Stream Your First Response

Streaming returns the reply token by token over Server-Sent Events (SSE): a long-lived HTTP response where each message begins with data:. The SDKs handle the framing.

Turn It On

Set stream: true on the same chat request:

curl -N "https://app.trynito.ai/v1/chat/completions" \
  -H "Authorization: Bearer $NITO_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "z-ai/glm-5.2",
    "stream": true,
    "messages": [{ "role": "user", "content": "Write a haiku about privacy." }]
  }'

from openai import OpenAI

client = OpenAI(api_key="z_sk_your_key_here", base_url="https://app.trynito.ai/v1")

stream = client.chat.completions.create(
    model="z-ai/glm-5.2",
    messages=[{"role": "user", "content": "Write a haiku about privacy."}],
    stream=True,
)
for chunk in stream:
    delta = chunk.choices[0].delta
    if delta.content:
        print(delta.content, end="", flush=True)

import OpenAI from "openai";

const client = new OpenAI({ apiKey: "z_sk_your_key_here", baseURL: "https://app.trynito.ai/v1" });

const stream = await client.chat.completions.create({
  model: "z-ai/glm-5.2",
  messages: [{ role: "user", content: "Write a haiku about privacy." }],
  stream: true,
});
for await (const chunk of stream) {
  const delta = chunk.choices[0]?.delta?.content;
  if (delta) process.stdout.write(delta);
}

-N turns off curl’s buffering so tokens show as they arrive.

The Stream

data: {"id":"...","object":"chat.completion.chunk","choices":[{"index":0,"delta":{"role":"assistant","content":"Sealed "},"finish_reason":null}]}

data: {"id":"...","object":"chat.completion.chunk","choices":[{"index":0,"delta":{"content":"and private"},"finish_reason":null}]}

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

data: [DONE]

New text is in choices[0].delta.content. Concatenate the chunks to rebuild the reply.
delta.role: "assistant" appears once, on the first chunk that carries content or reasoning.
The last content chunk sets finish_reason: stop, length, tool_calls, or error.
data: [DONE] ends the stream. It is not JSON.
: keep-alive comment lines can appear. Ignore them.

Usage

Usage is known only when generation finishes. Ask for it:

{ "model": "z-ai/glm-5.2", "stream": true, "stream_options": { "include_usage": true }, "messages": [ ... ] }

One extra chunk with an empty choices array and a usage object arrives before data: [DONE].

Mid-Stream Errors

Once the stream starts, HTTP 200 is already sent. A later failure arrives as a final chunk with a top-level error object, an empty delta, and finish_reason: "error", and the stream ends without data: [DONE]. Check each chunk for error as well as the status. A failure before any byte is written returns an ordinary HTTP error.

Stream Format Reference

Every chunk field.

Make Your First Request

Non-streaming.

Tool Call Streaming

Tool calls in fragments.

Basics

Features

Server-Side Tools

Multimodal

Best Practices

SDKs

API Reference

Stream Your First Response

Turn It On

The Stream

Usage

Mid-Stream Errors

Stream Format Reference

Make Your First Request

Tool Call Streaming

​Turn It On

​The Stream

​Usage

​Mid-Stream Errors

​Related

Stream Format Reference

Make Your First Request

Tool Call Streaming

Turn It On

The Stream

Usage

Mid-Stream Errors

Related