Generate Text

The basic LLM function is called generate(). It is imported from the models Polyfire module. By default generate() calls the OpenAI gpt-3.5-turbo endpoint, which is the model running ChatGPT.

An example of a simple AI-generated greeting in a React app:

function GreetingBox({ userName }) {
  const [greeting, setGreeting] = useState();
  const { models } = usePolyfire();

  useEffect(() => {
      models
        .generate(`Write a fun and simple greeting for ${userName}`)
        .then(setGreeting);
  }, []);

  return <div>{greeting}</div>;
}

Customizing the Model

You can change the the model the generate() function calls by including the options object. The complete list of available models and their names can be found here or on the Dashboard.

The same example but using GPT-4:

function GreetingBox({ userName }) {
  const [greeting, setGreeting] = useState();
  const { models } = usePolyfire();

  useEffect(() => {
    models
      .generate(`Write a fun and simple greeting for ${userName}`, { model: "gpt-4" })
      .then(setGreeting);
  }, []);

  return <div>{greeting}</div>;
}

Uncensored Generation

To generate uncensored text, call the model wizard-mega-13b-awq. It can be very handy if you are trying to build, let's say, an AI girlfriend.

function DirtyAnswer({ userMessage, girlfriendPrompt }) {
  const [dirtyTalk, setDirtyTalk] = useState();
  const { models } = usePolyfire();

  useEffect(() => {
    const prompt = `${girlfriendPrompt}. Write a dirty answer to this message: ${userMessage}.`
    
    models
      .generate(prompt, { model: "wizard-mega-13b-awq" })
      .then(setDirtyTalk);
  }, []);

  return <div>{dirtyTalk}</div>;
}