app.py

import warnings
import numpy as np
import pandas as pd
import os
import json
import random
import gradio as gr
import torch
from sklearn.preprocessing import OneHotEncoder
from transformers import AutoModelForSequenceClassification, AutoTokenizer, AutoModelForCausalLM, pipeline
from deap import base, creator, tools, algorithms
import nltk
from nltk.sentiment import SentimentIntensityAnalyzer
from nltk.tokenize import word_tokenize
from nltk.tag import pos_tag
from nltk.chunk import ne_chunk
from textblob import TextBlob
import matplotlib.pyplot as plt
import seaborn as sns

warnings.filterwarnings('ignore', category=FutureWarning, module='huggingface_hub.file_download')

# Download necessary NLTK data
nltk.download('vader_lexicon', quiet=True)
nltk.download('punkt', quiet=True)
nltk.download('averaged_perceptron_tagger', quiet=True)
nltk.download('maxent_ne_chunker', quiet=True)
nltk.download('words', quiet=True)

# Initialize Example Dataset (For Emotion Prediction)
data = {
    'context': [
        'I am overjoyed', 'I am deeply saddened', 'I am seething with rage', 'I am exhilarated', 'I am tranquil',
        'I am brimming with joy', 'I am grieving profoundly', 'I am at peace', 'I am frustrated beyond measure',
        'I am determined to succeed', 'I feel resentment burning within me', 'I am feeling glorious and triumphant',
        'I am motivated and inspired', 'I am utterly surprised', 'I am gripped by fear', 'I am trusting and open',
        'I feel a sense of disgust', 'I am optimistic and hopeful', 'I am pessimistic and gloomy', 'I feel bored and listless',
        'I am envious and jealous'
    ],
    'emotion': [
        'joy', 'sadness', 'anger', 'joy', 'calmness', 'joy', 'grief', 'calmness', 'anger',
        'determination', 'resentment', 'glory', 'motivation', 'surprise', 'fear', 'trust',
        'disgust', 'optimism', 'pessimism', 'boredom', 'envy'
    ]
}
df = pd.DataFrame(data)

# Encoding the contexts using One-Hot Encoding (memory-efficient)
try:
    encoder = OneHotEncoder(handle_unknown='ignore', sparse_output=True)
except TypeError:
    encoder = OneHotEncoder(handle_unknown='ignore', sparse=True)
contexts_encoded = encoder.fit_transform(df[['context']])

# Encoding emotions
emotions_target = pd.Categorical(df['emotion']).codes
emotion_classes = pd.Categorical(df['emotion']).categories

# Load pre-trained BERT model for emotion prediction
emotion_prediction_model = AutoModelForSequenceClassification.from_pretrained("bhadresh-savani/distilbert-base-uncased-emotion")
emotion_prediction_tokenizer = AutoTokenizer.from_pretrained("bhadresh-savani/distilbert-base-uncased-emotion")

# Load pre-trained large language model and tokenizer for response generation with increased context window
response_model_name = "gpt2-xl"
response_tokenizer = AutoTokenizer.from_pretrained(response_model_name)
response_model = AutoModelForCausalLM.from_pretrained(response_model_name)

# Set the pad token
response_tokenizer.pad_token = response_tokenizer.eos_token

# Enhanced Emotional States
emotions = {
    'joy': {'percentage': 10, 'motivation': 'positive and uplifting', 'intensity': 0},
    'sadness': {'percentage': 10, 'motivation': 'reflective and introspective', 'intensity': 0},
    'anger': {'percentage': 10, 'motivation': 'passionate and driven', 'intensity': 0},
    'fear': {'percentage': 10, 'motivation': 'cautious and protective', 'intensity': 0},
    'love': {'percentage': 10, 'motivation': 'affectionate and caring', 'intensity': 0},
    'surprise': {'percentage': 10, 'motivation': 'curious and intrigued', 'intensity': 0},
    'neutral': {'percentage': 40, 'motivation': 'balanced and composed', 'intensity': 0},
}

total_percentage = 100
emotion_history_file = 'emotion_history.json'
global conversation_history
conversation_history = []
max_history_length = 1000  # Increase the maximum history length

def load_historical_data(file_path=emotion_history_file):
    if os.path.exists(file_path):
        with open(file_path, 'r') as file:
            return json.load(file)
    return []

def save_historical_data(historical_data, file_path=emotion_history_file):
    with open(file_path, 'w') as file:
        json.dump(historical_data, file)

emotion_history = load_historical_data()

def update_emotion(emotion, percentage, intensity):
    emotions[emotion]['percentage'] += percentage
    emotions[emotion]['intensity'] = intensity

    # Normalize percentages
    total = sum(e['percentage'] for e in emotions.values())
    for e in emotions:
        emotions[e]['percentage'] = (emotions[e]['percentage'] / total) * 100

def normalize_context(context):
    return context.lower().strip()

creator.create("FitnessMulti", base.Fitness, weights=(-1.0, -0.5, -0.2))
creator.create("Individual", list, fitness=creator.FitnessMulti)

def evaluate(individual):
    emotion_values = individual[:len(emotions)]
    intensities = individual[len(emotions):]
    
    total_diff = abs(100 - sum(emotion_values))
    intensity_range = max(intensities) - min(intensities)
    emotion_balance = max(emotion_values) - min(emotion_values)
    
    return total_diff, intensity_range, emotion_balance

def evolve_emotions():
    toolbox = base.Toolbox()
    toolbox.register("attr_float", random.uniform, 0, 100)
    toolbox.register("attr_intensity", random.uniform, 0, 10)
    toolbox.register("individual", tools.initCycle, creator.Individual,
                     (toolbox.attr_float,) * len(emotions) +
                     (toolbox.attr_intensity,) * len(emotions), n=1)
    toolbox.register("population", tools.initRepeat, list, toolbox.individual)
    toolbox.register("mate", tools.cxTwoPoint)
    toolbox.register("mutate", tools.mutGaussian, mu=0, sigma=1, indpb=0.2)
    toolbox.register("select", tools.selNSGA2)
    toolbox.register("evaluate", evaluate)

    population = toolbox.population(n=100)
    algorithms.eaMuPlusLambda(population, toolbox, mu=50, lambda_=100, cxpb=0.7, mutpb=0.2, ngen=50,
                               stats=None, halloffame=None, verbose=False)

    best_individual = tools.selBest(population, k=1)[0]
    emotion_values = best_individual[:len(emotions)]
    intensities = best_individual[len(emotions):]

    for i, (emotion, data) in enumerate(emotions.items()):
        data['percentage'] = emotion_values[i]
        data['intensity'] = intensities[i]

   # Normalize percentages
    total = sum(e['percentage'] for e in emotions.values())
    for e in emotions:
        emotions[e]['percentage'] = (emotions[e]['percentage'] / total) * 100

def update_emotion_history(emotion, percentage, intensity, context):
    entry = {
        'emotion': emotion,
        'percentage': percentage,
        'intensity': intensity,
        'context': context,
        'timestamp': pd.Timestamp.now().isoformat()
    }
    emotion_history.append(entry)
    save_historical_data(emotion_history)

# Adding 443 features
additional_features = {}
for i in range(443):
    additional_features[f'feature_{i+1}'] = 0

def feature_transformations():
    global additional_features
    for feature in additional_features:
        additional_features[feature] += random.uniform(-1, 1)

def generate_response(input_text, ai_emotion, conversation_history):
    # Prepare a prompt based on the current emotion and input
    prompt = f"As an AI assistant, I am currently feeling {ai_emotion}. My response will reflect this emotional state. Human: {input_text}\nAI:"
    
    # Add conversation history to the prompt
    for entry in conversation_history[-100:]:  # Use last 100 entries for context
        prompt = f"Human: {entry['user']}\nAI: {entry['response']}\n" + prompt
    
    inputs = response_tokenizer(prompt, return_tensors="pt", padding=True, truncation=True, max_length=8192)
    
    # Adjust generation parameters based on emotion
    temperature = 0.7
    if ai_emotion == 'anger':
        temperature = 0.9  # More randomness for angry responses
    elif ai_emotion == 'joy':
        temperature = 0.5  # More focused responses for joyful state

    with torch.no_grad():
        response_ids = response_model.generate(
            inputs.input_ids,
            attention_mask=inputs.attention_mask,
            max_length=8192,
            num_return_sequences=1,
            no_repeat_ngram_size=2,
            do_sample=True,
            top_k=50,
            top_p=0.95,
            temperature=temperature,
            pad_token_id=response_tokenizer.eos_token_id
        )
    response = response_tokenizer.decode(response_ids[0], skip_special_tokens=True)
    
    # Extract only the AI's response
    response = response.split("AI:")[-1].strip()
    return response

def predict_emotion(context):
    inputs = emotion_prediction_tokenizer(context, return_tensors="pt", truncation=True, max_length=512)
    with torch.no_grad():
        outputs = emotion_prediction_model(**inputs)
    probabilities = torch.nn.functional.softmax(outputs.logits, dim=-1)
    predicted_class = torch.argmax(probabilities, dim=-1).item()
    emotion_labels = ["sadness", "joy", "love", "anger", "fear", "surprise"]
    return emotion_labels[predicted_class]

def sentiment_analysis(text):
    sia = SentimentIntensityAnalyzer()
    sentiment_scores = sia.polarity_scores(text)
    return sentiment_scores

def extract_entities(text):
    chunked = ne_chunk(pos_tag(word_tokenize(text)))
    entities = []
    for chunk in chunked:
        if hasattr(chunk, 'label'):
            entities.append(((' '.join(c[0] for c in chunk)), chunk.label()))
    return entities

def analyze_text_complexity(text):
    blob = TextBlob(text)
    return {
        'word_count': len(blob.words),
        'sentence_count': len(blob.sentences),
        'average_sentence_length': len(blob.words) / len(blob.sentences) if len(blob.sentences) > 0 else 0,
        'polarity': blob.sentiment.polarity,
        'subjectivity': blob.sentiment.subjectivity
    }

def visualize_emotions():
    emotions_df = pd.DataFrame([(e, d['percentage'], d['intensity']) for e, d in emotions.items()],
                               columns=['Emotion', 'Percentage', 'Intensity'])
    
    plt.figure(figsize=(12, 6))
    sns.barplot(x='Emotion', y='Percentage', data=emotions_df)
    plt.title('Current Emotional State of the AI')
    plt.xticks(rotation=45, ha='right')
    plt.tight_layout()
    plt.savefig('emotional_state.png')
    plt.close()

    return 'emotional_state.png'

def interactive_interface(input_text):
    global conversation_history
    try:
        evolve_emotions()
        predicted_emotion = predict_emotion(input_text)
        sentiment_scores = sentiment_analysis(input_text)
        entities = extract_entities(input_text)
        text_complexity = analyze_text_complexity(input_text)
        
        # Update AI's emotional state based on input
        update_emotion(predicted_emotion, random.uniform(5, 15), random.uniform(0, 10))
        
        # Determine AI's current dominant emotion
        ai_emotion = max(emotions, key=lambda e: emotions[e]['percentage'])
        
        # Generate response based on AI's emotion
        response = generate_response(input_text, ai_emotion, conversation_history)
        
        # Update conversation history
        conversation_history.append({
            'user': input_text,
            'response': response
        })
        
        # Trim conversation history if it exceeds the maximum length
        if len(conversation_history) > max_history_length:
            conversation_history = conversation_history[-max_history_length:]
        
        update_emotion_history(ai_emotion, emotions[ai_emotion]['percentage'], emotions[ai_emotion]['intensity'], input_text)
        feature_transformations()
        
        emotion_visualization = visualize_emotions()
        
        analysis_result = {
            'predicted_user_emotion': predicted_emotion,
            'sentiment_scores': sentiment_scores,
            'entities': entities,
            'text_complexity': text_complexity,
            'ai_emotion': ai_emotion,
            'ai_emotion_percentage': emotions[ai_emotion]['percentage'],
            'ai_emotion_intensity': emotions[ai_emotion]['intensity'],
            'emotion_visualization': emotion_visualization
        }
        
        return analysis_result
    except Exception as e:
        print(f"Error: {e}")
        return {'error': str(e)}

# Define the Gradio interface
demo = gr.Interface(
    fn=interactive_interface,
    inputs=gr.Textbox(label="Enter your message"),
    outputs=[
        gr.Textbox(label="Predicted User Emotion"),
        gr.Textbox(label="Sentiment Scores"),
        gr.Textbox(label="Extracted Entities"),
        gr.Textbox(label="Text Complexity"),
        gr.Textbox(label="AI Emotion"),
        gr.Textbox(label="AI Emotion Percentage"),
        gr.Textbox(label="AI Emotion Intensity"),
        gr.Image(label="Emotional State Visualization")
    ],
    title="Emotional AI Assistant",
    description="Interact with an AI assistant that adapts its responses based on its emotional state."
)

# Launch the Gradio interface
demo.launch()