grinev · cgmark101 · Jul 10, 2026 · Jul 11, 2026
diff --git a/.env.example b/.env.example
@@ -123,6 +123,15 @@ OPENCODE_MODEL_ID=big-pickle
 # If STT_API_URL is not set, voice messages will get a "not configured" reply.
 # STT_API_URL=
 # STT_API_KEY=
+
+# Document Extractor API (optional)
+# If set, the bot will extract text from PDF, DOCX, PPTX, and other documents
+# using an external API when the current model doesn't natively support them.
+# The API should accept multipart/form-data POST with a "file" field and
+# return JSON with a "text" field containing the extracted content.
+# Only DOC_EXTRACTOR_URL is required; API key is optional for self-hosted setups.
+# DOC_EXTRACTOR_URL=
+# DOC_EXTRACTOR_API_KEY=
 # STT_MODEL=
 # STT_LANGUAGE=
 # STT_NOTE_PROMPT="The following text is transcribed from voice. It may contain homophone or phonetic errors. Infer the intended meaning from context."

diff --git a/README.md b/README.md
@@ -239,6 +239,8 @@ When installed via npm, the configuration wizard handles the initial setup. The
 | `STT_MODEL`                                | STT model name passed to `/audio/transcriptions`                                                                      |    No    | `whisper-large-v3-turbo` |
 | `STT_LANGUAGE`                             | Optional language hint (empty = provider auto-detect)                                                                 |    No    | —                        |
 | `STT_NOTE_PROMPT`                          | Optional note prepended to the LLM prompt as `[Note: ...]` for voice transcriptions; empty / `false` / `0` disable it |    No    | —                        |
+| `DOC_EXTRACTOR_URL`                        | Document text extraction API URL (enables PDF/DOCX/PPTX extraction)                                                    |    No    | —                        |
+| `DOC_EXTRACTOR_API_KEY`                    | API key for the document extractor (optional for self-hosted extractors)                                                |    No    | —                        |
 | `TTS_PROVIDER`                             | TTS provider: `openai` for OpenAI-compatible APIs, `elevenlabs` for ElevenLabs, or `google` for Google Cloud TTS      |    No    | `openai`                 |
 | `TTS_API_URL`                              | TTS API base URL for OpenAI-compatible APIs or ElevenLabs                                                             |    No    | —                        |
 | `TTS_API_KEY`                              | TTS API key for OpenAI-compatible APIs or ElevenLabs                                                                  |    No    | —                        |
@@ -365,6 +367,20 @@ Supported provider examples (Whisper-compatible):
 
 If STT variables are not set, voice/audio transcription is disabled and the bot will ask you to configure STT.
 
+### Document Text Extraction (Optional)
+
+If `DOC_EXTRACTOR_URL` is set, the bot will extract text from PDF, DOCX, PPTX, and other document files using an external API when the current model does not natively support document input.
+
+The API contract is:
+
+- **Endpoint:** `POST {DOC_EXTRACTOR_URL}`
+- **Content-Type:** `multipart/form-data`
+- **Field:** `file` — the document binary
+- **Authorization:** `Bearer {DOC_EXTRACTOR_API_KEY}` (only sent when a key is configured)
+- **Response:** JSON `{ "text": "extracted content..." }`
+
+If the extractor is not configured and the model doesn't support documents, the bot replies with a notice and forwards only the caption text.
+
 ### Model Configuration
 
 The model picker uses OpenCode local model state (`favorite` + `recent`):

diff --git a/src/app/services/document-extractor-service.ts b/src/app/services/document-extractor-service.ts
@@ -0,0 +1,72 @@
+import { config } from "../../config.js";
+import { logger } from "../../utils/logger.js";
+
+const REQUEST_TIMEOUT_MS = 60_000;
+
+export interface DocExtractorResult {
+  text: string;
+}
+
+export function isDocExtractorConfigured(): boolean {
+  return Boolean(config.docExtractor.apiUrl);
+}
+
+export async function extractDocument(
+  fileBuffer: Buffer,
+  mimeType: string,
+  filename: string,
+): Promise<DocExtractorResult> {
+  if (!isDocExtractorConfigured()) {
+    throw new Error(
+      "Document extractor is not configured: DOC_EXTRACTOR_URL is required",
+    );
+  }
+
+  const url = config.docExtractor.apiUrl!;
+
+  const formData = new FormData();
+  formData.append("file", new Blob([new Uint8Array(fileBuffer)], { type: mimeType }), filename);
+
+  logger.debug(
+    `[DocExtractor] Sending extraction request: url=${url}, mime=${mimeType}, filename=${filename}, size=${fileBuffer.length} bytes`,
+  );
+
+  const controller = new AbortController();
+  const timeout = setTimeout(() => controller.abort(), REQUEST_TIMEOUT_MS);
+
+  try {
+    const headers: Record<string, string> = {};
+    if (config.docExtractor.apiKey) {
+      headers["Authorization"] = `Bearer ${config.docExtractor.apiKey}`;
+    }
+
+    const response = await fetch(url, {
+      method: "POST",
+      headers,
+      body: formData,
+      signal: controller.signal,
+    });
+
+    if (!response.ok) {
+      const errorBody = await response.text().catch(() => "");
+      throw new Error(
+        `Document extractor API returned HTTP ${response.status}: ${errorBody || response.statusText}`,
+      );
+    }
+
+    const data = (await response.json()) as { text?: string };
+
+    if (typeof data.text !== "string") {
+      throw new Error("Document extractor API response does not contain a text field");
+    }
+
+    return { text: data.text };
+  } catch (err) {
+    if (err instanceof DOMException && err.name === "AbortError") {
+      throw new Error(`Document extractor request timed out after ${REQUEST_TIMEOUT_MS}ms`);
+    }
+    throw err;
+  } finally {
+    clearTimeout(timeout);
+  }
+}
diff --git a/src/bot/handlers/document-handler.ts b/src/bot/handlers/document-handler.ts
@@ -7,6 +7,7 @@ import {
   isTextMimeType,
   isFileSizeAllowed,
 } from "../../app/services/file-download-service.js";
+import { isDocExtractorConfigured, extractDocument } from "../../app/services/document-extractor-service.js";
 import { getModelCapabilities, supportsInput } from "../../app/services/model-capabilities-service.js";
 import { getStoredModel } from "../../app/services/model-selection-service.js";
 import { logger } from "../../utils/logger.js";
@@ -112,18 +113,55 @@ export async function handleDocumentMessage(
       return;
     }
 
-    if (mimeType === "application/pdf") {
+    const DOCUMENT_MIME_TYPES = [
+      "application/pdf",
+      "application/msword",
+      "application/vnd.openxmlformats-officedocument.wordprocessingml.document",
+      "application/vnd.ms-powerpoint",
+      "application/vnd.openxmlformats-officedocument.presentationml.presentation",
+      "application/vnd.ms-excel",
+      "application/vnd.openxmlformats-officedocument.spreadsheetml.sheet",
+      "application/vnd.oasis.opendocument.text",
+      "application/vnd.oasis.opendocument.presentation",
+      "application/vnd.oasis.opendocument.spreadsheet",
+      "text/rtf",
+    ];
+
+    if (DOCUMENT_MIME_TYPES.includes(mimeType)) {
       const storedModel = getStored();
       const capabilities = await getCapabilities(storedModel.providerID, storedModel.modelID);
 
       if (!supportsInput(capabilities, "pdf")) {
-        logger.warn(
-          `[Document] Model ${storedModel.providerID}/${storedModel.modelID} doesn't support PDF input`,
-        );
-        await ctx.reply(t("bot.model_no_pdf"));
-
-        if (caption.trim().length > 0) {
-          await processPrompt(ctx, caption, deps);
+        if (isDocExtractorConfigured()) {
+          logger.warn(
+            `[Document] Model doesn't support PDF input, delegating document to DOC_EXTRACTOR_URL`,
+          );
+          await ctx.reply(t("bot.file_downloading"));
+          const downloadedFile = await downloadFile(ctx.api, doc.file_id);
+
+          try {
+            const result = await extractDocument(downloadedFile.buffer, mimeType, filename);
+            const promptWithFile = `--- Content of ${filename} ---\n${result.text}\n--- End of file ---\n\n${caption}`;
+            logger.info(
+              `[Document] Sending extracted document text from ${filename} (${result.text.length} chars) as prompt`,
+            );
+            await processPrompt(ctx, promptWithFile, deps);
+          } catch (extractErr) {
+            const errMsg = extractErr instanceof Error ? extractErr.message : String(extractErr);
+            logger.error(`[Document] Document extraction failed: ${errMsg}`);
+            await ctx.reply(t("bot.document_extraction_error"));
+            if (caption.trim().length > 0) {
+              await processPrompt(ctx, caption, deps);
+            }
+          }
+        } else {
+          logger.warn(
+            `[Document] Model doesn't support PDF input and DOC_EXTRACTOR_URL is not configured`,
+          );
+          await ctx.reply(t("bot.model_no_pdf"));
+          if (caption.trim().length > 0) {
+            await processPrompt(ctx, caption, deps);
+          }
         }
         return;
       }
@@ -141,7 +179,7 @@ export async function handleDocumentMessage(
       };
 
       logger.info(
-        `[Document] Sending PDF (${downloadedFile.buffer.length} bytes, ${filename}) with prompt`,
+        `[Document] Sending document (${downloadedFile.buffer.length} bytes, ${filename}, ${mimeType}) with prompt`,
       );
 
       await processPrompt(ctx, caption, deps, [filePart]);

diff --git a/src/config.ts b/src/config.ts
@@ -181,6 +181,10 @@ export const config = {
     language: getEnvVar("STT_LANGUAGE", false),
     notePrompt: getEnvVar("STT_NOTE_PROMPT", false),
   },
+  docExtractor: {
+    apiUrl: getEnvVar("DOC_EXTRACTOR_URL", false),
+    apiKey: getEnvVar("DOC_EXTRACTOR_API_KEY", false),
+  },
   tts: (() => {
     const provider = getOptionalTtsProviderEnvVar("TTS_PROVIDER", "openai");
     const defaultVoice =

diff --git a/src/i18n/ar.ts b/src/i18n/ar.ts
@@ -106,11 +106,12 @@ export const ar: I18nDictionary = {
   "bot.file_too_large": "⚠️ حجم الملف أكبر من الحد المسموح ({maxSizeMb}MB)",
   "bot.file_download_error": "🔴 تعذر تنزيل الملف",
   "bot.file_type_unsupported":
-    "⚠️ نوع الملف غير مدعوم. أرسل صورة أو ملف PDF أو ملفًا نصيًا أو برمجيًا.",
+    "⚠️ نوع الملف غير مدعوم. أرسل صورة أو مستندًا (PDF، DOCX، PPTX) أو ملفًا نصيًا أو برمجيًا.",
   "bot.media_group_not_processed":
     "⚠️ تعذر معالجة ملف أو أكثر في هذه المجموعة. لم يتم إرسال أي ملف إلى OpenCode.",
   "bot.media_group_download_error": "🔴 تعذر تنزيل أحد الملفات. لم يتم إرسال أي ملف إلى OpenCode.",
   "bot.model_no_pdf": "⚠️ النموذج الحالي لا يدعم ملفات PDF. سيتم إرسال النص فقط.",
+  "bot.document_extraction_error": "🔴 فشل استخراج نص المستند.",
   "bot.text_file_too_large": "⚠️ حجم الملف النصي أكبر من الحد المسموح ({maxSizeKb}KB)",
 
   "status.header_running": "🟢 خادم OpenCode يعمل",

diff --git a/src/i18n/de.ts b/src/i18n/de.ts
@@ -110,12 +110,13 @@ export const de: I18nDictionary = {
   "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.",
+    "⚠️ Dieser Dateityp wird nicht unterstützt. Sende ein Bild, Dokument (PDF, DOCX, PPTX) oder eine Text-/Code-Datei.",
   "bot.media_group_not_processed":
     "⚠️ Eine oder mehrere Dateien in diesem Album können nicht verarbeitet werden. Es wurde nichts an OpenCode gesendet.",
   "bot.media_group_download_error":
     "🔴 Eine der Dateien konnte nicht heruntergeladen werden. Es wurde nichts an OpenCode gesendet.",
   "bot.model_no_pdf": "⚠️ Das aktuelle Modell unterstützt keine PDF-Eingabe. Sende nur Text.",
+  "bot.document_extraction_error": "🔴 Dokumenttext konnte nicht extrahiert werden.",
   "bot.text_file_too_large": "⚠️ Textdatei ist zu groß (max. {maxSizeKb}KB)",
 
   "status.header_running": "🟢 OpenCode-Server läuft",

diff --git a/src/i18n/en.ts b/src/i18n/en.ts
@@ -103,12 +103,13 @@ export const en = {
   "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.",
+    "⚠️ This file type is not supported. Send an image, document (PDF, DOCX, PPTX), or text/code file.",
   "bot.media_group_not_processed":
     "⚠️ One or more files in this album cannot be processed. Nothing was sent to OpenCode.",
   "bot.media_group_download_error":
     "🔴 Failed to download one of the files. Nothing was sent to OpenCode.",
   "bot.model_no_pdf": "⚠️ Current model doesn't support PDF input. Sending text only.",
+  "bot.document_extraction_error": "🔴 Failed to extract document text.",
   "bot.text_file_too_large": "⚠️ Text file is too large (max {maxSizeKb}KB)",
 
   "status.header_running": "🟢 OpenCode Server is running",

diff --git a/src/i18n/es.ts b/src/i18n/es.ts
@@ -111,12 +111,13 @@ export const es: I18nDictionary = {
   "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.",
+    "⚠️ Este tipo de archivo no es compatible. Envía una imagen, documento (PDF, DOCX, PPTX) o archivo de texto/código.",
   "bot.media_group_not_processed":
     "⚠️ Uno o más archivos de este álbum no se pueden procesar. No se envió nada a OpenCode.",
   "bot.media_group_download_error":
     "🔴 No se pudo descargar uno de los archivos. No se envió nada a OpenCode.",
   "bot.model_no_pdf": "⚠️ El modelo actual no admite entrada PDF. Enviaré solo texto.",
+  "bot.document_extraction_error": "🔴 No se pudo extraer el texto del documento.",
   "bot.text_file_too_large": "⚠️ El archivo de texto es demasiado grande (max {maxSizeKb}KB)",
 
   "status.header_running": "🟢 OpenCode Server está en ejecución",

diff --git a/src/i18n/fr.ts b/src/i18n/fr.ts
@@ -111,13 +111,15 @@ export const fr: I18nDictionary = {
   "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.",
+    "⚠️ Ce type de fichier n'est pas pris en charge. Envoyez une image, un document (PDF, DOCX, PPTX) ou un fichier texte/code.",
   "bot.media_group_not_processed":
     "⚠️ Un ou plusieurs fichiers de cet album ne peuvent pas être traités. Rien n'a été envoyé à OpenCode.",
   "bot.media_group_download_error":
     "🔴 Impossible de télécharger l'un des fichiers. Rien n'a été envoyé à OpenCode.",
   "bot.model_no_pdf":
     "⚠️ Le modèle actuel ne prend pas en charge les PDF. Envoi du texte uniquement.",
+  "bot.document_extraction_error":
+    "🔴 Échec de l'extraction du texte du document.",
   "bot.text_file_too_large": "⚠️ Le fichier texte est trop volumineux (max {maxSizeKb}KB)",
 
   "status.header_running": "🟢 Le serveur OpenCode est en cours d'exécution",

diff --git a/src/i18n/ru.ts b/src/i18n/ru.ts
@@ -103,12 +103,13 @@ export const ru: I18nDictionary = {
   "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.",
+    "⚠️ Этот тип файла не поддерживается. Отправьте изображение, документ (PDF, DOCX, PPTX) или текстовый/кодовый файл.",
   "bot.media_group_not_processed":
     "⚠️ Один или несколько файлов в альбоме нельзя обработать. В OpenCode ничего не отправлено.",
   "bot.media_group_download_error":
     "🔴 Не удалось скачать один из файлов. В OpenCode ничего не отправлено.",
   "bot.model_no_pdf": "⚠️ Текущая модель не поддерживает PDF. Отправляю только текст.",
+  "bot.document_extraction_error": "🔴 Не удалось извлечь текст из документа.",
   "bot.text_file_too_large": "⚠️ Текстовый файл слишком большой (макс. {maxSizeKb}КБ)",
 
   "status.header_running": "🟢 OpenCode Server запущен",

diff --git a/src/i18n/zh.ts b/src/i18n/zh.ts
@@ -93,11 +93,13 @@ export const zh: I18nDictionary = {
   "bot.files_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.file_type_unsupported":
+    "⚠️ 不支持此文件类型。请发送图片、文档（PDF、DOCX、PPTX）或文本/代码文件。",
   "bot.media_group_not_processed":
     "⚠️ 此相册中有一个或多个文件无法处理。未向 OpenCode 发送任何内容。",
   "bot.media_group_download_error": "🔴 无法下载其中一个文件。未向 OpenCode 发送任何内容。",
   "bot.model_no_pdf": "⚠️ 当前模型不支持PDF输入。将仅发送文本。",
+  "bot.document_extraction_error": "🔴 文档文本提取失败。",
   "bot.text_file_too_large": "⚠️ 文本文件过大（最大 {maxSizeKb}KB）",
 
   "status.header_running": "🟢 OpenCode 服务器正在运行",