max-tokens 5

DSH 插件：LLM 自动重试设置卡片——重试次数/退避/抖动与 25 个错误码可勾选（也能自己输入），可按 provider/model 单独设策略、回答被输出 token 上限截断时自动续写且提示词可改，内置重试观测面板，改完实时生效。

Model Capability Manager for DSH Web: per-model thinking levels, context window, output caps, input modalities, gateway compat, route defaults, one-click presets, EN/中文 UI.

DSH plugin: auto-fill metadata (context length, output cap, display name, input modalities) for custom-provider models from the Model Metadata Registry (models.dev).

DSH plugin: safe auto-continue-once recovery for native continuable subagents that end with explicit max-tokens termination. No timers, no polling, no private APIs.

DeepSeek Harness plugin to audit and correct llm-pi-ai model capability declarations by probing the endpoint itself: out-of-range maxTokens, contextWindow, reasoning levels, image input.