Worked on the microsoft/agent-lightning repository to address a critical issue with local model naming, focusing on improving reliability in model selection workflows. Delivered a targeted bug fix by implementing standardized model path parsing and enforcing consistent naming conventions, which resolved incorrect model name reporting when using local models. This change ensured compatibility with the vLLM async server and reduced runtime errors in local deployments. The work involved Python development and required careful attention to detail in bug fixing, emphasizing robust handling of file paths and integration with asynchronous server components. The update enhanced overall stability for users deploying local models.