import streamlit as st
import requests
import os
from streamlit_chat import message


@st.cache
def query(payload):
    api_token = os.getenv("api_token")
    model_id = "deepset/roberta-base-squad2"
    headers = {"Authorization": f"Bearer {api_token}"}
    API_URL = f"https://api-inference.huggingface.co/models/{model_id}"
    response = requests.post(API_URL, headers=headers, json=payload)
    return response.json(), response




context = "To extract information from documents, use sentence similarity task. To do sentiment analysis from tweets, use text classification task. To detect masks from images, use object detection task. To extract information from invoices, use named entity recognition from token classification task."


message_history = [{"text":"Let's find out the best task for your use case! Tell me about your use case :)", "is_user":False}]


for msg in message_history:
    message(msg["text"], is_user = msg["is_user"])   # display all the previous message

placeholder = st.empty()  # placeholder for latest message

#with placeholder.container():
#    last_message = message_history[-1]
#    if last_message != "":
    
#        message(last_message["text"], last_message["is_user"]) # display the latest message
        
input = st.text_input("Ask me 🤗")
message(input, is_user=True)  # align's the message to the right
message_history.append({"text":input, "is_user" : True})

data, resp = query(
    {
        "inputs": {
            "question": input,
            "context": context,
        }
    }
)
if resp.status_code == 200:

    model_answer = data["answer"]
    response_templates = [f"{model_answer} is the best task for this 🤩", f"I think you should use {model_answer} 🪄", f"I think {model_answer} should work for you 🤓"]
    
    bot_answer = random.choice(response_templates)
    
    message_history.append({"text":bot_answer, "is_user" : False})