Worked on backend stability and reliability for the LightRAG repository, focusing on refining caching and initialization logic using Python. Addressed two critical bugs by stabilizing the hashing_kv initialization process across all operational modes, including hybrid, through targeted refactoring and improved configuration checks. Enhanced the cache handling mechanism to respect the global_config enable_llm_cache flag, ensuring that cache operations are bypassed when disabled, which prevents unnecessary processing. This work involved backend development, code optimization, and LLM integration, resulting in a more robust and maintainable system foundation that supports future feature development while minimizing runtime errors and configuration-related issues.