infercrane target add gpu-a \
  --url http://gpu-a:8000 \
  --runtime vllm \
  --upstream-model Qwen/Qwen3-8B

infercrane apply Qwen/Qwen3-8B \
  --name qwen-prod \
  --targets gpu-a \
  --idempotency-key qwen-prod-initial \
  --wait