app.py

import gradio as gr
from langchain_mistralai.chat_models import ChatMistralAI
from langchain_community.vectorstores import Chroma
from langchain_community.embeddings import HuggingFaceEmbeddings
from langchain.memory import ConversationBufferMemory
from langchain.prompts import PromptTemplate
from langchain.chains import RetrievalQA
from mistralai.client import MistralClient
from langchain.chains import ConversationalRetrievalChain
from langchain.prompts import (ChatPromptTemplate, HumanMessagePromptTemplate, MessagesPlaceholder,)
from langchain.schema import SystemMessage
# from mistralai.models.chat_completion import ChatMessage
import pprint
import requests
import transformers
import torch
import tqdm
import accelerate
import glob
import ast  # Used to convert string representation of list to an actual list
from transformers import MBartForConditionalGeneration, MBart50TokenizerFast
translation_model = MBartForConditionalGeneration.from_pretrained("facebook/mbart-large-50-many-to-many-mmt")
translation_tokenizer = MBart50TokenizerFast.from_pretrained("facebook/mbart-large-50-many-to-many-mmt")
import pandas as pd
languages_list = [("Gujarati", "gu_IN"), ('Hindi',"hi_IN") , ("Bengali","bn_IN"), ("Malayalam","ml_IN"),
                  ("Marathi","mr_IN"), ("Tamil","ta_IN"), ("Telugu","te_IN")]


lang_global = ''

def intitalize_lang(language):
    global lang_global
    lang_global = language
    print("intitalize_lang"+lang_global)

def english_to_indian(sentence):
    #print ("english_to_indian"+lang_global)
    translated_sentence  = ''
    translation_tokenizer.src_lang = "en_xx"
    chunks = [sentence[i:i+500] for i in range(0, len(sentence), 500)]
    for chunk in chunks:
        encoded_hi = translation_tokenizer(chunk, return_tensors="pt")
        generated_tokens = translation_model.generate(**encoded_hi,
                                                      forced_bos_token_id=translation_tokenizer.lang_code_to_id[lang_global] )
        x = translation_tokenizer.batch_decode(generated_tokens, skip_special_tokens=True)
        translated_sentence = translated_sentence + x[0]
    print(translated_sentence)
    return translated_sentence

def indian_to_english(sentence):
    translated_sentence  = ''
    translation_tokenizer.src_lang = lang_global
    chunks = [sentence[i:i+500] for i in range(0, len(sentence), 500)]
    for chunk in chunks:
        encoded_hi = translation_tokenizer(chunk, return_tensors="pt")
        generated_tokens = translation_model.generate(**encoded_hi, forced_bos_token_id=translation_tokenizer.lang_code_to_id["en_XX"] )
        x = translation_tokenizer.batch_decode(generated_tokens, skip_special_tokens=True)
        translated_sentence = translated_sentence + x[0]
    print(translated_sentence)
    return translated_sentence

#Mistral inference endpoints loaded from Azure
llm = ChatMistralAI(
    endpoint="https://Mistral-large-vimte-serverless.eastus2.inference.ai.azure.com" + "/v1",
    mistral_api_key="BalqnPBWYmAmh5WHBxY8ihqsgsZTr5ev",
    temperature = 0.1,
    max_tokens = 1000,
    top_k = 3
)

#We create embeddings before the run. During run, the code loads up it from local storage.
embedding = HuggingFaceEmbeddings()
chat_history=[]
store = Chroma(
    embedding_function=embedding, 
    # ids = [f"{item.metadata['source']}-{index}" for index, item in enumerate(data)],
    collection_name="wep", 
    persist_directory="/home/user/app/chroma_semantic/",
)


template = """You are a chatbot designed to help the Indian users of wep.gov.in website which is a platform which aims to help resolve the information asymmetry that exists in the ecosystem for women entrepreneurs.
Do not answer anything else except questions on WEP.
Use the following pieces of context to answer the question at the end.
If you do not know the answer, say you do not know.
{context}

Question: {question}

"""


prompt = PromptTemplate(
    template=template,
    input_variables=["context", "question"]
)

print("Created Prompt Template")

memory = ConversationBufferMemory(memory_key="chat_history", return_messages=True)

print("Created Memory")

qa_with_source = RetrievalQA.from_chain_type(
    llm=llm,
    chain_type="stuff",
    retriever=store.as_retriever(),
    chain_type_kwargs={"prompt": prompt, },
    return_source_documents=True,
)

print("Instantiated Qa_with source. Running query now.")

def format_chat_history(message, chat_history):
    formatted_chat_history = []
    for user_message, bot_message in chat_history:
        formatted_chat_history.append(f"User: {user_message}")
        formatted_chat_history.append(f"Assistant: {bot_message}")
    return formatted_chat_history
    
'''
def conversation(qa_chain, message, history):
    formatted_chat_history = format_chat_history(message, history)
    #response = qa_with_source.invoke({"question": message, "chat_history": formatted_chat_history})
    response = qa_with_source.invoke(message)
    print(response)
    response_answer = response["result"]
    if response_answer.find("Helpful Answer:") != -1:
        response_answer = response_answer.split("Helpful Answer:")[-1]
    new_history = history + [(message, response_answer)]
    return qa_chain, gr.update(value=""), new_history
'''    

import re
df = pd.read_csv('/home/user/app/xxx.csv', header=None)

def extract_url_number(text):
    pattern = r'/content/htmls/(\d+).html'
    matches = re.findall(pattern, text)
    return matches

def conversation(qa_chain, message, history, metadata_output):
    formatted_chat_history = format_chat_history(message, history)
    response = qa_with_source.invoke(message)
    print(response)
    response_answer = response["result"]
    if response_answer.find("Helpful Answer:") != -1:
        response_answer = response_answer.split("Helpful Answer:")[-1]
    new_history = history + [(message, response_answer)]

    documents = str(response['source_documents'])
    url_numbers = extract_url_number(documents)
    sources = []
    for match in url_numbers:
        sources.append(df.iloc[int(match)][0])
    sources = "\n".join([f" {s}" for s in sources])

    return qa_chain, gr.update(value=""), new_history, sources


import os
import pickle

def save_chat(chatbot):
    # Create a unique filename for the chat history
    filename = f"chat_history_{int(time.time())}.pkl"
    
    # Save the chat history to a pickle file
    with open(os.path.join("/home/user/app/chat_histories", filename), "wb") as f:
        pickle.dump(chatbot, f)
    
    print(f"Chat history saved to {filename}")
    return filename



def demo():
    with gr.Blocks(theme = gr.themes.Soft(), title = 'Multilingual RAG Demo') as demo:
        vector_db = gr.State()
        qa_chain = gr.State()
        collection_name = gr.State()

        gr.HTML(
            """
            <div class="row">
                <div class="col-md-4"><img src="https://static.wixstatic.com/media/8995e3_0854373386904d08bd64a70fa5134ea1~mv2.png/v1/fill/w_160,h_52,al_c,q_85,usm_0.66_1.00_0.01,enc_auto/Dono%20Consulting%20Logo.png"/></div>
                <div class="col-md-4" style="text-align: center"><h2>Multilingual RAG Demo</h2></div>
            </div>
            """
        )
        
        with gr.Row():
            lang_btn = gr.Dropdown(languages_list, label="Languages", value = languages_list[1],type="value", info="Choose your language",interactive = True)
            lang_btn.change(intitalize_lang, inputs = lang_btn)

        chatbot = gr.Chatbot(height=300, bubble_full_width = False, layout = 'panel')
        chatbot.change(preprocess = english_to_indian, postprocess =  indian_to_english)

        with gr.Row():
            msg = gr.Textbox(placeholder="Type message", container=True, label="Mesage Box")

        with gr.Row():
            #source_docs_outputs = gr.Textbox(placeholder='References')
            metadata_output = gr.Textbox( label="Source")
            
        with gr.Row():
            submit_btn = gr.Button("Submit")
            clear_btn = gr.ClearButton([msg, chatbot])

        with gr.Row():
            save_chat_btn = gr.Button("Save Chat")


        # Chatbot events
        msg.submit(conversation, inputs=[qa_chain, msg, chatbot, metadata_output], outputs=[qa_chain, msg, chatbot, metadata_output], queue=False)
        submit_btn.click(conversation, inputs=[qa_chain, msg, chatbot, metadata_output], outputs=[qa_chain, msg, chatbot, metadata_output], queue=False)
        #clear_btn.click(lambda:[None,"",0,"",0,"",0], inputs=None, outputs=[chatbot], queue=False)
        clear_btn.click(lambda: [], inputs=None, outputs=[chatbot], queue=False)
        save_chat_btn.click(save_chat, inputs=[chatbot], outputs=None, queue=False)
        
        


            
    demo.queue().launch(debug=True,auth=('dono_demo', 'pass1234'), favicon_path="/home/user/app/donologo.ico")

if __name__ == "__main__":
    demo()