RepoLens AI — Understand Any GitHub Repository in Seconds

Dário Pimentel — Sun, 15 Feb 2026 16:57:53 +0000

This is a submission for the GitHub Copilot CLI Challenge

What I Built

RepoLens AI is an AI-powered web application that instantly analyzes any public GitHub repository and delivers a comprehensive technical report — complete with stunning, zero-dependency SVG visualizations you won't find anywhere else.

Paste a GitHub URL, and in seconds you get:

Architectural Summary
Full Stack Breakdown
Strengths & ⚠️ Weaknesses
Improvement Suggestions
Beginner-Friendly Tasks

But what truly sets RepoLens AI apart are its 5 unique visualizations, all built in pure SVG with zero chart libraries:

Visualization	What it does
Language Ring	Animated donut chart with official GitHub language colors
Health Radar	Pentagon radar chart scoring 5 health dimensions
Architecture Graph	Interactive node graph of the project's folder structure with collapse/expand breathe animation
Repo Personality	Myers-Briggs style analysis — assigns an archetype (Pioneer, Guardian, Architect, Sprinter, Scholar, Community) with traits, radar chart, and dimension breakdown
Hero Orb	Atmospheric animated background orb for visual depth

The app also features:

Full i18n — English (default) + Portuguese, with AI responses respecting the selected language
Responsive design — Desktop 2-column grid for Strengths/Weaknesses, mobile-first everywhere
Copy & Export — Copy analysis to clipboard or download as .md
Analysis timer — Shows how long the AI took
History — Recently analyzed repos saved locally
Smart fallback — Graceful degradation when AI quota is exceeded

Tech Stack: React 19 + Vite 7.3 + TailwindCSS v4 | Express 5 + ES Modules | Google Gemini AI with automatic model discovery and fallback chain | GitHub REST API

Demo

🔗 Live App: repolens-ai.vercel.app

Repository: https://github.com/DARIO-engineer/RepoLens.AI

Screenshots

The clean landing page with animated orb background

AI-generated analysis with responsive 2-column layout on desktop

Language DNA donut ring and Health Radar pentagon — pure SVG, zero libraries

Interactive Architecture Graph with radial layout — click the center node to trigger the collapse/expand breathe animation

Repo Personality — Myers-Briggs for repositories. Archetype, traits, radar chart, and dimension breakdown

Built with Copilot CLI showcase — real commands used during development

Full i18n: one click to switch between English and Portuguese

My Experience with GitHub Copilot CLI

GitHub Copilot CLI wasn't just a helper in this project — it was my pair programming partner from start to finish. Every major feature, every tricky bug, every architectural decision went through gh copilot in the terminal.

How I Used It

1. Architecture & Refactoring

$ gh copilot suggest "refactor express server into modular architecture with services and routes"

Copilot suggested creating services/github.js, services/gemini.js, and routes/analyze.js with clean separation of concerns. What would have taken me 30+ minutes of planning happened in seconds.

2. Critical Bug Resolution

$ gh copilot explain "why is API_KEY undefined when using ES modules with dotenv"

This was a showstopper. ES module imports are hoisted, so dotenv.config() runs after the imports that need the env vars. Copilot immediately identified the issue and suggested import "dotenv/config" as the first import. Bug fixed in under 10 seconds.

3. Zero-Dependency Visualizations

$ gh copilot suggest "create an SVG pentagon radar chart in React without any chart library"

The Health Radar visualization was born from this single command. Copilot generated polar coordinate calculations, requestAnimationFrame animation, and interactive hover — all in pure SVG. I then refined it, but the foundation was solid from the start.

4. AI Output Normalization

$ gh copilot suggest "build a normalizeAnalysis function that transforms any AI output format into canonical sections"

Gemini's output format is unpredictable. Copilot generated a fuzzy-matching system with accent stripping, universal header regex, and variation mapping that makes the analysis 100% consistent regardless of how Gemini formats its response.

5. Performance Optimization

$ gh copilot suggest "smart file tree filtering for GitHub repos to reduce prompt size"

Created an intelligent filter system with CODE_EXTENSIONS, IMPORTANT_FILES, and IGNORED_PATTERNS to keep the file tree compact. This reduced prompt size by ~60%, making analysis faster and cheaper.

6. Frontend Polish

$ gh copilot suggest "animated SVG donut chart for language breakdown with GitHub official colors"

The Language Ring visualization came from this — cubic ease-out animation, hover glow effects, 35+ official GitHub language colors with HSL fallback.

Impact by the Numbers

Metric	Value
CLI commands used	40+
Bugs resolved	12
Features generated	8
Estimated time saved	~6 hours

What I Learned

Copilot CLI fundamentally changed how I approach development:

Think in prompts, not code — Instead of writing boilerplate, I describe what I want and refine the output
Debug faster — gh copilot explain is incredibly powerful for understanding why something fails, not just what failed
Architectural confidence — Using gh copilot suggest for structure decisions gives you well-reasoned patterns immediately
Zero-dependency mindset — Copilot helped me build complex SVG visualizations that would normally require Chart.js or D3 — keeping the bundle lean (96KB gzipped JS)

The CLI workflow is addictive. No context switching to a browser, no copy-pasting from StackOverflow. Just terminal → prompt → solution → build. This is how I want to develop from now on.

Built with ❤️ by DARIO-engineer using GitHub Copilot CLI v0.0.410

Forem: Dário Pimentel