Worked on stabilizing Azure OpenAI deployments within the HKUDS/LightRAG repository by addressing a deployment configuration issue that previously caused connection and model-selection errors. Implemented a fix in Python that prioritizes environment variables for model and deployment names, ensuring the Azure OpenAI client initializes with the correct deployment identifier. This technical approach improved the reliability and robustness of Azure-based LLM integrations, reducing downtime and operational risk. The work focused on cloud deployment and LLM integration, resulting in smoother production rollouts and more predictable onboarding for Azure OpenAI services, ultimately enhancing the dependability of AI-driven responses in production environments.