## Problem

`litellm config-validate` catches structural problems in a proxy config (typos in `routing_strategy`, duplicate `model_name`, missing `litellm_params.model`, unset `os.environ/VAR` references). It does not explain what the config does.

Operators reading their own `config.yaml` for the first time have no offline way to see the resolved shape of the router: how many deployments, which providers are configured, how the model_list is grouped, which `model_name` has only one deployment (often a fallback typo), which `model_name` is not referenced by any group. They either start the proxy and tail logs, or write a one-off Python script.

## Motivation

A sibling CLI in the same `litellm/cli/` family would give operators a single offline command that summarises the resolved router. The output is the same shape they would derive by hand from the config plus a 30-line Python snippet, but shipped as a tested, versioned tool.

`config-validate` answers "is this config well-formed?". `router-explain` answers "what does this config DO?". Both are pure static analysis against the config file. No network, no credentials, no proxy start.

## Background

`litellm/cli/` already hosts fo…