Streaming

POST /v1/chat/completions with "stream": true returns Server-Sent Events, OpenAI-format (data: {chunk} lines terminated by data: [DONE]). Both official OpenAI SDKs work unchanged with base_url/baseURL pointed at us.

stream = client.chat.completions.create(model="qwen3.6-35b-a3b-awq",
    messages=[...], stream=True)
for chunk in stream:
    print(chunk.choices[0].delta.content or "", end="")

TTS (/v1/audio/speech) returns raw audio bytes; everything else is JSON.