I have written a prompt to generate custom code as step in my workflow, it is throwing an error when I attempt to run the block. Here is the prompt:
"Write an openai function call based on the provided step by step instructions. Function calling enable models to generate function arguments which adhere to the provided specifications. Note that the API will not actually execute any function calls. It is up to developers to execute function calls using model outputs. In our experience, the real benefit of that feature is the ability for ChatGPT to support structured input and/or output. Here is an example of what a node.js script that uses function calling looks like."
// Import necessary packages and modules
require(‘dotenv’).config(); // Load environment variables from a .env file
const { Configuration, OpenAIApi } = require(“openai”); // Import classes from the OpenAI API module
const axios = require(“axios”); // Import the axios library for making HTTP requests
const moment = require(“moment-timezone”); // Import the moment-timezone library for time formatting
//Create a new OpenAI API configuration object with an API key loaded from the environment variables
const configuration = new Configuration({
apiKey: process.env.OPENAI_API_KEY,
});
const openai = new OpenAIApi(configuration); // Create an instance of the OpenAI API with the configuration object
/** Define an asynchronous function that takes a location string as a parameter,
retrieves the current time at that location using the World Time API,
then formats and logs the time to the console **/
async function lookupTime(location, name) {
try {
const response = await axios.get(http://worldtimeapi.org/api/timezone/${location});
// Make a GET request to the World Time API with the ‘location’ value as the timezone.
const { datetime } = response.data; // Destructure ‘datetime’ from the response to extract it.
const dateTime = moment.tz(datetime, location).format(“h:mmA”);
// Create a new date object using the specified timezone and format it in 12-hour time with AM/PM label.
console.log(The current time in ${name} is ${dateTime}.);
// Log the formatted time to the console.
} catch (error) {
console.error(error); //Log any errors that occur to the console.
}
}
/** Define an async function called ‘main’
which uses the OpenAI API to generate a chatbot-like response to a given input message
and includes the ability to call the ‘lookupTime’ function **/
async function main() {
const completion = await openai.createChatCompletion({
model: “gpt-3.5-turbo-0613”, // Choose the GPT model to use for generating responses
messages: [
{role: “system”, content: “You are a helpful assistant.”}, // Specify the role and content of the messages
{role: “user”, content: “What time is it in London, England?”} // sent to the GPT model for generating the response.
],
functions: [
{ // Define the ‘lookupTime’ function
name: “lookupTime”,
description: “get the current time in a given location”,
parameters: { // Define the input parameters for the function
type: “object”, // The parameter is an object
properties: {
location: { // The ‘location’ property of the object is a required string
type: “string”, // The ‘location’ property is a string value
description: “The location, e.g. Beijing, China. But it should be written in a timezone name like Asia/Shanghai”
},
name: {
type: “string”,
description: “The location mentioned in the prompt. Example: Beijing, China.”
}
},
required: [“location”, “name”] // The ‘location’ property is required
}
}
],
function_call: “auto” // Specify that the API should automatically detect and execute function calls within the response
})
console.log("completion.data: ", completion.data)
const completionResponse = completion.data.choices[0].message; // Extract the generated completion from the OpenAI API response
console.log("completionResponse: ", completionResponse); // Output the generated response to the console
if(!completionResponse.content) { // Check if the generated response includes a function call
const functionCallName = completionResponse.function_call.name;
console.log("functionCallName: ", functionCallName);
if(functionCallName === "lookupTime") { // If the function being called is 'lookupTime'
const completionArguments = JSON.parse(completionResponse.function_call.arguments); // Extract the argument for the function call
console.log("completionArguments: ", completionArguments);
lookupTime(completionArguments.location, completionArguments.name) // Call the 'lookupTime' function with the specified location argument.
}
}
return completionResponse.content
}
main();
// Call the ‘main’ function to run the program and generate the chatbot response.
Here are your step by step instructions to write a custom function:
{{steps.task_classification_and_instruction_function.code_generation_output.job_to_be_done}}
{{steps.task_classification_and_instruction_function.code_generation_output.code_generation_instructions}}
Here are the related docs.
{{steps.TESTING_source_doc}}
{{steps.TESTING_context_doc}}
{{steps.TESTING_template_doc}}
Generate a script that sucesfully defines an output schema to achieve the job to be be done.