Streaming

Set stream: true to receive the response as server-sent events (SSE): tokens arrive as they're generated. This is identical to the OpenAI streaming protocol, so any OpenAI-compatible client handles it out of the box.

Python

from openai import OpenAI

client = OpenAI(
    api_key="pgsk_your_key_here",
    base_url="https://api.pgsgrove.com/v1",
)

stream = client.chat.completions.create(
    model="deepseek-v4-pro",
    messages=[{"role": "user", "content": "Explain gradient descent in two sentences."}],
    stream=True,
)

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

JavaScript / TypeScript

const stream = await client.chat.completions.create({
  model: "deepseek-v4-pro",
  messages: [{ role: "user", content: "Explain gradient descent in two sentences." }],
  stream: true,
});

for await (const chunk of stream) {
  process.stdout.write(chunk.choices[0]?.delta?.content ?? "");
}

curl

curl https://api.pgsgrove.com/v1/chat/completions \
  -H "Authorization: Bearer $PGS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek-v4-pro",
    "stream": true,
    "messages": [{ "role": "user", "content": "Count to five." }]
  }'

The stream ends with a final data: [DONE] line.

Usage on streamed calls

By default a streamed response does not include a usage summary. To get token counts back on the final chunk, pass stream_options:

stream = client.chat.completions.create(
    model="deepseek-v4-pro",
    messages=[...],
    stream=True,
    stream_options={"include_usage": True},
)

Billing does not depend on this flag; you are metered accurately either way. It's purely so your client can read the token totals.

Cancelling a stream

If your client disconnects, generation stops immediately and you are billed only for what was produced up to that point. There's no need to "drain" a stream you no longer need; just close the connection (in the OpenAI SDKs, break out of the loop or abort the request).

Closing early is the recommended way to cap spend on an over-long response.