Mu-scorpio/dsh-reasoning-effort0

dsh-reasoning-effort

Provider-aware reasoning effort settings for DeepSeek Harness.

AI Analysis

核心用途是精细化控制支持推理模型的思考强度。适合需要根据任务复杂度或成本预算调整模型推理深度的用户。

Package
dsh-reasoning-effort
Version
0.1.0
License
MIT
Last updated
Aug 25, 2026

Install

$npx -p @deepseek-ai/dsh dsh plugin --profile web add github:Mu-scorpio/dsh-reasoning-effort

Configuration rules

FieldMeaning
apiSelects the default wire-value map for a provider protocol.
reasoningSets a provider-level default reasoning level when one is missing.
effortsOverrides the values sent for the generic Harness levels.
models..effortsOverrides the map for one model.
models..disabledExplicitly marks one model as non-reasoning.

The available Harness levels are off, minimal, low, medium, high, xhigh, and max. The built-in protocol maps cover the five common levels from low through max; custom gateways can provide their own values.

Providers must already exist in the llm-pi-ai settings. The plugin enhances an existing provider; it does not recreate one that has been removed.