Henta111/dsh-more-models-thinking-level ↗★ 0
dsh-more-models-thinking-level
多模型思考程度声明插件：为 DeepSeek Harness 中的 GPT、Gemini 及 OpenAI 兼容提供商声明单模型推理强度。
AI 分析
核心用途是让自定义网关能够根据模型声明选择匹配的推理强度（如 off/low/high），适合使用混合模型、需要精细控制各模型思考深度的用户。
安裝
$
npx -p @deepseek-ai/dsh dsh plugin --profile web add github:Henta111/dsh-more-models-thinking-level說明文件
閱讀完整 README ↗DSH Cross-model Reasoning Adapter
Installable and removable DSH bundle. Provider-native parameter conversion remains owned by the host's dsh-llm-pi-ai.
The plugin exposes per-model reasoning levels (off / minimal / low / medium / high / xhigh) for GPT, Codex, Gemini and OpenAI-compatible models in DeepSeek Harness, so your custom gateway can select a reasoning intensity that matches the model.
Install (plugin market)
Once published to npm:
dsh plugin add dsh-more-models-thinking-level
Or from this repository directly:
dsh plugin add https://github.com/Henta111/dsh-more-models-thinking-level
Remove
dsh plugin remove dsh-more-models-thinking-level
How it works
- The bundle layer registers the plugin into the profile's
dsh.profile.bundles. - The web client adds a settings section (
推理适配 / Reasoning adapter). - Provider-native wire values are declared per model via
reasoningEfforts; the host (dsh-llm-pi-ai) performs the actual conversion.
License
MIT