Streaming

Set "stream": true and the response arrives as Server-Sent Events — token deltas as they are generated, in the OpenAI chunk shape.

Enabling streaming

curl

curl https://api.astrodyne.ai/v1/chat/completions \
  -H "Authorization: Bearer $ASTRODYNE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
       "model": "YOUR_MODEL_ID",
       "stream": true,
       "messages": [{"role": "user", "content": "Hello"}]
     }'

Python

import os
from openai import OpenAI

client = OpenAI(
    base_url="https://api.astrodyne.ai/v1",
    api_key=os.environ["ASTRODYNE_API_KEY"],
)

stream = client.chat.completions.create(
    model="YOUR_MODEL_ID",
    messages=[{"role": "user", "content": "Hello"}],
    stream=True,
)
for chunk in stream:
    delta = chunk.choices[0].delta.content
    if delta:
        print(delta, end="", flush=True)

JavaScript

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://api.astrodyne.ai/v1",
  apiKey: process.env.ASTRODYNE_API_KEY,
});

const stream = await client.chat.completions.create({
  model: "YOUR_MODEL_ID",
  messages: [{ role: "user", content: "Hello" }],
  stream: true,
});
for await (const chunk of stream) {
  process.stdout.write(chunk.choices[0]?.delta?.content ?? "");
}

How chunks arrive

The response has Content-Type: text/event-stream. Each event is a data: line containing a JSON chunk; text arrives at choices[0].delta.content. A final chunk carries finish_reason, then the stream ends with:

SSE

data: {"object":"chat.completion.chunk","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}

data: [DONE]

What is available during the stream

The X-Astrodyne-Request-Id header is on the response from the start — capture it before reading the body.
Token usage is not included in stream chunks. Authoritative usage and the settled charge are recorded when the stream completes, and appear with the request in the Request Explorer.

Mid-stream failures and disconnects

If a failure occurs after streaming has begun, Astrodyne emits one sanitized error event in the standard error shape with "object": "error", then closes the stream without [DONE]. If your client disconnects mid-stream, the request is settled for what was actually delivered — you are never charged when nothing was delivered, and the final state is visible in the Request Explorer either way.

Cancelling from a client

Aborting the HTTP request (for example with an AbortController) is the supported way to stop a stream early. The Playground's Stop button does exactly this.

Request IDs → Track a streamed request end to end. Playground → Watch a stream live in your browser.