XimilalaXiang/TrustRAG — reverse-engineered prompt

Build me a polished cross platform AI knowledge workbench called TrustRAG. I want people to upload documents, organize them into workspaces, ask questions in plain language, and get answers that are clearly grounded in the source material. Every answer should show precise citations back to the original document, chunk, page, and heading, and there should be an easy review flow where users can approve, reject, or flag citations and keep a history of those decisions.

It should work really well as a self contained desktop app that runs fully on device with no server setup, using local storage and fast full text search. It should also support a server mode for teams, with multi user workspaces, role based access, and stronger document processing for things like PDF and DOCX. Please include document search, chat, and a knowledge graph style view for exploring entities and relationships across files. Keep the experience clean, trustworthy, and easy for non technical users. Feel free to check the docs online if you need to fill in details.