"Request contains an invalid argument" when using OpenAI compatibility

Gemini API
, , ,
1

The following code (on node v22.2.0) uses fetch to call the OpenAI-compatible Gemini API

async function main() {
  const resp = await fetch(
    "https://generativelanguage.googleapis.com/v1beta/openai/chat/completions",
    {
      method: "POST",
      headers: {
        Authorization: "Bearer $API_KEY",
        "Content-Type": "application/json",
      },
      body: JSON.stringify({
        model: "gemini-1.5-flash",
        messages: [{ role: "user", content: "Explain to me how AI works" }],
      }),
    },
  );
  const data = await resp.json();
  console.log(data);
}

main().catch(console.error);

but it gives

[
  {
    error: {
      code: 400,
      message: 'Request contains an invalid argument.',
      status: 'INVALID_ARGUMENT'
    }
  }
]

However, sending a raw request:

printf "POST /v1beta/openai/chat/completions HTTP/1.1\r\nHost: generativelanguage.googleapis.com\r\nAuthorization: Bearer $API_KEY\r\nContent-Type: application/json\r\nContent-Length: 96\r\n\r\n{\"model\":\"gemini-1.5-flash\",\"messages\":[{\"role\":\"user\",\"content\":\"Explain to me how AI works\"}]}\r\n" | openssl s_client -quiet -connect generativelanguage.googleapis.com:443

works.

What is going wrong when using node’s fetch and how can I address the problem? Due to some restrictions, I can’t use an external library (such as axios or node-fetch, which do work).

4

I think this is the same as another thread: there seems to be something about the request sent by undici/Node that the API server doesn’t like (but should, probably, be able to handle).