src/lib/agents/tools/textToSpeech.ts

import { getInference } from '$lib/agents/getInference';
import type { TextToSpeechArgs, TextToSpeechOutput } from '@huggingface/inference';

import type { Tool } from './tool';

export const textToSpeechTool: Tool<TextToSpeechArgs['inputs'], TextToSpeechOutput> = {
	name: 'textToSpeech',
	description: 'This tool takes a text input and turns it into an audio file.',
	examples: [
		{
			prompt: 'Say the following out loud:"Hello world!"',
			command: "textToSpeech('Hello world!')"
		},
		{ prompt: 'Say the content of the string txt out loud', command: 'textToSpeech(txt)' }
	],
	call: async (input) => {
		return await getInference().textToSpeech(
			{
				inputs: await input,
				model: 'espnet/kan-bayashi_ljspeech_vits'
			},
			{ wait_for_model: true }
		);
	}
};