Streaming

Server-sent events

Pass "stream": true. The API emits OpenAI-compatible chat.completion.chunk frames as SSE data: lines, ending with data: [DONE].

data: {"id":"…","object":"chat.completion.chunk","choices":[{"delta":{"content":"Hi"}}]}

data: [DONE]

Final chunks may include a usage object with token counts (and credits / TPS when available).