RepoPilot: Understand, run and validate any codebase

**Title: I spent months building an AI tool that understands your GitHub repo and generates tests. Finally launched it — looking for honest feedback!** Hey everyone! 👋 I'm a software engineer with 10+ years of experience, currently working full-time, and I've been building a side project called **RepoPilot**. The idea came from a problem I've faced repeatedly: joining an unfamiliar codebase, spending hours understanding dependencies, figuring out how everything connects, and then writing tests for code I barely know. I wanted to see whether AI could make that process easier — but without just dumping an entire repository into an LLM. So I built RepoPilot. **Here's what it does:** - 🔗 Connect GitHub, GitLab, or Bitbucket repositories (including private repos). - 🧠 Analyze code structure using AST parsing and dependency graphs. - 💬 Chat with your codebase using repository-aware context. - 🧪 Generate unit and integration tests for specific files. - 🎯 Describe business scenarios in plain English and generate relevant tests. **Tech stack:** Vue 3, FastAPI, TypeScript Compiler API, Python, and LLM integrations. One of the most interesting challenges was building a context engine that identifies relevant files and dependencies instead of blindly passing everything to the AI model. It's now live, and I'm looking for developers willing to try it. 🔗 **https://repopilotapp.com** It's still an early release, and I'm sure there are things that need improvement. **I'd genuinely appreciate feedback on three things:** 1. Does the repository-aware chat actually help you understand unfamiliar code? 2. Are the generated tests useful, or would you rather write them yourself? 3. What's the biggest missing feature that would make you use something like this regularly? I'm particularly interested in hearing from developers who maintain large or unfamiliar codebases. Happy to answer technical questions about the architecture or implementation. Thanks for checking it out! 🙌