Worked on backend reliability improvements for the infiniflow/ragflow repository, focusing on API integration and Python development. Addressed a specific bug by standardizing the GPUStack Chat Model’s base URL to include the required /v1 suffix, ensuring consistency with other model types and preventing misrouting issues. This targeted change reduced runtime errors and improved the developer experience for GPU-based chat workflows. The solution was implemented through a traceable, issue-linked commit, enhancing maintainability and code clarity. By aligning URL handling across models, the work contributed to more reliable customer-facing workflows and streamlined future backend development within the project’s API infrastructure.