xiaoxingdelabi1/dsh-command-retry-count1

@deepseek-ai/dsh-command-retry-count

Slash command to adjust LLM-provider retry count (maxRetries)

包名
@deepseek-ai/dsh-command-retry-count
版本
0.1.0-rc.7
许可证
MIT
最近更新
2026年8月16日

安装

此插件尚未提供可验证的 bundle,或兼容性检查未通过。请先阅读仓库说明。 阅读完整 README ↗

Usage

Type the command in your DSH Web GUI chat input, just like any other slash command:

/retry-count 
 

Examples

CommandWhat it does
/retry-count my-provider 5Set max retries to 5
/retry-count my-provider 0Disable retries entirely
/retry-count my-provider 20Set max retries to 20 (maximum allowed)
/retry-count my-provider infUnlimited retries (always mode) — keep retrying until success or interruption (-1, infinite, infinity and work too)
/retry-countShow usage help (no arguments)

After you run it

The plugin will respond with one of these messages:

Success (limited):

Provider "my-provider" retry policy updated to maxRetries=5. Changes take effect immediately.

Success (unlimited):

Provider "my-provider" retry policy updated to unlimited (always mode — retries until success or interruption). Changes take effect immediately.

No change needed (same value already set):

Provider "my-provider" already has maxRetries=5.
Provider "my-provider" already has unlimited (always mode — retries until success or interruption).

Provider not found:

Provider "whoami" is not registered. Available: my-provider, another-provider

Invalid number:

maxRetries must be an integer from 0 to 20, or "inf" for unlimited retries

What happens next

The next time your provider fails a model request, the retry count will use your new setting. If you set it to 5, you'll see:

已重试模型请求(1/5) · 1s
已重试模型请求(2/5) · 2s
...

With inf, retries never give up on their own — the counter keeps climbing (1/∞, 2/∞, ...) until the request succeeds or you stop it. The change is saved to settings.yaml and persists across DSH restarts.

Example: unlimited retry showing 1/∞

With unlimited retries enabled, the retry counter shows "1/无限" — attempt 1 of infinity. The counter keeps climbing and each retry waits with exponential backoff (capped at 60s) until the request succeeds or you manually stop it.

⚠️ Unlimited retries can hammer a provider that is down. Every retry waits with exponential backoff (capped), and stopping the request cancels any in-flight retry — but a persistently failing provider will keep being retried indefinitely. Use inf deliberately and switch back to a finite number when you're done.