项目实战：文档问答 Bot

// qa-bot.ts
import "dotenv/config";
import { TextLoader } from "langchain/document_loaders/fs/text";
import { RecursiveCharacterTextSplitter } from "@langchain/textsplitters";
import { MemoryVectorStore } from "langchain/vectorstores/memory";
import { OpenAIEmbeddings } from "@langchain/openai";
import { ChatOpenAI } from "@langchain/openai";
import { ChatPromptTemplate } from "@langchain/core/prompts";
import { RunnableSequence, RunnableMap } from "@langchain/core/runnables";
import { StringOutputParser } from "@langchain/core/output_parsers";
 
async function buildChain(docsPath: string) {
  // 1) 加载
  const raw = await new TextLoader(docsPath).load();
  // 2) 切分
  const splits = await new RecursiveCharacterTextSplitter({
    chunkSize: 500,
    chunkOverlap: 50,
  }).splitDocuments(raw);
  // 3) 嵌入 + 入库
  const vectorStore = await MemoryVectorStore.fromDocuments(
    splits,
    new OpenAIEmbeddings()
  );
  const retriever = vectorStore.asRetriever({ k: 4 });
 
  // 4) 组装 RAG 链
  const prompt = ChatPromptTemplate.fromTemplate(
    "只根据下面的上下文回答，若上下文无答案就说不知道。\n" +
    "上下文：\n{context}\n\n问题：{question}"
  );
 
  return RunnableSequence.from([
    RunnableMap.from({
      context: retriever,
      question: (i: { question: string }) => i.question,
    }),
    prompt,
    new ChatOpenAI({ model: "gpt-4o-mini", temperature: 0 }),
    new StringOutputParser(),
  ]);
}
 
// 使用
const chain = await buildChain("./docs/handbook.md");
const answer = await chain.invoke({ question: "如何申请测试环境？" });
console.log(answer);

// app/api/qa/route.ts
export const runtime = "nodejs";
const chain = await buildChain("./docs/handbook.md"); // 启动时构建一次
 
export async function POST(req: Request) {
  const { question } = await req.json();
  const stream = await chain.stream({ question });
  return new Response(stream as any, {
    headers: { "Content-Type": "text/plain; charset=utf-8" },
  });
}