Update app.py

app.py

@@ -328,23 +328,58 @@ textarea::placeholder {
 }
 """
 
-def process_question(question: str) -> Iterator[str]:
-    """
-    Process the question and return a response generator for streaming.
-    """
+# def process_question(question: str) -> Iterator[str]:
+#     """
+#     Process the question and return a response generator for streaming.
+#     """
+#     if question in question_cache:
+#         yield question_cache[question][0]
+#         return
+
+#     relevant_docs = retriever(question)
+#     context = "\n".join([doc.page_content for doc in relevant_docs])
+
+#     prompt = prompt_template.format_messages(
+#         context=context,
+#         question=question
+#     )
+
+#     full_response = ""
+#     try:
+#         for chunk in llm.stream(prompt):
+#             if isinstance(chunk, str):
+#                 current_chunk = chunk
+#             else:
+#                 current_chunk = chunk.content
+
+#             full_response += current_chunk
+#             yield full_response  # Send the updated response in streaming
+
+#         question_cache[question] = (full_response, context)
+#     except Exception as e:
+#         yield f"Erreur lors du traitement : {str(e)}"
+
+def process_question(question: str) -> tuple[str, list[str]]:
+    # Check cache first
     if question in question_cache:
-        yield question_cache[question][0]
-        return
+        return question_cache[question]
 
+    # Get relevant documents using the retriever
     relevant_docs = retriever(question)
-    context = "\n".join([doc.page_content for doc in relevant_docs])
-
-    prompt = prompt_template.format_messages(
+    
+    # Extract the content and sources
+    context = "\n".join(doc.page_content for doc in relevant_docs)
+    sources = [doc.metadata["source"] for doc in relevant_docs]
+    sources = os.path.splitext(sources[0])[0] if sources else "غير معروف"
+    
+    # Generate the prompt with the context
+    prompt = prompt_template.format(
         context=context,
         question=question
     )
 
     full_response = ""
+    
     try:
         for chunk in llm.stream(prompt):
             if isinstance(chunk, str):
@@ -353,13 +388,14 @@ def process_question(question: str) -> Iterator[str]:
                 current_chunk = chunk.content
 
             full_response += current_chunk
-            yield full_response  # Send the updated response in streaming
+            yield full_response, sources   # Send the updated response in streaming
 
-        question_cache[question] = (full_response, context)
+        question_cache[question,sources] = (full_response, context)
     except Exception as e:
         yield f"Erreur lors du traitement : {str(e)}"
 
 
+
 def gradio_stream(question: str, chat_history: list) -> Iterator[list]:
     """
     Format the output for Gradio Chatbot component with streaming.