sameerthawait/conflict-aware-rag — reverse-engineered prompt

Reverse engineered prompt

Build me a research assistant for scientific, clinical, or legal documents that focuses on conflicting evidence instead of giving a single blended answer.

I want to be able to drop in PDFs, markdown, or text files, index them, then ask questions in a chat style workspace. The app should pull relevant passages, group contradictory or different viewpoints side by side, and show which sources support each stance. Please include clear citations I can inspect, a visible contradiction strength signal, and some simple credibility scoring so I can tell how trustworthy the answer is. If the evidence is too weak, it should say so instead of guessing.

Make it feel like a clean academic dashboard with a frontend chat UI and a backend API. Add an admin page where I can save an API key, plus basic controls for managing keys and clearing cache. I also want it easy to run locally with Docker, and include monitoring and logging so I can see system health. Look up current docs online if you need to.

Want more depth? Deep Reverse