Streaming responses with Velqa — SSE, OpenAI-compatible

Velqa can return chat responses as a stream: tokens arrive as they're generated, via Server-Sent Events (SSE). This is ideal for a chat-style UI where you want to display text progressively. Just add "stream": true to your /v1/chat/completions request.

Enable streaming (curl)

curl https://api.velqa.dev/v1/chat/completions \
  -H "Authorization: Bearer sk-..." \
  -H "Content-Type: application/json" \
  -d '{
    "model": "kimi-k2.6",
    "messages": [{"role": "user", "content": "Write a haiku about Casablanca."}],
    "stream": true
  }'

SSE chunk format

The response is a stream of data: lines. Each chunk is a JSON object whose incremental text is in choices[0].delta.content. The stream ends with a data: [DONE] line.

data: {"choices":[{"delta":{"content":"Grey"}}]}

data: {"choices":[{"delta":{"content":" waves"}}]}

data: {"choices":[{"delta":{"content":" at"}}]}

data: [DONE]

Streaming in Python (openai SDK)

With the OpenAI SDK, pass stream=True and iterate over the chunks:

from openai import OpenAI

client = OpenAI(base_url="https://api.velqa.dev/v1", api_key="sk-...")

stream = client.chat.completions.create(
    model="kimi-k2.6",
    messages=[{"role": "user", "content": "Write a haiku about Casablanca."}],
    stream=True,
)

for chunk in stream:
    delta = chunk.choices[0].delta.content
    if delta:
        print(delta, end="", flush=True)
print()

Good to know

Some chunks have an empty delta (for example the very first, which carries the role) — always check that delta.content is not None before printing it.
Streaming works with tool calling: tool_calls also arrive in fragments inside delta.
Usage/billing is identical to a non-streamed request.