.

help_processor.py

@@ -1001,27 +1001,107 @@ def _file_url_to_path(url: str) -> Optional[Path]:
     return Path(s)
 
 
-def _load_external_image_bytes(url: str) -> Optional[tuple[bytes, str]]:
-    """Чете външна картинка (r:link / TargetMode=External); връща (data, ext)."""
-    path = _file_url_to_path(url)
-    if path is None:
-        return None
+def _image_ext_from_path(path: Path) -> str:
+    ext = (path.suffix or "").lstrip(".").lower() or "png"
+    return "jpg" if ext == "jpeg" else ext
+
+
+def _read_image_file(path: Path) -> Optional[tuple[bytes, str]]:
+    """Чете локален/UNC файл като (data, ext); None при липса/грешка."""
     try:
         if not path.is_file():
             return None
         data = path.read_bytes()
-    except OSError:
+    except OSError as e:
+        log.warning(f"    Cannot read image file {path}: {e}")
         return None
     if not data:
         return None
-    ext = (path.suffix or "").lstrip(".").lower() or "png"
-    if ext == "jpeg":
-        ext = "jpg"
-    return data, ext
+    return data, _image_ext_from_path(path)
 
 
-def _resolve_docx_image_rid(doc, rId: str) -> Optional[tuple[bytes, str]]:
-    """r:embed → related_parts; r:link → външен файл. Връща (data, ext)."""
+def _docx_linked_media_dir(docx_path: Path) -> Path:
+    """Папка до .docx за опаковани r:link картинки: <stem>_media/."""
+    return docx_path.parent / f"{docx_path.stem}_media"
+
+
+def _sidecar_image_candidates(
+    docx_path: Optional[Path], linked_path: Optional[Path], url: str = ""
+) -> list[Path]:
+    """Кандидати до .docx, ако UNC/file линкът е недостъпен (Coolify, офлайн)."""
+    if docx_path is None:
+        return []
+    name = ""
+    if linked_path is not None:
+        name = linked_path.name
+    if not name:
+        from urllib.parse import unquote
+
+        tail = unquote((url or "").replace("\\", "/").rstrip("/").split("/")[-1])
+        name = tail.split("?")[0] if tail else ""
+    if not name:
+        return []
+    parent = docx_path.parent
+    stem = docx_path.stem
+    return [
+        parent / name,
+        _docx_linked_media_dir(docx_path) / name,
+        parent / "media" / name,
+        parent / "images" / name,
+    ]
+
+
+def _pack_linked_image(docx_path: Path, filename: str, data: bytes) -> Optional[Path]:
+    """Копира успешно прочетена r:link картинка до <stem>_media/ за следващи сканирания."""
+    if not filename or not data:
+        return None
+    dest_dir = _docx_linked_media_dir(docx_path)
+    dest = dest_dir / filename
+    try:
+        if dest.is_file() and dest.stat().st_size == len(data):
+            return dest
+        dest_dir.mkdir(parents=True, exist_ok=True)
+        dest.write_bytes(data)
+        log.info(f"    Packed linked image → {dest}")
+        return dest
+    except OSError as e:
+        log.warning(f"    Cannot pack linked image {filename}: {e}")
+        return None
+
+
+def _load_external_image_bytes(
+    url: str, docx_path: Optional[Path] = None
+) -> Optional[tuple[bytes, str]]:
+    """Чете външна картинка (r:link); UNC/file, после sidecar до .docx; лог при провал."""
+    path = _file_url_to_path(url)
+    tried: list[Path] = []
+
+    if path is not None:
+        tried.append(path)
+        loaded = _read_image_file(path)
+        if loaded:
+            if docx_path is not None:
+                _pack_linked_image(docx_path, path.name, loaded[0])
+            return loaded
+
+    for cand in _sidecar_image_candidates(docx_path, path, url):
+        if any(cand == t for t in tried):
+            continue
+        tried.append(cand)
+        loaded = _read_image_file(cand)
+        if loaded:
+            log.info(f"    External image from sidecar {cand}")
+            return loaded
+
+    targets = ", ".join(str(p) for p in tried) if tried else (url or "(empty)")
+    log.warning(f"    Cannot read linked image (r:link): {targets}")
+    return None
+
+
+def _resolve_docx_image_rid(
+    doc, rId: str, docx_path: Optional[Path] = None
+) -> Optional[tuple[bytes, str]]:
+    """r:embed → related_parts; r:link → външен файл / sidecar. Връща (data, ext)."""
     if not rId:
         return None
     try:
@@ -1039,14 +1119,15 @@ def _resolve_docx_image_rid(doc, rId: str) -> Optional[tuple[bytes, str]]:
     if not getattr(rel, "is_external", False):
         return None
     target = getattr(rel, "target_ref", None) or ""
-    loaded = _load_external_image_bytes(target)
-    return loaded
+    return _load_external_image_bytes(target, docx_path=docx_path)
 
 
-def _extract_docx_paragraph_images(para, doc) -> list[ImageRef]:
+def _extract_docx_paragraph_images(
+    para, doc, docx_path: Optional[Path] = None
+) -> list[ImageRef]:
     """Намира drawing-и в параграф; връща ImageRef-и за филтрираните по размер.
 
-    Поддържа r:embed (вградени) и r:link (външни file:/// / UNC пътища).
+    Поддържа r:embed (вградени) и r:link (външни file:/// / UNC + sidecar до .docx).
     """
     from docx.oxml.ns import qn
     imgs: list[ImageRef] = []
@@ -1058,10 +1139,16 @@ def _extract_docx_paragraph_images(para, doc) -> list[ImageRef]:
     embed_attr = qn("r:embed")
     link_attr = qn("r:link")
     for blip in blips:
+        # Предпочитаме вградено копие, ако Word е запазил и двете
         rId = blip.get(embed_attr) or blip.get(link_attr)
         if not rId:
             continue
-        resolved = _resolve_docx_image_rid(doc, rId)
+        resolved = _resolve_docx_image_rid(doc, rId, docx_path=docx_path)
+        if not resolved:
+            # Ако има само link и се провали — опитай обратното атрибутче
+            other = blip.get(link_attr) if blip.get(embed_attr) else blip.get(embed_attr)
+            if other and other != rId:
+                resolved = _resolve_docx_image_rid(doc, other, docx_path=docx_path)
         if not resolved:
             continue
         data, ext = resolved
@@ -1225,7 +1312,8 @@ def _table_lines(table) -> list[str]:
 
 
 def parse_docx(path: Path) -> list[Section]:
-    doc = Document(path)
+    docx_path = Path(path)
+    doc = Document(docx_path)
     sections: list[Section] = []
     current_title, current_level = "", 1
     buf: list[str] = []
@@ -1279,7 +1367,7 @@ def parse_docx(path: Path) -> list[Section]:
         para = block
         style_name = para.style.name.lower() if para.style else ""
         text = para.text.strip()
-        para_imgs = _extract_docx_paragraph_images(para, doc)
+        para_imgs = _extract_docx_paragraph_images(para, doc, docx_path=docx_path)
 
         if not text and not para_imgs:
             continue