【OpenAi】串接openAi - chatGpt Api

利用套件來串接chatGpt api

composer require orhanerday/open-ai

到OpenAi官方網站進行註冊以及創造api token

Open Ai Api參數文件 [官方]

先設定env

OPENAI_API_KEY={OPENAI_API_KEY}

use Orhanerday\OpenAi\OpenAi;

// 建議使用config
$openAi = new OpenAi(env('OPENAI_API_KEY'));
// 跟基本串接API一樣
$chat = $open_ai->chat([
 'model' => 'gpt-3.5-turbo',
'messages' => [
 [
 "role" => "user",
"content" => "今天天氣如何"
 ],
],
'max_tokens' => 1000,
]);
// message
dd(Arr::get(json_decode($chat,1)),'choices.0.message.content');

會發現回應時間隨著傳送過去的文字複雜度以及長度而增加

查了網路資料後來終於發現一個東西

串流

使用串流套件的範例如下

php

use Orhanerday\OpenAi\OpenAi;

$openAi = new OpenAi(env('OPEN_AI_API_KEY'));

$opts = [
 'prompt' => "Hello",
 'temperature' => 0.9,
 "max_tokens" => 150,
 "frequency_penalty" => 0,
 "presence_penalty" => 0.6,
 "stream" => true,
];

// set header
header('Content-type: text/event-stream');
header('Cache-Control: no-cache');
header('Access-Control-Allow-Origin: *');

$openAi->chat($opts, function ($curl_info, $data) {
 echo $data . "

";
 echo PHP_EOL;
 ob_flush();
 flush();
 return strlen($data);
});

html

html>

 SSE Example

Hello

 var eventSource = new EventSource("{endPoint}");
 var div = document.getElementById('divID');
 eventSource.onmessage = function (e) {
 if (e.data == "[DONE]") {
 eventSource.close();
 console.log('done');
 return false;
 } else {
 const content = JSON.parse(e.data).choices[0].delta.content;
 if (content != undefined) {
 div.innerHTML += JSON.parse(e.data).choices[0].delta.content;
 }
 }
 };
 eventSource.onerror = function (e) {
 console.log(e);
 };

demo