Analyze images using a vision-capable GGUF language model.

1. Reads an image file (JPEG or PNG) from disk
2. Loads a vision-capable GGUF model via inference.LoadFile
3. Sends the image as part of an inference.Message with the Images field
4. Generates a text description or analysis using model.Chat

This uses the same multimodal API that powers the OpenAI-compatible /v1/chat/completions endpoint for vision requests.

Requires a vision-capable model (e.g., LLaVA, Gemma 3 with vision encoder). Text-only models will ignore the image data.

go build -o vision-analysis ./examples/vision-analysis/

# Describe an image
./vision-analysis --model path/to/vision-model.gguf --image photo.jpg

# Ask a specific question about an image
./vision-analysis --model path/to/vision-model.gguf --image chart.png \
    --prompt "What trend does this chart show?"

# With GPU
./vision-analysis --model path/to/vision-model.gguf --device cuda --image photo.jpg