Lyra - Cognitive Code Intelligence

Lyra is a Flask web app for code analysis and lightweight execution. It estimates time/space complexity, detects risks and edge cases, and surfaces strategy insights. A small ML model adds a quality label for Python code.

Features

• Code analysis: cyclomatic complexity (Python), nested loop depth, time/space estimates.
• Risk and edge-case heuristics.
• Strategy detection and optimization suggestions.
• Built-in code execution for Python, Java, C, C++, and JavaScript.
• Chat assistant powered by Google Gemini API.

How It Works (High-Level Steps)

1. User selects a language and submits code in the web UI.
2. The backend runs analysis in analyzers/ and optional execution in executor/.
3. Results are rendered in the UI (complexity, risks, strategy, ML quality).
4. The chat widget calls /chat to generate responses from Gemini.

Tech Stack

• Backend: Flask
• Frontend: HTML/CSS/JS
• Analysis: Radon, custom heuristics
• ML: scikit-learn + joblib
• Execution: subprocess + temp files

Project Structure

Lyra/  
│
├── app.py
├── main.py
├── complexity_analyzer.py
├── requirements.txt
│
├── analyzers/
│   ├── __init__.py
│   ├── alignment_engine.py
│   ├── complexity_engine.py
│   ├── generic_engine.py
│   ├── risk_engine.py
│   └── strategy_engine.py
│
├── executor/
│   ├── code_runner.py
│   ├── java_executor.py
│   ├── c_executor.py
│   ├── cpp_executor.py
│   └── js_executor.py
│
├── ml/
│   ├── quality_model.py
│   └── train_model.py
│
├── static/
│   ├── style.css
│   ├── animations.css
│   ├── chatbot.js
│   └── code-executor.js
│
├── templates/
│   └── index.html
│
└── sample_codes/
    └── example1.py

Setup

1) Create and activate a virtual environment

python -m venv .venv
# Windows PowerShell
.\.venv\Scripts\Activate.ps1

2) Install dependencies

pip install -r requirements.txt

3) Configure environment variables

Create a .env or set variables in your shell:

• GEMINI_API_KEY (required for chat)
• GEMINI_MODEL (optional, default: gemini-2.5-flash)

4) Run the app

python app.py

Open https://lyra-yylx.onrender.com/ in your browser.

Code Execution Requirements

The /execute endpoint uses local compilers/runtimes. Install them and ensure they are in PATH.

• Python: bundled with your environment
• Java: JDK (java + javac)
• C: gcc
• C++: g++
• JavaScript: Node.js (node)

On Windows, MSYS2 UCRT64 works for C/C++:

C:\msys64\ucrt64\bin

API Endpoints

• GET / - UI
• POST / - Analyze code and render results
• POST /execute - Run code (Python/Java/C/C++/JS)
• POST /chat - Chat assistant (Gemini)

ML Quality Model

The ML model is stored at ml/quality_model.pkl. If you want to retrain:

python ml/train_model.py

Notes

• Python analysis uses AST + Radon for cyclomatic complexity.
• Non-Python languages use heuristic analysis in analyzers/generic_engine.py.
• Execution is time-limited to reduce runaway code.