Update app.py

app.py

@@ -66,23 +66,28 @@ def create_db(splits):
     vectordb = FAISS.from_documents(splits, embeddings)
     return vectordb
 
-def initialize_llmchain(llm_model, temperature, max_tokens, top_k, vector_db):
-    """Initialize the LLM chain"""
+def initialize_llmchain(llm_model, temperature, max_tokens, top_k, vector_db, api_token):
+    """Initialize the LLM chain with a HuggingFace model"""
+    # Use valid Hugging Face parameters. `max_length` might be the correct field instead of `max_new_tokens`
     llm = HuggingFaceEndpoint(
         repo_id=llm_model,
         huggingfacehub_api_token=api_token,
         temperature=temperature,
-        max_new_tokens=max_tokens,
-        top_k=top_k,
+        max_length=max_tokens,  # Adjusted from max_new_tokens to max_length
+        # Remove top_k as it may not be valid or handled differently
     )
 
+    # Set up memory for conversation
     memory = ConversationBufferMemory(
         memory_key="chat_history",
         output_key='answer',
         return_messages=True
     )
 
+    # Ensure vector_db is used as a retriever
     retriever = vector_db.as_retriever()
+
+    # Initialize ConversationalRetrievalChain using LLM and the retriever
     qa_chain = ConversationalRetrievalChain.from_llm(
         llm,
         retriever=retriever,
@@ -93,6 +98,7 @@ def initialize_llmchain(llm_model, temperature, max_tokens, top_k, vector_db):
     )
     return qa_chain
 
+
 def format_chat_history(message, chat_history):
     """Format chat history for the LLM"""
     formatted_chat_history = []