dsh-reasoning-merge

DSH renders model responses block by block, as returned by the provider. When a single model call splits its reasoning into several segments, the UI stacks one thinking area after another, making the output feel noisy and hard to follow. dsh-reasoning-merge solves exactly that: within a single model call it joins adjacent reasoning segments into one cohesive block, inserting a single blank line between them, so the interface shows just one clean thinking area.\n\nThe plugin operates purely at the chunk-joining level. The reasoning text is neither added to nor trimmed, and everything else—text output, tool calls, token usage, finish reason—passes through untouched. There are no settings and no extra UI controls; install it, restart DSH, and it simply works. One boundary to keep in mind: merging only applies to adjacent reasoning segments inside the same model call. The moment a text block or tool call appears in between, the current segment closes and subsequent reasoning starts a fresh block, which feels perfectly natural.\n\nIt is a good fit for anyone who uses DSH as a daily client and whose provider tends to emit reasoning in multiple chunks. Zero configuration, zero interface overhead—just one clean, continuous thinking section where there used to be several.