In January 2026, J.H. Wessling enhanced the ai-dynamo/aiperf repository by developing OpenAI chat streaming and token counting features using Python. Wessling implemented dual-format support in the BaseEndpoint, enabling both streaming and non-streaming OpenAI chat responses while improving token counting accuracy. The work included adding comprehensive unit tests for streaming formats to ensure correctness and reliability. Additionally, Wessling addressed output token counting in the TemplateEndpoint, which improved cost visibility and performance for chat workloads. This focused backend development effort provided a more robust foundation for scalable streaming features, with an emphasis on API development and thorough test coverage.