fix dm commands, add url support for ocr

package.json

@@ -7,7 +7,7 @@
     "dev": "bun --watch src/index.ts",
     "start": "bun run src/index.ts",
     "sync": "bun run src/scripts/sync.ts",
-    "test": "bun run src/scripts/test.ts",
+    "sandbox": "bun run src/scripts/sandbox.ts",
     "deploy": "src/scripts/deploy.sh"
   },
   "dependencies": {

src/client.ts

@@ -1,4 +1,10 @@
-import { Client, Collection, GatewayIntentBits, REST } from "discord.js";
+import {
+  Client,
+  Collection,
+  GatewayIntentBits,
+  Partials,
+  REST,
+} from "discord.js";
 import { env } from "./env";
 import { ActivityType, API } from "@discordjs/core";
 import type { Command } from "./types/command";
@@ -28,6 +34,7 @@ export class ArtemisClient extends Client {
       presence: {
         activities: [{ name: "🩷", type: ActivityType.Custom }],
       },
+      partials: [Partials.Channel],
     });
 
     const rest = new REST().setToken(env.DISCORD_TOKEN);

src/commands/language/translate.ts

@@ -1,6 +1,6 @@
 import {
-  Attachment,
   AutocompleteInteraction,
+  hyperlink,
   SlashCommandBuilder,
   type InteractionEditReplyOptions,
 } from "discord.js";
@@ -39,7 +39,7 @@ export async function translateImpl(
   text: string,
   source: string | null,
   target: string,
-  attachment?: Attachment
+  imageUrl?: string
 ): Promise<InteractionEditReplyOptions> {
   const {
     text: translatedText,
@@ -56,6 +56,7 @@ export async function translateImpl(
 
   if (translatedText.length > 4096) {
     return {
+      content: imageUrl ? `${hyperlink("Image", imageUrl)}\n\n` : undefined,
       files: [
         {
           name: `${displaySource}-${displayTarget}.txt`,
@@ -63,7 +64,6 @@ export async function translateImpl(
             `--- From ${displaySource} to ${displayTarget} ---\n${translatedText}`
           ),
         },
-        ...(attachment ? [attachment] : []),
       ],
     };
   }
@@ -74,6 +74,7 @@ export async function translateImpl(
         title: `From ${displaySource} to ${displayTarget}`,
         description: translatedText,
         color: 0x0f2b46,
+        ...(imageUrl ? { image: { url: imageUrl } } : {}),
         author: {
           name: "DeepL",
           icon_url: "https://www.google.com/s2/favicons?domain=deepl.com&sz=64",
@@ -83,7 +84,6 @@ export async function translateImpl(
         },
       },
     ],
-    files: attachment ? [attachment] : [],
   };
 }
 

src/commands/language/translateMenu.ts

@@ -1,6 +1,7 @@
 import { ApplicationCommandType, ContextMenuCommandBuilder } from "discord.js";
 import { defineCommand } from "..";
 import { translateImpl } from "./translate";
+import { abort } from "../../utils/error";
 
 export default defineCommand({
   data: new ContextMenuCommandBuilder()
@@ -11,6 +12,8 @@ export default defineCommand({
     if (!interaction.isMessageContextMenuCommand()) return;
 
     const text = interaction.targetMessage.content;
+    if (!text) abort("No text to translate");
+
     await interaction.deferReply();
 
     const payload = await translateImpl(text, null, "en-US");

src/commands/ocr/ocr.ts

@@ -1,6 +1,6 @@
 import {
-  Attachment,
   codeBlock,
+  hyperlink,
   inlineCode,
   SlashCommandBuilder,
   type InteractionEditReplyOptions,
@@ -10,41 +10,73 @@ import { downloadFile } from "../../utils/http";
 import { abort } from "../../utils/error";
 import { yandexOcr } from "../../utils/yandex";
 import sharp from "sharp";
+import { getImageFromAttachmentOrString, run } from "../../utils/functions";
 
 export function buildOcrPayload(
   text: string,
   detected_lang: string,
-  attachment?: Attachment
+  imageUrl?: string
 ): InteractionEditReplyOptions {
-  const languageName =
+  const languageName = run(() => {
+    try {
+      return (
         new Intl.DisplayNames(["en"], { type: "language" }).of(detected_lang) ??
-    "unknown";
+        "unknown"
+      );
+    } catch {
+      return "unknown";
+    }
+  });
 
-  const content = `Detected language: ${inlineCode(languageName)}\n${codeBlock(
+  const content =
-    text
+    `Detected language: ${inlineCode(languageName)}\n${codeBlock(text)}` +
-  )}`;
+    (imageUrl ? `\n${hyperlink("Image", imageUrl)}` : "");
 
-  if (content.length > 2000) {
+  if (content.length > 4096) {
     return {
-      content: `Detected language: ${inlineCode(languageName)}`,
+      content:
+        `Detected language: ${inlineCode(languageName)}` +
+        (imageUrl ? `\n${hyperlink("Image", imageUrl)}` : ""),
       files: [
         {
           name: "ocr.txt",
           attachment: Buffer.from(text),
         },
-        ...(attachment ? [attachment] : []),
       ],
     };
   }
 
   return {
-    content,
+    embeds: [
-    files: attachment ? [attachment] : [],
+      {
+        description: codeBlock(text),
+        color: 0xffdb4d,
+        fields: [
+          {
+            name: "Detected language",
+            value: inlineCode(languageName),
+          },
+        ],
+        ...(imageUrl ? { image: { url: imageUrl } } : {}),
+        author: {
+          name: "Yandex",
+          icon_url:
+            "https://www.google.com/s2/favicons?domain=yandex.com&sz=64",
+        },
+      },
+    ],
   };
 }
 
-export async function ocrImpl(attachment: Attachment) {
+export async function ocrImpl(url: string) {
-  const { data, type } = await downloadFile(attachment.url);
+  const { data, type } = await run(async () => {
+    try {
+      return await downloadFile(url);
+    } catch {
+      abort("Failed to download the image");
+    }
+  });
+
   if (!type?.mime.startsWith("image/")) {
     abort("The file must be an image!");
   }
@@ -54,7 +86,12 @@ export async function ocrImpl(attachment: Attachment) {
     .jpeg({ quality: 90 })
     .toBuffer();
 
-  return yandexOcr(compressed, type.mime);
+  const result = await yandexOcr(compressed, type.mime);
+  if (!result.text) {
+    result.text = "No text detected";
+  }
+
+  return result;
 }
 
 export default defineCommand({
@@ -62,25 +99,24 @@ export default defineCommand({
     .setName("ocr")
     .setDescription("OCR an image using Yandex")
     .addAttachmentOption((option) =>
-      option
+      option.setName("image").setDescription("The image to OCR")
-        .setName("image")
+    )
-        .setDescription("The image to OCR")
+    .addStringOption((option) =>
-        .setRequired(true)
+      option.setName("url").setDescription("The image URL to OCR")
     ),
 
   async execute(interaction) {
-    const attachment = interaction.options.getAttachment("image", true);
+    const attachment = interaction.options.getAttachment("image");
-    if (!attachment.contentType?.startsWith("image/")) {
+    const url = interaction.options.getString("url");
-      abort("The file must be an image!");
+    const imageUrl = getImageFromAttachmentOrString(attachment, url);
-    }
 
     await interaction.deferReply();
 
-    const result = await ocrImpl(attachment);
+    const result = await ocrImpl(imageUrl);
     const payload = buildOcrPayload(
       result.text,
       result.detected_lang,
-      attachment
+      imageUrl
     );
     await interaction.editReply(payload);
   },

src/commands/ocr/ocrMenu.ts

@@ -1,7 +1,7 @@
 import { ApplicationCommandType, ContextMenuCommandBuilder } from "discord.js";
 import { defineCommand } from "..";
-import { abort } from "../../utils/error";
 import { buildOcrPayload, ocrImpl } from "./ocr";
+import { getImageFromAttachmentOrString } from "../../utils/functions";
 
 export default defineCommand({
   data: new ContextMenuCommandBuilder()
@@ -12,16 +12,14 @@ export default defineCommand({
     if (!interaction.isMessageContextMenuCommand()) return;
 
     const attachment = interaction.targetMessage.attachments.first();
-    if (!attachment) {
+    const imageUrl = getImageFromAttachmentOrString(
-      abort("No attachment found");
+      attachment,
-    }
+      interaction.targetMessage.content
-    if (!attachment.contentType?.startsWith("image/")) {
+    );
-      abort("The file must be an image!");
-    }
 
     await interaction.deferReply();
 
-    const result = await ocrImpl(attachment);
+    const result = await ocrImpl(imageUrl);
     const payload = buildOcrPayload(result.text, result.detected_lang);
     await interaction.editReply(payload);
   },

src/commands/ocr/ocrTranslate.ts

@@ -10,6 +10,7 @@ import {
   translateImpl,
 } from "../language/translate";
 import { ocrImpl } from "./ocr";
+import { getImageFromAttachmentOrString } from "../../utils/functions";
 
 export default defineCommand({
   data: new SlashCommandBuilder()
@@ -18,10 +19,10 @@ export default defineCommand({
       "OCR an image using Yandex and translate the result using DeepL"
     )
     .addAttachmentOption((option) =>
-      option
+      option.setName("image").setDescription("The image to OCR")
-        .setName("image")
+    )
-        .setDescription("The image to OCR")
+    .addStringOption((option) =>
-        .setRequired(true)
+      option.setName("url").setDescription("The image URL to OCR")
     )
     .addStringOption((option) =>
       option
@@ -39,13 +40,13 @@ export default defineCommand({
   autocomplete: translateAutocompleteImpl,
 
   async execute(interaction) {
-    const attachment = interaction.options.getAttachment("image", true);
+    const attachment = interaction.options.getAttachment("image");
+    const url = interaction.options.getString("url");
+
     const source = interaction.options.getString("source") ?? null;
     const target = interaction.options.getString("target") ?? "en-US";
 
-    if (!attachment.contentType?.startsWith("image/")) {
+    const imageUrl = getImageFromAttachmentOrString(attachment, url);
-      abort("The file must be an image!");
-    }
 
     await interaction.deferReply();
 
@@ -56,8 +57,8 @@ export default defineCommand({
       abort("Target language not supported");
     }
 
-    const { text } = await ocrImpl(attachment);
+    const { text } = await ocrImpl(imageUrl);
-    const payload = await translateImpl(text, source, target, attachment);
+    const payload = await translateImpl(text, source, target, imageUrl);
     await interaction.editReply(payload);
   },
 });

src/commands/ocr/ocrTranslateMenu.ts

@@ -1,8 +1,8 @@
 import { ApplicationCommandType, ContextMenuCommandBuilder } from "discord.js";
 import { defineCommand } from "..";
-import { abort } from "../../utils/error";
 import { translateImpl } from "../language/translate";
 import { ocrImpl } from "./ocr";
+import { getImageFromAttachmentOrString } from "../../utils/functions";
 
 export default defineCommand({
   data: new ContextMenuCommandBuilder()
@@ -13,16 +13,14 @@ export default defineCommand({
     if (!interaction.isMessageContextMenuCommand()) return;
 
     const attachment = interaction.targetMessage.attachments.first();
-    if (!attachment) {
+    const imageUrl = getImageFromAttachmentOrString(
-      abort("No attachment found");
+      attachment,
-    }
+      interaction.targetMessage.content
-    if (!attachment.contentType?.startsWith("image/")) {
+    );
-      abort("The file must be an image!");
-    }
 
     await interaction.deferReply();
 
-    const { text } = await ocrImpl(attachment);
+    const { text } = await ocrImpl(imageUrl);
     const payload = await translateImpl(text, null, "en-US");
     await interaction.editReply(payload);
   },

src/scripts/sandbox.ts

@@ -0,0 +1,6 @@
+import { findFirstUrl } from "../utils/functions";
+
+const text =
+  "I ended up doing https://cdn.discordapp.com/attachments/338689901111541760/1338884762785288253/image.png?ex=67acb51a&is=67ab639a&hm=92ed060dabffccf9544157da2922bce79386eca444a182db92f19d833d66fba6&b because yah";
+
+console.log(findFirstUrl(text));

src/scripts/test.ts

@@ -1,3 +0,0 @@
-import { getWikipediaEditions } from "../utils/wikipedia";
-
-console.log(await getWikipediaEditions());

src/utils/constants.ts

@@ -8,6 +8,9 @@ export const USER_AGENT = `artemis (discord.js ${version})`;
 export const FAKE_USER_AGENT =
   "Mozilla/5.0 (X11; Linux x86_64; rv:135.0) Gecko/20100101 Firefox/135.0";
 
+export const URL_REGEX =
+  /\bhttps?:\/\/(www\.)?[-a-zA-Z0-9@:%._+~#=]{1,256}\.[a-zA-Z0-9()]{1,6}\b([-a-zA-Z0-9()@:%_+.~#?&//=]*)\b/gi;
+
 export const GOOD_BOT_EMOJIS = [
   "<:teehee:825098257742299136>",
   "<:teehee2:825098258741067787>",

src/utils/functions.ts

@@ -1,6 +1,9 @@
 import * as cheerio from "cheerio";
 import { execa } from "execa";
 import { customAlphabet } from "nanoid";
+import { URL_REGEX } from "./constants";
+import type { Attachment } from "discord.js";
+import { abort } from "./error";
 
 export const nanoid = customAlphabet("1234567890abcdef");
 export const shell = execa({ reject: false });
@@ -63,3 +66,29 @@ export function lazy<T>(cb: () => T) {
 export function trim(str: string, maxLength: number) {
   return str.length > maxLength ? str.slice(0, maxLength) + "…" : str;
 }
+
+export function findUrls(text: string) {
+  return text.match(URL_REGEX) ?? [];
+}
+
+export function findFirstUrl(text: string) {
+  return findUrls(text)[0];
+}
+
+export function getImageFromAttachmentOrString(
+  attachment?: Attachment | null,
+  str?: string | null
+) {
+  if (attachment) {
+    if (!attachment.contentType?.startsWith("image/")) {
+      abort("The file must be an image!");
+    }
+    return attachment.url;
+  } else if (str) {
+    const match = findFirstUrl(str);
+    if (!match) abort("The URL is invalid!");
+    return match;
+  } else {
+    abort("You must provide an image or an image URL!");
+  }
+}