Neal-Salian/ai-interview-analyzer — reverse-engineered prompt

Build me a working AI interview analyzer for recruiters. I want a recruiter to log in, create candidates and interview sessions, connect a Zoom live stream, and watch a live dashboard while the interview is happening.

The dashboard should show facial emotion trends, live speech transcript chunks, sentiment over time, Big Five personality estimates, and 3 smart follow up questions after each candidate answer. The recruiter should be able to mark suggested questions as asked. After the interview ends, generate a PDF report with the transcript, emotion timeline, psychometric scores, sentiment summary, suggested questions, and an AI written summary.

Keep candidate video and audio private. Process emotion detection and transcription locally with DeepFace and Whisper, and only send anonymized text to Claude for question suggestions and summaries. Use the existing backend and frontend style from this repo, include auth, database storage, WebSocket live updates, Zoom webhook support, and Docker setup. Look up current docs online if you need to.