AI Consultant — Empathetic Chat + Emotion Recognition

This is a self-contained prototype web app that simulates an AI mental-health consultant with:

• Empathetic chatbot trained on 200+ mock data points (yoga, breathing, consoling).
• Face emotion recognition (using face-api.js in-browser).
• Voice emotion heuristics (Web Audio API + optional SpeechRecognition for transcripts).
• Session aggregation and an end-session report (negative or normal depending on sadness signals).

Files added:

• index.html — Front-end UI
• static/style.css — Styles
• static/client.js — App logic (chat, emotion detection, report)
• scripts/generate_data.js — Generates data/mock_data.json with 200 entries
• data/mock_data.json — (generated by script)
• server.js — Lightweight Node static server
• package.json — dev dependencies and start script

Quick start (Windows PowerShell):

1. Install Node.js (if not installed).
2. Generate mock data:

cd d:/AI_CONSULTANT
node scripts/generate_data.js

3. Start the server:

node server.js

4. Open http://localhost:3000 in a Chromium-based browser (Chrome/Edge) to access camera/mic.

Notes:

• Face emotion detection uses models loaded from CDN — an internet connection is required to download the weights on first use.
• SpeechRecognition (voice-to-text) works best in Chrome. If unavailable, use typed input.

If you want I can run the generator and start the server for you, or adjust thresholds and UX.

LLM integration

This project supports proxying chat requests to an external LLM (e.g., Gemini). To enable it:

1. Copy .env.example to .env and set GEMINI_API_URL and either GEMINI_API_KEY or GEMINI_BEARER_TOKEN.
2. Restart the server so server.js can pick up the environment variables.

NOTE: Do NOT commit your real .env into version control. Store keys securely (environment variables, secrets manager, etc.).