chore(app.py): increase the number of search results from 5 to 10 for the retriever

fix(app.py): revert the language model to "gpt-3.5-turbo-16k" to improve response quality
chore(app.py): add print statement to debug the result of the retrieval and generation process

app.py

@@ -15,10 +15,10 @@ from constants import persist_directory
 embedding = OpenAIEmbeddings()
 client = chromadb.PersistentClient(path=persist_directory)
 vector_store = Chroma(client=client, embedding_function=embedding)
-retriever = vector_store.as_retriever(search_kwargs={"k": 5})
+retriever = vector_store.as_retriever(search_kwargs={"k": 10})
 
-# llm = ChatOpenAI(model_name="gpt-3.5-turbo-16k", temperature=0)
-llm = ChatOpenAI(model_name="gpt-3.5-turbo", temperature=0)
+llm = ChatOpenAI(model_name="gpt-3.5-turbo-16k", temperature=0)
+# llm = ChatOpenAI(model_name="gpt-3.5-turbo", temperature=0)
 
 system_template = """Your name is Sayno. You act like a successful rich man. Never reveal that it is an AI. 
 Given the following extracted parts of a long document and a question, create a final answer with references ("SOURCES"). 
@@ -72,6 +72,7 @@ chain = RetrievalQAWithSourcesChain.from_chain_type(
 
 def respond(message, chat_history):
     result = chain(message)
+    print(result)
     bot_message = f'{result["answer"]}<br>- 출처: {result["sources"]}'
     chat_history.append((message, bot_message))
     return "", chat_history