Back to Projects
Case Study
Carto
AI Codebase Bundler for VS Code
TypeScriptVS Code APIGoogle GeminiOpenAIGroqOllama
Overview
Carto bundles your entire codebase into a single, beautifully structured Markdown file — ready to paste into any AI assistant. It supports multi-provider AI (Gemini, OpenAI, Groq, Ollama), auto-detects your tech stack, and runs entirely locally on your machine. Features intelligent context bundling with a table of contents, tech stack detection, directory tree, and full file contents grouped by folder.
Features
- Multi-provider AI integration — Gemini, OpenAI, Groq, Ollama
- Intelligent context bundling with TOC, tech stack detection, and directory tree
- AI toggle for senior-engineer-level codebase analysis
- Full output preview with rendered markdown inside VS Code sidebar
- Auto tech stack detection from package.json, requirements.txt, go.mod, Cargo.toml, and more
- Security scanning — detects and excludes .env, private keys, and credentials
- Local AI support via Ollama for 100% offline operation
Challenges
- Abstracting multiple AI providers behind a unified interface with different streaming behaviors
- Building a performant markdown preview renderer within VS Code WebView constraints
- Implementing accurate tech stack detection across 10+ ecosystem file formats
- Designing security scanning that catches sensitive files without false positives
Learnings
- Deep understanding of VS Code extension architecture and WebView communication
- Multi-provider LLM integration patterns and streaming response handling
- File system scanning optimization for large codebases
- Security pattern detection and file exclusion strategies