Velvora Nexus

ReLife AI: give every spare electronic part a second life THE PROBLEM Students and makers keep drawers of salvaged resistors, LEDs, transistors and old boards, but can't tell what they have, what they can build with it, or whether a part is safe to reuse. Perfectly good parts get binned. According to the latest reports from the environmental managements all over the world, the electronic waste are more in weight than any other waste and it leads to severe environmental pollution. OUR CORE IDEA To help students and all electronic and electrical project builders to reuse the components by selling and buying in our app and also with the guided build with AI Copilot which will help in step by step in the project building. WHAT IT DOES 1. Point your live camera at your parts. Muse Spark identifies each one: type, value or markings, quantity, visible damage, plus a confidence score and the evidence it used. 2. You confirm or correct every proposal. Nothing is saved until you do. 3. ReLife AI matches your confirmed parts against reliable and doable projects and shows what you can build now and exactly what is missing. 4. A voice-guided build walks you through the steps. The AI Copilot is available everywhere in the app and takes typed questions, spoken questions and photos of your build. 5. Hold the components in your and don't worry about the next steps, our Voice Assistant will continuously assist you in each step. 6. Spare parts can be listed in a marketplace, with suggested titles and prices. Missing parts can be found from other listings. HOW MUSE SPARK IS USED - Vision: identifies components from photos with strict structured JSON output, retries and a fallback. - Copilot: multimodal (text and image and voice) and multilingual, with the current build and your inventory as context. Voice conversation uses speech input and output. - Salvage assessment of old devices and build-photo review. Model: ARDUINO UNO. TRUST BY DESIGN: AI proposes, you confirm, code decides Project matching, spec constraints, stock allocation, calculators, test verdicts, marketplace rules and checkout are deterministic code with automated tests. The AI never marks a part as working: photo-based parts stay "unverified" until you test them. Parts suspected damaged are excluded from builds and cannot be sold. Fixed safety responses cover mains power, smoke and swollen batteries. EVIDENCE On every real photos of 30 parts: 97% fully correct including values, 1% flagged for review, 2% wrong without a flag. Only the blurred images failed to correctly identify the connection. IMPACT The dashboard shows parts reused and resold and estimated e-waste diverted, computed from completed builds and sales using typical per-part masses. These are estimates, and the demo environment includes seeded sample data. HONEST LIMITS Simulated: payments, delivery. TECH React 18, TypeScript, Vite, Tailwind, FastAPI, SQLAlchemy, PostgreSQL; Meta Model API(MUSE SPARK 1.3).