Analyze a Screenshot in 3 Steps

Step 1: Prepare Your Image

You need an image URL (publicly accessible) or a base64-encoded image.

• URL method: Upload your screenshot to any image host or use an existing URL
• Base64 method: Encode locally: base64 -w0 screenshot.png

The API accepts PNG, JPEG, WebP, and GIF formats.

Step 2: Send Your Request

Send a POST request to the analyze endpoint with your image and a prompt.

curl -X POST https://vision.kim8.s4s.host/api/analyze \
  -H "Content-Type: application/json" \
  -d '{
    "model": "google/gemma-3-27b-it",
    "prompt": "Describe what you see in this image.",
    "image": "https://example.com/screenshot.png",
    "temperature": 0.3,
    "max_tokens": 500
  }'

You can also use the web playground to upload and analyze without code.

Step 3: Read the Result

The API returns JSON with the analysis content, model used, token usage, and timing:

{
  "success": true,
  "message": "Analysis complete",
  "data": {
    "content": "The image shows a dashboard with...",
    "model": "google/gemma-3-27b-it",
    "usage_prompt": 128,
    "usage_completion": 85,
    "time_ms": 3421.5
  }
}

Extract the content: curl ... | jq -r '.data.content'

Next Steps

Choose Your Model

Quick Tips

• Use temperature: 0.1-0.3 for consistent, factual analysis
• Use temperature: 0.5-0.8 for creative descriptions
• Be specific in your prompt: "Check for broken elements" > "What do you see?"
• Use multi-model mode to cross-check results