Hook up the API to GPT-3

[sirkitree]

Note: the following is not an endpoint but is a process we will have to figure out in how to send and receive data from GPT-3.

When we receive the quantum data we will need to then send this data to GPT-3.

• [ ] push data to GPT-3

And then record the response from GPT-3

• [ ] write data from GPT-3

[sirkitree]

I found this library we may be able to use, or at least look at further to see if ti will suit our needs: https://github.com/openai/GPT-3-Encoder

[sirkitree]

Also found some sample code here: https://www.twilio.com/blog/lyrics-javascript-openai-gpt3-twilio-functions

looks like that uses a library called 'got' to make the requests: https://www.npmjs.com/package/got

const got = require('got');

exports.handler = function(context, event, callback) {
    let message = "";
    let twiml = new Twilio.twiml.MessagingResponse();
    const prompt = event["prompt"]

    got.post('https://api.openai.com/v1/engines/davinci/completions', {
            json: {
                "prompt": prompt,
                "max_tokens": 50,
                "temperature": 0.9,
                "top_p": 1,
                "presence_penalty": 0.4,
                "frequency_penalty": 0.75,
                "stop": "\n",
            },
            headers: {
                'Authorization': `Bearer ${context.OPENAI_SECRET_KEY}`,
            }
    })
    .json()
    .then(response => {
        message = response.choices[0].text
    })
    .catch(error => {
        console.log(error);
        message = "Oops! Something went wrong.";
    })
    .finally(() => {
        twiml.message(message);
        callback(null, twiml);
    });
};

[sirkitree]

Here's a Node library to help: https://www.npmjs.com/package/openai-api