use Ollama to host a vision language model, by default using the llava vlm. it can grab a frame from opencv and describe it based on the system prompt. (#1443)

Author: kwatters