Required and Arguments in function calling

[windoze95]

Hi, I am having trouble trying to provide a required field in the function calling, if I am understanding this correctly, when you set a field as required, it is a field that you are required to provide before making the request, and you provide it using the arguments, correct?
Am I going about this correctly in the snippet below? I'm getting mixed results in my commonRecipeDefs (second code snippet).

	// Define the function for use in the API call
	var functionDef = openai.FunctionDefinition{
		Name: "create_recipe",
		Parameters: jsonschema.Definition{
			Type: jsonschema.Object,
			Properties: map[string]jsonschema.Definition{
				"title":       {Type: jsonschema.String},
				"main_recipe": commonRecipeDef,
				"sub_recipes": {
					Type:        jsonschema.Array,
					Description: "Additional recipes like sauces, sides, buns, tortillas, etc",
					Items:       &commonRecipeDef,
				},
				"dall_e_prompt": {
					Type:        jsonschema.String,
					Description: "Prompt to generate an image for the recipe",
				},
				"unit_system": {
					Type:        jsonschema.String,
					Enum:        []string{"us customary", "metric"},
					Description: "Unit system to be used (us customary or metric)",
				},
				"hashtags": {
					Type:        jsonschema.Array,
					Description: "Relevant hashtags for the main recipe and any associated sub-recipes (Alphanumeric characters only. No #. Exclude terms like 'recipe', 'homemade', 'DIY', or similar words, as they are understood to be implied. Omit the '#' symbol. camelCase (if it starts with a letter, the first letter is always lowercase) formatting if more than one word.)",
					Items:       &jsonschema.Definition{Type: jsonschema.String},
				},
			},
			Required: []string{"unit_system"},
		},
	}
        
	// Use the functionDef in the list of function definitions for the API call
	functions := []openai.FunctionDefinition{functionDef}

	maxRetries := 5
	var resp openai.ChatCompletionResponse
	var err error

	for i := 0; i < maxRetries; i++ {
		resp, err = c.Client.CreateChatCompletion(
			context.Background(),
			openai.ChatCompletionRequest{
				Model:     openai.GPT4,
				Messages:  messages,
				Functions: functions,
				FunctionCall: &openai.FunctionCall{
					Name:      functionDef.Name,
					Arguments: "{\"unit_system\":\"us customary\"}",
				},
			},
		)

	// Common recipe definition
	var commonRecipeDef = jsonschema.Definition{
		Type: jsonschema.Object,
		Properties: map[string]jsonschema.Definition{
			"recipe_name": {Type: jsonschema.String},
			"ingredients": {
				Type:        jsonschema.Array,
				Description: "List of ingredients used in the recipe",
				Items: &jsonschema.Definition{
					Type: jsonschema.Object,
					Properties: map[string]jsonschema.Definition{
						"name":   {Type: jsonschema.String, Description: "Name of the ingredient"},
						"unit":   {Type: jsonschema.String, Description: "Unit for the ingredient, comply with UnitSystem field if specified. Omit for whole ingredients.", Enum: []string{"grams", "ml", "cups", "pieces", "teaspoons", "tablespoons", "ounces", "pounds", "pinch", "dash", "quarts", "gallons", "liters"}},
						"amount": {Type: jsonschema.Number, Description: "Amount of the ingredient"},
					},
				},
			},
			"instructions": {
				Type:        jsonschema.Array,
				Description: "Steps to prepare the recipe (no numbering)",
				Items:       &jsonschema.Definition{Type: jsonschema.String},
			},
			"time_to_cook": {
				Type:        jsonschema.Number,
				Description: "Total time to prepare the recipe(s) in minutes",
			},
		},
	}

[windoze95]

After reviewing the openai documentation again, specifically the weather examples, I may have overthought the process. Given my use case isn't designed to have an interactive experience, I believe injecting the unit system directly into the user prompt will suffice and the arguments are not relevant in this scenario.