The incredible generative capabilities of Large Language Models (LLMs) have ushered in a new era of automation in coding tasks. Applications like Amazon Code Whisperer, GitHub Copilot, and Replit have become ubiquitous, harnessing LLMs to effortlessly complete code segments or execute code modifications based on natural language instructions. Despite their remarkable performance, these tools encounter challenges when tackling repository-level coding tasks.