Aegis：以本地開源 AI 驅動的私隱優先 DevSecOps 副業項目

Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle Submission 💸

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle

Team Submission:

• Krish (@krishtrue)
• Keshav (@keshav)

What I Built

Every developer and freelancer knows the feeling: you run a security audit before handing off a project to a client, and suddenly your terminal floods with dozens of alerts — leaked API keys, outdated CVEs, SQL injection risks, and unescaped inputs.

Triaging and manually writing code patches for all of them takes hours of tedious context switching.

Large enterprises spend thousands of dollars on tools like Snyk or GitHub Advanced Security. But for freelancers, boutique development agencies, and solo founders, existing solutions present two massive roadblocks:

1. The Privacy & Compliance Dilemma: You cannot feed confidential client code or unpatched zero-day vulnerabilities into closed cloud AI APIs (like OpenAI or Anthropic). Doing so violates NDAs, enterprise security policies, and client trust.
2. The Per-Token Margin Killer: Closed AI APIs charge per token. Analyzing whole files and generating diffs for hundreds of security alerts across dozens of repos destroys the economics of a security side hustle.

Enter Aegis 🛡️

Aegis is a privacy-first, automated DevSecOps platform and security remediation bot with open-source AI at its very core.

It connects to your GitHub repositories via a GitHub App, triggers asynchronous security scans (Semgrep for static analysis, Gitleaks for secrets, Trivy for dependencies), and uses local open-source AI models via Ollama optimized with Direct Preference Optimization (DPO) to automatically generate surgical code fixes and open pull requests directly on GitHub.

mermaid
flowchart LR
    A[GitHub Repo] --> B[Aegis Microservices]
    B --> C[Transient Scanners\nSemgrep / Gitleaks / Trivy]
    C --> D[Local Ollama Engine\nOpen-Weight Coding LLM + DPO]
    D --> E[Automated GitHub PR\nPatch + AI Code Review]