Ridwannurudeen/fixfirst-edge — reverse-engineered prompt

Build me an offline maintenance copilot for factory technicians called FixFirst Edge. I want a simple web app where a tech can type an error code, paste a manual snippet, upload a photo of a broken part, or add a voice note, then get a practical diagnosis with three clear pieces of evidence, the matching manual section, the closest past incident, and the likely replacement part.

It should run locally on a laptop with no internet needed during diagnosis. Use a FastAPI backend, a Next.js frontend, and Actian VectorAI DB for the search. Please support text, image, and audio inputs, plus filters like machine type, model number, fault code, severity, and part number. The answers should be traceable to stored manuals, incidents, and parts, not made up AI text.

Include ingest flows for PDFs, CSVs, images, and voice notes, plus a way to save a new incident. Make the UI clean and demo friendly with upload, search, filters, result cards, and an offline status banner. Add Docker Compose, seed data, and start and stop scripts. Look up current docs online if needed.