Data Generator#

This is the heart of Suppress. It implements an algorithm that takes a prompt and a format and generates data based on the prompt and format.

How it works#

By default the user should interact with the object just by passing the parameters to the constructor. and then passing the object to the server.

The object has a method called generate that takes the passed parameters and generates the data. The server normally calls this method and returns the generated data to the client over the WEB.

Constructor Parameters#

The parameters are passed to the constructor of the object. The parameters are:

prompt|string - The prompt that the user wants to generate data for.
format|string - The format of the data that the user wants to generate
llm|SuppressModel - The model that the user wants to use to generate the data.

`generate` method#

This method only takes the data which is supposed to be parsed into the prompt. It uses the parameters that were passed to the constructor. It returns an object that is the generated data with some metadata. The response format depends on the LLM used.

What should the `data` look like?#

The data should be a json, where the keys are the names of the variables in the prompt and the values are the values of the variables. For example, if the prompt is I have {number} apples and the data is {"number": 5}, then the generated data will be based on I have 5 apples.

Example Response#

In this example lets use OpenAILLM as the LLM.

{
    "response": "This is the generated string or JSON or whatever the format is.",
    "about": {
        "model": "OpenAI GPT-3",
        "omniID": "text-davinci-003"
    }
}

Again, the result depends on the LLM used, but this is the general structure of the SuppressModel. You can find more information about the LLMs in the LLM documentation.

Properties of the DataGenerator#

The DataGenerator has the following properties:

parseJson|boolean - This property is used to tell the DataGenerator to parse the JSON that is returned from the formatter. The default value is true.
doFormat|boolean - This property is used to tell the DataGenerator to format the data that is returned from the LLM. This is useful when the LLM returns a string. The default value is true.

You can customize these properties when creating the DataGenerator object. Use the set method to set the properties.

const dg = new DataGenerator(prompt, format, llm).set({
    parseJson: false,
    doFormat: true
});

The set method returns the DataGenerator object so you can chain the methods.

Data Generator

Contents