// Import LLMChain, ChatOpenAI and prompt templates from langchain library
import { LLMChain } from "langchain";
import { ChatOpenAI } from "langchain/chat_models";
import {
	ChatPromptTemplate,
	HumanMessagePromptTemplate,
	SystemMessagePromptTemplate,
} from "langchain/prompts";

// Create a chat model instance using your OpenAI API key from an environment variable
const chat = new ChatOpenAI({
	openAIApiKey: process.env.NEXT_PUBLIC_OPENAI_API_KEY,
	temperature: 0,
});

// Create an array of mental models as objects with name and description properties
const mentalModels = [
	{
		name: "Pareto Principle",
		description:
			"The Pareto principle states that 80% of outcomes result from 20% of causes.",
	},
	{
		name: "Occam's Razor",
		description:
			"Occam's razor states that the simplest explanation is usually the correct one.",
	},
	{
		name: "Confirmation Bias",
		description:
			"Confirmation bias is the tendency to seek out or interpret information that confirms our existing beliefs.",
	},
];

// Create a function that returns a random element from an array
const randomElement = (array) => {
	return array[Math.floor(Math.random() * array.length)];
};

// Create a function that returns a prompt template for introducing a mental model
const introduceMentalModel = () => {
	const model = randomElement(mentalModels); // Pick a random mental model
	const message = `Let me introduce you to ${model.name}. ${model.description}.`; // Compose a message explaining the mental model
	const prompt = new SystemMessagePromptTemplate(message); // Create a system message prompt template with the message
	return prompt; // Return the prompt template
};

// Create a function that returns a prompt template for asking user about their scenario or problem
const askUserScenario = () => {
	const message =
		"Now tell me about your scenario or problem where you want to apply this mental model."; // Compose a message asking user about their scenario or problem
	const prompt = new SystemMessagePromptTemplate(message); // Create a system message prompt template with the message
	return prompt; // Return the prompt template
};

// Create a function that returns a human message prompt template based on user input
const getUserScenario = (userInput) => {
	const input = userInput; // Get user input as string
	const prompt = new HumanMessagePromptTemplate(input); // Create human message prompt template with user input as text property
	return prompt; // Return the prompt template
};

// Create a function that returns an empty system message prompt template for generating response based on user input and mental model name
const generateResponse = (userInput, modelName) => {
	const input = userInput; // Get user input as string
	const name = modelName; // Get mental model name as string
	const textProperty = `Here is how you can apply ${name} to your scenario or problem: ${input}`; // Compose text property for system message by appending user input after colon
	const textPropertyWithLLM = `<llm>${textProperty}</llm>`; // Wrap text property with <llm> tags for LLM processing
	const emptyTextProperty = ""; // Set empty text property for system message
	const llmProperty = true; // Set llm property as true for system message
	const llmOptionsProperty = {}; // Set llm options property as empty object for system message
	const prompt = new SystemMessagePromptTemplate(
		emptyTextProperty,
		llmProperty,
		llmOptionsProperty
	); // Create an empty system message prompt template with llm properties
	prompt.text = textPropertyWithLLM; // Assign text property with LLM tags to prompt
	return prompt; // Return the prompt template
};