amari/src/commands/altText.ts

/**
 * @copyright NHCarrigan
 * @license Naomi's Public License
 * @author Naomi Carrigan
 */

import { MessageFlags, type ChatInputCommandInteraction } from "discord.js";
import { ids } from "../config/ids.js";
import { anthropic } from "../utils/anthropic.js";
import { logger } from "../utils/logger.js";
import { amariPersonality } from "../utils/makeAiRequest.js";
import type { Amari } from "../interfaces/amari.js";

type ValidImageMediaType =
  | "image/gif"
  | "image/jpeg"
  | "image/png"
  | "image/webp";

const validImageTypes = new Set<string>([
  "image/gif",
  "image/jpeg",
  "image/png",
  "image/webp",
]);

const isValidContentType = (type: string): type is ValidImageMediaType => {
  return validImageTypes.has(type);
};

const altTextSystemPrompt = `${amariPersonality}\n\nYour role in this`
  + " conversation is to generate descriptive and accessible alt-text for the"
  + " user's image. Be as descriptive as possible. Do not include ANYTHING in"
  + " your response EXCEPT the actual alt-text. Wrap the text in a multi-line"
  + " code block for easy copying.";

/**
 * Downloads an image and asks Claude to generate alt-text for it.
 * @param imageUrl - The URL of the image to process.
 * @param contentType - The validated MIME type of the image.
 * @returns The generated alt-text, or null if the request fails.
 */
const generateAltText = async(
  imageUrl: string,
  contentType: ValidImageMediaType,
): Promise<string | null> => {
  const downloadRequest = await fetch(imageUrl);
  const blob = await downloadRequest.arrayBuffer();
  const base64 = Buffer.from(blob).toString("base64");

  const response = await anthropic.messages.create({
    // eslint-disable-next-line @typescript-eslint/naming-convention -- Anthropic API field.
    max_tokens: 2000,
    messages:   [
      {
        content: [
          {
            source: {
              data:       base64,
              // eslint-disable-next-line @typescript-eslint/naming-convention -- Anthropic API field.
              media_type: contentType,
              type:       "base64",
            },
            type: "image",
          },
        ],
        role: "user",
      },
    ],
    model:  "claude-sonnet-4-6",
    system: altTextSystemPrompt,
  });

  const textContent = response.content.find((block) => {
    return block.type === "text";
  });

  return textContent?.type === "text"
    ? textContent.text
    : null;
};

/**
 * Validates the attachment is a supported image, downloads it, and sends it
 * to Claude to generate descriptive alt-text.
 * @param _amari - The Amari instance (unused but kept for handler consistency).
 * @param interaction - The Discord slash command interaction.
 */
// eslint-disable-next-line max-lines-per-function, complexity -- Image validation requires multiple checks.
export const altText = async(
  _amari: Amari,
  interaction: ChatInputCommandInteraction,
): Promise<void> => {
  if (interaction.user.id !== ids.users.naomi) {
    await interaction.reply({
      content: "This command is restricted to Naomi.",
      flags:   [ MessageFlags.Ephemeral ],
    });
    return;
  }

  const image = interaction.options.getAttachment("image", true);
  const { contentType, height, width, size, url } = image;

  if (
    contentType === null
    || !isValidContentType(contentType)
    || height === null
    || width === null
  ) {
    await interaction.reply({
      content: "That does not appear to be a valid image.",
      flags:   [ MessageFlags.Ephemeral ],
    });
    return;
  }

  if (size > 5 * 1024 * 1024) {
    await interaction.reply({
      // eslint-disable-next-line stylistic/max-len -- Long user-facing string.
      content: "That image is too large. Please provide an image that is less than 5MB.",
      flags:   [ MessageFlags.Ephemeral ],
    });
    return;
  }

  if (height > 8000 || width > 8000) {
    await interaction.reply({
      // eslint-disable-next-line stylistic/max-len -- Long user-facing string.
      content: "That image is too large. Please provide an image less than 8000 pixels in either dimension.",
      flags:   [ MessageFlags.Ephemeral ],
    });
    return;
  }

  await interaction.deferReply({ flags: [ MessageFlags.Ephemeral ] });

  try {
    const result = await generateAltText(url, contentType);

    await interaction.editReply({
      content: result
        ?? "I'm sorry, I wasn't able to generate alt-text for that image.",
    });
  } catch (error) {
    await logger.error("alt-text command", error instanceof Error
      ? error
      : new Error(String(error)));
    await interaction.editReply({
      content: "Something went wrong whilst generating the alt-text.",
    });
  }
};