Update app.py

app.py

@@ -71,7 +71,8 @@ retriever=faiss_db.as_retriever(search_type="similarity", search_kwargs={"k":3})
 llm = ChatGroq(
     model="llama-3.3-70b-versatile",
     temperature=0.0,
-    max_tokens=1200
+    max_tokens=1200,
+    streaming=True
     )
 
 prompt = ChatPromptTemplate.from_messages([
@@ -120,14 +121,31 @@ convers_chain = RunnableWithMessageHistory(
 
 SESSION_ID = str(uuid.uuid4())  # session globale
 
+# def chat_fn(message, history):
+#     result = convers_chain.invoke(
+#         {"input": message},
+#         config={"configurable": {"session_id": SESSION_ID}}
+#     )
+#     return result.get("answer", str(result))
+
 def chat_fn(message, history):
-    result = convers_chain.invoke(
+    """Chat avec streaming optimisé"""
+    response = ""
+    
+    for chunk in convers_chain.stream(
         {"input": message},
         config={"configurable": {"session_id": SESSION_ID}}
-    )
-    return result.get("answer", str(result))
-
-
+    ):
+        # Gestion des différents types de chunks possibles
+        if isinstance(chunk, dict):
+            token = chunk.get("answer") or chunk.get("content") or str(chunk)
+        elif hasattr(chunk, "content"):   # Pour AIMessage
+            token = chunk.content
+        else:
+            token = str(chunk)
+        
+        response += token
+        yield response
 
 
 # ================= GRADIO ====================