Only gpt-3.5-turbo works. Older models return 404 error.

[jacksutherland]

I've change the model to multiple options gpt-3.5-turbo, text-davinci-003, text-davinci-002, etc, and gpt-3.5-turbo is the only one that returns a valid response. I get a 404 for everything else. I'm happy to accept it's something on my part, except I've gone trough the docs verbatim.

This wouldn't be a issue except that gpt-3.5-turbo is extremely slow. Sometimes a 5-10 second response, which will result in a bounce for most users. So I'd like to use an older/faster model where possible.

The Stack Trace:

**Tectalic\OpenAi\ClientException**
Unsuccessful response. HTTP status code: 404 (Not Found).

1. in .../tectalic/openai/src/Handlers/ChatCompletions.phpat line 150
141142143144145146147148149150151152153154155156157158159     *
     * @api
     * @return AbstractModel|AbstractModelCollection
     * @throws ClientException if an unsuccessful HTTP response occurs.
     * @throws ClientException if the response body cannot be parsed.
     */
    public function toModel(): object
    {
        if ($this->getResponse()->getStatusCode() < 200 || $this->getResponse()->getStatusCode() >= 300) {
            throw new ClientException(
                \sprintf(
                    'Unsuccessful response. HTTP status code: %s (%s).',
                    $this->getResponse()->getStatusCode(),
                    $this->getResponse()->getReasonPhrase()
                )
            );
        }
        $class = $this->modelType;
        try {

[thejamescollins]

Hi there @jacksutherland,

Some of those models are only available via the chatCompletions handler, and others are available via the completions handler.

Please see the Examples in the readme for details on which models are available on which endpoints.

You may also find this official information from OpenAI useful too.

Regarding API endpoint speeds (latencies), you may like to review these suggestions from OpenIA themselves.

All the best,

James

[jacksutherland]

Ah, I see. Your readme lists text-davinci-003 and text-davinci-002 as supported for the completions and chatCompletions handlers, but the Open AI doc only shows them available for the completions endpoint.

[thejamescollins]

Thanks @jacksutherland, we have now corrected this error in our readme (#27).