RepoPulse AI

AI-powered engineering intelligence for GitHub repositories.

Paste a repository URL to measure documentation quality, community health, development activity, dependency hygiene, metadata completeness, and technical debt — then read Gemini-assisted explanations on a professional dashboard. Sign in with GitHub to analyze private repos you can access and keep personal history.

Live stack: React · Vite · Tailwind · Recharts · Express · MongoDB · GitHub REST API · Google Gemini

Demo Link: https://repo-pulse-ai-pi.vercel.app/

---

Project Overview

RepoPulse AI is a full-stack MERN application designed for portfolio and production demos. It combines deterministic heuristic scoring with optional generative AI explanations, persists analyses in MongoDB, and lets you reopen or compare past runs without re-hitting GitHub.

Concern	Approach
Scoring	Rule-based heuristics (explainable, testable)
AI	Gemini explains metrics — it does not invent scores
Persistence	MongoDB Atlas (soft-fail if unavailable)
Deploy	Frontend → Vercel · Backend → Render

---

Features

• Analyze public GitHub repositories without an account
• GitHub OAuth for end users — private repos + personal history
• Weighted engineering health score + letter grade
• Five dimension score cards with reasons
• Technical debt hotspot table (risk-ranked files)
• Gemini AI insights panel
• Interactive Recharts visualizations (lazy-loaded)
• Analysis history with search, owner filter, and sort
• Compare two saved analyses (improved / declined / unchanged)
• Production hardening: Helmet, rate limits, compression, CORS, timeouts, logging
• Accessible UI: skip link, ARIA labels, focus styles, offline banner, 404/500 pages

---

Architecture

Browser (Vercel)                API (Render)                 Externals
─────────────────               ────────────                 ─────────
React pages  ──Axios──►  Routes → Controllers
Context store            Services (GitHub, scoring,    ◄── GitHub REST + OAuth
Charts / History         debt, Gemini, auth)           ◄── Gemini
                         Models (Mongoose)             ◄── MongoDB Atlas

Request flow (analyze): validate URL → fetch GitHub bundle → score → debt scan → Gemini explain → save → return JSON → Dashboard via Context (no second fetch).

---

Tech Stack

Layer	Technologies
Frontend	React 19, Vite, Tailwind CSS 4, React Router, Axios, Recharts
Backend	Node.js 20+, Express 5, Mongoose 9
Data	MongoDB Atlas
Integrations	GitHub REST API, GitHub OAuth, Google Gemini (@google/genai)
Ops	Helmet, express-rate-limit, compression, Morgan, CORS

---

Installation

Prerequisites: Node.js 20+, npm, MongoDB Atlas account, GitHub PAT, GitHub OAuth App (for private repos / login), Gemini API key (optional but recommended).

git clone https://github.com/manogna7s/RepoPulse-AI.git
cd RepoPulse-AI

npm install
npm --prefix client install
npm --prefix server install

---

Running Locally

# from repo root — starts Vite + Express together
npm run dev

• App: http://localhost:5173
• Health: http://localhost:5000/api/health

Useful scripts:

Command	Description
npm run dev	Client + server (development)
npm run build	Production frontend build
npm start	Start Express (server)
npm run lint	Lint the client

---

Deployment

See docs/DEPLOYMENT.md for full steps.

Quick summary

1. Render (API): root directory server, start npm start, health /api/health, set env vars.
2. Vercel (UI): root directory client, build npm run build, output dist, set VITE_API_URL to the Render API /api URL.
3. Point Render CLIENT_URL at your Vercel origin for CORS.

Optional Blueprint: root render.yaml.

---

Folder Structure

RepoPulse-AI/
├── client/                      # Vite + React frontend
│   ├── public/                  # Favicon, OG image, web manifest
│   ├── src/
│   │   ├── charts/              # Lazy-loaded Recharts
│   │   ├── components/          # UI + dashboard + common
│   │   ├── constants/           # Shared app constants
│   │   ├── context/             # Analysis + Auth context
│   │   ├── hooks/
│   │   ├── layouts/
│   │   ├── pages/               # Home, Dashboard, History, Compare, Auth callback
│   │   ├── services/            # Axios API clients
│   │   └── utils/
│   ├── vercel.json              # SPA rewrites for Vercel
│   └── index.html               # SEO / Open Graph / theme
├── server/                      # Express API
│   └── src/
│       ├── config/              # Env + MongoDB
│       ├── controllers/
│       ├── middleware/          # Security, logging, timeout, errors
│       ├── models/
│       ├── routes/
│       ├── services/            # GitHub, scoring, debt, AI, auth, persistence
│       └── utils/
├── docs/
│   ├── API.md                   # Endpoint documentation
│   ├── DEPLOYMENT.md
│   ├── RESUME.md                # Resume / LinkedIn / portfolio copy
│   └── INTERVIEW_GUIDE.md
├── render.yaml
├── package.json                 # Root orchestration scripts
└── README.md

---

API Documentation

Full reference: docs/API.md

Method	Path	Purpose
GET	/api/auth/github	Start GitHub OAuth
GET	/api/auth/me	Current user
POST	/api/repository/analyze	Full analysis pipeline
GET	/api/history	List saved analyses (signed-in)
GET	/api/history/:id	Load one analysis
DELETE	/api/history/:id	Delete one analysis
GET	/api/health	Liveness + DB status

---

---

Future Roadmap

• GitHub OAuth and private repository support
• Background job queue for very large repositories
• Deeper dependency / lint-based debt signals
• Historical trend charts across many runs
• Automated tests + CI
• Redis caching for repeated analyzes

---

License

MIT — see LICENSE if present, or treat this repository as MIT for portfolio use unless otherwise noted.

---