Vision

Send images alongside text. Routing is automatic.

Both raven models read images. There is no separate vision endpoint: include an image part in a message and the system routes the request for visual understanding automatically. The rest of the request is unchanged.

#Sending an image

An image travels as a content part of type image_url next to your text. The URL can be a base64 data URL (no upload step needed) or a publicly reachable https URL.

shell

B64=$(base64 -w0 image.png)

curl https://api.dipoleml.com/v1/chat/completions \
  -H "Authorization: Bearer $RAVEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d "{
    \"model\": \"Raven Flash\",
    \"messages\": [{
      \"role\": \"user\",
      \"content\": [
        {\"type\": \"text\", \"text\": \"In one sentence, what is in this image?\"},
        {\"type\": \"image_url\", \"image_url\": {\"url\": \"data:image/png;base64,$B64\"}}
      ]
    }]
  }"

python

import base64, os
from openai import OpenAI

client = OpenAI(
    base_url="https://api.dipoleml.com/v1",
    api_key=os.environ["RAVEN_API_KEY"],
)

with open("image.png", "rb") as f:
    b64 = base64.b64encode(f.read()).decode()

resp = client.chat.completions.create(
    model="Raven Flash",
    messages=[{
        "role": "user",
        "content": [
            {"type": "text", "text": "In one sentence, what is in this image?"},
            {"type": "image_url",
             "image_url": {"url": f"data:image/png;base64,{b64}"}},
        ],
    }],
)
print(resp.choices[0].message.content)

#How routing works

Image handling is automatic and per request. If any message in the request carries an image part, the whole request is served by the vision path; otherwise it takes the text path. You never pick a vision model or change the model name. Mixed requests, where one message has text and another has an image, work the same way.

#Practical notes

Supported formats include PNG and JPEG; send the correct MIME type in the data URL.
Images count as input tokens and bill at the input rate. Large images use more tokens; downscale before sending when detail is not critical.
The vision path has its own quality tuning, so screenshots, diagrams, and photos of documents all read correctly. Ask specific questions and you get specific answers.