dsh-llm-retry-settings

A DSH plugin that adds a settings card for the DSH LLM auto-retry engine (@deepseek-ai/dsh-llm-retry). From Settings → General you can tune retry count and backoff, pick or type error codes, set per provider/model overrides, and optionally auto-continue replies cut off by the output-token limit.

• Settings card in Settings → General; ships its own client bundle (lib/client.js), so no separate UI package is needed.
• Overrides maxRetries, initialDelayMs, maxDelayMs and jitterRatio on the agent/request-error retry policy; default enabled: false means fully bypassed.
• retryableCodes merges extra failure codes into each provider's own list (never replaces it); defaults to INVALID_REQUEST and PI_AI_ERROR.
• Error-code chips grouped into six categories ordered by whether retrying helps; a 自定义 group accepts free-form codes validated as A-Z 0-9 _ - .
• Auto-continue on output truncation (off by default): queues one follow-up continuation turn per truncation, capped by maxContinuations.
• continuationPrompt is editable, with built-in wording when left empty; continueOnError can continue once after retries are exhausted on transient errors.
• Per provider/model overrides: first matching row wins, '*' wildcards are supported, and -1 inherits the global value.
• Read-only observation panel with retried failures, auto-continues, cap hits and skipped rounds by error code and provider/model, plus a host log at ~/.dsh/logs/