CircularOS

CircularOS turns a photo into a practical, explainable circularity action plan. People often recognize an unwanted object but do not know whether to reuse, repair, donate, recycle, or seek specialized handling. The right answer depends on its condition, material and available local channels—not appearance alone. Using Meta Model API, Muse Spark identifies objects, proposes materials and reads visible text when possible. Users can assess up to eight objects across multiple photos, correct the classification and confirm what they know. SAM 3.1 highlights the selected object in its original photo, connecting the assessment to visible evidence. Users then add household or business context. A versioned, deterministic decision engine produces recommendations with reasons, supporting evidence and outstanding checks. Electronics are routed toward specialized handling rather than ordinary recycling. Users can track their actions and download a readable PDF plan. Accessibility is central: camera capture, larger controls, English and Spanish voice guidance, and optional audio preparation help users navigate one step at a time. Natural spoken guidance is powered by ElevenLabs. Our proposed free community pilot starts in Lince, Lima, Peru, with a focus on including older adults in meaningful circular-economy participation. The intended journey is: identify, review, prepare, choose a receiving point, deliver and weigh, verify receipt, and record partner-agreed compensation. This remains a pilot proposal: no receiving or paying partners are currently verified, and the demo does not offer payments, redeemable credits or pickup bookings. A photograph never establishes accepted weight, commercial value or earnings. CircularOS makes uncertainty visible and keeps human confirmation separate from verified delivery. The working MVP is publicly deployed at https://circularos.circulapp.la, with open-source code, 46 passing tests and a narrated demo under three minutes.