Spaces:
Sleeping
Apply for a GPU community grant: Personal project
GitRead is an open-source AI-powered developer tool that lets users chat with any GitHub repository in plain English. Developers paste a GitHub URL and can instantly ask questions like "How do I run this?", "What does this file do?", "Find the bugs", or "Explain the architecture" — without reading a single line of documentation.
The tool uses a RAG pipeline built with LangGraph, sentence transformers for local embeddings, ChromaDB for vector search, and Google Gemini as the LLM. The entire pipeline runs locally with no external embedding APIs, which makes it fast on GPU but slow on CPU.
GitRead is aimed at developers onboarding to new codebases, open source contributors trying to understand unfamiliar repos, and anyone who wants to understand code faster.
The project is fully open source at github.com/vidit1920/GitRead and live at huggingface.co/spaces/toasted1920/GitRead.
ZeroGPU would make a significant difference — the embedding and vector search steps that currently take 5-10 minutes on CPU complete in under 30 seconds on GPU, which is essential for a good user experience on large repositories.
@huggingface-team @huggingface
Hi! I would appreciate if the team could review this GPU grant request for GitRead.
GitRead is an open-source tool that helps developers understand GitHub repositories using AI and RAG.
Thank you for your time!