Update app.py

app.py

@@ -102,14 +102,14 @@ else:
     print("Created new vector store and persisted embeddings.")
 
 # Create a BM25 retriever from the document splits.
-bm25_retriever = BM25Retriever.from_documents(all_splits)
+bm25_retriever = BM25Retriever.from_documents(all_splits, search_kwargs={"k": 5})
 ensemble_retriever = EnsembleRetriever(
-        retrievers=[
-            vectorstore.as_retriever(),
-            bm25_retriever
-        ],
-        weights=[0.8, 0.2]
-    )
+    retrievers=[
+        vectorstore.as_retriever(search_kwargs={"k": 5}),
+        bm25_retriever
+    ],
+    weights=[0.8, 0.2]
+)
 retriever = ensemble_retriever
 
 # -------------------------------
@@ -197,7 +197,7 @@ def message_and_history(message, history):
 
     # Combine the most recent conversation turns, excluding the assistant's prefix.
     conversation_context = "\n".join(
-        [f"{msg['role']}: {msg['content'].replace('<b>LA2050 Navigator:</b><br>', '')}" for msg in history[-2:]]
+        [f"{msg['role']}: {msg['content'].replace('<b>LA2050 Navigator:</b><br>', '')}" for msg in history[-3:]]
     )
 
     chain_input = {"input": conversation_context}