This is a submission for the Hacktoberfest Weekend Challenge: Build for a Friend
What I Built
Night-Owl is an AI-powered systems debugging and learning platform. You paste code, logs, errors, or networking output, pick a domain (C/C++, OS/Concurrency, or Networking), and it explains what went wrong, why it happened, the evidence, the root cause, how to fix it, and the systems concept behind it. It also generates an Exam Mode summary with a definition, a short exam answer, and viva questions.
Who I built it for: my fellow students and friends studying Operating Systems, Computer Networks, and C/C++. Most of us have hit the same wall: you fix a bug, but you can't explain it when a professor asks in a viva.
The problem it solves: systems bugs are hard because the visible error is rarely the real root cause. Take this:
int main() {
int* p = new int(10);
delete p;
*p = 20;
}
The interesting part isn't "it might crash". It's the chain: delete p frees the memory, p becomes a dangling pointer, and *p = 20 touches memory whose lifetime has ended. That's a use-after-free, which is undefined behavior. Generic AI fixers hand you a patch. Night-Owl walks you through that whole chain so you can understand and explain it.
Demo
Live demo: https://night-owl-woad.vercel.app
Select C / C++, paste the snippet above, and submit. On the production deployment, Night-Owl detected it as a Use-After-Free with Critical severity, and returned the root cause, evidence, why it happened, a suggested fix, the systems concept (dangling pointers and RAII), and Exam Mode with viva questions.
The backend is hosted on Render, so the first request after inactivity may be slow.
Code
Devansh270
/
Night-Owl
First Hacktober 2026 project
🦉 Night-Owl — AI Systems Debugger
Debug the systems. Understand the problem.
Night-Owl is an AI-powered systems debugging and learning platform for students and developers working with C/C++, Operating Systems, Concurrency, and Computer Networking. Paste code, logs, errors, or network output, pick a domain, and Night-Owl explains what went wrong, why it happened, the evidence, the root cause, how to fix it, and how to explain it in an exam or viva.
Unlike a generic "AI code fixer", Night-Owl is built around understanding. It combines deterministic analyzers that collect concrete evidence with Google Gemini that turns that evidence into a structured, teachable explanation.
🔗 Links
| 🚀 Live Demo | night-owl-woad.vercel.app |
| ⚙️ Backend API | night-owl-4rg7.onrender.com |
| 💻 Repository | github.com/Devansh270/Night-Owl |
| 🐛 Issues | github.com/Devansh270/Night-Owl/issues |
📑 Table of Contents
- Repository: https://github.com/Devansh270/Night-Owl
- Backend API: https://night-owl-4rg7.onrender.com
How I Built It
I want to be upfront: the AI layer uses Google Gemini through the Vercel AI SDK (@ai-sdk/google). That is a hosted API, not an open-weight model or local inference.
What I focused on instead is making the project not depend on the AI to find everything. Night-Owl uses a hybrid architecture: deterministic analysis plus AI explanation.
User Input → Domain Selection → Deterministic Analyzer → Concrete Evidence
→ Gemini AI → Structured Root-Cause Explanation → Exam Mode → Database / History
-
Deterministic analyzers run first and collect concrete evidence. Today there are analyzers for deadlocks (
deadlockDetector.ts), race conditions (raceConditionDetector.ts), memory errors (memoryErrorDetector.ts), and networking (networkAnalyzer.ts). - Gemini turns that evidence into a structured explanation: problem, severity, root cause, evidence, why it happened, suggested fix, concept, and Exam Mode.
- History stores the structured analysis with Prisma and PostgreSQL (Supabase), so you can revisit past sessions.
Stack:
| Layer | Technologies |
|---|---|
| Frontend | React, TypeScript, Vite, React Router, Axios, Lucide React |
| Backend | Node.js, TypeScript, Express, Mastra |
| AI | Google Gemini, Vercel AI SDK, @ai-sdk/google
|
| Database | PostgreSQL, Supabase, Prisma ORM |
| Deployment | Vercel (frontend), Render (backend), Supabase (database) |
Why Does Open Innovation Matter?
Honestly, Night-Owl's AI layer currently relies on a closed API, so I can't claim it is fully open end to end. But open innovation still shaped it:
- The project itself is open source under the MIT License, so any student can read how it works, run it, and learn from the code.
- The deterministic analyzers are plain, readable code. Contributors can add new ones (buffer overflow, double-free, memory leaks, starvation, livelock, DNS and ARP troubleshooting) without touching the AI at all.
- It's built on open-source tooling: React, Express, Prisma, the Vercel AI SDK, and more.
- There's room to go further. The Vercel AI SDK supports multiple model providers, so adding support for open-weight models would be a natural contribution for someone who wants to take that on.
For an educational tool, openness matters because students should be able to see how a bug was diagnosed, and the community should be able to extend it.
Night-Owl is open to contributions and is suitable for developers participating in Hacktoberfest and other open-source initiatives. Ideas include new analyzers, deadlock graphs, race-condition timelines, better prompts, tests, and docs. The README has setup steps and a contribution workflow.
If you try it, I'd love to hear which systems bug always confused you the most. And a star on GitHub means a lot.
Debug the systems. Understand the problem.
Top comments (0)