Update app.py

app.py

@@ -107,10 +107,10 @@ async def start():
         youtube_docs = await create_youtube_transcription(youtube_link['content']) # create the youtube transcription
         transcription = youtube_docs[0].page_content # get the transcription of the first document
         await cl.Message(content=f"youtube docs: {transcription}").send() # display the transcription of the first document to show that we have the correct data
-        split_docs = create_text_splitter(youtube_docs) # split the documents into chunks
-        vector_db = create_faiss_vector_store(split_docs) # create the vector db
+        split_docs = await create_text_splitter(youtube_docs) # split the documents into chunks
+        vector_db = await create_faiss_vector_store(split_docs) # create the vector db
         bm25 = create_bm25_retreiver(split_docs) # create the BM25 retreiver
-        ensemble_retriever = await create_ensemble_retriever(vector_db, bm25) # create the ensemble retriever
+        ensemble_retriever = create_ensemble_retriever(vector_db, bm25) # create the ensemble retriever
         cl.user_session.set("ensemble_retriever", ensemble_retriever) # store the ensemble retriever in the user session for our on message function
     except Exception as e:
         await cl.Message(content=f"failed to load model: {e}").send() # display the error if we failed to load the model
@@ -132,7 +132,8 @@ async def message(message: cl.Message):
     cl.Message(content=f"Displaying Relevant Docs").send() # we display the relevant documents to the user
     for doc in relevant_docs: # loop through the relevant documents and display each one!
         await cl.Message(content=doc.page_content).send() 
-    rag_chain = RunnableSequence({"context": ensemble_retriever, "question": RunnablePassthrough()} | prompt_template | llm)
-    response = rag_chain.invoke(message.content) # we invoke the rag chain with the user's message
     await cl.Message(content="Done Displaying Relevant Docs").send()
+    # question -> retrieve relevant docs -> format the question and context and add it to the prompt template -> pass to LLM 
+    rag_chain = RunnableSequence({"context": ensemble_retriever, "question": RunnablePassthrough()} | prompt_template | llm) 
+    response = rag_chain.invoke(message.content) # we invoke the rag chain with the user's message
     await cl.Message(content=f"LLM Response: {response.content}").send() # we display the response to the user