[P] I built a system that lets you ask questions about any GitHub repo and get answers grounded in the actual source code [P]
A developer has created GitRAG, a system that allows users to query any public GitHub repository and receive answers directly grounded in the source code. The tool utilizes a hybrid retrieval pipeline combining semantic chunking with AST parsing, dense embeddings, and BM25 keyword indexing. It then employs Reciprocal Rank Fusion and Cohere reranking before sending the top candidates to Groq's Llama-3.3-70b model for response generation, supporting over 15 programming languages. AI
IMPACT Provides a novel way for developers to interact with and understand codebases, potentially improving code comprehension and debugging.