Developed comprehensive documentation for the aws-neuron/aws-neuron-sdk repository, focusing on the vLLM Online Inference Bucketing Guide. This work introduced a new section in the vLLM user guide, detailing how to specify context and token buckets for online inference and configure the OpenAI-compatible server using override_neuron_config for prefill and decode workloads. The documentation, written in reStructuredText (rst), guides users in optimizing inference performance and achieving predictable latency through explicit bucketing parameters. Emphasizing clarity and actionable steps, the contribution advanced the SDK’s documentation quality, supporting customers in tuning their AWS Neuron workloads for improved efficiency and operational consistency.