import streamlit as st
from langchain_community.llms import HuggingFaceEndpoint

#When deployed on huggingface spaces, this values has to be passed using Variables & Secrets setting, as shown in the video :)
#import os
#os.environ["OPENAI_API_KEY"] = "sk-PLfFwPq6y24234234234FJ1Uc234234L8hVowXdt"

#from api import Api
#import streamlit as st 
from langchain.prompts import PromptTemplate
from langchain.chains import LLMChain

#external class for api integrations, 
#api = Api()

#llm default OpenAPI
#llm = api.llm
llm = HuggingFaceEndpoint(repo_id="mistralai/Mistral-7B-Instruct-v0.2", Temperature=0.9)

#streamlit view components
with st.form("my_form"):
    st.title('Sentiment Analysis')
    text_review = st.text_area('Write me a review') 

    option = st.selectbox(
    'Select the language to evaluate:',
    ('Italian', 'Spanish', 'English'))
    submitted = st.form_submit_button("Submit")
    if submitted:
        
        #1 prompt template
        template = """
        Please act as a machine learning model trained for perform a supervised learning task, 
        for extract the sentiment of a review in '{option}' Language.

        Give your answer writing a Json evaluating the sentiment field between the dollar sign, the value must be printed without dollar sign.
        The value of sentiment must be "Positive" , "Negative" or "Neutral"  otherwise if the text is not valuable write "null".

        Example:

        field 1 named :
        text_review with value: {text_review}
        field 2 named :
        sentiment with value: $sentiment$
        Field 3 named : 
        language with value: {option}
        Review text: '''{text_review}'''

        """

        prompt = PromptTemplate(template=template, input_variables=["text_review","option"])

        llm_chain = LLMChain(prompt=prompt, llm=llm)

        if prompt:
            response = llm_chain.run({"text_review": text_review, "option": option})
            #json printed
            print(response)
            st.text(response)