TypeError: fetch failed #29

iProgramme · 2023-12-19T02:16:38Z

Actual Behavior

node:internal/process/promises:288
            triggerUncaughtException(err, true /* fromPromise */);
            ^

TypeError: fetch failed
    at Object.fetch (node:internal/deps/undici/undici:11576:11)
    at process.processTicksAndRejections (node:internal/process/task_queues:95:5)
    at async makeRequest (/Users/yubowen/Desktop/MyPrivate/node_modules/@google/generative-ai/dist/index.js:195:20)
    at async generateContentStream (/Users/yubowen/Desktop/MyPrivate/node_modules/@google/generative-ai/dist/index.js:519:22)

Node.js v18.18.0

Steps to Reproduce the Problem

npm install @google/generative-ai

const { GoogleGenerativeAI } = require("@google/generative-ai");

// Access your API key as an environment variable (see "Set up your API key" above)
const genAI = new GoogleGenerativeAI('xxxxxxxxxxxxxxxx');

async function run() {
  // For text-only input, use the gemini-pro model
  const model = genAI.getGenerativeModel({ model: "gemini-pro" });

  const prompt = "Write a story about a magic backpack."

  const chat = await model.startChat('');
  const result = await chat.sendMessageStream(prompt);
  let text = ''
  for await (const chunk of result.stream) {
    text += chunk.text();
  }
  console.log(text);
}

run();

node index.js

Specifications

Version: Node.js v18.18.0, "@google/generative-ai": "^0.1.3"
Platform: macOS 12.6.5

objectIsNotDefined · 2023-12-19T03:45:58Z

got same errorw

const { GoogleGenerativeAI } = require('@google/generative-ai')

const genAI = new GoogleGenerativeAI('XXX')

async function run () {
  const model = genAI.getGenerativeModel({ model: 'gemini-pro'})

  const prompt = 'Write a story about a AI.'

  const result = await model.generateContent(prompt)
  const response = await result.response
  const text = response.text()
  console.log(text)

}

run()

output:

node:internal/process/promises:289
triggerUncaughtException(err, true /* fromPromise */);
^

TypeError: fetch failed
at Object.fetch (node:internal/deps/undici/undici:11730:11)
at process.processTicksAndRejections (node:internal/process/task_queues:95:5)
at async makeRequest (/Users/janber/Documents/demo/test/node_modules/@google/generative-ai/dist/index.js:195:20)
at async generateContent (/Users/janber/Documents/demo/test/node_modules/@google/generative-ai/dist/index.js:525:22)
at async handle (/Users/janber/Documents/demo/test/ai/test.js:10:18)

Node.js v20.10.0

lzghades · 2023-12-19T08:46:48Z

same error, how to fix it?

xuzhiyang · 2023-12-19T11:11:48Z

我知道了，有这个错误的应该都是在中国吧。谷歌代码能力不行fetch没用到代理。
也就是说虽然电脑科学上网了，但fetch科学不了。
虽然知道这个问题，但不知道怎么解决，总不能去修改“node_modules/@google/generative-ai/dist/index.js:195”这文件吧，npm install就被覆盖了。
就是要解决GoogleGenerativeAI使用HttpsProxyAgent方法问题。谁解决了告诉我一声[email protected]谢了。

Hello, GoogleGenerativeAI module how to use HttpsProxyAgent? thx.

e-roy · 2023-12-19T11:25:44Z

const { GoogleGenerativeAI } = require("@google/generative-ai");

// Access your API key as an environment variable (see "Set up your API key" above)
const genAI = new GoogleGenerativeAI('xxxxxxxxxxxxxxxx');

async function run() {
  // For text-only input, use the gemini-pro model
  const model = genAI.getGenerativeModel({ model: "gemini-pro" });

  const prompt = "Write a story about a magic backpack."

  const chat = await model.startChat('');
  const result = await chat.sendMessageStream(prompt);
  let text = ''
  for await (const chunk of result.stream) {
    text += chunk.text();
  }
  console.log(text);
}

run();

Try replacing:

const chat = await model.startChat("");

with:

const chat = model.startChat({});

hsubox76 · 2023-12-19T18:47:55Z

I've copy-pasted the entire sample code as provided in the original post, inserted my API key, and run it using @google/[email protected] and Node 18.18.0 and was unable to reproduce it.

This seems odd because the fetch call is wrapped in a try/catch where any error should be caught and wrapped in GoogleGenerativeAIError (line 91) with some additional text, but in these error logs, fetch seems to be throwing an uncaught exception that is bypassing this try/catch, as I don't see any of the error wrapper text that the catch block should add.
https://github.com/google/generative-ai-js/blob/main/packages/main/src/requests/request.ts#L68

According to nodejs/undici#1248 there should be a cause property on the error, if you have a way to catch it and inspect it.

For anyone who wants to make small modifications to the node_modules code themselves that you think will fix it, there's always https://www.npmjs.com/package/patch-package.

alx13 · 2023-12-19T19:19:50Z

@objectIsNotDefined Please revoke your API key. I edited the comment, but since it's leaked you need to invalidate it.

@iProgramme @lzghades @objectIsNotDefined from which region you are trying to access Gemini API?
Available regions: https://ai.google.dev/available_regions

Not important, but if you are running async function from the main module, then you should handle Promise rejection e.g., like that:

run().catch((e) => {
  console.error(e);
  process.exit(1)
})

lzghades · 2023-12-20T03:33:18Z

@xuzhiyang

试了一下只能改sdk里的indexjs了，先安装npm i undici，然后修改代码
const { ProxyAgent } = require('undici');

async function makeRequest(url, body) {
let response;
try {
const proxyHost = 'xxx';
const proxyPort = xxx;
const proxyUrl = http://${proxyHost}:${proxyPort};
const client = new ProxyAgent(proxyUrl);
response = await fetch(url.toString(), {
method: "POST",
headers: {
"Content-Type": "application/json",
"x-goog-api-client": getClientHeaders(),
"x-goog-api-key": url.apiKey,
},
body,
dispatcher: client
});

sunnylqm · 2023-12-20T04:47:02Z

@xuzhiyang @lzghades
There are a lot of programs that do not use the system proxy (such as node fetch). In such cases, you need to use tun mode or deploy a proxy on your router.
And it is also common to patch node_modules with tools like patch-package, yarn patch, or pnpm patch

iProgramme · 2023-12-20T05:18:13Z

const { GoogleGenerativeAI } = require("@google/generative-ai");

// Access your API key as an environment variable (see "Set up your API key" above)
const genAI = new GoogleGenerativeAI('xxxxxxxxxxxxxxxx');

async function run() {
  // For text-only input, use the gemini-pro model
  const model = genAI.getGenerativeModel({ model: "gemini-pro" });

  const prompt = "Write a story about a magic backpack."

  const chat = await model.startChat('');
  const result = await chat.sendMessageStream(prompt);
  let text = ''
  for await (const chunk of result.stream) {
    text += chunk.text();
  }
  console.log(text);
}

run();

Try replacing:

const chat = await model.startChat("");

with:

const chat = model.startChat({});

got the same error.

I use it in China，and I still got the same error when using VPN to proxy to Taiwan or Japan

iProgramme · 2023-12-20T06:26:23Z

@xuzhiyang @lzghades There are a lot of programs that do not use the system proxy (such as node fetch). In such cases, you need to use tun mode or deploy a proxy on your router. And it is also common to patch node_modules with tools like patch-package, yarn patch, or pnpm patch

The VPN software I am using is Clash Verge, and I have selected a Japanese node. I have tried enabling the tun mode in the terminal once, and also attempted to directly activate the tun mode using Clash Verge (reference link: Clash Verge教程). However, I still encounter the same error

sunnylqm · 2023-12-20T09:06:16Z

what do you mean by enabling the tun mode in the terminal once? i feel you did not enable it at all. anyway i am confident that's the solution to your problem and you should search more tutorials about that, which is totally off the topic here. @iProgramme

iProgramme · 2023-12-21T03:25:20Z

what do you mean by enabling the tun mode in the terminal once? i feel you did not enable it at all. anyway i am confident that's the solution to your problem and you should search more tutorials about that, which is totally off the topic here. @iProgramme

So, I reported that error in China because the Chinese region does not support accessing Google. Is it possible to use VPN and enable tun mode to be effective?

sunnylqm · 2023-12-21T04:20:02Z

我用中文再重复一遍吧，对于墙一类的问题，在任何repo提都不合适，因为他们什么也做不了解决不了。而你要做的是使用合理的代理工具和代理模式。正如我上面所言，很多程序和应用并不走系统代理，所以一般的proxy对于他们并不生效，这里也是这样一种情况。tun模式和路由代理是常见的更底层的透明代理，可以用于解决这类情况，具体怎么配置，碰到哪些环境问题，怎么调试等等，远远超出我几个字能解释的范围，所以建议你去别处搜索和学习。

xuzhiyang · 2023-12-21T13:30:25Z

楼主问题还没解决？下面是解决方法
google api:
const { setGlobalDispatcher, ProxyAgent } = require("undici");
const dispatcher = new ProxyAgent({ uri: new URL(process.env.https_proxy).toString() });
//全局fetch调用启用代理
setGlobalDispatcher(dispatcher);

随便跟你说下openai的
const { HttpsProxyAgent } = require('https-proxy-agent');
const openai = new OpenAI({
apiKey: XXXXXXXXX,
httpAgent: new HttpsProxyAgent(process.env.https_proxy) // 设置代理
});

PS：记得在控制台设置https_proxy变量
export https_proxy=http://127.0.0.1:7890 http_proxy=http://127.0.0.1:7890 all_proxy=socks5://127.0.0.1:7890

iProgramme · 2023-12-22T07:59:16Z

楼主问题还没解决？下面是解决方法 google api: const { setGlobalDispatcher, ProxyAgent } = require("undici"); const dispatcher = new ProxyAgent({ uri: new URL(process.env.https_proxy).toString() }); //全局fetch调用启用代理 setGlobalDispatcher(dispatcher);

随便跟你说下openai的 const { HttpsProxyAgent } = require('https-proxy-agent'); const openai = new OpenAI({ apiKey: XXXXXXXXX, httpAgent: new HttpsProxyAgent(process.env.https_proxy) // 设置代理 });

PS：记得在控制台设置https_proxy变量 export https_proxy=http://127.0.0.1:7890 http_proxy=http://127.0.0.1:7890 all_proxy=socks5://127.0.0.1:7890

我成功了，如下图。

我的解决方案是把 process.env.https_proxy 直接改成我的地址 http://127.0.0.1:7890。所以说，如果是运行在一个项目中，可以在运行时将地址设置到环境变量中来解决该问题

完整代码如下

const { GoogleGenerativeAI } = require("@google/generative-ai");
const { setGlobalDispatcher, ProxyAgent } = require("undici"); 
const dispatcher = new ProxyAgent({ uri: new URL('http://127.0.0.1:7890').toString() });
setGlobalDispatcher(dispatcher); 
// Access your API key as an environment variable (see "Set up your API key" above)
const genAI = new GoogleGenerativeAI('xxxxxxxxxxxxx');

async function run() {

  // For text-only input, use the gemini-pro model
  const model = genAI.getGenerativeModel({ model: "gemini-pro"});

  const prompt = "Write a story about a magic backpack."

  const result = await model.generateContent(prompt);
  const response = await result.response;
  const text = response.text();
  console.log(text);
}

run();

1zilc · 2024-03-29T05:27:34Z

参考：https://www.hehehai.cn/posts/set-nextjs-fetch-proxy-agent

@google/[email protected] 版本支持传递 requestOptions 来设置请求参数, 虽然这次更新主要为了支持 timeout ，但是依然可以传递其他参数。

import { GoogleGenerativeAI } from '@google/generative-ai'
import { ProxyAgent } from 'undici'

const genAI = new GoogleGenerativeAI(ENV.GEMINI_AI_API_KEY || '')
const geminiStream = await genAI
      .getGenerativeModel({ model: 'gemini-pro' }, {
        agent:
          ENV.LOCAL_FETCH_PROXY
            ? new ProxyAgent(ENV.LOCAL_FETCH_PROXY)
            : undefined,
      } as any)
      .generateContentStream(...)

In fact, the fetchOptions parameter is currently not initialized with parameters other than timeout
https://github.com/google/generative-ai-js/blob/c230732baa03ea19d146b598154ebbc5c87e1708/packages/main/src/requests/request.ts#L110-L119

xhaeffer · 2024-05-17T15:46:02Z

I got a similar error too, is there any solution? (still error if using gemini-1.5-pro).
This error appears when I send pdf and video files (images work fine, audio isn't tested)

duke79 · 2024-06-11T15:02:32Z

I was facing the same issue.

I switched to node version 20 from 18, it started to work.

wujohns · 2024-06-24T08:10:49Z

this repo support the proxy feature: https://github.com/wujohns/google-generative-ai

sanyers · 2024-11-12T06:39:16Z

My solution to not breaking the code structure is:

const http = require('http')
const httpProxy = require('http-proxy')
const { HttpsProxyAgent } = require('https-proxy-agent')

const proxy = httpProxy.createProxyServer()
const agent = new HttpsProxyAgent('http://127.0.0.1:7890')

const server = http.createServer(function (req, res) {
  //   req.headers.host = 'api.openai.com'
  req.headers.host = 'generativelanguage.googleapis.com'
  proxy.web(req, res, {
    target: 'https://generativelanguage.googleapis.com', // google
    // target: 'https://api.openai.com/v1', // openai
    agent,
  })

  proxy.on('error', function (err) {
    console.log(err)
  })
})
server.listen(8080, '0.0.0.0')

Then use it in the code:

const { GoogleGenerativeAI } = require("@google/generative-ai");

// Access your API key as an environment variable (see "Set up your API key" above)
const genAI = new GoogleGenerativeAI('xxxxxxxxxxxxxxxx');

async function run() {
  // For text-only input, use the gemini-pro model
  const model = genAI.getGenerativeModel({ model: "gemini-pro" }, { baseUrl:'http://127.0.0.1:8080' }); // setup baseUrl

  const prompt = "Write a story about a magic backpack."

  const chat = await model.startChat('');
  const result = await chat.sendMessageStream(prompt);
  let text = ''
  for await (const chunk of result.stream) {
    text += chunk.text();
  }
  console.log(text);
}

run();

isnolan mentioned this issue Dec 28, 2023

feat: support https proxy agent by node-fetch #38

Closed

2 tasks

iwestlin mentioned this issue Jan 1, 2024

how to apply an http proxy ? EvanZhouDev/gemini-ai#2

Closed

mingyue0094 mentioned this issue Mar 19, 2024

[Request] 谷歌模型无法直接访问时 lobehub/lobe-chat#1493

Closed

timothycarambat mentioned this issue Apr 14, 2024

[BUG]: Error fetching from gemini Mintplex-Labs/anything-llm#1082

Closed

junyanxu self-assigned this Oct 14, 2024

junyanxu added the status:wontfix This will not be worked on label Oct 14, 2024

XIAO-FAN-5257 mentioned this issue Dec 23, 2024

[GoogleGenerativeAI Error]: Error fetching from https://generativelanguage.googleapis.com/v1beta/models/gemini-2.0-flash-exp:streamGenerateContent?alt=sse: fetch failed cline/cline#918

Open

asukaminato0721 mentioned this issue Jan 8, 2025

能添加一个代理地址参数 ammaarreshi/Gemini-Search#11

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

TypeError: fetch failed #29

TypeError: fetch failed #29

iProgramme commented Dec 19, 2023

objectIsNotDefined commented Dec 19, 2023 •

edited by alx13

Loading

lzghades commented Dec 19, 2023

xuzhiyang commented Dec 19, 2023 •

edited

Loading

e-roy commented Dec 19, 2023

hsubox76 commented Dec 19, 2023

alx13 commented Dec 19, 2023

lzghades commented Dec 20, 2023

sunnylqm commented Dec 20, 2023 •

edited

Loading

iProgramme commented Dec 20, 2023

iProgramme commented Dec 20, 2023

sunnylqm commented Dec 20, 2023

iProgramme commented Dec 21, 2023

sunnylqm commented Dec 21, 2023

xuzhiyang commented Dec 21, 2023 •

edited

Loading

iProgramme commented Dec 22, 2023

1zilc commented Mar 29, 2024

xhaeffer commented May 17, 2024 •

edited

Loading

duke79 commented Jun 11, 2024 •

edited

Loading

wujohns commented Jun 24, 2024

sanyers commented Nov 12, 2024

TypeError: fetch failed #29

TypeError: fetch failed #29

Comments

iProgramme commented Dec 19, 2023

Actual Behavior

Steps to Reproduce the Problem

Specifications

objectIsNotDefined commented Dec 19, 2023 • edited by alx13 Loading

lzghades commented Dec 19, 2023

xuzhiyang commented Dec 19, 2023 • edited Loading

e-roy commented Dec 19, 2023

hsubox76 commented Dec 19, 2023

alx13 commented Dec 19, 2023

lzghades commented Dec 20, 2023

sunnylqm commented Dec 20, 2023 • edited Loading

iProgramme commented Dec 20, 2023

iProgramme commented Dec 20, 2023

sunnylqm commented Dec 20, 2023

iProgramme commented Dec 21, 2023

sunnylqm commented Dec 21, 2023

xuzhiyang commented Dec 21, 2023 • edited Loading

iProgramme commented Dec 22, 2023

1zilc commented Mar 29, 2024

xhaeffer commented May 17, 2024 • edited Loading

duke79 commented Jun 11, 2024 • edited Loading

wujohns commented Jun 24, 2024

sanyers commented Nov 12, 2024

objectIsNotDefined commented Dec 19, 2023 •

edited by alx13

Loading

xuzhiyang commented Dec 19, 2023 •

edited

Loading

sunnylqm commented Dec 20, 2023 •

edited

Loading

xuzhiyang commented Dec 21, 2023 •

edited

Loading

xhaeffer commented May 17, 2024 •

edited

Loading

duke79 commented Jun 11, 2024 •

edited

Loading