Show HN: RepoReaper – AST-aware, JIT-loading code audit agent (Python/AsyncIO) (github.com)

RepoReaper has launched a groundbreaking open-source project that introduces an intelligent, autonomous code audit agent built with Python and AsyncIO. This innovative system automates architectural analysis and semantic code searches, surpassing conventional code assistants by using a Large Language Model (LLM) as its processing engine and a Vector Store for dynamic context caching. Instead of using static indexing, RepoReaper employs a unique approach to build a lightweight symbol map of the repository's code using Abstract Syntax Trees (AST), selecting critical files for analysis in real-time. If context gaps appear during user queries, the agent smartly retrieves missing data from GitHub through Just-In-Time (JIT) reads, ensuring accurate and context-rich responses. This development is significant for the AI/ML community as it combines advanced techniques like Retrieval-Augmented Generation (RAG), dynamic prompt engineering, and a sophisticated Reasoning + Acting (ReAct) loop, providing a powerful tool for developers seeking insights into their codebases. The system also emphasizes memory efficiency and error resilience, operationalizing a streamlined user experience. By integrating structured chunking and a hybrid retrieval method that balances dense and sparse data, RepoReaper sets a new standard for intelligent code analysis tools, potentially transforming how software engineering teams interact with and understand their code.