llama-swap has been deployed to namespace "{{ .Release.Namespace }}".

Head-end:
  kubectl -n {{ .Release.Namespace }} get deploy {{ include "llama-swap.fullname" . }}

Backend workloads the head-end manages (kubeswap ships in the same image):
  kubectl -n {{ .Release.Namespace }} exec deploy/{{ include "llama-swap.fullname" . }} \
    -- /usr/local/bin/kubeswap status --namespace {{ .Release.Namespace }}

Try it (from a machine with cluster access):
  kubectl -n {{ .Release.Namespace }} port-forward \
    svc/{{ include "llama-swap.fullname" . }} 8080:{{ .Values.service.port }}
  curl http://localhost:8080/v1/models
  curl http://localhost:8080/v1/chat/completions \
    -H 'Content-Type: application/json' \
    -d '{"model":"<model id from config>","messages":[{"role":"user","content":"hello"}]}'

Models are provisioned on demand: the first request for a model takes as
long as its pod takes to start and load (healthCheckTimeout in config
bounds that wait).
{{- if .Values.ingress.enabled }}

Ingress is enabled:
{{- range .Values.ingress.hosts }}
  http://{{ .host }}/
{{- end }}
{{- else if eq .Values.service.type "LoadBalancer" }}

The service is a LoadBalancer; once the cloud provider assigns an address:
  kubectl -n {{ .Release.Namespace }} get svc {{ include "llama-swap.fullname" . }}
{{- end }}
