fix: recognize images in attached files

Author: grinev

src/bot/handlers/document.ts

@@ -76,6 +76,42 @@ export async function handleDocumentMessage(
       return;
     }
 
+    if (mimeType.startsWith("image/")) {
+      const storedModel = getStored();
+      const capabilities = await getCapabilities(storedModel.providerID, storedModel.modelID);
+
+      if (!supportsInput(capabilities, "image")) {
+        logger.warn(
+          `[Document] Model ${storedModel.providerID}/${storedModel.modelID} doesn't support image input`,
+        );
+        await ctx.reply(t("bot.photo_model_no_image"));
+
+        if (caption.trim().length > 0) {
+          await processPrompt(ctx, caption, deps);
+        }
+        return;
+      }
+
+      await ctx.reply(t("bot.file_downloading"));
+      const downloadedFile = await downloadFile(ctx.api, doc.file_id);
+
+      const dataUri = toDataUri(downloadedFile.buffer, mimeType);
+
+      const filePart: FilePartInput = {
+        type: "file",
+        mime: mimeType,
+        filename: filename,
+        url: dataUri,
+      };
+
+      logger.info(
+        `[Document] Sending image (${downloadedFile.buffer.length} bytes, ${filename}, ${mimeType}) with prompt`,
+      );
+
+      await processPrompt(ctx, caption, deps, [filePart]);
+      return;
+    }
+
     if (mimeType === "application/pdf") {
       const storedModel = getStored();
       const capabilities = await getCapabilities(storedModel.providerID, storedModel.modelID);
@@ -112,7 +148,8 @@ export async function handleDocumentMessage(
       return;
     }
 
-    logger.debug(`[Document] Unsupported document MIME type: ${mimeType}, ignoring`);
+    logger.warn(`[Document] Unsupported document MIME type: ${mimeType}, filename=${filename}`);
+    await ctx.reply(t("bot.file_type_unsupported"));
   } catch (err) {
     logger.error("[Document] Error handling document message:", err);
     await ctx.reply(t("bot.file_download_error"));

src/i18n/de.ts

@@ -96,6 +96,8 @@ export const de: I18nDictionary = {
   "bot.file_downloading": "⏳ Lade Datei herunter...",
   "bot.file_too_large": "⚠️ Datei ist zu groß (max. {maxSizeMb}MB)",
   "bot.file_download_error": "🔴 Datei konnte nicht heruntergeladen werden",
+  "bot.file_type_unsupported":
+    "⚠️ Dieser Dateityp wird nicht unterstützt. Sende ein Bild, PDF oder eine Text-/Code-Datei.",
   "bot.model_no_pdf": "⚠️ Das aktuelle Modell unterstützt keine PDF-Eingabe. Sende nur Text.",
   "bot.text_file_too_large": "⚠️ Textdatei ist zu groß (max. {maxSizeKb}KB)",
 

src/i18n/en.ts

@@ -89,6 +89,8 @@ export const en = {
   "bot.file_downloading": "⏳ Downloading file...",
   "bot.file_too_large": "⚠️ File is too large (max {maxSizeMb}MB)",
   "bot.file_download_error": "🔴 Failed to download file",
+  "bot.file_type_unsupported":
+    "⚠️ This file type is not supported. Send an image, PDF, or text/code file.",
   "bot.model_no_pdf": "⚠️ Current model doesn't support PDF input. Sending text only.",
   "bot.text_file_too_large": "⚠️ Text file is too large (max {maxSizeKb}KB)",
 

src/i18n/es.ts

@@ -96,6 +96,8 @@ export const es: I18nDictionary = {
   "bot.file_downloading": "⏳ Descargando archivo...",
   "bot.file_too_large": "⚠️ El archivo es demasiado grande (max {maxSizeMb}MB)",
   "bot.file_download_error": "🔴 No se pudo descargar el archivo",
+  "bot.file_type_unsupported":
+    "⚠️ Este tipo de archivo no es compatible. Envía una imagen, PDF o archivo de texto/código.",
   "bot.model_no_pdf": "⚠️ El modelo actual no admite entrada PDF. Enviaré solo texto.",
   "bot.text_file_too_large": "⚠️ El archivo de texto es demasiado grande (max {maxSizeKb}KB)",
 

src/i18n/fr.ts

@@ -97,6 +97,8 @@ export const fr: I18nDictionary = {
   "bot.file_downloading": "⏳ Téléchargement du fichier...",
   "bot.file_too_large": "⚠️ Le fichier est trop volumineux (max {maxSizeMb}MB)",
   "bot.file_download_error": "🔴 Impossible de télécharger le fichier",
+  "bot.file_type_unsupported":
+    "⚠️ Ce type de fichier n'est pas pris en charge. Envoyez une image, un PDF ou un fichier texte/code.",
   "bot.model_no_pdf":
     "⚠️ Le modèle actuel ne prend pas en charge les PDF. Envoi du texte uniquement.",
   "bot.text_file_too_large": "⚠️ Le fichier texte est trop volumineux (max {maxSizeKb}KB)",

src/i18n/ru.ts

@@ -89,6 +89,8 @@ export const ru: I18nDictionary = {
   "bot.file_downloading": "⏳ Скачиваю файл...",
   "bot.file_too_large": "⚠️ Файл слишком большой (макс. {maxSizeMb}МБ)",
   "bot.file_download_error": "🔴 Не удалось скачать файл",
+  "bot.file_type_unsupported":
+    "⚠️ Этот тип файла не поддерживается. Отправьте изображение, PDF или текстовый/кодовый файл.",
   "bot.model_no_pdf": "⚠️ Текущая модель не поддерживает PDF. Отправляю только текст.",
   "bot.text_file_too_large": "⚠️ Текстовый файл слишком большой (макс. {maxSizeKb}КБ)",
 

src/i18n/zh.ts

@@ -80,6 +80,7 @@ export const zh: I18nDictionary = {
   "bot.file_downloading": "⏳ 正在下载文件...",
   "bot.file_too_large": "⚠️ 文件过大（最大 {maxSizeMb}MB）",
   "bot.file_download_error": "🔴 下载文件失败",
+  "bot.file_type_unsupported": "⚠️ 不支持此文件类型。请发送图片、PDF 或文本/代码文件。",
   "bot.model_no_pdf": "⚠️ 当前模型不支持PDF输入。将仅发送文本。",
   "bot.text_file_too_large": "⚠️ 文本文件过大（最大 {maxSizeKb}KB）",
 

tests/bot/handlers/document.test.ts

@@ -250,41 +250,103 @@ describe("bot/handlers/document", () => {
     });
   });
 
-  describe("unsupported file types", () => {
-    it("ignores unsupported MIME types silently", async () => {
+  describe("image files", () => {
+    it("downloads and sends image documents when model supports images", async () => {
       const { ctx, replyMock } = createDocumentContext({
         document: {
-          file_id: "zip-file-id",
-          file_unique_id: "zip-unique-id",
-          file_name: "archive.zip",
-          mime_type: "application/zip",
+          file_id: "image-file-id",
+          file_unique_id: "image-unique-id",
+          file_name: "photo.png",
+          mime_type: "image/png",
           file_size: 5000,
         },
+        caption: "Describe this image",
       });
       const { deps, processPromptMock, downloadMock } = createDocumentDeps();
 
       await handleDocumentMessage(ctx, deps);
 
-      expect(replyMock).not.toHaveBeenCalled();
+      expect(replyMock).toHaveBeenCalledWith(t("bot.file_downloading"));
+      expect(downloadMock).toHaveBeenCalled();
+      expect(processPromptMock).toHaveBeenCalledWith(
+        ctx,
+        "Describe this image",
+        deps,
+        expect.arrayContaining([
+          expect.objectContaining({
+            type: "file",
+            mime: "image/png",
+            filename: "photo.png",
+            url: expect.stringMatching(/^data:image\/png;base64,/),
+          }),
+        ]),
+      );
+    });
+
+    it("shows error when model does not support images", async () => {
+      const { ctx, replyMock } = createDocumentContext({
+        document: {
+          file_id: "image-file-id",
+          file_unique_id: "image-unique-id",
+          file_name: "photo.png",
+          mime_type: "image/png",
+          file_size: 5000,
+        },
+      });
+      const { deps, processPromptMock, downloadMock } = createDocumentDeps({
+        getModelCapabilities: vi.fn().mockResolvedValue({
+          input: { image: false },
+        }),
+      });
+
+      await handleDocumentMessage(ctx, deps);
+
+      expect(replyMock).toHaveBeenCalledWith(t("bot.photo_model_no_image"));
       expect(downloadMock).not.toHaveBeenCalled();
       expect(processPromptMock).not.toHaveBeenCalled();
     });
 
-    it("ignores image files", async () => {
-      const { ctx, replyMock } = createDocumentContext({
+    it("sends caption-only when model does not support images but caption exists", async () => {
+      const { ctx } = createDocumentContext({
         document: {
           file_id: "image-file-id",
           file_unique_id: "image-unique-id",
           file_name: "photo.png",
           mime_type: "image/png",
           file_size: 5000,
         },
+        caption: "Describe this image",
+      });
+      const { deps, processPromptMock, downloadMock } = createDocumentDeps({
+        getModelCapabilities: vi.fn().mockResolvedValue({
+          input: { image: false },
+        }),
       });
-      const { deps, processPromptMock } = createDocumentDeps();
 
       await handleDocumentMessage(ctx, deps);
 
-      expect(replyMock).not.toHaveBeenCalled();
+      expect(downloadMock).not.toHaveBeenCalled();
+      expect(processPromptMock).toHaveBeenCalledWith(ctx, "Describe this image", deps);
+    });
+  });
+
+  describe("unsupported file types", () => {
+    it("shows error for unsupported MIME types", async () => {
+      const { ctx, replyMock } = createDocumentContext({
+        document: {
+          file_id: "zip-file-id",
+          file_unique_id: "zip-unique-id",
+          file_name: "archive.zip",
+          mime_type: "application/zip",
+          file_size: 5000,
+        },
+      });
+      const { deps, processPromptMock, downloadMock } = createDocumentDeps();
+
+      await handleDocumentMessage(ctx, deps);
+
+      expect(replyMock).toHaveBeenCalledWith(t("bot.file_type_unsupported"));
+      expect(downloadMock).not.toHaveBeenCalled();
       expect(processPromptMock).not.toHaveBeenCalled();
     });
   });