UnderstandYour

RAG-powered code understanding for JavaScript & TypeScript repositories. Ask questions. Get answers with citations.

// code analysis in progress

>

GitIt is a RAG-powered code understanding engine built for JavaScript and TypeScript repositories. It clones your repo, parses every file into an AST, chunks the code into retrievable units with embeddings, and answers natural-language questions grounded in your actual codebase.

// ask "How does caching work?" and get cited answers

++++

Git Integration

Clone repos, track branches, parse diffs. Seamless connection to your GitHub workflow.

++++

AST Parsing

Full abstract syntax tree analysis. Understand code structure, not just text.

++++

Hybrid Search

BM25 keyword search combined with vector similarity. Best of both retrieval worlds.

++++

RAG Pipeline

Retrieval-augmented generation grounded in your actual codebase. No hallucinations.

++++

Future Roadmap

  • >>Multi-language support (Python, Go, Rust)
  • >>Team knowledge base
  • >>IDE plugin
  • >>Custom embedding models
  • >>Self-hosted deployment

>

I kept wondering — when an AI agent works on your code, how does it actually know which files to look at? Out of thousands of files in a repo, something has to figure out which ones matter for the task at hand.

That question pulled me into a rabbit hole: how do you turn source code into something a model can reason over? How do indexes, embeddings, and retrieval ranking actually work under the hood — not as a black box, but piece by piece?

This is a pure curiosity project. No product requirements, no deadlines. Just building each layer from scratch — tokenizers, parsers, embeddings, search — to understand how it all connects.

>

GitHub Integration

Connect your repos, track branches, and analyze code changes in real-time.

AST Parsing

Full abstract syntax tree analysis using a Pratt parser to generate symbol tables and understand code structure.

Hybrid Search

Combines BM25 lexical search with vector embeddings and AST-structural signals for precise retrieval.

RAG Pipeline

Code-aware chunking transforms files and functions into retrievable units with embeddings.

Natural Language Q&A

Ask questions in plain English and get grounded answers with source citations from your codebase.

Code Search

Advanced search with filters, boolean queries, and field-scoped syntax for precise code discovery.

Code Preview

Monaco editor integration with syntax highlighting and line number navigation for seamless code exploration.

Context Citations

See exactly where answers come from with file paths and line ranges linked directly to source.

>

RAG SystemsBM25 & Inverted IndexHybrid RetrievalAbstract Syntax TreesSymbol TablesRAG SystemsBM25 & Inverted IndexHybrid RetrievalAbstract Syntax TreesSymbol Tables
Code ChunkingVector EmbeddingsANN SearchDeveloper Search UXRetrieval MetricsCode ChunkingVector EmbeddingsANN SearchDeveloper Search UXRetrieval Metrics
{ }

Readyto

Connect your GitHub. Ask questions about your codebase. Get answers grounded in your actual code.

Free and open source. Your code stays yours.