Deploy a fine-tuned model with vLLM
Render + apply the vLLM manifests and register the model on readiness.
Authorizations
The access token received from the authorization server in the OAuth 2.0 flow.
Body
Deploy a fine-tuned model behind vLLM.
Deployment name (DNS-1123 label). Used for the K8s resources (serve-vllm-) and the router endpoint. Unique per served model.
A COMPLETE fine-tune training job whose adapter to serve.
Model id exposed via the router. Defaults to the job's registered model name.
200Override GPU count for the serving pod.
1 <= x <= 64KEDA min replicas (0 enables scale-to-zero).
x >= 0x >= 1vLLM --max-model-len override.
x >= 1vLLM --dtype (e.g. 'bfloat16', 'auto').
Router quality score for this model.
Response
Deployment created; auto-registered into the router when ready.
pending, deploying, active, failed, stopped, archived 
