images.edit() fails with Zod validation error — multipart `n` parameter sent as string

[keu]

When using the OpenAI Python SDK with the orq.ai proxy (base_url="https://api.orq.ai/v2/router"), calls to images.edit() fail with a 400 error because the n parameter is
received as a string instead of a number.

This happens because images.edit() uploads files, so the OpenAI SDK sends the request as multipart/form-data. Per the HTTP spec, all fields in multipart form data are encoded
as strings. OpenAI's API handles this correctly by coercing "1" → 1, but the orq.ai proxy runs strict Zod validation before forwarding and rejects the string value.

images.generate() is not affected because it sends a JSON body where n is a proper integer.

Error

{
  "success": false,
  "error": {
    "name": "ZodError",
    "message": [
      {
        "expected": "number",
        "code": "invalid_type",
        "path": ["n"],
        "message": "Invalid input: expected number, received string"
      }
    ]
  }
}

Reproduction

  from openai import AsyncOpenAI

  client = AsyncOpenAI(
      base_url="https://api.orq.ai/v2/router",
      api_key="<ORQ_API_KEY>",
  )

  # ✅ This WORKS (JSON body)
  await client.images.generate(
      model="openai/gpt-image-1.5",
      prompt="A red circle",
      n=1,
      size="1024x1024",
  )

  # ❌ This FAILS (multipart/form-data body)
  await client.images.edit(
      model="openai/gpt-image-1.5",
      image=open("test.png", "rb"),
      prompt="Make this brighter",
      n=1,              # <-- sent as string "1" in multipart, rejected by Zod
      size="1024x1024",
  )

Expected behavior

The proxy should coerce numeric string values from multipart form fields to numbers before validation, matching OpenAI's API behavior.