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 using AzureOpenAI without a client-level deployment and passing the deployment via the per-request model param (the documented pattern, which works for chat completions, embeddings, audio transcriptions/translations and images.generate), images.edit() does not get its path rewritten to /deployments/{model}/images/edits. The request is sent to https://{endpoint}/openai/images/edits, which returns 404 on Azure.

Root cause: AzureOpenAI.buildRequest resolves the deployment with

const model = this.deploymentName || options.body['model'] || options.__metadata?.['model'];

/images/edits is a multipart endpoint, so by the time buildRequest runs, options.body is already a FormData instance — options.body['model'] is undefined. The other multipart endpoints handle this by passing the model through request metadata (see audio/transcriptions.ts and audio/translations.ts, which both set __metadata: { model: body.model }), but the images.edit method does not, so the fallback is also undefined and the path rewrite is skipped.

Note images.generate is unaffected because its body stays a plain JSON object. images.edit is the only deployments-routed endpoint (_deployments_endpoints includes /images/edits) with this gap, which is probably why it went unnoticed — the Azure test suite covers model→deployment mapping for generate/transcriptions/translations but has no images.edit case.

To Reproduce

import { AzureOpenAI } from 'openai';

const client = new AzureOpenAI({
  endpoint: 'https://my-resource.openai.azure.com',
  apiKey: '...',
  apiVersion: '2025-04-01-preview',
  // no client-level `deployment`
});

await client.images.edit({
  model: 'my-gpt-image-deployment', // per-request deployment
  image: fs.createReadStream('input.png'),
  prompt: 'add a hat',
});
// → POST https://my-resource.openai.azure.com/openai/images/edits?api-version=...
// → 404 (missing /deployments/my-gpt-image-deployment/)

Expected: POST https://my-resource.openai.azure.com/openai/deployments/my-gpt-image-deployment/images/edits?api-version=...

Workaround: instantiate the client with a client-level deployment — but that forces a separate client instance per deployment.

OS

macOS / Linux

Node version

Node v26.0.0

Library version

openai v6.42.0 (current master)