VisionBridge is a tiny, OpenAI-compatible proxy that enables text-only large language models to process images. It allows reasoning models like DeepSeek, Qwen, and GLM to see by querying a separate vision model through tools such as look, OCR, scan, crop, and compare. The project requires no training or additional weights and is released under the MIT license.
VisionBridge proxy gives text LLMs vision via separate model
from English