Skip to main content
Use the chat completions API to describe or reason about images. Add both the prompt and the image URL to the same user message so the model can reference the visual context.

​
Streaming example

from openai import OpenAI

client = OpenAI(
    base_url="https://api.llm7.io/v1",
    api_key="unused",  # replace with your token for higher limits
)

image_url = "https://images.weserv.nl/?url=wsrv.nl/lichtenstein.jpg&w=600&output=webp"

messages = [
    {
        "role": "user",
        "content": [
            {"type": "text", "text": "Describe this image for alt text."},
            {"type": "image_url", "image_url": {"url": image_url}},
        ],
    }
]

stream = client.chat.completions.create(
    model="default",
    messages=messages,
    temperature=0.5,
    stream=True,
)

model_used = None
for chunk in stream:
    model_used = chunk.model
    delta = chunk.choices[0].delta.content or []
    if isinstance(delta, list):
        delta = "".join(
            part.get("text", "") for part in delta if part.get("type") == "text"
        )
    print(delta, end="", flush=True)

print(f"\nModel selected: {model_used}")
Use a publicly reachable https image (or a signed URL) and keep text + image together in one content array. The streamed response is text only.

​
Non-streaming call

from openai import OpenAI

client = OpenAI(
    base_url="https://api.llm7.io/v1",
    api_key="unused",
)

image_url = "https://images.weserv.nl/?url=wsrv.nl/lichtenstein.jpg&w=600&output=webp"

result = client.chat.completions.create(
    model="pro",
    messages=[
        {
            "role": "user",
            "content": [
                {"type": "text", "text": "Summarize the scene in one sentence."},
                {"type": "image_url", "image_url": {"url": image_url}},
            ],
        }
    ],
    temperature=0.3,
)

print(result.choices[0].message.content)
print(f"Model selected: {result.model}")

​
What to remember

  • messages[*].content is an array: include both text and image_url parts in the same user message.
  • Image URLs must be reachable over HTTPS; signed URLs work if they stay valid for the request duration.
  • response.model (or chunk.model while streaming) tells you which underlying model handled the call.