Stream responses

POST /api/arena/chat/stream takes the same body as Chat and streams the reply as Server-Sent Events. Use it when you want text on screen as soon as the model starts writing.

The event stream

Each event is one data: line holding a JSON object, followed by a blank line.

data: {"object":"ai.stream.delta","delta":{"type":"output_text","text":"There once"}}

data: {"object":"ai.stream.delta","delta":{"type":"output_text","text":" was a packet"}}

data: {"object":"ai.stream.done","finishReason":"stop"}

`object`	Meaning	Fields
`ai.stream.delta`	A chunk of text	`delta.text`
`ai.stream.done`	Generation finished	`finishReason`
`ai.error`	Something went wrong	`code`, `message`

The stream endpoint answers with one model and doesn’t save the turn to a thread. For battles, or to keep history, use Start a battle or Chat with a threadId.

Read the stream

The endpoint is a POST, so use fetch (or your HTTP client’s streaming mode) rather than the browser’s EventSource.

curl -N https://api.dualmindlab.tech/api/arena/chat/stream \
  -H "Authorization: Bearer $DUALMIND_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"prompt": "Write a limerick about latency", "model": "auto"}'

const res = await fetch("https://api.dualmindlab.tech/api/arena/chat/stream", {
  method: "POST",
  headers: { Authorization: `Bearer ${process.env.DUALMIND_TOKEN}`, "Content-Type": "application/json" },
  body: JSON.stringify({ prompt: "Write a limerick about latency", model: "auto" }),
});
if (!res.ok) throw new Error(`HTTP ${res.status}`);

const reader = res.body.pipeThrough(new TextDecoderStream()).getReader();
let buffer = "";
for (;;) {
  const { value, done } = await reader.read();
  if (done) break;
  buffer += value;
  const events = buffer.split("\n\n");
  buffer = events.pop(); // keep a partial event for the next read
  for (const evt of events) {
    if (!evt.startsWith("data: ")) continue;
    const msg = JSON.parse(evt.slice(6));
    if (msg.object === "ai.stream.delta") process.stdout.write(msg.delta.text);
    else if (msg.object === "ai.error") throw new Error(`${msg.code}: ${msg.message}`);
  }
}

import json, os, requests

with requests.post(
    "https://api.dualmindlab.tech/api/arena/chat/stream",
    headers={"Authorization": f"Bearer {os.environ['DUALMIND_TOKEN']}"},
    json={"prompt": "Write a limerick about latency", "model": "auto"},
    stream=True,
    timeout=120,
) as r:
    r.raise_for_status()
    for line in r.iter_lines(decode_unicode=True):
        if not line or not line.startswith("data: "):
            continue
        msg = json.loads(line[6:])
        if msg["object"] == "ai.stream.delta":
            print(msg["delta"]["text"], end="", flush=True)
        elif msg["object"] == "ai.error":
            raise RuntimeError(f"{msg['code']}: {msg['message']}")

Tips

Cancel early

Abort the request (for example with an AbortController) to stop generation. The server stops when the connection closes.

Errors arrive in-band

Once streaming starts the status is already 200, so failures come as an ai.error event. Always handle it.

Render Markdown at the end

Chunks can split Markdown and code fences. Show raw text while streaming and render Markdown once ai.stream.done arrives.

Pick a model

model takes a modelName, a modelId or auto (a random Groq or Google AI Studio model).

Overview

Tutorials

Arena

Voting & leaderboard

Models

Threads

Utilities

The event stream

Read the stream

Tips

Cancel early

Errors arrive in-band

Render Markdown at the end

Pick a model

​The event stream

​Read the stream

​Tips

Cancel early

Errors arrive in-band

Render Markdown at the end

Pick a model

The event stream

Read the stream

Tips