Vision Image Analysis via Chat Completions — KiosAPI

Send images to vision-capable models via POST /v1/chat/completions. Supports public image URLs, web links, and inline base64-encoded data URIs.

Vision-capable models like gpt-4o can analyze images you pass directly inside the messages array of a standard chat completions request. You send images the same way you send text — by including them as content objects within a user message — and the model returns a natural-language description, answer, or analysis. Both public web URLs and base64-encoded data URIs are supported, so you can reference hosted images or embed raw image data without any separate upload step.

Request

Endpoint: POST https://kiosapi.com/v1/chat/completions

Headers

Header	Value	Description
Content-Type	application/json	Required for all requests
Authorization	Bearer sk-xxx	Your KiosAPI key

Request Body

model · string · required

A vision-capable model ID, e.g. gpt-4o. Not all models support image input — check the model’s documentation before sending image content.

messages · array · required

An array of message objects. For vision requests, include at least one user message whose content is an array of text and image objects.

Show messages[]

messages[].role · string · required

For vision requests, use "user".

messages[].content · array · required

An array of content part objects. Each object must have a type field. Vision requests combine text parts and image_url parts in the same array.

max_tokens · number

The maximum number of tokens to generate in the response.

temperature · number

Sampling temperature between 0 and 2. Higher values produce more varied output.

stream · boolean

When true, the response is streamed back as server-sent events. Defaults to false.

Content Part Types

Each object in the content array must specify a type. The supported types for vision requests are:

Type	Required Fields	Description
text	text	A plain-text prompt or instruction accompanying the image
image_url	image_url.url	An image to analyze — accepts both base64 data URIs and public web URLs

image_url.url accepts both public web URLs (e.g. https://example.com/image.png) and base64 data URIs. Base64 format: data:image/png;base64,iVBOR...

Examples

cURL

curl https://kiosapi.com/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer sk-xxx" \
  -d '{
    "model": "gpt-4o",
    "messages": [
      {
        "role": "user",
        "content": [
          {
            "type": "text",
            "text": "What is in this image?"
          },
          {
            "type": "image_url",
            "image_url": {
              "url": "https://example.com/image.png"
            }
          }
        ]
      }
    ],
    "max_tokens": 300
  }'

Python

import openai

openai.api_key = "sk-xxx"
openai.base_url = "https://kiosapi.com/v1/"

response = openai.chat.completions.create(
    model="gpt-4o",
    messages=[
        {
            "role": "user",
            "content": [
                {"type": "text", "text": "What is in this image?"},
                {
                    "type": "image_url",
                    "image_url": {"url": "https://example.com/image.png"}
                }
            ]
        }
    ],
    max_tokens=300,
)

print(response.choices[0].message.content)

Response

A successful request returns a standard chat completion object. The model’s analysis or description of the image appears in choices[0].message.content.

{
  "id": "chatcmpl-A7ETdn5hnbTjpw9dtRifjMFZYcdFW",
  "object": "chat.completion",
  "created": 1726287309,
  "model": "gpt-4o-2024-05-13",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "This image shows a cartoon-style portrait of an elderly man...",
        "refusal": null
      },
      "logprobs": null,
      "finish_reason": "stop"
    }
  ],
  "usage": {
    "prompt_tokens": 265,
    "completion_tokens": 46,
    "total_tokens": 311
  },
  "system_fingerprint": "fp_992d1ea92d"
}

Vision requests consume significantly more prompt_tokens than text-only requests because image content is tokenized as pixel patches. Monitor your usage.prompt_tokens accordingly when processing large or high-resolution images.