Manny2706/NIMIT-AI — reverse-engineered prompt

Reverse engineered prompt

GitHub

Build me a simple web app called NimitAI that helps sales people review meeting transcripts. I want a page where I can paste a full customer call or meeting transcript, click an Analyse button, and then see clear cards showing the important sales signals the AI found, like buying intent, objections, risks, urgency, decision makers, next steps, and anything the rep should follow up on.

Each card should include the signal type, the relevant quote or summary from the transcript, and a short coaching tip in plain English. Keep the interface clean and easy to use, with a textarea, loading state, error handling, and nice empty states.

Use a Python backend that exposes an analyse endpoint accepting JSON with a transcript field. The backend should call OpenRouter using an environment API key and return clean JSON for the frontend to render. If there is no API key, it should not crash, just return an empty signals list. Make it runnable locally on port 8000 and ready to deploy on Vercel.

Want more depth? Deep Reverse