DeepSeek Harness Plugin: dsh-plugin-abtest
dsh-plugin-abtest is a plugin for DeepSeek Harness (DSH) designed to address the challenge of validating plugin changes. It performs paired A/B experiments in isolated environments and supports evaluating task success rate, token consumption, latency, and tool error rate. It provides auditable evidence, deterministic release gates, and configurable rules to help teams determine whether a change is release-ready before publishing. It only generates offline recommendations and does not automatically modify systems. It requires a Node.js environment and is intended for DSH plugin developers.
Read More