Devora analyzes any GitHub profile in seconds — scoring repositories across 5 dimensions, detecting technical depth signals, and generating AI-powered insights with a personalized roadmap. No setup. No backend. Runs entirely in your browser.
Devora is not a star counter. It is an intelligent portfolio analyzer that goes deep into every repository to detect what you actually built, how technically complex it is, and what it would take to score higher.
It identifies project types, detects 10 technical depth signals, and surfaces your strongest and weakest repos — then generates AI-powered insights using Groq's llama3-8b-8192 model to give you real, actionable feedback.
Everything runs in the browser. No backend. No build step. No npm install.
Enter Username → Fetch All Repos → Analyze Each Repo → Score 5 Dimensions → AI Insights
Step 1 — Repository Discovery
All public repositories are fetched. Profile README repos (same name as username) are automatically excluded from scoring so they do not skew your results.
Step 2 — Per-Repo Analysis
Each repository is scored across 4 sub-dimensions:
| Sub-Dimension | What It Measures | Weight |
|---|---|---|
| Quality | README depth, LICENSE, .gitignore, tests, CI/CD, deployment, topics | 28% |
| Activity | Recency, commit patterns, project age, codebase size | 22% |
| Depth | Technical signal detection (see below) | 30% |
| Popularity | Stars, forks, original vs fork status | 20% |
Step 3 — Portfolio Score
Individual repo scores are aggregated into a 5-dimension portfolio score:
| Dimension | Weight | What It Measures |
|---|---|---|
| Reputation | 30% | Stars, forks, and follower count across all repos |
| Activity | 20% | Recent commits, active repo ratio, account longevity |
| Technical Depth | 25% | Advanced signal detection across all repos |
| Diversity | 15% | Language variety and project type range |
| Quality | 10% | Documentation and structure health |
Devora scans file names, directory names, and README text for each repository:
| Signal | How It Is Detected | Score Boost |
|---|---|---|
| Auth | auth.js, passport.js, bcrypt, jsonwebtoken, nextauth in files or README |
+15 pts |
| Database | prisma, mongoose, models/ dir, mongodb, postgresql in README |
+15 pts |
| API Design | routes/ dir, graphql, openapi, REST patterns |
+12 pts |
| AI / ML | openai, langchain, tensorflow, pytorch, groq in README |
+20 pts |
| Testing | tests/ dir, jest.config.js, pytest.ini, cypress.config.js |
+12 pts |
| Docker | Dockerfile, docker-compose.yml in root |
+10 pts |
| CI/CD | .github/workflows/, .travis.yml, jenkinsfile |
+8 pts |
| TypeScript | tsconfig.json present, or language === TypeScript |
+8 pts |
| Real-time | socket.io, websocket, SSE patterns in README |
+12 pts |
| State Management | redux, zustand, vuex, recoil in README |
+8 pts |
Last commit today → 100 points
Last commit 1 week → 88 points
Last commit 1 month → 72 points
Last commit 3 months → 48 points
Last commit 6 months → 24 points
Last commit 1+ year → 8 points
Active repo ratio → 35% of activity score
Account longevity → 20% of activity score (2+ years = 100)
Scores above 85 require real-world community impact. The ceiling is:
Prestige Ceiling = 85 + prestige points (max ~18)
Best repo stars: 100k+ = +7, 20k+ = +6, 5k+ = +5, 1k+ = +4
Follower count: 100k+ = +5, 10k+ = +4, 1k+ = +3
Language tier: C/Rust/Zig = +3, Go/Java = +2, TypeScript/Python = +1
Fork count: 10k+ owned forks = +2, 500+ = +1
Quality repos: 4+ repos scoring 70+ = +1
Devora detects 7 project types from file structure and directory names:
| Type | Detection Method |
|---|---|
fullstack |
Both frontend and backend directories present |
backend |
server.js, routes/, controllers/, models/ present |
frontend-app |
vite.config, next.config, components/ present |
static |
HTML-only project, no server files |
cli |
bin/ directory, CLI-specific entry points |
ml-ai |
.ipynb files, ML library references in README |
library |
Published package patterns |
Powered by Groq llama3-8b-8192 with 4 built-in API keys and automatic rotation:
What the AI generates
- Strengths — 3 specific strengths identified from your portfolio data
- Areas to Improve — 3 specific weaknesses with context
- Recommendations — 5 actionable items with specific guidance
- Career Readiness —
Not Ready|Partially Ready|Job Ready|Senior Ready - AI Roadmap — 5-step personalized plan with priority levels and timeframe estimates
Key rotation and fallback
If a Groq key hits its rate limit, Devora automatically tries the next built-in key. If all keys are exhausted, an error screen guides you through getting your own free key from console.groq.com in under 60 seconds.
Each portfolio gets a pentagon radar chart across 5 dimensions. Dots are interactive — hover or tap to see exactly what each score means and how to improve it. No numbers are shown by default to keep the chart clean.
Suggestions are always generated — the system never shows an empty card. Priority levels:
| Priority | Color | Examples |
|---|---|---|
| Critical | 🔴 | Missing README, no LICENSE file |
| Recommended | 🟡 | No deployment, inactive for 6+ months |
| Optional | 🔵 | Missing topics, no CONTRIBUTING.md |
Every analysis run while signed in is saved automatically. Re-analyze the same profile after making improvements and your score history builds into a timeline. The progress tab shows:
- Score on each date analyzed
- First score vs latest score
- Total change (positive or negative)
The AI roadmap provides:
- A specific title and description for each step
- Priority level —
high,medium, orlow - Timeframe estimate —
1–2 weeks,4–8 weeks,Ongoing - A focus sentence — the single most important next action
Guests receive a rule-based roadmap. Signed-in users get the full AI-generated version.
| Service | Purpose |
|---|---|
| Supabase | Email + password authentication, analysis history storage |
| GitHub REST API | Public repo data, README content, file tree |
| Groq API | AI insights and roadmap generation |
| GitHub Pages | Static hosting — zero server cost |
Inspired by Linear — minimal, fast, sharp, and keyboard-friendly.
- Hard
1pxborders everywhere — no blur, no soft shadows - Wave theme transition — old theme shrinks away from click point, revealing new theme underneath
- Fully responsive — sidebar collapses to a slide-in drawer on mobile
- Horizontal-scrollable tabs on small screens
| Layer | Technology |
|---|---|
| Frontend | Vanilla JavaScript ES2022 — no frameworks, no build step |
| Fonts | Inter + JetBrains Mono (Google Fonts) |
| AI | Groq API — llama3-8b-8192 |
| Auth + Database | Supabase (PostgreSQL + Row Level Security) |
| Hosting | GitHub Pages — free, zero configuration |
| CI | GitHub Actions — syntax checks + unit tests on every push |
node tests/scoring.test.js # 15 scoring algorithm tests
node tests/detection.test.js # 18 signal detection testsAll tests run automatically on every push via GitHub Actions.
# Clone
git clone https://github.com/shaikhshahnawaz13/devora.git
# Open in browser — no install needed
open devora/index.htmlNo npm install. No webpack. No framework. Open index.html directly and it works.
| Section | Description |
|---|---|
| Overview | Score ring, dimension scores, pentagon radar, language distribution |
| AI Insights | Strengths, weaknesses, recommendations, career readiness score |
| Progress | Score history timeline with delta tracking |
| Roadmap | 5-step AI or rule-based improvement plan |
| All Repos | Every repository with full scores — click any card for deep analysis |
| Top 6 | Highest-scoring 6 repositories highlighted |
- Side-by-side comparison of two GitHub profiles
- Exportable PDF score report
- GitHub organization analysis — score a whole team
- Weekly email digest of score changes
- Public opt-in leaderboard
git clone https://github.com/shaikhshahnawaz13/devora.git
cd devora
# No build step — edit JS/CSS and refresh the browser
# Run tests before submitting a PR
node tests/scoring.test.js
node tests/detection.test.jsMIT — see LICENSE
Built by Shaikh Shahnawaz
BSc Information Technology · Akbar Peerbhoy College · University of Mumbai
