Premium rentals

Auth: same Bearer API key as the rest of /v1.

Create / list / stop

bash
# Attach an existing dedicated deployment
curl "$GATEWAY/v1/premium/rentals" \
  -H "Authorization: Bearer $OPENDOOR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "deployment_id": "DEPLOYMENT_UUID",
    "model_id": "flux-1-schnell",
    "hours": 2
  }'

curl "$GATEWAY/v1/premium/rentals" \
  -H "Authorization: Bearer $OPENDOOR_API_KEY"

curl -X DELETE "$GATEWAY/v1/premium/rentals/RENTAL_UUID" \
  -H "Authorization: Bearer $OPENDOOR_API_KEY"

Local Metal start (Ollama + image-process probe) is on the dashboard: POST /api/premium/rentals with { "target": "local" }. The gateway attach path requires deployment_id.

Generate an image

Pins a private OpenAI-compatible image worker when PRIVATE_IMAGE_GEN_URL or a rental FQDN is set. 503 Studio GPU offline if that process is down.

bash
curl "$GATEWAY/v1/premium/images" \
  -H "Authorization: Bearer $OPENDOOR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "premium:RENTAL_UUID",
    "prompt": "A red door on a white wall",
    "size": "1024x1024"
  }'

model may also be custom:<deployment-uuid>.

Chat on the same box

If the rental’s deployment is a chat endpoint (Ollama / vLLM):

bash
curl "$GATEWAY/v1/chat/completions" \
  -H "Authorization: Bearer $OPENDOOR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "premium:RENTAL_UUID",
    "messages": [{"role":"user","content":"hello"}]
  }'

custom:<deployment-uuid> is equivalent. Neither id is routed to Together or Vertex MaaS.

Dedicated deployments

Overview