Back to directory

dsh-speak-plain

In a curated listMaintenance: Active

exception-h/dsh-speak-plain

Rewrites completed DSH answers in plain language while keeping the original answer available for inspection.

View on GitHub
$ dsh plugin --profile desktop add -w "github:Exception-H/dsh-speak-plain"

14

stars

2

forks

JavaScript

Language

MIT

License

2026-11-01

Created

2026-11-03

Last push

MIT-licensed DSH plugin that rewrites finished answers in plain language while keeping the untouched original, code and GenUI fences one click away.

DSH integration

Native runtime

Author-claimed

Safety audit

Unaudited

Last verified

2026-11-14

License

MIT

01What can it help you accomplish?

  • Read a finished DSH answer without decoding report-speak

    The same answer rewritten in plain language and shown as the default body, with the untouched original folded behind a "view original answer" toggle

    DeepSeek Harness users who want the final answer understandable at a glance but still auditable

  • Rewrite only the one answer that was hard to read, instead of every turn

    A per-answer "polish this answer" action that produces the plain-language version on demand

    Cost-conscious users who do not want an extra model call after every single turn

  • Keep the original wording, code and evidence for anything that matters

    The native DSH rendering of the original answer (code, images, thinking included) stays one click away, and execution traces plus later model context still use the original text

    Engineers and reviewers who must verify technical details instead of trusting a paraphrase

02How to install into DeepSeek Harness

Prerequisites

  • A DeepSeek Harness profile to install into (`desktop`, or `web` for the web profile)
  • A channel and model already configured in DSH — the plugin reuses DSH's model routing and never asks for an API key

Installation steps

  1. 01

    Install from GitHub: `dsh plugin --profile desktop add -w "github:Exception-H/dsh-speak-plain"`

    $ dsh plugin --profile desktop add -w "github:Exception-H/dsh-speak-plain"

  2. 02

    Or from a local checkout: `git clone https://github.com/Exception-H/dsh-speak-plain.git`, then `dsh plugin --profile desktop add -w ./dsh-speak-plain`

    $ git clone https://github.com/Exception-H/dsh-speak-plain.git

  3. 03

    Or from the release tarball: `dsh plugin --profile desktop add -w "完整路径/dsh-speak-plain-0.3.1.tgz"`

    $ dsh plugin --profile desktop add -w "完整路径/dsh-speak-plain-0.3.1.tgz"

  4. 04

    Restart DSH, open Settings → 说人话, pick the channel and model, tick "enable" and save

Verify the integration

  • After the restart the 说人话 settings page is reachable from DSH settings
  • Newly finished ordinary main-session answers get polished automatically; answers already in history are not re-run

03DSH integration and capability boundaries

DSH integrationNative runtime

DSH plugin installed with `dsh plugin --profile desktop add`; it hooks `agent/turn-stopping` and renders a `conversation.chat.turnTail` panel that rewrites the finished answer and hides the native one behind a toggle

  • Final-answer rewrite in plain language

    the last user question, the rewritable text of the finished answer, and one expression preference→a polished answer rendered by the turn-tail panel and shown as the default body of that turn

    spends quota on the selected polishing modelhides the native answer row with a CSS class until the toggle is usedstores the saved polish result in the profile's plugin storage-domain
  • On-demand polish of a single answer

    one chosen answer in the chat→the plain-language version of just that answer, leaving the rest of the session untouched

    spends quota only for the answers you ask to polish
  • Protected code, links and GenUI fences

    answer text containing code fences, inline code, Markdown links or `dsh-ui` fences→the same constructs restored after polishing; `dsh-ui` fences in the polished area rendered by the public GenUI embed component

    unparsable components keep their code plus an explanatory note

04Who is it for? When not to use it?

Good for

  • DeepSeek Harness users who want the final answer understandable at a glance but still auditable
  • Cost-conscious users who do not want an extra model call after every single turn
  • Engineers and reviewers who must verify technical details instead of trusting a paraphrase

Not for

  • Only normal, uninterrupted final answers with no tool calls and a body are polished; sub-agent output, failures, truncated output and empty answers are skipped, and tool steps or running answers are never touched.
  • The plugin targets DSH builds that expose `agent/turn-stopping`, `ctx.llm.stream()`, `storageDomain` and `conversation.chat.turnTail`; if those interfaces change, compatibility has to be re-verified.

05Compatibility, maintenance and safety notes

  • Only normal, uninterrupted final answers with no tool calls and a body are polished; sub-agent output, failures, truncated output and empty answers are skipped, and tool steps or running answers are never touched.
  • Polishing costs extra quota on the selected model, waits at most 60 seconds, and skips answers longer than 60,000 characters; a failed or cancelled polish keeps the original answer and is not retried.
  • The plugin targets DSH builds that expose `agent/turn-stopping`, `ctx.llm.stream()`, `storageDomain` and `conversation.chat.turnTail`; if those interfaces change, compatibility has to be re-verified.
  • Saved polish results live in the current profile's plugin storage-domain and survive restarts, but uninstalling the plugin does not clean them up.
2026-11-012026-11-03v0.3.1

MIT · README evidence only, latest release v0.3.1, no security audit

06Frequently asked questions

How do I install dsh-speak-plain into DeepSeek Harness?

Run `dsh plugin --profile desktop add -w "github:Exception-H/dsh-speak-plain"` (use `--profile web` for the web profile), or install from a local clone with `-w ./dsh-speak-plain`, or from the release tgz. Then restart DSH and open Settings → 说人话 to pick the channel and model, tick "enable" and save.

Does it need its own API key or a second provider account?

No. The plugin does not store or ask for an API key; polish requests go through the channel and model already configured in DSH, which means they consume extra quota on that model. Only the last user question, the rewritable text of the answer and your expression preference are sent — not the thinking process or the full tool log.

Which answers get polished in a DSH session?

Only normal, uninterrupted final answers that stopped normally, made no tool calls and have a body. Sub-agent output, failed turns, truncated output and empty answers are skipped, and answers already in history are not re-run. Tool steps and running answers keep their native rendering.

What happens to code blocks, links and GenUI panels after polishing?

Code fences, inline code and Markdown links are swapped for protection markers before the request and restored afterwards; real `dsh-ui` fences in the polished area are rendered by the public GenUI embed component, and unparsable components keep their code with a note. `panel:true` fences are not re-published in the polish area — view those in the original session panel.

Can I polish just one answer instead of every turn?

Yes. In the 0.3.1 settings there are two switches: "enable final-answer polish" and "auto-polish all answers". Auto mode is off by default, so nothing runs automatically — you click "polish this answer" under a single answer and only that segment is rewritten, which saves quota.

08Data and sources

  • Author-claimedgithub.com05832d49c475…

    让 DSH 的最终回答更容易听懂:用一个你选的模型润色最终回答,**默认显示润色版,原回答折叠、可随时点开查看**。主模型照常执行任务;润色只改表达,不改事实、条件与技术细节。

  • Author-claimedgithub.com05832d49c475…

    dsh plugin --profile desktop add -w "github:Exception-H/dsh-speak-plain"

  • Author-claimedgithub.com05832d49c475…

    面向当前具有 `agent/turn-stopping`、`ctx.llm.stream()`、`storageDomain` 和 `conversation.chat.turnTail` 接口的 DSH。升级后如果这些接口发生变化,需要重…

This page is generated from the project’s public documentation, repository metadata and a structured parse of DSH Plugins; last verified on 2026-11-14. Found an error? Submit a correction.

🏆

Best DeepSeek Harness Plugins

Twelve plugins worth installing first — picked from the whole catalog, across every category.

DSH Plugins is an independent community directory of DeepSeek Harness plugins. Not affiliated with or endorsed by DeepSeek. Third-party plugins are not security-audited — review the source before installing.

New DeepSeek Harness plugins, weekly. No spam.