RouterR1 (inference)

RouterR1 is an agentic router with a special runtime (vLLM + external routing service).

Notebook: https://github.com/ulab-uiuc/LLMRouter/blob/main/notebooks/router_r1/01_router_r1_inference.ipynb

Router docs: https://github.com/ulab-uiuc/LLMRouter/blob/main/llmrouter/models/router_r1/README.md

Config

• Test config: https://github.com/ulab-uiuc/LLMRouter/blob/main/configs/model_config_test/router_r1.yaml

Important behavior

• router_r1 does not support --route-only in the CLI.
• It requires hparam.api_base and hparam.api_key in the YAML config.

Run (CLI)

llmrouter infer --router router_r1 --config configs/model_config_test/router_r1.yaml --query "Explain transformers."

Warning

RouterR1 requires CUDA (vLLM GPU runtime) and optional dependencies. Follow the notebook and the router README for setup.