import streamlit as st
from openai import OpenAI
import os
import pandas as pd
import numpy as np
from sentence_transformers import SentenceTransformer
from sklearn.metrics.pairwise import cosine_similarity
import torch
import requests

# Set up OpenAI client
client = OpenAI(api_key=os.getenv("OPENAI_API_KEY"))

# Check if GPU is available
device = "cuda" if torch.cuda.is_available() else "cpu"
print(f"Using device: {device}")

# Load metadata and embeddings (ensure these files are in your working directory or update paths)
metadata_path = 'question_metadata.csv'
embeddings_path = 'question_dataset_embeddings.npy'

metadata = pd.read_csv(metadata_path)
embeddings = np.load(embeddings_path)

# Load the SentenceTransformer model
model = SentenceTransformer("all-MiniLM-L6-v2").to(device)

# Load prompts from files
with open("question_generation_prompt.txt", "r") as file:
    question_generation_prompt = file.read()

with open("technical_interviewer_prompt.txt", "r") as file:
    technical_interviewer_prompt = file.read()

st.title("Real-World Programming Question Mock Interview")

# Initialize session state variables
if "messages" not in st.session_state:
    st.session_state.messages = []

if "generated_question" not in st.session_state:
    st.session_state.generated_question = None

if "code_output" not in st.session_state:
    st.session_state.code_output = ""

if "evaluation_output" not in st.session_state:
    st.session_state.evaluation_output = ""

# Sidebar layout for Generated Question and Code Box
st.sidebar.markdown("## Generated Question")
if st.session_state.generated_question:
    st.sidebar.markdown(st.session_state.generated_question)
else:
    st.sidebar.markdown("_No question generated yet._")

st.sidebar.markdown("---")
st.sidebar.markdown("## Code Box")

code_input = st.sidebar.text_area(
    label="Write your Python code here:",
    height=200,
    placeholder="Enter your code...",
)

col1, col2 = st.sidebar.columns(2)

# Button to run code and display output
if col1.button("Run Code"):
    try:
        exec_globals = {}
        exec(code_input, exec_globals)
        st.session_state.code_output = exec_globals.get("output", "Code executed successfully.")
    except Exception as e:
        st.session_state.code_output = f"Error: {str(e)}"

# Button to evaluate code using OpenAI API
if col2.button("Evaluate Code"):
    if not st.session_state.generated_question:
        st.sidebar.error("Generate a question first!")
    else:
        try:
            evaluation_prompt = (
                f"Question: {st.session_state.generated_question}\n\n"
                f"Code:\n{code_input}\n\n"
                f"Evaluate this code's correctness, efficiency, and style."
            )
            response = client.chat.completions.create(
                model="gpt-4",
                messages=[{"role": "user", "content": evaluation_prompt}],
            )
            evaluation_response = response.choices[0].message.content
            st.session_state.evaluation_output = evaluation_response

            # Add evaluation output to follow-up conversation
            st.session_state.messages.append({"role": "assistant", "content": evaluation_response})
        except Exception as e:
            st.sidebar.error(f"Error during evaluation: {str(e)}")

# Display outputs below the main app content
st.subheader("Code Output")
st.text(st.session_state.code_output)

st.subheader("Evaluation Output")
st.text(st.session_state.evaluation_output)

# Main app logic for generating questions and follow-up conversation remains unchanged.
with st.form(key="input_form"):
    company = st.text_input("Company", value="Google")
    difficulty = st.selectbox("Difficulty", ["Easy", "Medium", "Hard"], index=1)
    topic = st.text_input("Topic (e.g., Backtracking)", value="Backtracking")
    
    generate_button = st.form_submit_button(label="Generate")

if generate_button:
    query = f"{company} {difficulty} {topic}"
    
    def find_top_question(query):
        query_embedding = model.encode(query, convert_to_tensor=True, device=device).cpu().numpy()
        query_embedding = query_embedding.reshape(1, -1)
        similarities = cosine_similarity(query_embedding, embeddings).flatten()
        top_index = similarities.argsort()[-1]
        top_result = metadata.iloc[top_index].copy()
        top_result['similarity_score'] = similarities[top_index]
        return top_result
    
    top_question = find_top_question(query)
    
    detailed_prompt = (
        f"Transform this LeetCode question into a real-world interview scenario:\n\n"
        f"**Company**: {top_question['company']}\n"
        f"**Question Name**: {top_question['questionName']}\n"
        f"**Difficulty Level**: {top_question['difficulty level']}\n"
        f"**Tags**: {top_question['Tags']}\n"
        f"**Content**: {top_question['Content']}\n"
        f"\nPlease create a real-world interview question based on this information."
    )
    
    response_text = client.chat.completions.create(
        model="gpt-4",
        messages=[{"role": "assistant", "content": question_generation_prompt}, {"role": "user", "content": detailed_prompt}],
    ).choices[0].message.content
    
    st.session_state.generated_question = response_text

for message in st.session_state.messages:
    with st.chat_message(message["role"]):
        st.markdown(message["content"])

if user_input := st.chat_input("Continue your conversation or ask follow-up questions here:"):
    with st.chat_message("user"):
        st.markdown(user_input)
    
    assistant_response_text = client.chat.completions.create(
        model="gpt-4",
        messages=[{"role": "assistant", "content": technical_interviewer_prompt}] + [{"role": msg["role"], "content": msg["content"]} for msg in st.session_state.messages],
    ).choices[0].message.content
    
    with st.chat_message("assistant"):
        st.markdown(assistant_response_text)
    
    # Append to session state messages for persistence
    st.session_state.messages.append({"role": "user", "content": user_input})
    st.session_state.messages.append({"role": "assistant", "content": assistant_response_text})