core/chat.ts

import { OpenAI } from "openai";
import { loadEnv } from "../utils/general";

// Initialize the OpenAI client
const client = new OpenAI({
	baseURL: "https://api-inference.huggingface.co/v1/",
	apiKey: loadEnv("HF_TOKEN"), // Load the API key
});

/**
 * Generate a response based on a message, either describing an image or engaging in a chat.
 * @param {string} message - The text message or image URL to send to the model.
 * @param {boolean} isImage - Flag to determine if the input is an image URL or plain text.
 * @returns {Promise<string>} A promise that resolves with the generated response.
 */
export async function generateResponse(message, isImage = false) {
	let output = "";

	// Prepare the message structure based on whether it's an image or text
	const content = isImage
		? [
				{
					type: "text",
					text: "Describe this image in one sentence.",
				},
				{
					type: "image_url",
					image_url: { url: message },
				},
		  ]
		: [
				{
					type: "text",
					text: message,
				},
		  ];

	// Create a streaming completion request
	const stream = await client.chat.completions.create({
		model: "meta-llama/Llama-3.2-11B-Vision-Instruct",
		messages: [
			{
				role: "user",
				content: content,
			},
		],
		max_tokens: 512,
		stream: true,
	});

	// Process the streamed response
	for await (const chunk of stream) {
		if (chunk.choices && chunk.choices.length > 0) {
			const newContent = chunk.choices[0].delta.content;
			output += newContent;
			// console.log(newContent); // Log the streamed content (optional)
		}
	}

	return output;
}