Langchain部署至Firebase Function报错问题求助

问题概述 #

将基于LangChain的函数部署到Firebase Functions时持续出现编译错误：

• 使用TypeScript 4.9.0时，出现3个错误：找不到名称'this'、ChatOpenAI类型无法匹配RunnableLike参数
• 升级至TypeScript 5后，出现IterableReadableStreamInterface接口冲突的新错误
  当前使用LangChain ^0.1.18，相关配置及代码如下。

一、TypeScript 4.9.0时的错误信息 #

node_modules/@langchain/openai/node_modules/@langchain/core/dist/language_models/chat_models.d.ts:64:129 - error TS2304: Cannot find name 'this'.

64     _generateCached({ messages, cache, llmStringKey, parsedOptions, handledOptions, }: ChatModelGenerateCachedParameters<typeof this>): Promise<LLMResult & {
                                                                                                                                   ~~~~

node_modules/@langchain/openai/node_modules/@langchain/core/dist/language_models/llms.d.ts:66:122 - error TS2304: Cannot find name 'this'.

66     _generateCached({ prompts, cache, llmStringKey, parsedOptions, handledOptions, }: LLMGenerateCachedParameters<typeof this>): Promise<LLMResult & {
                                                                                                                            ~~~~

src/index.ts:36:31 - error TS2345: Argument of type 'ChatOpenAI<ChatOpenAICallOptions>' is not assignable to parameter of type 'RunnableLike<ChatPromptValueInterface, BaseMessageChunk>'.
  Type 'ChatOpenAI<ChatOpenAICallOptions>' is missing the following properties from type 'RunnableMapLike<ChatPromptValueInterface, BaseMessageChunk>': concat, text, content, additional_kwargs, and 3 more.

36     const chain = prompt.pipe(chatModel).pipe(outputParser);
                                 ~~~~~~~~~


Found 3 errors in 3 files.

Errors  Files
     1  node_modules/@langchain/openai/node_modules/@langchain/core/dist/language_models/chat_models.d.ts:64
     1  node_modules/@langchain/openai/node_modules/@langchain/core/dist/language_models/llms.d.ts:66
     1  src/index.ts:36

Error: functions predeploy error: Command terminated with non-zero exit code 2

相关配置及代码 #

tsconfig.json

{
  "compilerOptions": {
    "module": "NodeNext",
    "noImplicitReturns": true,
    "noUnusedLocals": true,
    "outDir": "lib",
    "sourceMap": true,
    "strict": true,
    "target": "ES2020",
    "esModuleInterop": true,
    "moduleResolution": "node"
  },
  "compileOnSave": true,
  "include": ["src"]
}

package.json

{
  "name": "functions",
  "scripts": {
    "build": "tsc",
    "build:watch": "tsc --watch",
    "serve": "npm run build && firebase emulators:start --only functions",
    "shell": "npm run build && firebase functions:shell",
    "start": "npm run shell",
    "deploy": "firebase deploy --only functions",
    "logs": "firebase functions:log"
  },
  "engines": {
    "node": "20"
  },
  "main": "lib/index.js",
  "dependencies": {
    "@langchain/openai": "^0.0.14",
    "firebase-admin": "^11.8.0",
    "firebase-functions": "^4.3.1",
    "langchain": "^0.1.18",
    "stream-chat": "^8.16.0"
  },
  "overrides": {
    "@langchain/core": "0.1.5"
  },
  "devDependencies": {
    "firebase-functions-test": "^3.1.0",
    "typescript": "^4.9.0"
  },
  "private": true
}

函数代码

import { onRequest } from "firebase-functions/v2/https";
import * as logger from "firebase-functions/logger";
import { StreamChat } from "stream-chat";
import { ChatOpenAI } from "@langchain/openai";
import { ChatPromptTemplate } from "@langchain/core/prompts";
import { StringOutputParser } from "@langchain/core/output_parsers";

export const myFunction = onRequest(
  { timeoutSeconds: 60, region: ["europe-west2"] },
  (request, response) => {
    const client = StreamChat.getInstance(
      "api_key",
      "api_secret"
    );
    const signature = request.headers["x-signature"] as string;
    const valid = client.verifyWebhook(request.rawBody, signature);
    if (!valid) response.status(204).end();

    const chatModel = new ChatOpenAI({
      openAIApiKey: "openai_key",
    });

    const outputParser = new StringOutputParser();

    const prompt = ChatPromptTemplate.fromMessages([
      [
        "system",
        "You are a helpful assistant",
      ],
      ["user", "{input}"],
    ]);

    const chain = prompt.pipe(chatModel).pipe(outputParser);

    if (request.body.user.id !== "system_user") {
      chain.invoke({ input: request.body.message.text }).then((aiRes) => {
        const channel = client.channel(
          "messaging",
          "channel_id"
        );
        return channel.sendMessage({
          user_id: "my-app",
          text: aiRes,
        });
      });
    }

    response.status(200).end();
  }
);

二、升级TypeScript 5后的新错误 #

node_modules/@langchain/core/dist/utils/stream.d.ts:1:18 - error TS2320: Interface 'IterableReadableStreamInterface<T>' cannot simultaneously extend types 'ReadableStream<T>' and 'AsyncIterable<T>'.
  Named property '[Symbol.asyncIterator]' of types 'ReadableStream<T>' and 'AsyncIterable<T>' are not identical.

1 export interface IterableReadableStreamInterface<T> extends ReadableStream<T>, AsyncIterable<T> {
                   ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

node_modules/@langchain/core/dist/utils/stream.d.ts:9:5 - error TS2425: Class 'ReadableStream<T>' defines instance member property '[Symbol.asyncIterator]', but extended class 'IterableReadableStream<T>' defines it as instance member function.

9     [Symbol.asyncIterator](): this;
      ~~~~~~~~~~~~~~~~~~~~~~


Found 2 errors in the same file, starting at: node_modules/@langchain/core/dist/utils/stream.d.ts:1


Error: functions predeploy error: Command terminated with non-zero exit code 2

三、解决方案 #

1. 修复LangChain依赖版本冲突 #

核心问题是@langchain/openai ^0.0.14与langchain ^0.1.18版本不兼容，且手动指定的@langchain/core 0.1.5存在类型定义缺陷。修改package.json如下：

• 移除overrides字段，让npm自动处理兼容依赖
• 升级@langchain/openai到^0.1.6（与langchain 0.1.18匹配的版本）
• 升级TypeScript到^5.3.3（兼容Node.js 20及LangChain最新类型）

修改后的package.json：

{
  "name": "functions",
  "scripts": {
    "build": "tsc",
    "build:watch": "tsc --watch",
    "serve": "npm run build && firebase emulators:start --only functions",
    "shell": "npm run build && firebase functions:shell",
    "start": "npm run shell",
    "deploy": "firebase deploy --only functions",
    "logs": "firebase functions:log"
  },
  "engines": {
    "node": "20"
  },
  "main": "lib/index.js",
  "dependencies": {
    "@langchain/openai": "^0.1.6",
    "firebase-admin": "^11.8.0",
    "firebase-functions": "^4.3.1",
    "langchain": "^0.1.18",
    "stream-chat": "^8.16.0"
  },
  "devDependencies": {
    "firebase-functions-test": "^3.1.0",
    "typescript": "^5.3.3"
  },
  "private": true
}

2. 调整TypeScript配置 #

修改tsconfig.json以适配Node.js 20和LangChain的类型要求：

• 将target设为ES2022（Node.js 20原生支持）
• 保持module和moduleResolution为NodeNext以支持ES模块
• 添加skipLibCheck: true跳过第三方库类型检查，避免类型冲突

修改后的tsconfig.json：

{
  "compilerOptions": {
    "module": "NodeNext",
    "noImplicitReturns": true,
    "noUnusedLocals": true,
    "outDir": "lib",
    "sourceMap": true,
    "strict": true,
    "target": "ES2022",
    "esModuleInterop": true,
    "moduleResolution": "NodeNext",
    "skipLibCheck": true
  },
  "compileOnSave": true,
  "include": ["src"]
}

3. 修复函数代码中的异步问题 #

原代码存在异步操作未等待、未处理错误的问题，修改为：

import { onRequest } from "firebase-functions/v2/https";
import * as logger from "firebase-functions/logger";
import { StreamChat } from "stream-chat";
import { ChatOpenAI } from "@langchain/openai";
import { ChatPromptTemplate } from "@langchain/core/prompts";
import { StringOutputParser } from "@langchain/core/output_parsers";

export const myFunction = onRequest(
  { timeoutSeconds: 60, region: ["europe-west2"] },
  async (request, response) => {
    const client = StreamChat.getInstance(
      "api_key",
      "api_secret"
    );
    const signature = request.headers["x-signature"] as string;
    const valid = client.verifyWebhook(request.rawBody, signature);
    
    if (!valid) {
      response.status(204).end();
      return; // 终止后续执行
    }

    const chatModel = new ChatOpenAI({
      openAIApiKey: "openai_key",
    });

    const outputParser = new StringOutputParser();

    const prompt = ChatPromptTemplate.fromMessages([
      ["system", "You are a helpful assistant"],
      ["user", "{input}"],
    ]);

    const chain = prompt.pipe(chatModel).pipe(outputParser);

    if (request.body.user.id !== "system_user") {
      try {
        const aiRes = await chain.invoke({ input: request.body.message.text });
        const channel = client.channel("messaging", "channel_id");
        await channel.sendMessage({
          user_id: "my-app",
          text: aiRes,
        });
      } catch (err) {
        logger.error("消息处理失败:", err);
      }
    }

    response.status(200).end();
  }
);

4. 重新安装依赖并部署 #

执行以下命令：

rm -rf node_modules package-lock.json
npm install
npm run deploy

内容的提问来源于stack exchange，提问作者Victor