Streaming

Set "stream": true Live to receive text/event-stream chunks in OpenAI format, terminated by data: [DONE]. Chunks flush as the provider emits them — nothing is buffered server-side.

curl -N https://ciptamodel.com/v1/chat/completions \
  -H "Authorization: Bearer sk-cm-live-..." \
  -H "Content-Type: application/json" \
  -d '{"model": "deepseek-v4.1-flash", "stream": true,
       "messages": [{"role": "user", "content": "Count to five"}]}'

Mid-stream provider failures arrive as an OpenAI-shaped error event followed by [DONE] — the HTTP status stays 200 once streaming has started.