Developed a flexible data-part conversion hook for model input processing in the vercel/ai repository, enabling custom data types such as URLs and code files to be seamlessly converted into text or file parts suitable for AI models. Leveraged TypeScript and full stack development skills to introduce an optional convertDataPart callback within the convertToModelMessages function, enhancing data ingestion reliability and reducing manual preparation. The implementation included comprehensive automated tests covering multiple conversion scenarios, along with thorough documentation and API reference updates. This work improved the adaptability of model workflows, supporting richer data inputs and facilitating quicker, more reliable model deployment.