liustack/modlens

ModLens is a vision plugin and CLI that turns pasted or referenced images into structured evidence and answers for text-only models. In DSH, the browser plugin captures pasted image bytes, the host side stores them as a private temporary file, and the CLI analyzes that image; alternatively, users can invoke the CLI directly with an image path or URL. The analyzer selects providers and fails over across built-in engines and reusable local CLIs. Auto-discovered routes, paste recovery, model guarding, and configuration/health commands are distinct supporting capabilities. Source excerpts are partial, so uncertain internal wiring is omitted.