Codect: AI-Generated Code Detection

Overview

Codect is a free and open-source tool designed to detect whether a piece of code was written by an AI or a human. It currently supports Python and JavaScript, and reports both a classification and supporting heuristics such as token entropy, comment density, structural complexity, and signal strength.

This monorepo contains three main packages:

• @codect/core: Core detection algorithms and language analysis
• @codect/cli: Beautiful command-line interface with interactive mode
• @codect/api: Fast REST API server built with FastAPI

Features

• Multi-Language Support: Detect AI-generated code in Python and JavaScript
• Chunk-Level Detection: Splits longer files into chunks and aggregates their results instead of relying on a single whole-file score
• Low-Signal Abstention: Returns Uncertain (Mixed Signals) for snippets that do not contain enough evidence
• Entropy & Complexity Analysis: Uses token entropy, nesting depth, naming patterns, and formatting consistency as part of the score
• Human vs AI Signals: Tracks both AI-like and human-like indicators such as example-heavy code, TODO comments, debug statements, and repeated structure
• Beautiful CLI: Interactive terminal UI with colors and ASCII art
• Fast API: High-performance REST API with FastAPI
• Modular Architecture: Clean separation of concerns with monorepo structure

Star History

Installation

Prerequisites

• Node.js 18+ and npm
• Python 3.8+
• Git

Setup

# Clone the repository
git clone https://github.com/GustyCube/Codect.git
cd Codect

# Install dependencies
npm install
npm run build

# Install Python dependencies for the API
cd packages/api
pip install -r requirements.txt
cd ../..

Usage

CLI Tool

The CLI provides a beautiful terminal interface for code analysis:

# Run interactive mode
npx codect

# Analyze a specific file
npx codect analyze path/to/file.py --detailed

# Get help
npx codect --help

Features:

• Interactive mode with menu navigation
• Beautiful ASCII art logo with gradient colors
• Colored output and progress spinners
• Detailed analysis with feature breakdown and confidence
• Support for both file and snippet analysis

API Server

Start the API server:

cd packages/api
python main.py

The API will be available at http://localhost:8000

Endpoints

Health Check

curl http://localhost:8000/health

Basic Analysis (/basic)

curl -X POST "http://localhost:8000/basic" \
  -H "Content-Type: application/json" \
  -d '{"code": "def add(x, y): return x + y", "language": "python"}'

Detailed Analysis (/premium)

curl -X POST "http://localhost:8000/premium" \
  -H "Content-Type: application/json" \
  -d '{"code": "def add(x, y): return x + y", "language": "python"}'

Development

# Run all packages in development mode
npm run dev

# Run tests
npm test

# Lint code
npm run lint

# Build all packages
npm run build

How It Works

1. The code is split into chunks so larger files can be evaluated on local signal instead of one global pass.
2. Each chunk is tokenized and measured for entropy, comment ratio, naming patterns, formatting consistency, and structural complexity.
3. Language-specific analyzers extract Python AST metrics or JavaScript structural metrics such as function count, loop count, try/catch usage, and nesting depth.
4. The detector computes both AI-oriented and human-oriented heuristic scores for each chunk.
5. Results are aggregated across chunks into a final classification, confidence score, and supporting signals.
6. If the snippet is too short or the heuristics conflict, the detector abstains with Uncertain (Mixed Signals).

Testing

The repository now includes regression tests for the detector in packages/core/tests. The current suite covers:

• AI-like and human-like Python samples
• AI-like and human-like JavaScript samples
• Low-signal abstention behavior
• End-to-end validation of the Python analysis bridge used by the TypeScript wrapper

Run the full workspace tests with:

npm test

Or run the detector regression suite directly with:

python3 -m unittest discover -s packages/core/tests -p 'test_*.py'

Contributing

We welcome contributions! Feel free to:

• Submit issues & feature requests.
• Improve multi-language support.
• Optimize AI detection heuristics.
• Expand API functionality.

License

Codect is licensed under the GNU General Public v3.0 license. You are free to modify and distribute the project as needed.

Contact

For questions or contributions, open an issue on GitHub or reach out to [email protected].

---

Let us know if you have any suggestions! 🚀