dsh-llm-retry-settings

When using DeepSeek LLM for model inference, network jitter or specific server errors (such as HTTP 400) often lead to request interruptions. While the default retry strategy is basic, it often struggles to handle complex business scenarios, resulting in unnecessary failures or wasted resources.

This plugin provides an intuitive settings card that allows you to finely manage retry behavior. It not only allows overriding core parameters like maximum retry count and initial delay but also supports customizing "retryable error codes," incorporating specific failure states (such as reasoning stream errors) into the retry logic, thereby significantly enhancing system fault tolerance.

This is ideal for developers and operations personnel who need to handle unstable network environments or require high service stability. By adjusting and saving configurations in real-time within the settings, you can quickly optimize the robustness of the inference service without modifying code.