Preface¶

In the DeepSeek Harness (DSH) Web chat page, hastily written prompts are often rough, lacking clear goals, context, or expected output formats. dsh-ai-prompt-optimizer provides an entry point for such scenarios: adding an “✨ AI Optimize” button next to the send button in the input box, helping turn the current draft into a clearer, more complete prompt ready to send directly to AI.

What This Is¶

dsh-ai-prompt-optimizer is a prompt optimization plugin for the DeepSeek Harness (DSH) Web chat page, maintained by wuchubuzai2018, licensed under MIT.

It solves the problem of “organizing prompts before sending”: it reads the raw draft from the input box, calls the default model already configured in DSH to generate an optimized version, and fills the result back into the input box for user confirmation before sending.

Core Features¶

• One-click optimization: Reads the current input box draft and generates an optimized version.
• Reuses the current model: Uses the default model already configured in DSH, no separate API Key configuration needed.
• Confirm before sending: The optimized result is not submitted automatically; you can continue editing.
• Visible progress: The button displays “Optimizing…” during the request.
• Failures are noticeable: Clear notifications appear when input is empty, the model is unavailable, or the call fails.
• Broader model compatibility: Only sends basic model request fields, avoiding failures caused by some Providers not supporting optional parameters.
• Optimization results are only filled back into the input box and are not sent automatically; the final content remains under your control.

Installation and Activation¶

First, confirm the environment meets the following requirements:

• DeepSeek Harness Web 0.1.0-rc.6
• A usable default model is configured in DSH
• Node.js 22+ and pnpm are required when installing from source

The installation command is as follows; the plugin will be installed into the web profile:

dsh plugin --profile web add dsh-ai-prompt-optimizer

After installation, restart DSH Web:

dsh web

If the system manages pnpm via Corepack, you can first execute:

corepack enable pnpm

If pnpm blocks Git dependencies from running build scripts, follow the terminal prompts to add the corresponding key to allowBuilds in the pnpm-workspace.yaml file under the DSH profile directory, then re-run the installation command.

Typical Usage¶

1. Open the DSH Web chat page.

2. Write your raw request in the input box, for example, “Help me write a weekly report.”

3. Click the “✨ AI Optimize” button next to the send button.

4. Wait for the optimization to complete.

5. Review the filled-in prompt; you can continue editing, then send after confirmation.

Note: The optimization result will replace the current input box draft. If the original text is important, please copy and save it first.

Applicable Scenarios and Notes¶

It is suitable for users who need to organize rough ideas into more complete prompts on the DSH Web chat page. Before using, please confirm that a usable default model is configured in DSH.

The plugin runs with the current dsh process permissions. Before installation, it is recommended to review the plugin source code and confirm that the MIT license fits your use case.

If you do not see the button after installation, check the following in order:

1. The plugin was installed into the web profile.

2. dsh web has been restarted after installation.

3. The current page is a chat session page.

4. You can see ai-prompt-optimizer when running the following command:

dsh --profile web --dump-config

Links¶

The DSH community directory is an independent site and does not imply official affiliation with DeepSeek / High-Flyer. This article does not provide a verifiable directory page URL.

GitHub repository: dsh-ai-prompt-optimizer