TypeError: Cannot read private member from an object whose class did not declare it at __classPrivateFieldGet

[FaresKi]

Confirm this is a Node library issue and not an underlying OpenAI API issue

• This is an issue with the Node library

Describe the bug

When running my code in production, I got this very random error:

TypeError: Cannot read private member from an object whose class did not declare it at __classPrivateFieldGet 

(file:///app/node_modules/.pnpm/openai@5.10.1_ws@8.18.3_zod@3.25.76/node_modules/openai/internal/tslib.mjs:14:15) at Proxy.buildURL
(/app/node_modules/.pnpm/openai@5.10.1_ws@8.18.3_zod@3.25.76/node_modules/openai/src/client.ts:446:23) at Proxy.buildRequest 
(/app/node_modules/.pnpm/openai@5.10.1_ws@8.18.3_zod@3.25.76/node_modules/openai/src/client.ts:818:22) at Proxy.makeRequest 
(/app/node_modules/.pnpm/openai@5.10.1_ws@8.18.3_zod@3.25.76/node_modules/openai/src/client.ts:532:46) at process.processTicksAndRejections (node:internal/process/task_queues:105:5) at async file:///app/node_modules/.pnpm/@langchain+openai@0.6.2_@langchain+core@0.3.65_@opentelemetry+api@1.9.0_@opentelemetry+_6b14e933d3433c65dede7732832881c4/node_modules/@langchain/openai/dist/embeddings.js:182:29 at async RetryOperation._fn (/app/node_modules/.pnpm/p-retry@4.6.2/node_modules/p-retry/index.js:50:12)

What's strange is that this error typically happens in a container with a Linux environment. In my case, Ubuntu.
I've tried to replicate it but I simply cannot.
Considering downgrading for now but I'm not sure if it's a "langchain/openai" issue, or an "openai" one.

To Reproduce

I'm not even sure, I'm sorry.
i'm creating the openai client via langchain:

new ChatOpenAI({
     apiKey: process.env.OPENAI_API_KEY,
     modelName: "gpt-4o",
})

Code snippets

OS

Ubuntu

Node version

v22.17.1

Library version

openai 5.10.1

[RobertCraigie]

Thanks for the report, can you share more details about your setup? How are you running your code?

[UNLRN]

We've started to see this error as well. It started occurring when upgrading the package from 5.9.2 to 5.10.1, and oddly, downgrading didn't resolve this issue for us. This also only seems to be occurring for our calls that are using the images.generate functions. Our chat.completions.create are having no issues.

We're making these calls inside a typescript node server (v22). I'm not familiar with the typescript setup but if that's useful information, I can share. I will say that our source code is being compiled to commonjs using swc, so we're importing the openai .js files. Our code is simple enough where we're just creating the client and then calling the API:

export function createImage() {
  // create client
  const client = new OpenAI({ apiKey: 'XXX' });

  // generate image
  const response = client.images.generate(/* ... */); // errors here

  // process response
}

As far as I can tell, the issue (as reported), points to both of the "private" class fields causing issues, here and here.

Once I patched the package to modify the "__classPrivateFieldSet/Get", I know longer see this issue for the image calls.

[FaresKi]

@RobertCraigie this issue occurs in my langchain setup.
More specifically, in Langgraph, using the Tavily web search feature.

Now based on @UNLRN's comment, i too had to downgrade far down the road to get it to work as expected
Let me find the exact version
Edit: had to downgrade to a full major version below (4.10) for it work

[nbonamy]

I am seeing this too when using the Response API in an Electron app (calling the API from the renderer process).

This happens when doing await this.client.responses.create(request)

Call-stack on 5.10.2 is:

Error while generating text TypeError: Cannot read private member from an object whose class did not declare it
    at __classPrivateFieldGet (tslib.mjs:14:15)
    at Proxy.buildURL (client.ts:446:23)
    at Proxy.buildRequest (client.ts:818:22)
    at Proxy.makeRequest (client.ts:532:46)

Apparently offending line:

const baseURL = (!this.#baseURLOverridden() && defaultBaseURL) || this.baseURL;

Thanks

[marcelgerber]

Hi! I was also facing this issue, and like @FaresKi could not repro this locally, but only on the server.

Turns out, the culprit is a Sentry dependency update we had recently: @sentry/node@9.40 (which also explains why I couldn't repro locally, because I didn't have Sentry set up for instrumentation on localhost).

I tested three versions of @sentry/node:

• 9.39 works fine
• 9.40 (released a week ago) is broken, and leads to the above error
• 9.41 (released today) is also still broken

Interestingly, the 9.40 release notes are talking about an OpenAI integration, instrumenting openAI API calls. I guess that's the culprit, then (although I don't have this integration enabled currently).

[nbonamy]

Turns out, the culprit is a Sentry dependency update we had recently: @sentry/node@9.40 (which also explains why I couldn't repro locally, because I didn't have Sentry set up for instrumentation on localhost).

I don't depend on Sentry and still facing the issue.

[marcelgerber]

I could find a minimal repro for the Sentry problem, and have filed that with them now: getsentry/sentry-javascript#17159

[UNLRN]

@marcelgerber Thanks for that. It seems like this was also our issue. Downgrading sentry to 9.39 worked for us with using the latest openai version.

[nbonamy]

In my case downgrading openai to 4.104.0 fixes the problem. Tried to the smallest upgrade to 5.0.2 and things start to fail.

Interestingly, things work in an isolated node project (https://github.com/nbonamy/multi-llm-ts). In the electron project (https://github.com/nbonamy/witsy), things fail. Given the number of dependencies there, hard to know which one is the problem...