Streaming

Set "stream": true to get text/event-stream. Frames are data: {...} lines terminated by data: [DONE]. Non-data: lines never appear.

curl -N https://api.refuseless.dev/v1/chat/completions \
  -H "Authorization: Bearer $REFUSELESS_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Accept: text/event-stream" \
  -d '{
    "model": "glm-5.3-refuseless",
    "messages": [{"role": "user", "content": "Count to 5."}],
    "stream": true
  }'

Notes:

Raw upstream chunks pass through untouched, including reasoning_content deltas, so you can render thinking live.
stream_options: {"include_usage": true} is accepted and forwarded.
The model field always echoes your public id.
If credits run out mid-stream you get a terminal error event, then [DONE]:

{ "error": { "code": "insufficient_credits", "message": "Credits ran out. Generation stopped." } }

Disconnecting aborts the generation server-side; you are billed only for tokens produced.