import streamlit as st
from meta_ai_api import MetaAI
from urllib.parse import urlparse
import pandas as pd
import plotly.express as px
from nltk.sentiment.vader import SentimentIntensityAnalyzer
import nltk
import json


# Initialize Meta AI API
ai = MetaAI()

PAGE_CONFIG = {
    "page_title": "Meta AI Query Analysis - a Free SEO Tool by WordLift",
    "page_icon": "img/fav-ico.png",
    "layout": "centered"
}

def local_css(file_name):
    with open(file_name) as f:
        st.markdown(f'<style>{f.read()}</style>', unsafe_allow_html=True)


st.set_page_config(**PAGE_CONFIG)

local_css("style.css")


def fetch_response(query):
    response = ai.prompt(message=query)
    return response

def display_sources(sources):
    if sources:
        for source in sources:
            # Parse the domain from the URL
            domain = urlparse(source['link']).netloc
            # Format and display the domain and title
            st.markdown(f"- **{domain}**: [{source['title']}]({source['link']})", unsafe_allow_html=True)
    else:
        st.write("No sources available.")

# ---------------------------------------------------------------------------- #
# Sentiment Analysis Function
# ---------------------------------------------------------------------------- #

# Download the VADER lexicon for sentiment analysis
nltk.download('vader_lexicon')

# Initialize the Sentiment Intensity Analyzer
sid = SentimentIntensityAnalyzer()

def sentiment_analysis(text):
    # Split the text into sentences
    sentences = [sentence.strip() for sentence in text.split('.') if sentence]

    # Create a DataFrame to hold the content and sentiment scores
    df = pd.DataFrame(sentences, columns=['content'])

    # Calculate sentiment scores for each sentence
    df['sentiment_scores'] = df['content'].apply(lambda x: sid.polarity_scores(x))

    # Split sentiment_scores into separate columns
    df = pd.concat([df.drop(['sentiment_scores'], axis=1), df['sentiment_scores'].apply(pd.Series)], axis=1)

    # Determine the dominant sentiment and its confidence
    df['dominant_sentiment'] = df[['neg', 'neu', 'pos']].idxmax(axis=1)
    df['confidence'] = df[['neg', 'neu', 'pos']].max(axis=1)

    return df

# ---------------------------------------------------------------------------- #
# Advanced Analysis
# ---------------------------------------------------------------------------- #

def advanced_analysis(query, response_message):
    analysis_prompt = f"Analyze the question of the user '{query}', analyze the response '{response_message}' and provide a JSON that includes: the user's intent, up to four follow-up questions, the entities in the response."
    analysis_response = ai.prompt(message=analysis_prompt)
    return analysis_response   

def parse_analysis(analysis_message):
    try:
        # Check if the expected delimiter 'Here is the JSON:' is present
        if 'JSON:\n' in analysis_message:
            json_str = analysis_message.split('JSON:\n')[1].strip()
            analysis_data = json.loads(json_str)
            return analysis_data
        else:
            # If the delimiter is not found, return a default error message or handle it
            return {"error": "JSON delimiter not found in response"}
    except json.JSONDecodeError as e:
        # Handle cases where JSON decoding fails
        return {"error": "Failed to decode JSON", "details": str(e)}

        
# ---------------------------------------------------------------------------- #
# Main Function
# ---------------------------------------------------------------------------- #


def main():
    # Path to the image
    image_path = 'img/meta-ai-logo.png'  # Replace with your image's filename and extension

    # Create two columns
    col1, col2 = st.columns([1, 2])  # Adjust the ratio as needed for your layout

    # Use the first column to display the image
    with col1:
        st.image(image_path, width=60)

    # Use the second column to display the title and other content
    with col2:
        st.title("Meta AI SEO Tool")    

    # User input
    user_query = st.text_area("Enter your query:", height=150)
    submit_button = st.button("Analyze Query")

    if submit_button and user_query:
        # Fetching response from Meta AI
        response = fetch_response(user_query)
        msg = response.get('message', 'No response message.')    
        # Write response
        st.write(msg)

        # Run sentiment analysis
        df_sentiment = sentiment_analysis(msg)

        # Advanced analysis with second AI call
        advanced_response = advanced_analysis(user_query, msg)
        advanced_msg = advanced_response.get('message', 'No advanced analysis available.')
        
        # Parse the advanced analysis response
        analysis_data = parse_analysis(advanced_msg)
        
        if "error" in analysis_data:
            st.error("Error in analysis: " + analysis_data["error"])
            if "details" in analysis_data:
                st.error("Details: " + analysis_data["details"])
        else:
            # Display parsed data in a collapsible section
            with st.expander("Show Advanced Analysis"):
                st.write("### User Intent")
                st.write(analysis_data['user_intent'])
                st.write("### Follow-up Questions")
                for question in analysis_data['follow_up_questions']:
                    st.write("- " + question)
                st.write("### Identified Entities")
                for entity_type, entities in analysis_data['entities'].items():
                    st.write(f"**{entity_type.capitalize()}**: {', '.join(entities)}")
        
        # Display the sentiment in a collapsible section
        with st.expander("Show Sentiment"):
            # Display negative sentence locations
            fig = px.scatter(df_sentiment, y='dominant_sentiment', color='dominant_sentiment', size='confidence', 
                             hover_data=['content'], 
                             color_discrete_map={"neg": "firebrick", "neu": "navajowhite", "pos": "darkgreen"},
                             labels={'dominant_sentiment': 'Sentiment'},
                             title='Sentiment Analysis of the Response')
            fig.update_layout(width=800, height=300)
            st.plotly_chart(fig)
        
        # Display the AI response in a collapsible section
        with st.expander("Show Sources"):
            # Display sources with clickable links in a collapsible section
            display_sources(response.get('sources', []))
                    
if __name__ == "__main__":
    main()