Skip to main content

Chat Model

The Chat class is the main entry point for building Chat Completions requests. It extends the abstract Model class and produces a payload compatible with both OpenAI and Ollama (/v1 endpoint).

use ByJG\LlmApiObjects\Model\Chat;

Constructor​

new Chat(
    model: string,           // Model name, e.g. 'gpt-4o', 'llama3'
    messages: Message[],     // Conversation turns
    system: string,          // System prompt (prepended automatically)
    stream: bool,            // Whether to stream the response (default: false)
)

All constructor parameters are optional with defaults (model: 'llama3', empty messages, no system, stream: false).

toApi()​

Serializes the chat to the array expected by openai-php/client's chat()->create():

$result = $client->chat()->create($chat->toApi());

What toApi() produces​

[
    'model'    => 'gpt-4o',
    'messages' => [
        ['role' => 'system',    'content' => '...'],  // from $system
        ['role' => 'user',      'content' => '...'],
        ['role' => 'assistant', 'content' => '...'],
    ],
    'stream'   => false,
    // 'tools'      => [...],       // when tools are added
    // 'tool_choice'=> 'auto',      // when tool_choice is set
    // 'options'    => [...],       // when ModelFileParameters are set (Ollama)
]
info

When Ollama ModelFileParameters are set, the key is named options (not parameters) because that is the key Ollama's API expects. Chat::toApi() performs this remapping automatically.

System Prompt​

Set a system prompt via the system constructor argument. It is prepended to messages as the first entry with role=system:

$chat = new Chat(
    model: 'gpt-4o',
    messages: [
        new Message(role: Role::user, message: 'Hello!'),
    ],
    system: 'You are a helpful assistant. Always respond in English.',
);

Streaming​

Set stream: true to request a streamed response. Use createStreamed() on the client side:

$chat = new Chat(model: 'gpt-4o', messages: [...], stream: true);
$stream = $client->chat()->createStreamed($chat->toApi());

foreach ($stream as $response) {
    echo $response->choices[0]->delta->content;
}

Adding Tools​

See Tools for the full reference. Quick example:

use ByJG\LlmApiObjects\Enum\ToolChoice;
use ByJG\LlmApiObjects\Tool\Tool;
use ByJG\LlmApiObjects\Tool\ToolFunction;
use ByJG\LlmApiObjects\Tool\ToolParameter;

$chat = new Chat(model: 'gpt-4o', messages: [...]);

$chat->addTool(new Tool(
    new ToolFunction(
        name: 'get_weather',
        description: 'Get the current weather',
        parameters: [
            new ToolParameter(name: 'location', type: 'string', required: true),
        ],
    ),
));

$chat->setToolChoice(ToolChoice::auto);

Ollama Model Parameters​

Use setParameter() to set Ollama-specific generation parameters. These are serialized under the options key:

use ByJG\LlmApiObjects\Enum\ModelFileParameters;

$chat->setParameter(ModelFileParameters::temperature, 0.7);
$chat->setParameter(ModelFileParameters::numCtx, 4096);

See Model Parameters for the full list of available parameters and their valid ranges.