[ Feature ] Need a VISION_MODE flag to support most free models

[RMartires]

I Need a flag that will let Strix work with non vision LLMs i run all AI agents using the OpenRouter's free models to save costs since agents do use a bunch of tokens

I Want to add a VISION_MODE flag This flag will run the entire agent by not passing any images to the LLM, allowing the agent to run only on text data

these can then be used on cheaper LLM models

this also stops the agent or child agents from hanging on these errors