import os
import json
import streamlit as st
from huggingface_hub import InferenceClient
from dotenv import load_dotenv

# Load environment variables for local development
load_dotenv()

# Initialize the Hugging Face Inference client
hf_token = os.getenv("HUGGINGFACEHUB_API_TOKEN")
client = InferenceClient(provider="auto", api_key=hf_token)

# Streamlit configuration
st.set_page_config(
    page_title="Interview Prep Bot",
    page_icon="🧠",
    layout="centered"
)
st.title("🎓 Interview Preparation Chatbot")

# Session state initialization
if "questions" not in st.session_state:
    st.session_state.questions = []
if "topic" not in st.session_state:
    st.session_state.topic = "Machine Learning"
if "score" not in st.session_state:
    st.session_state.score = 0
if "correct_count" not in st.session_state:
    st.session_state.correct_count = 0
if "incorrect_count" not in st.session_state:
    st.session_state.incorrect_count = 0

# Sidebar: Topic and Score display
st.sidebar.header("Practice Topic")
st.session_state.topic = st.sidebar.selectbox(
    "Select a topic:",
    ["Machine Learning","Data Structures","Python","Generative AI","Computer Vision","Deep Learning"],
    index=["Machine Learning","Data Structures","Python","Generative AI","Computer Vision","Deep Learning"].index(st.session_state.topic)
)
st.sidebar.markdown("---")
st.sidebar.header("Your Score")
st.sidebar.markdown(f"**Total:** {len(st.session_state.questions)}")
st.sidebar.markdown(f"**Correct:** {st.session_state.correct_count}")
st.sidebar.markdown(f"**Incorrect:** {st.session_state.incorrect_count}")
st.sidebar.markdown(f"**Points:** {st.session_state.score}")

# Function to fetch a new MCQ question via HF Inference API

def fetch_question(topic):
    prompt = {
        "role": "system",
        "content": (
            "You are an expert interviewer. "
            "Generate a multiple-choice question on the topic of {topic}. "
            "Respond with a JSON object: {\"question\": str, \"options\": [str, ...], \"correct_index\": int}"
        ).format(topic=topic)
    }
    response = client.chat.completions.create(
        model="mistralai/Mistral-7B-Instruct-v0.1",
        messages=[prompt]
    )
    content = response.choices[0].message["content"].strip()
    try:
        data = json.loads(content)
        return data
    except json.JSONDecodeError:
        st.error("Failed to parse question. Please try again.")
        return None

# Buttons for initial and next questions
if not st.session_state.questions:
    if st.button("Get Question"):
        q = fetch_question(st.session_state.topic)
        if q:
            st.session_state.questions.append({**q, "selected": None, "submitted": False})
else:
    if st.button("Next Question"):
        q = fetch_question(st.session_state.topic)
        if q:
            st.session_state.questions.append({**q, "selected": None, "submitted": False})

# Display questions and capture answers
for idx, q in enumerate(st.session_state.questions):
    st.markdown(f"### Question {idx+1}")
    st.write(q["question"])

    # Radio options
    opts = q["options"]
    sel = st.radio(
        "Choose an answer:",
        options=list(range(len(opts))),
        format_func=lambda i: opts[i],
        index=q["selected"] if q["selected"] is not None else 0,
        key=f"radio_{idx}",
        disabled=q["submitted"]
    )
    st.session_state.questions[idx]["selected"] = sel

    # Submit button
    if not q["submitted"]:
        if st.button("Submit Answer", key=f"submit_{idx}"):
            st.session_state.questions[idx]["submitted"] = True
            if sel == q["correct_index"]:
                st.success("Correct!")
                st.session_state.score += 10
                st.session_state.correct_count += 1
            else:
                st.error(f"Incorrect! Correct answer: {opts[q['correct_index']]}.")
                st.session_state.score -= 10
                st.session_state.incorrect_count += 1

# Footer info
st.markdown("---")
st.markdown("*Points: +10 for correct, -10 for incorrect.*")