Visual Q&A (VQA)

Give an image and a question — the model "reads" the image and answers. OpenAI vision format — the image is embedded as an image_url content block inside the chat completion.

Available models

Model	Note
spook-vision (recommended, local)	Multimodal, fast, GPU-backed
gpt-4o	Via OpenAI — top quality, higher cost
claude-3-5-sonnet	Via Anthropic

curl

curl https://app.qevron.ai/v1/chat/completions \
  -H "Authorization: Bearer sk-..." \
  -H "Content-Type: application/json" \
  -d '{
    "model": "spook-vision",
    "messages": [{
      "role": "user",
      "content": [
        {"type": "text", "text": "What is in this image?"},
        {"type": "image_url", "image_url": {"url": "data:image/jpeg;base64,..."}}
      ]
    }]
  }'

Python (openai SDK)

import base64
from openai import OpenAI

client = OpenAI(api_key="sk-...", base_url="https://app.qevron.ai/v1")

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

resp = client.chat.completions.create(
    model="spook-vision",
    messages=[{
        "role": "user",
        "content": [
            {"type": "text", "text": "How many people are in this photo and what are they doing?"},
            {"type": "image_url", "image_url": {"url": f"data:image/jpeg;base64,{b64}"}},
        ],
    }],
)
print(resp.choices[0].message.content)

URL works too:

{"type": "image_url", "image_url": {"url": "https://example.com/photo.jpg"}}

Size limits

• Maximum payload: 50 MB (data URL + JSON, whole body).
• Recommended: client-side downscale to 1280 px max width + JPEG quality 0.85 → ~150–300 KB.
• Larger images may hit a 413 at the ingress (nginx/Caddy).

Which model for vision?

Need	Model
"What's in this image?" — general description	spook-vision
Detailed analysis / reasoning	gpt-4o (needs OpenAI key)
Extracting text from image (OCR)	spook-ocr — dedicated endpoint, OCR page
Bounding boxes / object counts	spook-detect — dedicated endpoint, Detect page

Troubleshooting

• "413 Request Entity Too Large" → image too big. Downscale + compress client-side.
• "image format not supported" → use data:image/png;base64,... or data:image/jpeg;base64,... prefix.
• Wrong answer → ask more specifically. "What's in the image?" is vague; "How many dogs?" is much more reliable.

Next: OCR · Object detection