LLM Router Guide

Help user manage and optimize multi-backend LLM routing.

When to use

When user asks about LLM settings, model switching, cost optimization, which model to use, or router configuration.

How to use

AI Agent supports up to 4 LLM backends with intelligent routing.

Check current status

run_shell "router_status" to see all backends, their metrics, and current profile.

Add/change backends

run_shell "router_set <api_key>" to add a backend. Presets: kimi, qwen, deepseek, glm, mimo, openai, claude, openrouter

Switch routing profile

run_shell "router_profile " where profile is:

• eco: prefer cheapest backend (mimo-flash, qwen-turbo)
• auto: balance cost and quality based on task complexity
• premium: prefer highest quality (claude, kimi, openai)

Quick model switch

run_shell "set_llm " for single-backend mode run_shell "set_llm model " to change model only

Recommendations

• For casual chat: eco profile with mimo or qwen
• For complex reasoning: premium profile with kimi or claude
• For code generation: deepseek or openai
• For cost-sensitive: openrouter with free models (list_models --free)

Example

User: "看看当前用的什么模型" → run_shell "router_status" → "当前路由状态： Slot 0: kimi (api.moonshot.cn, kimi-k2.5) — 活跃 Profile: auto 建议：如需省钱可切换 eco 模式：router_profile eco"

User: "切换到省钱模式" → run_shell "router_profile eco" → "已切换到 eco 模式，优先使用低成本模型。"