from openai import OpenAI
import streamlit as st
from streamlit import session_state
import numpy as np
client = OpenAI()
openai.api_key = os.getenv("OPENAI_API_KEY")
from sentence_transformers import SentenceTransformer, LoggingHandler, losses, util, InputExample
model = SentenceTransformer('paraphrase-MiniLM-L6-v2')
from scipy import spatial
def cosine_similarity(x,y):
    return 1 - spatial.distance.cosine(x,y)
def scorer(m,s):
    response = client.chat.completions.create(
      model="gpt-4",
      messages=[
        {
          "role": "system",
          "content": "You are a Grading assistant. Your task is to give marks to the student answers based on the model answer. You will be provided student answer and model answer. Please check both carefully and give marks out of 10.  \n\n<<REMEMBER>>\nThese answers are written by High school kids. Don't be too strict. Give marks generously. \nGive marks in the range of 0.5. \nIf the model answer and student answer is slightly matching then give marks generously.  Please give output in json format like this:\n{\"output\":5}"
        },
        {
          "role": "user",
          "content": f"Model answer: {m}"},
        {
          "role": "user",
          "content": f"Student answer: {s}"
        }
      ],
      temperature=0,
      max_tokens=256,
      top_p=1,
      frequency_penalty=0,
      presence_penalty=0
    )
    return response.choices[0].message.content
def embeddings_cosine(s1,s2,model):
    embeddings1 = model.encode(s1)
    embeddings2 = model.encode(s2)
    cosine_scores = cosine_similarity(embeddings1, embeddings2)
    return cosine_scores
def main_score(m,s):
    if len(s.split())>15:
        body_score = embeddings_cosine(preprocess_bert(m),preprocess_bert(s),model)
        score = body_score*10
        return np.abs(np.round(score * 2) / 2)
    else:
        score = scorer(m,s)
        return score
    
from st_pages import Page, Section, show_pages, add_page_title,add_indentation
st.set_page_config(page_title="Auto score Openai", page_icon="📈")

st.markdown("<h1 style='text-align: center; color: black;'> Welcome to Our App! 👋</h1>", unsafe_allow_html=True)

if 'result' not in session_state:
    session_state['result']= ""
    
st.title("Auto score")
text1= st.text_area(label= "Please write the text bellow", 
              placeholder="What does the tweet say?")
text2= st.text_area(label= "Please write the text bellow", 
              placeholder="What does the tweet say?")
def classify(text):
    session_state['result'] = main_score(text1,text2)


st.text_area("result", value=session_state['result'])

st.button("Classify", on_click=classify, args=[text1,text2])