Update app.py

app.py

@@ -1,12 +1,3 @@
-# -*- coding: utf-8 -*-
-"""api.ipynb
-
-Automatically generated by Colaboratory.
-
-Original file is located at
-    https://colab.research.google.com/drive/1XRryfVWG4d_ScN5ADvlZpKmREvTJN3mg
-"""
-
 import gradio as gr
 import os
 
@@ -58,7 +49,6 @@ def initialize_database(file_path):
     print('Collection name: ', collection_name)
     # Load document and create splits
     doc_splits = load_doc(file_path)
-    # Create or load vector database
     # global vector_db
     vector_db = create_db(doc_splits, collection_name)
     return vector_db, collection_name, "Complete!"
@@ -71,7 +61,6 @@ def create_db(splits, collection_name):
         embedding=embedding,
         client=new_client,
         collection_name=collection_name,
-        # persist_directory=default_persist_directory
     )
     return vectordb
 
@@ -89,7 +78,6 @@ def initialize_llmchain(temperature, max_tokens, top_k, vector_db):
 
     llm = HuggingFaceEndpoint(
             repo_id='mistralai/Mixtral-8x7B-Instruct-v0.1',
-            # model_kwargs={"temperature": temperature, "max_new_tokens": max_tokens, "top_k": top_k, "load_in_8bit": True},
             temperature = temperature,
             max_new_tokens = max_tokens,
             top_k = top_k,
@@ -101,14 +89,12 @@ def initialize_llmchain(temperature, max_tokens, top_k, vector_db):
         retriever=retriever,
         chain_type="stuff",
         memory=memory,
-        # combine_docs_chain_kwargs={"prompt": your_prompt})
         return_source_documents=True,
-        #return_generated_question=False,
         verbose=False,
     )
     return qa_chain
 
-qa = initialize_llmchain(0.7, 1024, 1, vec_cre)
+qa = initialize_llmchain(0.7, 1024, 1, vec_cre) #The model question answer
 
 def format_chat_history(message, chat_history):
     formatted_chat_history = []
@@ -119,26 +105,21 @@ def format_chat_history(message, chat_history):
 
 def conversation(message, history):
     formatted_chat_history = format_chat_history(message, history)
-    #print("formatted_chat_history",formatted_chat_history)
 
     # Generate response using QA chain
     response = qa({"question": message, "chat_history": formatted_chat_history})
     response_answer = response["answer"]
     if response_answer.find("Helpful Answer:") != -1:
         response_answer = response_answer.split("Helpful Answer:")[-1]
+    #You can also return from where the model got the answer to fine-tune or adjust your model mais ici c'est bon
     response_sources = response["source_documents"]
     response_source1 = response_sources[0].page_content.strip()
     response_source2 = response_sources[1].page_content.strip()
     response_source3 = response_sources[2].page_content.strip()
-    # Langchain sources are zero-based
     response_source1_page = response_sources[0].metadata["page"] + 1
     response_source2_page = response_sources[1].metadata["page"] + 1
     response_source3_page = response_sources[2].metadata["page"] + 1
-    # print ('chat response: ', response_answer)
-    # print('DB source', response_sources)
 
-    # Append user message and response to chat history
-    # return gr.update(value=""), new_history, response_sources[0], response_sources[1]
     return response_answer