refactor convert_documents_to_elements

backend/utils/elements.py

@@ -10,6 +10,8 @@
 from langchain_core.documents import Document
 from pydantic import BaseModel, PrivateAttr, validator
 
+from .image import local_image_to_base64
+
 
 class Element(BaseModel):
     """Abstract base class representing an element with a type, format, and metadata.
@@ -311,6 +313,69 @@ class TableImage(Table, Image):
     format: Literal["image"] = "image"
 
 
+def _create_text_element(doc: Document, element_class: type[Text]) -> Element:
+    """Create a text element from a Langchain Document object.
+
+    Args:
+        doc (Document): Langchain Document object.
+        element_class (Type[Text]): Text element class to create (Text, TableText).
+
+    Returns:
+        Element: Text element created from the Document object.
+    """
+    is_summary = doc.metadata.get("is_summary", False)
+    if is_summary:
+        element = element_class(
+            type=doc.metadata["type"],
+            format=doc.metadata["format"],
+            text="No content available",
+            metadata=doc.metadata,
+        )
+        element.set_summary(doc.page_content)
+    else:
+        element = element_class(
+            type=doc.metadata["type"],
+            format=doc.metadata["format"],
+            text=doc.page_content,
+            metadata=doc.metadata,
+        )
+    return element
+
+
+NO_IMAGE = local_image_to_base64("img/no_image.png")
+
+
+def _create_image_element(doc: Document, element_class: type[Image]) -> Element:
+    """Create an image element from a Langchain Document object.
+
+    Args:
+        doc (Document): Langchain Document object.
+        element_class (Type[Image]): Image element class to create (Image, TableImage).
+
+    Returns:
+        Element: Image element created from the Document object.
+    """
+    is_summary = doc.metadata.get("is_summary", False)
+    if is_summary:
+        element = element_class(
+            type=doc.metadata["type"],
+            format=doc.metadata["format"],
+            base64=NO_IMAGE,
+            mime_type="image/png",
+            metadata=doc.metadata,
+        )
+        element.set_summary(doc.page_content)
+    else:
+        element = element_class(
+            type=doc.metadata["type"],
+            format=doc.metadata["format"],
+            base64=doc.page_content,
+            mime_type=doc.metadata["mime_type"],
+            metadata=doc.metadata,
+        )
+    return element
+
+
 def convert_documents_to_elements(docs: list[Document]) -> list:
     """Convert a list of Langchain Document objects to a list of Element objects.
 
@@ -327,37 +392,19 @@ def convert_documents_to_elements(docs: list[Document]) -> list:
     for doc in docs:
         match doc.metadata["type"]:
             case "text":
-                text_format = doc.metadata["format"]
-                element = Text(
-                    text=doc.page_content, format=text_format, metadata=doc.metadata
-                )
+                element = _create_text_element(doc, Text)
             case "image":
-                element = Image(
-                    base64=doc.page_content,
-                    mime_type=doc.metadata["mime_type"],
-                    metadata=doc.metadata,
-                )
-
+                element = _create_image_element(doc, Image)
             case "table":
-                table_format = doc.metadata["format"]
-                match table_format:
+                match doc.metadata["format"]:
                     case "text" | "html" | "markdown":
-                        element = TableText(
-                            text=doc.page_content,
-                            format=table_format,
-                            metadata=doc.metadata,
-                        )
-
+                        element = _create_text_element(doc, TableText)
                     case "image":
-                        element = TableImage(
-                            base64=doc.page_content,
-                            mime_type=doc.metadata["mime_type"],
-                            metadata=doc.metadata,
-                        )
+                        element = _create_image_element(doc, TableImage)
                     case other:
                         raise ValueError(f"Unsupported table format: {other}")
             case other:
-                raise ValueError(f"Unsupported document type: {other['type']}")
+                raise ValueError(f"Unsupported document type: {other}")
         elements.append(element)
 
     return elements

backend/utils/image.py

@@ -2,6 +2,7 @@
 
 import base64
 import io
+from pathlib import Path
 
 from PIL import Image
 
@@ -32,3 +33,16 @@ def resize_base64_image(
 
     # Encode the resized image to Base64
     return base64.b64encode(buffered.getvalue()).decode("utf-8")
+
+
+def local_image_to_base64(image_path: str) -> str:
+    """Convert a local image to a Base64 string.
+
+    Args:
+        image_path (str): Path to the image.
+
+    Returns:
+        str: Base64 string.
+    """
+    with Path(image_path).open("rb") as image_file:
+        return base64.b64encode(image_file.read()).decode("utf-8")

backend/utils/retriever.py

@@ -79,7 +79,7 @@ def add_documents_multivector(
     summary_docs = [
         Document(
             page_content=s,
-            metadata=metadata_list[i],
+            metadata={**metadata_list[i], "is_summary": True},
         )
         for i, s in enumerate(summary_list)
     ]
@@ -88,7 +88,7 @@ def add_documents_multivector(
     content_docs = [
         Document(
             page_content=c if isinstance(c, str) else c.page_content,
-            metadata=metadata_list[i],
+            metadata={**metadata_list[i], "is_summary": False},
         )
         for i, c in enumerate(content_list)
     ]