Worked on the hsliuping/tradingagents-cn repository to enable local Llama3.1 inference and enhance configurability for QuickThink LLM within trading workflows. Developed a feature that allows users to select local models through a configurable option, replacing previous hard-coded references and reducing the risk of misconfiguration. Integrated Ollama to support local inference, which accelerates experimentation and offers potential cost savings by minimizing reliance on remote model calls. Utilized Python for both AI integration and API development, with a focus on robust data handling. Also improved user-facing documentation by correcting parameter descriptions, contributing to clearer configuration and easier model management.