Agent-callable vision tool that describes local images via any OpenAI-compatible vision endpoint you configure, with an optional multi-model cross-check and no built-in keys.
dsh plugin --profile web add dsh-vision-tool
View the source on GitHub