dsh-speak-plus: Vocalizing DSH's Answers and Intermediate Steps

dsh-speak-plus is an enhanced branch of dsh-speak 1.8.4 designed for the DSH client, enabling the reading aloud of both the agent's responses and intermediate processes. When the upstream version reads the intermediate process line by line, fragments between tool calls are fragmented, resulting in a poor listening experience. This branch accumulates continuous text into paragraphs before reading and defines interruption timing. New capabilities include merging the intermediate process into paragraphs, preemption protection for approvals or questions, content handover per turn, one-click stop via Alt+S, and falling back to reading the profile configuration when the settings schema is unavailable. All new behaviors are disabled by default or set to conservative values; without configuration, the behavior remains consistent with the upstream version. The plugin is mounted as a DSH bundle and configured via the profile's `cordis.patch.yml`; it uses SAPI voices on Windows and the `say` command on macOS. This article provides a reproducible installation and configuration guide, a comparison of new and inherited configuration options, typical usage patterns, and known limitations.