Streaming

Set stream: true in your Chat Completions request to receive Server-Sent Events (SSE).

How it works

Instead of one JSON response, the server sends a stream of data: lines. Each line contains a partial completion object with a delta field.

Example with cURL

bash

curl https://modelgates.ai/api/v1/chat/completions \  -H "Authorization: Bearer sk-mg-YOUR_KEY" \  -H "Content-Type: application/json" \  -d '{    "model": "anthropic/claude-sonnet-4",    "messages": [{"role": "user", "content": "Count to 10"}],    "stream": true  }'

Parsing the stream

Each chunk looks like this:

code

data: {"id":"gen-xxx","choices":[{"delta":{"content":"1"}}]} data: {"id":"gen-xxx","choices":[{"delta":{"content":"2"}}]} data: [DONE]

The final chunk is always data: [DONE].

The OpenAI SDK handles streaming automatically when you set `stream=True`.