Toolcall loop when using tool name in withToolChoice

[apm-tobias-duesterhus]

Hey There,

I am using Version 0.99.11 and probably found a bug or did something totally wrong.

I create the following test-script to verify the effect:

$weatherTool = Tool::as('weather')
    ->for('Get current weather conditions')
    ->withStringParameter('city', 'The city to get weather for')
    ->using(function (string $city): string {
        // Your weather API logic here
        Log::debug('weather tool called for city: ' . $city);
        return "The weather in {$city} is sunny and 72°F.";
    });

$pendingRequest = \Prism\Prism\Facades\Prism::text()->using(
    provider: Provider::OpenAI,
    model: ...,
    providerConfig: ...
);

$result = $pendingRequest
    ->withMessages([new UserMessage('What is the weather like in New York today?')])
    ->withTools([$weatherTool])
    ->withToolChoice('weather')
    ->withMaxSteps(5)
    ->asText();

dump($result->text);

In this case, the log from the weather tool is printed five times, the result test from the llm is empty. In case I comment out the tool choice, I get the correct answer. This seems to be an issue at least for OpenAI and Gemini.