Guardians of the Code

Code Guard combats the $2.4 trillion lost yearly to poor software quality. With 58% of devs losing 5+ hrs weekly to QA bottlenecks, it automates checks, cuts testing time 60%, and finds 90% of bugs.

Comment

Guardians of the Code

Inspiration

In Agile environments, QA testers often struggle with repetitive manual testing, unclear requirements, and missed loopholes. Developers spend hours explaining functionality, and when bugs slip past QA, fixing them consumes even more sprint time. These inefficiencies drag down velocity, inflate costs, and reduce overall sprint efficiency.

What it does

Guardians of the Code uses AI to automatically validate code against hyperspecific tasks derived from project documentation.
It checks updated or new files to ensure they align with functional goals, detect noncompliance, and highlight exactly which file, line, and rule failed — along with how to fix it.
Essentially, it acts as an intelligent QA assistant that boosts developer productivity and QA accuracy in Agile pipelines.

How we built it

We built the system using:

  • JavaScript and Python for core logic
  • OpenAI API for code understanding and reasoning
  • SQLite for lightweight data storage
  • FastAPI for backend endpoints
  • Next.js + React for the frontend dashboard
  • GitHub Actions, YAML, and webhooks for automated PR integration

A YAML-based rule engine defines hyperspecific tasks, while the AI agent analyzes incoming code via GitHub webhooks and generates structured, actionable outputs.

Challenges we ran into

  • Getting GitHub to send PR data to a hosted server-side endpoint.
  • Managing database connections between the frontend, backend, and hosted DB.
  • Designing structured agent outputs that consistently capture rule violations and fixes.

Accomplishments we’re proud of

  • Set up a fully functional GitHub workflow pipeline for automated validation.
  • Built an AI agent capable of structured reasoning and code compliance checks.
  • Implemented a YAML-driven configuration system for flexible, reusable rule definitions.

What we learned

  • How to create and fine-tune AI agents for structured reasoning.
  • How to work with YAML, GitHub Actions, and webhooks for seamless automation.
  • How to host and connect databases and server-side logic for scalable integrations.

What’s next for Guardians of the Code

Next, we aim to extend the platform to:

  • Automatically execute generated testing scenarios in isolated environments.
  • Further reduce manual QA effort and cost, preventing bugs from being deployed.
  • Improve the agent’s reasoning to adapt rules based on code patterns and documentation updates.

Built With

Share this project:

Updates