🧠 CodeLens — AI-Powered Code Review Platform

CodeLens is a modular, full-stack code review platform designed for analyzing and reviewing source code from multiple sources — including direct input, file uploads, and GitHub repositories. It uses asynchronous job handling for scalable performance, supports GPU acceleration, and is designed for easy future extension.

---

🚀 Features

🔧 Core Functionality

• Review raw source code, uploaded files, or entire GitHub repositories
• Asynchronous job queue with background task processing (asyncio)
• Real-time job tracking via job IDs
• Unified job creation and management
• Persistent job storage via database (SQLite3)
• Get Code reviews, Architecture Feedback with diagrams and Security findings

💻 Frontend (Next.js + TailwindCSS)

• Modern, minimal, responsive design using TailwindCSS
• Code editor built with Monaco Editor (@monaco-editor/react)
• Dynamic code input card with styled slider and animated transitions
• Snackbar notifications for job success / error states
• Enhanced buttons and inputs across pages (CodeInput, FileUpload, GithubRepoInput)
• Modular component design for reuse and scalability

⚙️ Backend (FastAPI + Python)

• Modular async architecture
• Unified job creation logic
• File, code, and GitHub URL analysis handled independently
• Built-in GPU support for ML-based review models (PyTorch)
• Centralized error handling and structured logging
• Integrated health checks and background workers

---

🧩 Directory Structure

CodeLens/
│
├── backend/
│   │
│   ├── app/                          
│   │   ├── __init__.py
│   │   │
│   │   ├── main.py                   
│   │   ├── config.py                 
│   │   ├── analyzer.py               
│   │   ├── db.py                     
│   │   ├── schemas.py                
│   │   ├── security.py               
│   │   ├── utils.py                  
│   │   ├── personalization.py        
│   │   ├── prepare_files.py          
│   │   │
│   │   ├── core/
│   │   │   ├── __init__.py
│   │   │   └── logging_config.py     
│   │   │
│   │   ├── models/
│   │   │   ├── __init__.py
│   │   │   └── job_status.py         
│   │   │
│   │   └── review_engines/
│   │       ├── __init__.py
│   │       ├── base.py               
│   │       └── python_engine.py      
│   │
│   ├── test_scripts/
│   │   ├── test_script1.py
│   │   └── test_script2.py
│   │
│   ├── main_modal.py                 
│   ├── requirements.txt
│   ├── reviews.db                   
│   └── personal.db                  
│
├── frontend/
│   ├── app/
│   │   ├── layout.tsx
│   │   ├── page.tsx
│   │   ├── layout.css
│   │   └── components/
│   │       ├── ArchitectureDiagram.tsx
│   │       ├── CodeInput.tsx
│   │       ├── FileUpload.tsx
│   │       ├── GithubRepoInput.tsx
│   │       ├── JobStatusLoader.tsx
│   │       ├── ReviewResult.tsx
│   │       └── StatusBadge.tsx
│   │
│   ├── lib/
│   │   └── api.ts
│   │
│   ├── styles/
│   │   └── globals.css
│   │
│   └── utils/
│       └── StarterCodes.ts
│
├── docker-compose.yml
├── Dockerfile.backend
├── Dockerfile.frontend
├── .gitignore
├── CODEOWNERS
├── LICENSE
└── README.md

---

⚡ Tech Stack

Layer	Technology
Frontend	Next.js 15, React, TailwindCSS, Monaco Editor
Backend	FastAPI, Python 3.10+, asyncio
AI/ML	PyTorch (GPU-accelerated)
Database	SQLite (dev) / PostgreSQL (prod)
Logging	Python logging module (configurable for Elastic/Kibana)
Job Management	Custom async queue with in-memory + DB persistence

---

🔁 Async Job Workflow

1. A new request (code, file, or repo) triggers schedule_job().
2. A unique job_id is created and stored in both memory and DB.
3. A background task (asyncio.create_task(process_job())) begins processing.
4. The frontend polls or queries /status/{job_id} to fetch updates.
5. Once completed, results or errors are updated in DB and memory.
6. Snackbar notifications inform the user on the frontend.

---

🧠 GPU Support

The backend supports GPU acceleration using PyTorch.

If you encounter CUDA issues, reload the UVM kernel modules:

sudo rmmod nvidia_uvm
sudo modprobe nvidia_uvm

Verify GPU availability:

import torch
print(torch.cuda.is_available())  # should return True

🧪 Local Development

Frontend

cd frontend
nvm use <version> # in case of multiple node versions
npm install
npm run dev

Backend

cd backend
python -m venv venv # To create virtual environment 
source venv/bin/activate
pip install -r requirements.txt

PyTorch Installation (Environment specific)

PyTorch packages are excluded from requirements.txt to avoid build and deployment conflicts.

For GPU environments:

pip install torch==1.13.1+cu117 torchvision==0.14.1+cu117 torchaudio==0.13.1+cu117 --index-url https://download.pytorch.org/whl/cu117

For CPU environments

pip install torch==1.13.1+cpu torchvision==0.14.1+cpu torchaudio==0.13.1+cpu --index-url https://download.pytorch.org/whl/cpu

Then, to run:

uvicorn app.main:app --reload

Next (possible) steps

• Multi-language static analysis engines
• Job queue persistence with Redis or RabbitMQ
• Authentication and user accounts
• Role-based dashboards for reviewers
• Live job progress via WebSockets