penguin-oo/dsh-delegate-router

README 摘要

dsh-delegate-router English · 中文 Automatic Flash/Pro routing for DeepSeek Harness subagent calls. Your main conversation keeps running on a strong model (e.g. V4 Pro). When the agent delegates a task to a subagent, this plugin classifies the task and sends light tasks to a cheap model (e.g. V4 Flash) while heavy tasks stay on the strong model — deterministically, without relying on the model's cooperation. - Auto routing — light tasks (search / summarize / list …) run on Flash; heavy tasks (refactor / implement / debug …) stay on Pro. - Manual overrides — per-call provider / model parameters are added to the subagent / subagent fork tools, so you can force a model on any call. - /delegate command — per-session mode: /delegate auto off flash-all . - Budget cap — (config ready, token accounting in a later release) once the session exceeds budgetCapTokens , new subagents downgrade to Flash. - Decision ledger — every routed delegation is recorded (task, chosen route, trigger) in a storage domain; a savings panel is planned. How it works The plugin wraps the official subagent / subagent fork tools on each agent's scoped context (same seam the proven dsh-reasoning-settings uses), classif…

分类