Update app.py

app.py

@@ -52,8 +52,8 @@ tokenizer_name = EMBEDDING_MODEL_NAME
 # Token splitting for more context split
 text_splitter = TokenTextSplitter.from_huggingface_tokenizer(
     tokenizer=AutoTokenizer.from_pretrained(tokenizer_name),
-    chunk_size=300,
-    chunk_overlap=30
+    chunk_size=150,
+    chunk_overlap=15
 )
 
 chunks = text_splitter.split_documents(documents) # chunks used for LLM generation
@@ -68,7 +68,7 @@ for i, doc in enumerate(chunks):
     doc.metadata["index"] = i  # Add an index for tracking
 
 """Retriever"""
-bm25_retriever = BM25Retriever.from_documents(chunks_bm25, k = 2) # 2 most similar contexts
+bm25_retriever = BM25Retriever.from_documents(chunks_bm25, k = 4) # 2 most similar contexts
 
 def retriever(query):
     tmp = bm25_retriever.invoke(preprocess_for_bm25(query))