What it does!
CodeSearch-Ext is a local, AI-powered semantic search tool for codebases that lets developers explore repositories using natural language instead of manual searching. The idea is simple: instead of guessing file names, you can ask questions like “where is authentication handled?” or “how does the app create a user session?” and instantly get relevant code sections ranked by meaning, not just keywords.
Inspiration
This project was inspired by a recurring frustration while working on medium-to-large codebases: even when you understand the general structure, finding where logic actually lives is slow and tedious. Tools like traditional search or even IDE “find in files” are limited because they only match text, not intent.
How we built it
We built the entire project using Cursor, which dramatically accelerated our development process. Cursor acted as an AI pair programmer, helping us iterate quickly on parsing logic, embedding pipelines, and search functionality while keeping us focused on system design rather than boilerplate code.
Challenges we ran into
One of the biggest challenges we faced was getting the system to actually work end-to-end. While each individual piece (parsing files, generating embeddings, and running similarity search) seemed straightforward on its own, integrating them revealed a lot of small issues. The code often broke due to mismatched data formats, missing dependencies, and inconsistent chunking of the codebase.
Accomplishments that we're proud of
Despite the setbacks, we’re proud that we got a fully working end-to-end system. CodeSearch-Ext successfully turns natural language queries into meaningful code navigation, and it actually works on real repositories without needing any API keys.
Log in or sign up for Devpost to join the conversation.