JuneLearn/dsh-reasoning-settings
50Last commit Aug 15, 2026
dsh-reasoning-settings DSH plugin
This plugin introduces an independent Reasoning effort settings page within DeepSeek Harness Web, allowing users to configure thinking intensity levels for custom llm-pi-ai providers. It automatically reads the custom providers and models already added on the official Models page, and provides per-model effort level selection along with provider-wide defaults. The server entry also fixes subagent model inheritance and adds optional provider/model/reasoning_effort fields to the subagent and subagent_fork tools.
How to install the dsh-reasoning-settings DSH plugin
dsh plugin --profile web add github:JuneLearn/dsh-reasoning-settingsCopying does not run this command. Review the repository and version before installing the dsh-reasoning-settings DSH plugin.
dsh-reasoning-settings DSH plugin data source
dsh-reasoning-settings DSH plugin snapshot date: Aug 16, 2026
discovered
What the dsh-reasoning-settings DSH plugin can do
- Auto-reads custom providers and models from the official Models page.
- Declares per-model reasoning effort levels: off, minimal, low, medium, high, xhigh, max.
- Customizes the actual wire value sent to the API for each effort level.
- Sets a provider-wide default effort restricted to levels supported by all models in that provider.
- Configures the reasoning wire format for openai-completions providers.
- Writes back to llm-pi-ai.providers.* through the official settings.mutate API without touching API keys.
- Fixes subagent model inheritance: children inherit parent's live provider/model instead of create-time official default.
- Adds provider, model, and reasoning_effort fields to the subagent and subagent_fork tools without replacing their original behavior.
Where the dsh-reasoning-settings DSH plugin fits
- A user with a third-party API relay wants to set different reasoning effort levels for different models.
- A developer using subagents needs the child agent to inherit the parent's current third-party provider and reasoning effort.
- An advanced user wants to fine-tune the wire values sent for each effort level to match the API relay's expected values.
- A user who wants to override the model in a subagent call while keeping the parent's reasoning effort default.
- A workflow or agent that uses the augmented subagent tool to explicitly target a specific provider/model with a specific reasoning effort.
Who the dsh-reasoning-settings DSH plugin is for
- Developers using DeepSeek Harness Web with custom third-party LLM providers.
- Advanced users who need granular control over reasoning effort for each model and provider.
- Users who rely on subagents or forked agents and need correct model/reasoning inheritance.
dsh-reasoning-settings DSH plugin limitations
- Only works with the Web profile (profile web) of DeepSeek Harness; other profiles are not supported.
- Requires DeepSeek Harness 0.1.0-rc.5 or compatible; Harness is still in Developer Preview and interfaces may change.
- The plugin itself does not enforce that the API relay honors the sent reasoning effort — HTTP 400 may occur if the relay does not support the value.
- Per-call targeting of provider/model/reasoning_effort only affects local in-process subagents; Codex, Claude Code, and ACP subagents run in separate processes and are unaffected.
- Network access to registry.npmjs.org and github.com is required for installation.
- Node.js 22.19.x or 24+, pnpm, and Git must be installed on the system.
dsh-reasoning-settings DSH plugin: from the repository README
Quoted from the JuneLearn/dsh-reasoning-settings README, the upstream source of the dsh-reasoning-settings DSH plugin. Copyright remains with the original authors.
A [DeepSeek Harness](https://github.com/deepseek-ai/deepseek-harness) Web settings plugin that adds an independent **reasoning effort** (thinking intensity) page for custom `llm-pi-ai` providers. The official UI only exposes effort selection for first-party DeepSeek models; this plugin makes it work for third-party / relay APIs too. [中文说明](./README.zh-CN.md) > **Third-party service recommendation (referral link)** > > If you are looking for an OpenAI-compatible API relay, you can take a look at [WPIronman API](https://api.wpironman.top/register?aff=JUNE). This is my referral link, and I may receive a referral benefit if you register through it; any promotion or benefit is subject to the provider's current terms. This plugin is independent of that service and does not require any particular relay. Evaluate pricing, reliability, and privacy terms before choosing a provider. > > New users can redeem the gift quota with code `99F509ABC6C38F77`. ## Features - Auto-reads the custom providers and models you already added on the official **Models** page. - Declares a per-model effort level from `off / minimal / low / medium / high / xhigh / max`. - Customizes the actual wire value sent
Read the full READMERepository license: MIT
dsh-reasoning-settings DSH plugin questions
How do I install the DSH Reasoning Settings plugin?
You can install it using the npx command: `npx --yes -p @deepseek-ai/dsh dsh plugin --profile web add github:JuneLearn/dsh-reasoning-settings`. This requires Node.js, pnpm, and Git to be installed first. Alternatively, if you have a DeepSeek Harness source checkout, run `pnpm dsh plugin --profile web add github:JuneLearn/dsh-reasoning-settings` from the source root.
How do I upgrade the plugin to the latest version?
Simply run the same install command again. For the npx method: `npx --yes -p @deepseek-ai/dsh dsh plugin --profile web add github:JuneLearn/dsh-reasoning-settings`. For the pnpm source method: `pnpm dsh plugin --profile web add github:JuneLearn/dsh-reasoning-settings`. The plugin will be updated to the latest version without needing to uninstall first.
How do I uninstall the plugin?
Use the uninstall command: for npx method run `npx --yes -p @deepseek-ai/dsh dsh plugin --profile web remove dsh-reasoning-settings`. For pnpm source method run `pnpm dsh plugin --profile web remove dsh-reasoning-settings` from the Harness source root. After removal, restart `dsh web` and the Reasoning effort settings page will be gone.
Why doesn't my subagent inherit the reasoning effort I set?
The plugin's server-side fix for subagent inheritance is enabled by default. However, make sure the plugin is properly installed and that you have restarted the Web server after installation. Also check that the subagent is a local in-process subagent (not Codex, Claude Code, or ACP). If you have disabled the fix via the `subagentRouting` config, set it to `true` in the plugin mount configuration.
What reasoning effort levels are supported?
The plugin supports seven levels: `off`, `minimal`, `low`, `medium`, `high`, `xhigh`, and `max`. You can assign these to each model and also set a provider-wide default that is restricted to levels supported by all models in that provider. The actual wire value sent to the API is customizable for each level.