What Inspired Me
My inspiration came directly from the "DevOps" Server example in the hackathon overview. As a computer science student, I know that when a production site fails, the first question is always, "What changed?" I wanted to build a tool that answers that instantly. My goal was to create a "Code Red" assistant that could intelligently correlate a system alert with the exact code deployment or feature toggle that caused it, solving a high-stakes, real-world problem.
How I Built It
I built the project as a two-part system. The backend is an MCP server built with Python and FastAPI. It uses "Service Connectors" to talk to external APIs, a "Correlation Engine" to find relationships between events, and a "Context Formatter" to create a clean summary for the AI. The frontend is a simple "Before vs. After" demo built with Streamlit, which shows how useless an AI is without context and how powerful it becomes with the context my server provides.
What I Learned
This hackathon was a huge learning experience. My biggest takeaway is the true value of "signal vs. noise." An AI is only as good as its context, and the real magic is summarizing—turning 1,000 log lines into one critical sentence. I also learned how to build a robust API with FastAPI, how to build a compelling demo with Streamlit, and, thanks to GitHub's Push Protection, I learned a lot about Git security.
Challenges I Faced
My biggest challenges were in the real-world workflow. I spent time fixing Python environment bugs like ModuleNotFoundError and case-sensitivity issues. My most significant hurdle was GitHub's security, which blocked my push because I had accidentally left a placeholder token in an example file. This forced me to learn the proper Git workflow for security, using git commit --amend to fix the commit before pushing. It was a stressful but incredibly valuable lesson in security-first development.
Log in or sign up for Devpost to join the conversation.