README

(Written by ChatGPT)

This is a library for interacting with Language Learning Models (LLMs), built with a focus on AI-driven chat functionalities. The library currently supports OpenAI's GPT-3.5-turbo, but it can be extended to support other LLMs as well.

Features

• Cross-API Compatibility: The library is designed to work with different language learning models, including OpenAI's GPT-3.5-turbo.

• Ease of Use: With simple configuration and intuitive API, this library enables developers to easily interact with complex language learning models.

• Streaming Support: The library offers streaming capabilities that allow real-time interactions with language learning models.

• Chat History: The library maintains a history of the chat session which can be retrieved and analyzed anytime.

• You can add support for other AI providers by extending the ApiClient class and implementing the required methods.

• Error Handling: The library provides robust error handling to deal with potential issues during interactions with language learning models.

• Documentation: Comprehensive documentation is available to guide developers in using the library.

• Community Support: Issues and feature requests can be raised in the GitHub repository, and the active community is open to discussions and improvements.

• License: The library is open source and is licensed under the MIT License.

• Support for Future AI Models: The architecture of the library is designed to support any future language learning models seamlessly.

• It is clean, well-structured, and includes comments for clarity.

Getting Started

Firstly, you need to install the library in your PHP project.

composer require blinq/openai

Usage

Chat with OpenAI's GPT-3.5-turbo

You need to initialize a client object using ApiClient configuration. The following example demonstrates a chat session with OpenAI's GPT-3.5-turbo.

<?php

use Blinq\LLM\Config\ApiConfig;
use Blinq\LLM\Entities\ChatMessage;
use Blinq\LLM\Client;

$config = new ApiConfig('openai', 'your-api-key');
$client = new Client($config);

// Set the system message. This is optional.
$client->setSystemMessage("You are a nice chatbot");

// User message
$client->chat("Hello, how are you?");

// Get the last message
$message = $client->getLastMessage();

echo $message->content; // Prints "I am fine, thank you. How are you?"

// Get the chat history
$history = $client->getHistory();

foreach ($history as $message) {
    echo $message->content;
}

The client by default holds a history of the chat session. You can retrieve the history using the getHistory() method. The history is an array of ChatMessage objects, which have the following properties:

• content: The content of the message
• role: The role of the message (user or system)

You can reset the chat history using the resetHistory() method.

Or use the getLastMessageAndClearHistory() method to get the last message and clear the history.

Chat with Stream

If the streaming option is used, the following example shows how to add a handler to deal with the stream of chat messages.

<?php

use Blinq\LLM\Config\ApiConfig;
use Blinq\LLM\Entities\ChatStream;
use Blinq\LLM\Client;

$config = new ApiConfig('openai', 'your-api-key');
$client = new Client($config);

$client->addStreamHandler(function (ChatStream $stream) {
    // Handle the stream data
    echo $stream->getMessage()?->content; // Prints the partial message content
    
    // The $stream object has a 'done' property to check if the stream is done
    if ($stream->done) {
        // Do something when the stream is done
    }
});

$client->chat("Hello, how are you?", "user", ['stream' => true]);

Customization

The ApiClient class is abstract and can be extended to create custom drivers for different APIs.

The current implementation is for OpenAI, but if you'd like to use a different AI provider, you can create a new class that extends ApiClient, and implement the required methods (chat(), completion(), etc.).

For any issues or additional features, feel free to open an issue in the GitHub repository.

License

This project is licensed under the MIT License. See the LICENSE file for details.