Back to ProjectsAI/ML

SentinelAI

AI-powered code security and quality scanner. Paste a GitHub repo URL and get a detailed report on bugs, vulnerabilities, and code quality — all running locally with no data sent to the cloud.

Tech Stack

Node.jsExpressMongoDBOllama (llama3)

Overview

SentinelAI is a local-first AI code analysis tool that scans GitHub repositories for security vulnerabilities, code smells, and quality issues. It fetches repository files via the GitHub API, feeds them to a locally running LLaMA 3 model through Ollama, and returns a structured report covering bug risks, security flaws, and improvement suggestions — all without sending your code to any external AI service.

Key Highlights

  • 100% local AI — no data leaves your machine
  • Scans any public GitHub repo by URL
  • Structured severity-based reports
  • MongoDB-backed scan history

How I Built It

Built a Node.js/Express backend that accepts a GitHub repo URL, uses the GitHub REST API to fetch file contents, then streams each file to Ollama's local LLaMA 3 model with a carefully crafted security-analysis prompt. Results are aggregated and stored in MongoDB for history. The frontend is a React SPA with a clean report viewer that highlights issues by severity.

Future Vision

Support for more languages, CI/CD integration (GitHub Actions), a VS Code extension, and a diff-based re-scan that only analyzes changed files.

Links