The latest articles on Forem by Derin Akay (@dxa204). https://forem.com/dxa204 https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F3745610%2F273f57e8-d1f5-418f-80c8-bdb16e27677c.png https://forem.com/dxa204 en Derin Akay Fri, 06 Feb 2026 06:07:36 +0000 https://forem.com/dxa204/git-cluster-rag-semantic-routing-for-git-history-copilot-cli-challenge-4fd0 https://forem.com/dxa204/git-cluster-rag-semantic-routing-for-git-history-copilot-cli-challenge-4fd0 <p><em>This is a submission for the <a href="https://dev.to/challenges/github-2026-01-21">GitHub Copilot CLI Challenge</a></em></p> <p>Repository: <a href="https://github.com/dxa204/git-cluster-rag.git" rel="noopener noreferrer">https://github.com/dxa204/git-cluster-rag.git</a></p> <h2> What I Built </h2> <p>I built Git Cluster RAG, a command-line tool that uses K-Means Clustering to "route" questions about a repository's history to the correct context.</p> <h2> The Problem </h2> <p>Standard RAG (Retrieval-Augmented Generation) applications are "flat." If you ask a question like "Why did we remove the notes file?", a standard vector search might retrieve unrelated commits just because they share keywords. It struggles to distinguish between Code Refactoring, Documentation Updates, and One-off Cleanups.</p> <h2> The Solution </h2> <p>My tool uses the GitHub Copilot CLI to build a pipeline that:</p> <ol> <li>Ingests commit history (messages + file diffs).</li> <li>Embeds the changes using sentence-transformers.</li> <li>Clusters the commits using K-Means.</li> <li>Routes user queries to the specific semantic cluster (e.g., "Cluster 0: Maintenance") before retrieving answers. This "Cluster-Guided" approach ensures that when I ask about a deleted file, the system prioritizes "Cleanup" commits over "Feature" commits.</li> </ol> <h2> Demo: Cluster-Guided Routing in Action </h2> <p>In this video, you can see the tool ingesting the git history, identifying the clusters, and then correctly routing a specific query about a deleted file to the "Maintenance" cluster.</p> <p><a href="https://youtu.be/FY4GY0uqMxI" rel="noopener noreferrer">https://youtu.be/FY4GY0uqMxI</a></p> <h2> My Experience with GitHub Copilot CLI </h2> <p>Building this project entirely with the Copilot CLI changed my workflow from "Stack Overflow searcher" to "Command Line architect."</p> <ol> <li><p>Scaffolding with Context<br> I used the <code>@workspace /new</code> command to generate the entire project structure (<code>ingest.py</code>, <code>cluster.py</code>, <code>chat.py</code>) in one go. Instead of writing boilerplate, I could focus on the logic of the K-Means algorithm.</p></li> <li><p>The "Agent" Workflow<br> The standout feature for me was the <code>/init</code> command. By running this, I was able to generate a <code>.github/copilot-instructions.md</code> file that taught Copilot the specific constraints of my project (e.g., "Always use 3 clusters", "Truncate diffs to 500 chars"). This effectively turned Copilot into a specialized teammate that knew my architecture, not just a generic code generator.</p></li> <li><p>Frictionless Debugging<br> When I hit syntax errors or needed to generate dummy git data for testing, I didn't leave the terminal. I used <code>gh copilot suggest</code> to generate complex shell commands that created dummy commits, enabling me to test the clustering algorithm in seconds.</p></li> </ol> devchallenge githubchallenge cli githubcopilot