Vision bridge plugin for DeepSeek Harness (dsh): image understanding for text-only models via luma-mcp, free Qwen3-VL-Flash by default.
dsh plugin --profile web add github:FuzzySoul/dsh-free-vision
View the source on GitHub