dsh-llm-retry-settings

Tweaking a specific LLM provider's retry mode, max retries, or timeout values in DSH used to mean digging into per-provider YAML namespaces and double-checking syntax by hand. dsh-llm-retry-settings moves that work into a clean settings page: retry policy (normal or always mode, maxRetries) alongside timeoutMs, websocketConnectTimeoutMs, and streamIdleTimeoutMs, all grouped by provider and editable in place.

Every save goes through the settings.mutate wire face, so the change is picked up by the very next model request without restarting the gateway or reloading your profile. Timeout fields are auto-discovered from each provider's settings schema when available, with a verified fallback map for the bundled providers such as llm-pi-ai and llm-deepseek, so new providers often light up automatically.

This plugin is a good fit for DSH users who run multiple LLM providers and want per-provider retry and timeout tuning without repeatedly hand-editing config files. It ships as a pure client-side bundle with an empty host half, hooks into your profile bundles on install, and requires no manual patch edits to cordis.patch.yml.