vision-tool
Image recognition using Ollama + qwen3.5:4b with think=False for reliable content extraction.
