Skip to main content
Logging images with Keywords AI follows the same process as regular logging, but includes image URLs in the message content.

​
Image Format

Include images in your prompt messages using this JSON structure: 
{
  "type": "image_url",
  "image_url": {
    "url": "https://keywordsai-static.s3.us-east-1.amazonaws.com/docs/integrations/integration_cards/openai_agent_v0.png",
    "detail": "auto"
  }
}

​
Example Request

{
  "model": "gpt-4-vision-preview",
  "prompt_messages": [
    {
      "role": "user",
      "content": [
        {
          "type": "text",
          "text": "What do you see in this image?"
        },
        {
          "type": "image_url",
          "image_url": {
            "url": "https://keywordsai-static.s3.us-east-1.amazonaws.com/docs/integrations/integration_cards/openai_agent_v0.png",
            "detail": "auto"
          }
        }
      ]
    }
  ],
  "completion_message": {
    "role": "assistant",
    "content": "This image is the logo for OpenAI Agent, featuring a circular design with a stylized 'O' shape."
  },
  "prompt_tokens": 1120,
  "completion_tokens": 23
}

​
Key Points

  • Use vision-capable models for evaluations.
  • Images are included as image_url objects within the message content array
  • The detail parameter can be set to auto, low, or high
  • All other logging parameters remain the same as regular text logging
For complete API reference, see the Request Logging Endpoint documentation.
⌘I