Update utils.py

utils.py

@@ -255,7 +255,7 @@ def load_pdf_with_metadata(file_path):
             "page": page_num + 1,
             "path": file_path
         }
-        documents.append({"content": content, "metadata": metadata})
+        documents.append({"page_content": content, "metadata": metadata})
     return documents
 
 def load_word_with_metadata(file_path):
@@ -267,7 +267,7 @@ def load_word_with_metadata(file_path):
     contents = []
     for para in document.paragraphs:
         content = para.text
-        contents.append({"content": content, "metadata": {**metadata, "page": 1}})
+        contents.append({"page_content": content, "metadata": {**metadata, "page": 1}})
     return contents
 
     
@@ -400,7 +400,7 @@ def rag_chain(llm, prompt,  retriever):
         #RAG_CHAIN_PROMPT = PromptTemplate(template="Context: {context}\n\nQuestion: {question}\n\nAnswer:")
 
         # Inahlte Abrufen der relevanten Dokumente
-        doc_contents = [doc["content"] for doc in extracted_docs]
+        doc_contents = [doc["page_content"] for doc in extracted_docs]
 
         #Berechne die Ähnlichkeiten und finde das relevanteste Dokument
         question_embedding = embedder_modell.encode(prompt, convert_to_tensor=True)
@@ -412,7 +412,7 @@ def rag_chain(llm, prompt,  retriever):
         most_relevant_docs = [extracted_docs[i] for i in most_relevant_doc_indices]
 
         #Kombiniere die Inhalte aller relevanten Dokumente
-        combined_content = " ".join([doc["content"] for doc in most_relevant_docs])
+        combined_content = " ".join([doc["page_content"] for doc in most_relevant_docs])
 
         #Formuliere die Eingabe für das Generierungsmodell
         input_text = f"frage: {prompt} kontext: {combined_content}"
@@ -444,7 +444,7 @@ def extract_document_info(documents):
     extracted_info = []
     for doc in documents:
         info = {
-            'content' : doc["content"],
+            'content' : doc["page_content"],
             'metadaten' : doc["metadata"],
             'titel' : metadaten.get("title", "Keine Überschrift"),
             'seite' : metadaten.get("page", "Unbekannte Seite"),