AWS Builder Center

Code Sherpa 🧗

Code Sherpa is the missing onboarding layer for open-source beginners. Instead of leaving you to drown in a 500-file repository, it takes a GitHub issue, isolates the exact files you need to look at, and explains the bug in plain English. We don't write the code for you—we just give you the roadmap to make your first real commit.

Demystifying Open Source: How We Built Code Sherpa to Guide First-Time Contributors

By: Team meridian
Open source is often called the heartbeat of the software world. But let’s be honest: for a first-time contributor, staring down a massive, established repository is absolutely terrifying.
You want to help. You find an issue labeled "good first issue," but when you clone the repo, you are hit with a wall of 500+ undocumented files. You don’t even know where to begin asking questions, let alone where to start writing code. That overwhelming wall is exactly the barrier my teammate and I hit, and it is the exact reason we built Code Sherpa.

The Problem: The "500-File Wall"

We realized that the gap between wanting to contribute and submitting that first pull request isn't a lack of coding skills—it's a lack of context. Beginners don't need an AI to write the code for them; they need a mentor to point to the right file and say, "The bug is happening in this function, and here is why."
Without that guidance, most students and junior developers simply give up, leaving thousands of open-source issues unresolved.

The Build: Enter Code Sherpa

We built Code Sherpa to be that mentor. It is an AI-driven, issue-based onboarding tool designed specifically for developers.
Instead of spoon-feeding the final code, Code Sherpa takes a GitHub issue URL, scans the repository, and isolates the exact 2 or 3 files where the problem lives. We designed the frontend to reflect a premium, high-contrast "brutalist" developer environment (inspired by tools like Vercel and Linear) so it feels like a native part of a developer's workflow.
It breaks down the root cause in plain English, pinpoints the bug's location, and outlines step-by-step hands-on debugging instructions. It teaches the user how to fix the bug, ensuring they actually learn the codebase in the process.

Architecture: Where AWS Fits (And How We Adapted)

Building this required a robust cloud backend to handle fetching repositories, processing code text, and generating AI insights in real-time.
Our initial architecture was heavily centered around AWS Bedrock. We wanted to leverage Bedrock's scalable foundation models to power the core intelligence of Code Sherpa. However, we ran into a classic real-world cloud engineering hurdle: as students working with a brand-new AWS account, we faced service limits and account verification delays that prevented us from accessing the Bedrock models in time for the hackathon deadline.
In the true spirit of a hackathon, we had to pivot gracefully. We quickly re-engineered our API pipeline to use the Gemini API as our generative intelligence engine. Because we had built a modular backend, swapping out the AI provider didn't break our system.
Despite the pivot, the AWS ecosystem remained central to our infrastructure design. We utilized AWS S3 (via LocalStack for rapid local iteration) to handle our backend telemetry, session logging, and receipt storage. Every time a user analyzes a repository, the metadata and run logs are structured and stored, allowing us to maintain a transparent, highly decoupled cloud architecture.

What We Learned

This project was a massive lesson in cloud resilience. Hitting the AWS Bedrock account limitation forced us to think about fallback systems, modular API design, and how to keep a project moving forward when a critical dependency is blocked.
Ultimately, Code Sherpa is exactly the tool we wished we had when we first started coding. By combining an intuitive, developer-first UI with a resilient cloud backend, we hope to lower the barrier to entry for open source, turning intimidated beginners into confident contributors.
Any opinions in this article are those of the individual author and may not reflect the opinions of AWS.
Enjoyed reading this content? Let the author know!

Your likes, comments, shares, and saves help creators reach more builders.

Loading recommendations

Loading article