docs(office): add section 19 to roadmap and changelog for two-phase handwritten ocr pipeline

2026-09-25 16:35:27 +03:00
parent 90656944c5
commit 11691701e1
8 changed files with 346 additions and 86 deletions
@@ -1,21 +1,22 @@
 """
 ===============================================================================
 FILE: modules/web_api/routers/chat.py
-ROLE: Полнофункциональный роутер чата с извлечением текста из PDF и сканов,
-      поддержкой Function Calling, Fast-Path и оптического распознавания OCR.
+ROLE: Роутер чата с чистым разделением:
+      - Диалог и команды СКУД/1С (через agent.py).
+      - Парсинг и извлечение документов без обрезания (через file_parser.py).
 ===============================================================================
 """

 import os
 import shutil
-import base64
 import logging
-from typing import Optional, List, Dict, Any
+from typing import Optional

 from fastapi import APIRouter, Header, HTTPException, UploadFile, File, Form
 from pydantic import BaseModel

 from llm.agent import process_chat_message
+from llm.file_parser import parse_uploaded_file
 from config import BASE_DIR

 logger = logging.getLogger("CHAT_API")
@@ -39,11 +40,6 @@ def resolve_user_id(authorization: Optional[str] = None, explicit_user_id: Optio
        token = authorization.replace("Bearer ", "").strip()
        if token.isdigit():
            return int(token)
-        elif token.startswith("dev_token_"):
-            try:
-                return int(token.replace("dev_token_", ""))
-            except ValueError:
-                pass
    return 1


@@ -56,8 +52,6 @@ async def chat_endpoint(payload: ChatMessageRequest, authorization: Optional[str
    if not user_msg:
        raise HTTPException(status_code=400, detail="Пустое сообщение")

-    logger.info(f"Сообщение от user_id={user_id}, session_id={session_id}: {user_msg}")
-
    reply_text, history, action_payload = process_chat_message(
        user_id=user_id,
        user_message=user_msg,
@@ -86,60 +80,60 @@ async def chat_upload_endpoint(
    with open(file_path, "wb") as buffer:
        shutil.copyfileobj(file.file, buffer)

-    file_context = ""
-    image_b64 = None
+    user_msg = (message or "").strip()
+    msg_lower = user_msg.lower()
    fn_lower = file.filename.lower()

-    # 1. Текстовые форматы
-    if fn_lower.endswith((".txt", ".csv", ".log", ".md")):
-        try:
-            with open(file_path, "r", encoding="utf-8", errors="ignore") as f:
-                file_context = f.read(6000)
-        except Exception as e:
-            logger.warning(f"Не удалось прочитать текст: {e}")
+    # ⭐️ 1. ПРЯМАЯ ИЗОЛИРОВАННАЯ ОБРАБОТКА PDF/СКАНОВ В WORD ЧЕРЕЗ OFFICE МОДУЛЬ
+    ocr_keywords = [
+        "распознай", "распознать", "в word", "в ворд", "для ворда", "для word", 
+        "отформатируй", "текст для вставки", "извлеки текст", "сделай документ", "переведи в ворд"
+    ]
+    
+    if fn_lower.endswith(".pdf") and (any(k in msg_lower for k in ocr_keywords) or not user_msg):
+        logger.info(f"[Office] Прямой запуск распознавания PDF в Word: {file.filename}")
+        from services.office.service import convert_pdf_to_word_service
+        from modules.web_api.llm.db_tools import db_save_chat_message, db_get_chat_history

-    # 2. Изображения (прямой OCR)
-    elif fn_lower.endswith((".png", ".jpg", ".jpeg", ".webp")):
-        try:
-            with open(file_path, "rb") as f:
-                image_b64 = base64.b64encode(f.read()).decode("utf-8")
-        except Exception as e:
-            logger.warning(f"Ошибка кодирования картинки в base64: {e}")
+        # Сохраняем вопрос пользователя в историю
+        prompt_text = user_msg or f"Распознать документ {file.filename} для MS Word"
+        db_save_chat_message(session_id, "user", prompt_text, is_ephemeral=0)

-    # 3. PDF документы (текстовый слой + рендеринг скана при необходимости)
-    elif fn_lower.endswith(".pdf"):
-        # Попытка извлечь встроенный текстовый слой
-        try:
-            import pypdf
-            reader = pypdf.PdfReader(file_path)
-            extracted = []
-            for page in reader.pages:
-                t = page.extract_text()
-                if t:
-                    extracted.append(t)
-            file_context = "\n".join(extracted).strip()
-        except Exception:
-            pass
+        # Вызываем офисный сервис постраничного OCR и сборки DOCX
+        office_res = convert_pdf_to_word_service(file_path, file.filename)

-        # Если текстового слоя мало (скан или фото документа), рендерим страницу в картинку для Vision OCR
-        if len(file_context) < 40:
-            try:
-                import fitz  # PyMuPDF
-                doc = fitz.open(file_path)
-                if len(doc) > 0:
-                    page = doc[0]
-                    pix = page.get_pixmap(dpi=150)
-                    img_bytes = pix.tobytes("png")
-                    image_b64 = base64.b64encode(img_bytes).decode("utf-8")
-                    file_context = ""
-            except Exception as e:
-                logger.warning(f"PyMuPDF не установлен или сбой рендеринга PDF: {e}")
+        reply_text = (
+            f"📄 **{office_res['message']}**\n\n"
+            f"**Фрагмент первой страницы документа:**\n"
+            f"```text\n{office_res['preview_text']}...\n```\n\n"
+            f"Файл готов к скачиванию и редактированию в MS Word."
+        )
+        db_save_chat_message(session_id, "assistant", reply_text, is_ephemeral=0)

-    user_msg = message.strip() or f"Распознай и проанализируй прикрепленный документ {file.filename}"
+        action_payload = {
+            "type": "FILE_DOWNLOAD_CARD",
+            "filename": office_res["filename"],
+            "download_url": office_res["download_url"],
+            "tasks_count": f"{office_res['total_pages']} стр."
+        }

+        return {
+            "status": "success",
+            "user_id": user_id,
+            "session_id": session_id,
+            "response": reply_text,
+            "action_payload": action_payload
+        }
+
+    # 2. Если это не задача OCR в Word — отправляем файл в обычный диалог агента
+    parsed = parse_uploaded_file(file_path, file.filename)
+    file_context = parsed.get("context_text", "")
+    image_b64 = parsed.get("image_b64")
+
+    prompt_for_agent = user_msg or f"Проанализируй документ {file.filename}"
    reply_text, history, action_payload = process_chat_message(
        user_id=user_id,
-        user_message=user_msg,
+        user_message=prompt_for_agent,
        file_context=file_context,
        image_b64=image_b64,
        session_id=session_id
@@ -11,16 +11,32 @@
    * { overflow-anchor: none !important; }
    #chat-messages-container, main { overflow-anchor: none !important; }   
    #chat-messages-container { padding-bottom: clamp(400px, 85vh, 900px) !important; }
+    
+    /* ⭐️ Комфортный размер шрифта в стиле Gemini */
    #chat-messages-container .message-content,
+    #chat-messages-container .text-sm,
    #chat-messages-container .text-xs,
-    #chat-messages-container .text-sm {
-        font-size: 14.5px !important;
-        line-height: 1.6 !important;
+    #chat-messages-container p,
+    #chat-messages-container li {
+        font-size: 16px !important;
+        line-height: 1.65 !important;
+        color: #1e293b !important; /* slate-800 */
    }
+    
+    #chat-messages-container .user-chat-bubble .text-sm,
+    #chat-messages-container .user-chat-bubble p,
+    #chat-messages-container .user-chat-bubble div {
+        font-size: 16px !important;
+        line-height: 1.6 !important;
+        color: #ffffff !important;
+    }
+
    #chat-messages-container pre, 
    #chat-messages-container code {
-        font-size: 13.5px !important;
+        font-size: 14.5px !important;
+        line-height: 1.5 !important;
    }
+
    .user-chat-bubble { scroll-margin-top: 24px !important; }
    </style>
 </head>
@@ -234,12 +250,12 @@
            <!-- ЛЕНТА ЧАТА -->
            <div id="chat-messages-container" class="flex-1 overflow-y-auto p-4 md:p-6 flex flex-col gap-4">
                <div class="flex gap-3 max-w-4xl mx-auto w-full">
-                    <div class="w-7 h-7 rounded-lg bg-indigo-600 text-white flex items-center justify-center shrink-0 shadow-sm mt-0.5">
-                        <i class="fa-solid fa-robot text-xs"></i>
+                    <div class="w-8 h-8 rounded-lg bg-indigo-600 text-white flex items-center justify-center shrink-0 shadow-sm mt-0.5">
+                        <i class="fa-solid fa-robot text-sm"></i>
                    </div>
                    <div class="flex-1 bg-white border border-slate-200 rounded-2xl rounded-tl-none p-4 shadow-sm">
-                        <div class="text-[10px] font-bold text-indigo-600 uppercase tracking-wider mb-1">ИИ-ассистент SCUD Orion AI</div>
-                        <div class="text-xs text-slate-700 leading-relaxed">
+                        <div class="text-[11px] font-bold text-indigo-600 uppercase tracking-wider mb-1">ИИ-ассистент SCUD Orion AI</div>
+                        <div class="text-sm text-slate-700 leading-relaxed">
                            Привет! Вы можете задавать вопросы ассистенту, управлять системным промптом, сверять кадровые нестыковки СКУД и 1С или формировать срезы и отчеты.
                        </div>
                    </div>
@@ -258,24 +274,35 @@
            </div>

            <!-- СТРОКА ВВОДА -->
-            <div class="px-4 py-2 bg-white border-t border-slate-200">
-                <div id="chat-input-box" class="max-w-4xl mx-auto w-full flex items-center gap-2 bg-slate-50 border border-slate-300 rounded-xl px-2.5 py-1 transition focus-within:border-indigo-500 focus-within:bg-white focus-within:ring-1 focus-within:ring-indigo-100">
-                    <button type="button" onclick="document.getElementById('file-upload-input').click()" class="text-slate-400 hover:text-indigo-600 p-1 transition shrink-0" title="Прикрепить файл">
-                        <i class="fa-solid fa-paperclip text-xs"></i>
+            <div class="px-4 py-3 bg-white border-t border-slate-200">
+                <div id="chat-input-box" class="max-w-4xl mx-auto w-full flex items-end gap-3 bg-slate-50 border border-slate-300 rounded-2xl px-4 py-2.5 transition focus-within:border-indigo-500 focus-within:bg-white focus-within:ring-2 focus-within:ring-indigo-100 shadow-xs">
+                    <button type="button" onclick="document.getElementById('file-upload-input').click()" class="text-slate-400 hover:text-indigo-600 p-1.5 transition shrink-0 mb-0.5" title="Прикрепить файл">
+                        <i class="fa-solid fa-paperclip text-base"></i>
                    </button>
                    <input type="file" id="file-upload-input" class="hidden" />

-                    <textarea id="user-input" rows="1" placeholder="Команда, вопрос (Enter - отправить, Shift+Enter - перенос строки)..." 
-                            class="flex-1 bg-transparent border-0 focus:outline-none text-xs text-slate-800 resize-none py-0 leading-5" style="height: 24px; line-height: 24px;"></textarea>
+                    <textarea id="user-input" rows="1" placeholder="Задайте вопрос, отправьте команду (Enter — отправить, Shift+Enter — перенос)..." 
+                            class="flex-1 bg-transparent border-0 focus:outline-none text-[15.5px] text-slate-800 placeholder:text-slate-400 resize-none py-1 leading-6 min-h-[32px] max-h-36" style="height: 32px;"></textarea>

-                    <button type="button" onclick="window.sendMessage()" class="w-6 h-6 rounded-lg bg-indigo-600 hover:bg-indigo-700 text-white flex items-center justify-center shrink-0 shadow-sm transition">
-                        <i class="fa-solid fa-paper-plane text-[10px]"></i>
+                    <button type="button" onclick="window.sendMessage()" class="w-8 h-8 rounded-xl bg-indigo-600 hover:bg-indigo-700 active:bg-indigo-800 text-white flex items-center justify-center shrink-0 shadow-sm transition mb-0.5">
+                        <i class="fa-solid fa-paper-plane text-xs"></i>
                    </button>
                </div>
            </div>
        </main>
    </div>

+    <!-- ПОЛНОЭКРАННЫЙ ОВЕРЛЕЙ DRAG & DROP -->
+    <div id="global-drag-overlay" class="fixed inset-0 bg-indigo-950/70 backdrop-blur-xs z-50 flex items-center justify-center p-8 hidden pointer-events-none transition-all">
+        <div class="border-3 border-dashed border-indigo-300 rounded-3xl w-full h-full flex flex-col items-center justify-center text-white gap-4 bg-indigo-900/40">
+            <div class="w-20 h-20 rounded-2xl bg-white/10 flex items-center justify-center shadow-lg backdrop-blur-md border border-white/20">
+                <i class="fa-solid fa-cloud-arrow-up text-4xl text-indigo-200"></i>
+            </div>
+            <div class="text-xl font-bold tracking-wide">Перетащите файл в окно для загрузки</div>
+            <div class="text-sm text-indigo-200">Поддерживаются PDF (документы и сканы), изображения, Excel, CSV, TXT</div>
+        </div>
+    </div>
+
    <!-- КОНТЕЙНЕР ДИНАМИЧЕСКИХ МОДАЛЬНЫХ ОКОН -->
    <div id="modals-container"></div>

@@ -48,7 +48,7 @@ function updateInputHeightAndFade(textarea) {
    if (!textarea) return;

    if (!textarea.value || textarea.value.trim() === '') {
-        textarea.style.height = '24px';
+        textarea.style.height = '32px';
        textarea.style.overflowY = 'hidden';
        textarea.style.maskImage = 'none';
        textarea.style.webkitMaskImage = 'none';
@@ -56,8 +56,8 @@ function updateInputHeightAndFade(textarea) {
    }

    textarea.style.height = 'auto';
-    const minHeight = 24;
-    const maxHeight = 120;
+    const minHeight = 32;
+    const maxHeight = 140;
    const currentScrollHeight = textarea.scrollHeight;

    if (currentScrollHeight <= minHeight + 2) {
@@ -143,23 +143,23 @@ function appendAssistantMessage(text, buttons = [], actionPayload = null) {
            payloadHtml = renderInteractiveTaskCard(actionPayload.tasks);
        } else if (actionPayload.type === 'FILE_DOWNLOAD_CARD') {
            const dlUrl = actionPayload.download_url || '#';
-            const fName = actionPayload.filename || 'ROADMAP.md';
+            const fName = actionPayload.filename || 'document.docx';
            const count = actionPayload.tasks_count || '';
            payloadHtml = `
-                <div class="mt-3 p-3.5 bg-indigo-50/80 border border-indigo-200 rounded-xl flex items-center justify-between gap-3 shadow-sm">
+                <div class="mt-3 p-3.5 bg-slate-50 border border-slate-200 rounded-xl flex items-center justify-between gap-3 shadow-xs">
                    <div class="flex items-center gap-3 min-w-0">
-                        <div class="w-9 h-9 rounded-lg bg-indigo-600 text-white flex items-center justify-center shrink-0 shadow-sm">
-                            <i class="fa-solid fa-file-lines text-base"></i>
+                        <div class="w-10 h-10 rounded-xl bg-blue-600 text-white flex items-center justify-center shrink-0 shadow-sm">
+                            <i class="fa-solid fa-file-word text-lg"></i>
                        </div>
                        <div class="min-w-0">
                            <div class="text-sm font-bold text-slate-800 truncate">${escapeHtml(fName)}</div>
-                            <div class="text-xs text-slate-500">Задач выгружено: ${count} шт. · Markdown</div>
+                            <div class="text-xs text-slate-500">${count ? count + ' · ' : ''}Документ MS Word (.docx)</div>
                        </div>
                    </div>
                    <a href="${dlUrl}" download="${escapeHtml(fName)}" target="_blank"
-                       class="px-4 py-2 bg-indigo-600 hover:bg-indigo-700 active:bg-indigo-800 text-white rounded-lg text-xs font-bold shadow-sm transition flex items-center gap-1.5 shrink-0">
-                        <i class="fa-solid fa-download text-xs"></i>
-                        <span>Скачать файл</span>
+                       class="px-4 py-2.5 bg-emerald-600 hover:bg-emerald-700 active:bg-emerald-800 text-white font-bold rounded-xl text-xs shadow-sm transition flex items-center gap-2 shrink-0">
+                        <i class="fa-solid fa-arrow-down-to-line text-sm text-white"></i>
+                        <span class="text-white tracking-wide">Скачать файл</span>
                    </a>
                </div>
            `;
@@ -363,7 +363,7 @@ window.sendMessage = async function() {
    saveCommandToHistory(messageText);

    input.value = "";
-    input.style.height = '24px';
+    input.style.height = '32px';
    input.style.overflowY = 'hidden';
    input.style.maskImage = 'none';
    input.style.webkitMaskImage = 'none';
@@ -478,7 +478,7 @@ document.addEventListener("DOMContentLoaded", () => {

    if (input) {
        input.style.lineHeight = '24px';
-        input.style.height = '24px';
+        input.style.height = '32px';

        input.addEventListener("keydown", (e) => {
            if (e.key === "Enter") {