Using SSE in AI Chat
ChatGPT has been popular for a year now。It has also driven the use of SSE.
Here, I summarize my understanding of it.
Advantages of SSE
- Uses the standard HTTP protocol, and not WebSocket. Relative to WebSocket, it has a smaller resource overhead.
- SSE transmits text data with simple overhead.
How to Use
-
1
2
3
4const evtSource = new EventSource("sse.php");
evtSource.onmessage = (e) => {
console.log(`message: ${e.data}`);
}; -
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53fetch('/api/openai/stream', {
method: 'post',
headers: {
'Content-Type': 'application/json;charset=utf-8',
},
body: JSON.stringify({
messages
})
}).then(res => {
const reader = res.body.getReader();
let buffer = '';
const readChunk = () => {
reader.read()
.then(({
value,
done
}) => {
if (done) {
console.log('Stream finished');
return;
}
const chunkString = new TextDecoder().decode(value);
buffer += chunkString;
let position;
while ((position = buffer.indexOf('\n\n')) !== -1) {
const completeMessage = buffer.slice(0, position);
buffer = buffer.slice(position + 2);
completeMessage.split('\n').forEach(line => {
if (line.startsWith('data:')) {
const jsonText = line.slice(5).trim();
if (jsonText === '[DONE]') {
console.log('done');
return;
}
try {
const dataObject = JSON.parse(jsonText);
console.log(dataObject);
} catch (error) {
console.error('JSON parse error:', error, jsonText);
}
}
});
}
readChunk();
})
.catch(error => {
console.error(error);
});
};
readChunk();
})
Fetch vs EventSource
- Both have roughly the same compatibility.
- A disadvantage of EventSource is that it cannot send a request body, so it typically requires setting up a request to send messages first, then initiating EventSource. Fetch, like XHR, can carry a request body. Thus, I think Fetch is the better choice.
ChatGPT’s Effect
ChatGPT has long supported stream returns, which give users the effect of a typewriter. Here is how GPT does it:
ChatGPT uses Fetch
, send request body, and the return type is text/event-stream; charset=utf-8
, meaning the server continuously returns response chunk until it ends.
You can find requests can be found in the network, and the request url is https://chat.openai.com/backend-api/conversation
done.