How kishan-solanki/intervue-ai works
Project type and purpose
• This is a full-stack web application built with Next.js (“src/app”, “next.config.mjs”)
• It provides an AI-driven interview platform (“Intervue AI”), integrating a Python-based nervousness detection model (“nervesness_model/…”) and external AI services (e.g., Google Gemini SDK)
• It supports user authentication, interview creation, interview results, and user data persistence
File structure
• Frontend (React/Next.js):
– src/app/(auth) and src/app/(root) contain UI routes and pages (login, register, dashboard, interview pages)
– src/app/component and src/components hold reusable UI components (Navbar, Button, Hero, custom animations)
– globals.css, layout.js, page.js configure global styles and page shells
• Backend (Next.js API routes):
– src/app/api/auth, src/app/api/interview, src/app/api/user define server-side endpoints (Route.js files)
– src/lib holds utilities: dbconfig.js (database connection), jwt.js (token handling), googleGemini.js (external LLM), vapi.sdk.js, utils.js
– src/helpers/mailer.js for email notifications
– src/model contains Mongoose/ORM models (userModel.js, interviewModel.js)
• Machine Learning microservice:
– nervesness_model folder with Python code and pickle model
– run.py likely exposes a REST or gRPC interface for stress-level inference
• Configuration & tooling:
– package.json, eslint.config.mjs, postcss.config.mjs, next.config.mjs, jsconfig.json
– middleware.js for cross-cutting concerns (authentication guard, logging)
README
• Minimal description: “This is Intervue AI”
Final advice
• Major pieces without over-partitioning – group related API routes together
• Async or long-running flows (e.g., model inference)