DeepSeek Harness 插件

dsh-llm-retry-settings

在 DSH 设置里调整 LLM 自动重试的次数与退避时间并实时生效的设置卡片。

跳到安装方式

来源信息

GitHub 仓库
zeng6125-rgb/dsh-llm-retry-settings
最近更新
2026年8月19日
分类
模型与服务商
GitHub stars
0

安装

默认先复制一段 Prompt,让 Agent 读页面和仓库;需要自己装时再切到命令。

复制这段 Prompt,发给 DSH、Codex 或其他 Agent,让它先读页面和仓库。

请先不要安装。阅读这个 DeepSeek Harness 插件,说明它解决什么问题、会访问哪些文件、网络或密钥,以及如何安装和卸载。

插件页面:https://deepseekplugins.org/zh/plugins/zeng6125-rgb/dsh-llm-retry-settings
GitHub:https://github.com/zeng6125-rgb/dsh-llm-retry-settings
插件名:dsh-llm-retry-settings
作者:zeng6125-rgb
安装命令:dsh plugin --profile web add github:zeng6125-rgb/dsh-llm-retry-settings

确认前不要执行安装命令。

检查来源文件

安装前先看这个插件目录里的 README 和其他文件。

文件资源管理器3 个文件
README.md来源说明 · 只读预览

@dsh-external/dsh-llm-retry-settings

A settings card for the DSH LLM auto-retry engine (@deepseek-ai/dsh-llm-retry). Tune the retry count and backoff from Settings → General; changes take effect immediately.

[中文说明](./README.zh-CN.md)

Features

  • Includes the settings UI (client bundle lib/client.js): a card in Settings → General — no separate UI package needed.
  • Overrides maxRetries, initialDelayMs, maxDelayMs, and jitterRatio on the agent/request-error retry policy.
  • Preserves each provider's own mode and retryableCodes — only the retry count and backoff timing are overridden.
  • Default enabled: false = fully bypassed; nothing changes until you enable the override.

Install

Prerequisite: a DSH Desktop profile (the web profile lives at ~/.dsh/profiles/web).

Option A — GitHub Release package (recommended)

# 1. download the packaged plugin tgz from the v0.0.1 release
gh release download v0.0.1 -R zeng6125-rgb/dsh-llm-retry-settings

# 2. unpack it into the profile's node_modules
mkdir -p ~/.dsh/profiles/web/node_modules/@dsh-external
tar -xzf dsh-external-dsh-llm-retry-settings-0.0.1.tgz -C ~/.dsh/profiles/web/node_modules/@dsh-external/
mv ~/.dsh/profiles/web/node_modules/@dsh-external/package \
   ~/.dsh/profiles/web/node_modules/@dsh-external/dsh-llm-retry-settings

# 3. register the bundle in the profile, then restart DSH
#    add "@dsh-external/dsh-llm-retry-settings" to dsh.profile.bundles
#    in ~/.dsh/profiles/web/package.json

Option B — dsh CLI / pnpm (requires git + network)

The dsh plugin command forwards its arguments to pnpm in the profile directory:

# from a git repo (pnpm clones it; the committed lib/ means no build needed)
dsh plugin --profile web add github:zeng6125-rgb/dsh-llm-retry-settings

# or from the release tarball URL
dsh plugin --profile web add https://github.com/zeng6125-rgb/dsh-llm-retry-settings/releases/download/v0.0.1/dsh-external-dsh-llm-retry-settings-0.0.1.tgz

Then enable the plugin in the profile: add "@dsh-external/dsh-llm-retry-settings" to dsh.profile.bundles (or use the Desktop plugin-inventory UI) and restart DSH.

> Note: the command is dsh plugin (a subcommand of the dsh CLI), not dsh-plugin. The dsh CLI is bundled with the Desktop app; if it is not on your PATH, invoke it via the app's node_modules bin, e.g. node "<app>/node_modules/@deepseek-ai/dsh/lib/bin.js" plugin --profile web ....

Option C — from source

git clone https://github.com/zeng6125-rgb/dsh-llm-retry-settings.git
cd dsh-llm-retry-settings
npm run build        # bash scripts/build.sh (requires DSH_CHECKOUT)

Link the local build into the profile and register the bundle:

dsh plugin --profile web link "$PWD"
# then add "@dsh-external/dsh-llm-retry-settings" to dsh.profile.bundles and restart DSH

Usage

1. Open DSH Settings → General. 2. Find the LLM 自动重试 card. 3. Toggle 开启覆盖 (enabled) to apply the override. 4. Set maxRetries / initialDelayMs / maxDelayMs / jitterRatio and click 保存.

Changes are written to the dsh-llm-retry settings namespace and picked up live by the retry engine.

Configuration

KeyTypeDefaultDescription
enabledbooleanfalseEnable the retry-policy override.
maxRetriesinteger (≥ 0)2Maximum retry count (0 = do not retry).
initialDelayMsinteger (≥ 1)500Initial backoff before the first retry (ms).
maxDelayMsinteger (≥ 1)10000Upper bound for backoff (ms).
jitterRationumber (0–1)0.1Random jitter applied to backoff (0 = none).

How it works

The host half registers the dsh-llm-retry settings namespace (schema validation + persistence + live sync) and prepends an agent/request-error listener that rewrites retryPolicy before the official @deepseek-ai/dsh-llm-retry recover runs:

agent/request-error  →  [this plugin: override count/backoff]  →  dsh-llm-retry recover

UI

Settings → General → LLM 自动重试 card. Edits are draft-based: click 保存 to commit or 放弃 to discard. After saving, the card verifies the write against the settings snapshot and shows 已保存 ✓ / 保存失败 ✗.

Requirements

  • Host plugin: @deepseek-ai/dsh-llm-retry
  • Client runtime: @deepseek-ai/dsh-client-web-react
  • DSH settings service

Build

npm run build        # bash scripts/build.sh (requires DSH_CHECKOUT)
npm run typecheck    # tsc --noEmit

License

[MIT](./LICENSE)