Ecosyste.ms: Awesome

An open API service indexing awesome lists of open source software.

Awesome Lists | Featured Topics | Projects

https://github.com/rkrehn/phpchatgpt

How to use ChatGPT with PHP
https://github.com/rkrehn/phpchatgpt

chatgpt chatgpt-api chatgpt-app php php7 php8 tutorial tutorial-code

Last synced: about 1 month ago
JSON representation

How to use ChatGPT with PHP

Awesome Lists containing this project

README 

        
# About

This guide is designed to teach people how to use ChatGPT using a basic "I want one response" in PHP. This code is used in my travel itinerary generation website [ReisPlan](https://www.reisplan.net) and built off my [C# tutorial](https://github.com/rkrehn/csharpchatgpt). I love how PHP handles JSON and XML data, so this tutorial is a bit easier.



# Setup



Declare your API key (insert your [OpenAI API key](https://platform.openai.com/account/api-keys) in the quotes)



```PHP

// API endpoint URL

$url = 'https://api.openai.com/v1/chat/completions';



// Your OpenAI API key

$apiKey = '[API KEY]';

```



# Setting up the request



The following code creates the JSON request to OpenAI. You can review all the options you can include here: [https://platform.openai.com/docs/api-reference/chat/create](https://platform.openai.com/docs/api-reference/chat/create).



It's worth noting that only the "model" and the "messages" are required in the body. There are [many models](https://platform.openai.com/docs/models), but I'm using **gpt-3.5-turbo** in this example. Likewise, there are a few roles you can use such as system, user, assistance, or function. Typically, the **system** role defines what you want to accomplish. In our example, we'll tell ChatGPT **sytem** role that is acting as a travel agent (used on my [ReisPlan](https://www.reisplan.net) site).



First, we prepare the prompt statement:



```PHP

  // Prompt for the conversation

  $messages = array(

      array('role' => 'system', 'content' => 'You are a travel advisor that will deliver a detailed itinerary based on the information provided by the user.'),

      array('role' => 'user', 'content' => 'I am traveling to Denver for 7 days. Things I\'m interested in include museums and hiking.')

  );

```



Next, we build a JSON array. In the below scenario, I'm only sending the prompt and the model using an **array** and then encoding it into JSON data using the **json_encode** function.



```PHP

// Convert messages to JSON

$postData = array(

    'messages' => $messages,

    'model' => 'gpt-3.5-turbo'  // Specify the model here

);



$jsonData = json_encode($postData);

```



Next, we set up the headers using an **array** to send the request using the Authorization header with our API Key and JSON content type. The **$options** array will set up the header, method (POST), and content using the **$jsonData** we built above. 



```PHP

// Set headers for the API request

$headers = array(

    'Content-Type: application/json',

    'Authorization: Bearer ' . $apiKey

);



$options = array(

    'http' => array(

        'header'  => implode("\r\n", $headers),

        'method'  => 'POST',

        'content' => $jsonData

    )

);

```



The **stream_context_create** function will create the send data to OpenAI using an HTTP request.



```PHP

$context = stream_context_create($options);

```



Finally, we'll execute the API request using **file_get_contents** function and check for errors.



```PHP

// Execute the API request

$response = file_get_contents($url, false, $context);



// Check for errors

if ($response === false) {

    $error = error_get_last()['message'];

    echo 'API request failed: ' . $error;

    exit;

}

```



# Receiving the response



We'll just the **json_decode** function to read the $response variable from OpenAI.



```PHP

// Decode the API response

$data = json_decode($response, true);

```

You will receive a JSON response like this in the **$data** variable (don't copy this):



```JSON

{

  "id": "chatcmpl-123",

  "object": "chat.completion",

  "created": 1677652288,

  "model": "gpt-3.5-turbo-0613",

  "choices": [{

    "index": 0,

    "message": {

      "role": "assistant",

      "content": "\n\nSure, I can recommend you check out Livin the Dream brewery in Littleton and Denver Beer Co in Englewood.",

    },

    "finish_reason": "stop"

  }],

  "usage": {

    "prompt_tokens": 9,

    "completion_tokens": 12,

    "total_tokens": 21

  }

}

```



* choices[0] is the first response

* message is the node within the choices[0]

* content is the response you're looking for



Next, we'll extract the **$data** using the above and then explode individual lines in the response. The **$reply** variable contains the whole response.



But, I used **$lines** to break the reply response so I could get multiple lines of data and have them display on the website.



```PHP

// Extract the assistant's reply

$reply = $data['choices'][0]['message']['content'];



// Output the reply

// Assuming $reply contains the response from the API

$lines = explode("\n", $reply);

```



The following loop examines each **$line** received and displays it on the website using the echo function.



```PHP

// Loop through the lines

$previousLineEmpty = false;

foreach ($lines as $line) {

    // Process each line as needed

    if (empty($line)) {

        $previousLineEmpty = true;

    } else {

        if ($previousLineEmpty) {

            echo "
" . $line . "
"; // Output the line in bold

            $previousLineEmpty = false;

        } else {

            echo $line . "
"; // Output each line

        }

    }

}

```



# Everything together



Now, put it together in all its glory:



```PHP

// API endpoint URL

$url = 'https://api.openai.com/v1/chat/completions';



// Your OpenAI API key

$apiKey = '[API KEY]';



// Prompt for the conversation

$messages = array(

    array('role' => 'system', 'content' => 'You are a travel advisor that will deliver a detailed itinerary based on the information provided the user.'),

    array('role' => 'user', 'content' => 'I am traveling to Denver for 7. Things I\'m interested in include breweries.')

);



// Convert messages to JSON

$postData = array(

    'messages' => $messages,

    'model' => 'gpt-3.5-turbo'  // Specify the model here

);



$jsonData = json_encode($postData);



// Set headers for the API request

$headers = array(

    'Content-Type: application/json',

    'Authorization: Bearer ' . $apiKey

);



$options = array(

    'http' => array(

        'header'  => implode("\r\n", $headers),

        'method'  => 'POST',

        'content' => $jsonData

    )

);



$context = stream_context_create($options);



// Execute the API request

$response = file_get_contents($url, false, $context);



// Check for errors

if ($response === false) {

    $error = error_get_last()['message'];

    echo 'API request failed: ' . $error;

    exit;

}



// Decode the API response

$data = json_decode($response, true);



// Extract the assistant's reply

$reply = $data['choices'][0]['message']['content'];

```



> Sure, I can recommend you check out Livin the Dream brewery in Littleton and Denver Beer Co in Englewood.