => {
    // 读取上一步的输出
    const prevData = context.getStepResult<{ data: string }>("fetch-data");
    const summary = `分析结果：${prevData?.data ?? ""}`;
    return { summary };
  },
});

// 组装工作流
const researchWorkflow = new Workflow({
  name: "research-workflow",
  triggerSchema: z.object({ topic: z.string() }),
})
  .step(fetchDataStep)
  .then(analyzeStep)
  .commit();

// 运行
const { runId, start } = researchWorkflow.createRun();
const result = await start({ triggerData: { topic: "LangGraph" } });
console.log(result.results["analyze"].output.summary);

import { Agent } from "@mastra/core/agent";
import { Memory } from "@mastra/memory";
import { openai } from "@ai-sdk/openai";

const memory = new Memory();

const agent = new Agent({
  name: "MemoryAgent",
  instructions: "你是一个有记忆的助手。",
  model: openai("gpt-4o-mini"),
  memory,
});

// 传入 threadId 关联同一个会话
const thread = { threadId: "user-123", resourceId: "resource-1" };

await agent.generate("我叫张三", { ...thread });
const r = await agent.generate("我叫什么名字？", { ...thread });
console.log(r.text); // 应该记得张三

import { MastraVector } from "@mastra/vector-pg"; // 或其他向量数据库
import { openai } from "@ai-sdk/openai";

// 创建向量存储
const vectorStore = new MastraVector({ connectionString: process.env.DB_URL! });

// 索引文档
await vectorStore.upsert({
  indexName: "docs",
  vectors: await embedDocuments(documents), // 你的 embedding 函数
});

// 在 Agent 里接入 RAG 工具
const ragTool = createTool({
  id: "search-docs",
  description: "搜索文档库",
  inputSchema: z.object({ query: z.string() }),
  execute: async ({ context }) => {
    const results = await vectorStore.query({
      indexName: "docs",
      queryVector: await embed(context.query),
      topK: 3,
    });
    return { results };
  },
});

// app/api/chat/route.ts
import { mastra } from "@/mastra";
import { NextRequest } from "next/server";

export async function POST(req: NextRequest) {
  const { message } = await req.json();
  const agent = mastra.getAgent("assistant");

  const stream = await agent.stream(message);
  return stream.toDataStreamResponse();
}

// src/mastra/index.ts
import { Mastra } from "@mastra/core";
import { assistant } from "./agents/assistant";

export const mastra = new Mastra({
  agents: { assistant },
});