Advanced code analysis tool that detects whether code is AI-generated or human-written using multi-layer detection pipeline
│
├── components/
│ ├── about.html
│ ├── dataset.html
│ ├── editor.html
│ ├── footer.html
│ ├── github.html
│ ├── header.html
│ └── models.html
│
├── css/
│ ├── about.css
│ ├── analysis.css
│ ├── base.css
│ ├── components.css
│ ├── editor.css
│ ├── header.css
│ ├── layout.css
│ ├── main.css
│ ├── pages.css
│
├── js/
│ ├── analysis.js
│ ├── app.js
│ ├── editor.js
│ ├── examples.js
│ ├── github.js
│ ├── loader.js
│ ├── patterns.js
│ └── render.js
│
├── index.html
└── README.md git clone https://github.com/RakkaEvandra06/TrackerSense.git|
🔍 Core Analysis
Multi-layer pipeline, 5 independent signals fused into one score Language-aware patterns, Python, JS, TS, Java, C++, Go, Rust, PHP, Ruby, C# Shannon entropy curve, visualized across 30 code segments AST structural proxy, counts functions, classes, docstrings, type annotations Identifier consistency, camelCase vs snake_case ratio analysis |
🎨 UI / UX
Live heatmap overlay, highlights AI vs Human patterns inline in the editor Typing animation, Claude AI streams its analysis character-by-character Scan line visual, animated sweep during analysis Confidence bar + split bar, dual visual for AI/Human breakdown Dark terminal aesthetic, JetBrains Mono + Syne typography |
|
🐙 GitHub Integration
Live repo scanning, fetches real files via GitHub REST API Parallel file processing, 5-thread concurrency pool Per-file AI scoring, lightweight heuristic optimized for speed Language breakdown, auto-detects language from file extension |
🛡️ Accuracy & Honesty
False-positive mitigation, Bayesian smoothing on low-signal inputs Code length factor, short snippets get conservative scores Deterministic attribution, no random noise in AI source estimates Clear simulation labels, ML model UI is labeled SIMULATED honestly |
Made with ❤️
Vibe Detect Team