dsh-llm-fallbacks

dsh-llm-fallbacks addresses a common failure mode in agent workflows: an LLM request keeps failing because of auth errors, exhausted retries, quota limits, or rate limiting, and the task stalls. With this plugin, dsh can walk a configured fallback chain and switch to another provider or model automatically, allowing the current step or turn to continue instead of interrupting the whole run.
Its core capabilities include root and role-based fallback chains, automatic failure switching, cooldown and revert behavior, explicit logging for every switch, and safety valves that prevent runaway switching. It also supports time-slot routing, letting different chains become active during peak or valley windows, and works across both the web settings surface and the dsh-tui terminal experience.
It is best suited for dsh users running multi-model or multi-provider agents who need more resilience than a single default model can provide. It is especially useful for teams managing subagents with different model policies, handling quota or rate-limit pressure, or needing continuous operation across changing provider conditions.