Building an efficient code search tool for Graphite Chat that supports searches across vast codebases and arbitrary commits posed significant challenges, leading to a reevaluation of traditional methods. While modern tools like grep can quickly search files on a local disk, scaling this functionality to handle millions of files across non-default branches was problematic. Initial attempts with AWS-based solutions demonstrated performance disparities, especially for large repositories, where caching limitations became apparent. Instead of indexing repository states for each commit, which proved unfeasible due to the massive data volume, the team drew inspiration from Git’s efficient storage model, utilizing "blobs" and "trees" to streamline searches. This method allowed for fast, parallel queries, drastically reducing search latency to under 100 milliseconds. The innovative approach, now operational within Graphite Chat, surpasses previous methods involving the GitHub API, offering targeted branch searches and improved file retrieval without rate limits. The exploration into Git’s storage strategies suggests further potential optimizations and foreshadows additional insights into repository management in future discussions.