from flask import Flask, request, render_template
# Import the relevant modules and functions from your code
from langchain.llms import GooglePalm
from langchain.llms import LlamaCpp
from transformers import pipeline
from langchain.llms import HuggingFacePipeline
from langchain.embeddings import GooglePalmEmbeddings
from langchain.document_loaders.csv_loader import CSVLoader
from langchain.embeddings import HuggingFaceInstructEmbeddings
from langchain.vectorstores import FAISS
from langchain.prompts import PromptTemplate
from langchain.chains import RetrievalQA
from langchain.text_splitter import RecursiveCharacterTextSplitter
import torch


app = Flask(__name__)

# Initialize Langchain components (put this code outside the app route)
#api_key = 'AIzaSyCZwsYvr3ht7ctxcrOLWvfppySP33ducmE'
#llm = GooglePalm(google_api_key=api_key, temperature=0.1)

from langchain.llms import CTransformers

llm = CTransformers(model=r"C:\Aiproject\mainmodel\llama-2-7b-chat.ggmlv3.q8_0.bin",model_type="llama",temperature=0.1,gpu_layers=50,do_sample=True)

# llm = CTransformers(
#         model = r"C:\Aiproject\mistral\Mistral-7B-v0.1",
#         model_type="mistral",
#         max_new_tokens = 1048,
#         temperature = 0.1,gpu_layers=50
#     )


# device=torch.device('cpu')
# from transformers import AutoTokenizer, AutoModelForSeq2SeqLM 

# checkpoint = "MBZUAI/LaMini-T5-738M"
# tokenizer = AutoTokenizer.from_pretrained(checkpoint)
# base_model = AutoModelForSeq2SeqLM.from_pretrained(
#     checkpoint,
#     device_map=device,
#     torch_dtype=torch.float32
# )


# pipe = pipeline(
#         'text2text-generation',
#         model = base_model,
#         tokenizer = tokenizer,
#         max_length = 256,
#         do_sample = True,
#         temperature = 0.1,
#         top_p= 0.95
#     )
# llm = HuggingFacePipeline(pipeline=pipe)





# Initialize instructor embeddings using the Hugging Face model
instructor_embeddings = HuggingFaceInstructEmbeddings(model_name=r"C:\Aiproject\SupportGpt\model")

# Load the data from CSV
loader = CSVLoader(file_path=r"C:\Aiproject\codebasics_faqs.csv", encoding='iso-8859-1', source_column="Question")
data = loader.load()
text_splitter=RecursiveCharacterTextSplitter(chunk_size=500,chunk_overlap=50)
texts=text_splitter.split_documents(data)


# Create a FAISS instance for vector database
vectordb = FAISS.from_documents(documents=texts, embedding=instructor_embeddings)

# Create a retriever for querying the vector database
retriever = vectordb.as_retriever(score_threshold=0.7)

# Define the prompt template
# prompt_template = """Use the following pieces of information to answer the user's question,dont give additional answers.
# If you don't know the answer, just say that you don't know, don't try to make up an answer.

prompt_template="""Use the following peices of information to answer the user's question and do not give any additional or extra answers that do not exist in the given information.If the answer doesn't exist in the given information, say "I dont know". If the answer exists in the information, give the exact answer as it exists in the given information.Users may use different keywords or synonyms, so use at least 10 keywords or synonyms from the user's question to identify exactly what the user wants to know from given information.Please refrain from making your own answers or modifying the answers.
CONTEXT: {context}

QUESTION: {question}"""

# Create a prompt template
PROMPT = PromptTemplate(template=prompt_template, input_variables=["context", "question"])
chain_type_kwargs = {"prompt": PROMPT}

# Create the Langchain RetrievalQA chain
chain = RetrievalQA.from_chain_type(llm=llm, chain_type="stuff", retriever=retriever, input_key="query",
                                   return_source_documents=True, chain_type_kwargs=chain_type_kwargs)



@app.route('/', methods=['GET', 'POST'])
def index():
    if request.method == 'POST':
        question = request.form['question']
        if question:
            result = chain(question)['result']
        else:
            result = None
        return render_template('index1.html', question=question, result=result)
    return render_template('index1.html', question='', result=None)


if __name__ == '__main__':
    app.run(host="0.0.0.0",debug=False)