{"repo":"junmo-kim/mac-local-vision","free":true,"listed":false,"github":"https://github.com/junmo-kim/mac-local-vision","clone":"git clone https://github.com/junmo-kim/mac-local-vision.git","description":"Zero-token, on-device vision for AI agents — OCR, pixel-exact UI targeting, and face grouping on Apple Vision. Pure Swift, single binary, no dependencies.","language":"Swift","stars":57,"topics":["ai-agents","apple-silicon","cli","computer-vision","e2e-testing","macos","mcp","ocr","on-device","swift"],"license":"MIT","category":"media-processing","readme_excerpt":"mac-local-vision ( macvis ) Zero-token, on-device vision for AI agents and E2E tests — built as a 100% Pure Swift single binary on Apple's native Vision and FoundationModels frameworks. No Node, no Python, no runtime dependencies: the OS is the dependency. - Zero-Token OCR — extract text/layout locally, never spending cloud vision tokens. - Fast E2E targeting — find returns the exact pixel center of a word for click/assert. - QR/barcode scanning — barcode decodes every symbology Vision supports (QR, Code128, EAN, PDF417, ...) in one call. - QR generation — make-qr writes a scannable QR code PNG (CoreImage, no Vision needed). - Document rectification — rectify-document finds a photographed document's boundary and flattens it into a straightened, top-down scan; document-bounds returns just the four corners. - Structured document OCR — document-ocr extracts title/paragraphs/tables/lists with layout preserved, not just flat lines of text. - Image classification — classify tags an image against a 1,303-label taxonomy (outdoor, document, people, ...). - On-device semantic ask (Beta) — multimodal reasoning via Apple Foundation Models (macOS 27 Beta). - Local face sorting — cluster photos by person without uploading anything. Tiny & fast A 550 KB stripped single binary (the frameworks are the OS, nothing is bundled — no Node, no Python, no node modules ), and every call is a fresh 0.3 s end-to-end — process launch plus recognition, no daemon to keep warm: command latency --- --- find","default_branch":null,"files":null,"tree":[],"storefront":"/r/junmo-kim","claimed":false,"request_supported":{"post":"https://gitbuyer.com/r/junmo-kim/mac-local-vision/request-supported","requests":0},"note":"indexed from public GitHub; nothing is for sale on this page. Clone it from GitHub. Paid listings live at /search."}