app.py

import streamlit as st
import time
import requests
from streamlit.components.v1 import html
import os
from dotenv import load_dotenv

# New imports for voice input
import torchaudio
import numpy as np
import torch
from io import BytesIO
import hashlib
from audio_recorder_streamlit import audio_recorder
from transformers import pipeline




######################################
# Voice Input Helper Functions
######################################

@st.cache_resource
def load_voice_model():
    # Loading the Whisper model (which automatically detects both English and Urdu)
    return pipeline("automatic-speech-recognition", model="openai/whisper-base")

def process_audio(audio_bytes):
    waveform, sample_rate = torchaudio.load(BytesIO(audio_bytes))
    if waveform.shape[0] > 1:  # Convert stereo to mono
        waveform = torch.mean(waveform, dim=0, keepdim=True)
    if sample_rate != 16000:  # Resample to 16kHz if needed
        resampler = torchaudio.transforms.Resample(orig_freq=sample_rate, new_freq=16000)
        waveform = resampler(waveform)
    return {"raw": waveform.numpy().squeeze(), "sampling_rate": 16000}

def get_voice_transcription(state_key):
    """Display audio recorder for a given key.
       If new audio is recorded, transcribe it and update the session state.
    """
    if state_key not in st.session_state:
        st.session_state[state_key] = ""
    # Use a unique key for the recorder widget
    audio_bytes = audio_recorder(key=state_key + "_audio", 
                                 pause_threshold=0.8,
                                 text="Speak to type",
                                 recording_color="#e8b62c",
                                 neutral_color="#6aa36f")
    if audio_bytes:
        current_hash = hashlib.md5(audio_bytes).hexdigest()
        last_hash_key = state_key + "_last_hash"
        if st.session_state.get(last_hash_key, "") != current_hash:
            st.session_state[last_hash_key] = current_hash
            try:
                audio_input = process_audio(audio_bytes)
                whisper = load_voice_model()
                transcribed_text = whisper(audio_input)["text"]
                st.info(f"📝 Transcribed: {transcribed_text}")
                # Append (or set) new transcription
                st.session_state[state_key] += (" " + transcribed_text).strip()
                st.experimental_rerun()
            except Exception as e:
                st.error(f"Voice input error: {str(e)}")
    return st.session_state[state_key]

######################################
# Existing Game Helper Functions
######################################

@st.cache_resource
def get_help_agent():
    from transformers import pipeline
    # Using BlenderBot 400M Distill as the public conversational model (used elsewhere)
    return pipeline("conversational", model="facebook/blenderbot-400M-distill")

def inject_custom_css():
    st.markdown("""
    <style>
        @import url('https://fonts.googleapis.com/css2?family=Inter:wght@400;500;600;700&display=swap');
        @import url('https://cdnjs.cloudflare.com/ajax/libs/font-awesome/6.0.0/css/all.min.css');

        * { font-family: 'Inter', sans-serif; }
        body { background: linear-gradient(135deg, #f8f9fa 0%, #e9ecef 100%); }
        .title { font-size: 2.8rem !important; font-weight: 800 !important;
                 background: linear-gradient(45deg, #6C63FF, #3B82F6);
                 -webkit-background-clip: text; -webkit-text-fill-color: transparent;
                 text-align: center; margin: 1rem 0; letter-spacing: -1px; }
        .subtitle { font-size: 1.1rem !important; text-align: center;
                    color: #64748B !important; margin-bottom: 2.5rem; animation: fadeInSlide 1s ease; }
        .question-box { background: white; border-radius: 20px; padding: 2rem; margin: 1.5rem 0;
                        box-shadow: 0 10px 25px rgba(0,0,0,0.08); border: 1px solid #e2e8f0;
                        position: relative; transition: transform 0.2s ease; color: black; }
        .question-box:hover { transform: translateY(-3px); }
        .question-box::before { content: "🕹️"; position: absolute; left: -15px; top: -15px;
                                 background: white; border-radius: 50%; padding: 8px;
                                 box-shadow: 0 4px 6px rgba(0,0,0,0.1); font-size: 1.2rem; }
        .input-box { background: white; border-radius: 12px; padding: 1.5rem; margin: 1rem 0;
                     box-shadow: 0 4px 6px rgba(0,0,0,0.05); }
        .stTextInput input { border: 2px solid #e2e8f0 !important; border-radius: 10px !important;
                              padding: 12px 16px !important; transition: all 0.3s ease !important; }
        .stTextInput input:focus { border-color: #6C63FF !important;
                                   box-shadow: 0 0 0 3px rgba(108, 99, 255, 0.2) !important; }
        button { background: linear-gradient(45deg, #6C63FF, #3B82F6) !important;
                 color: white !important; border: none !important; border-radius: 10px !important;
                 padding: 12px 24px !important; font-weight: 600 !important;
                 transition: all 0.3s ease !important; }
        button:hover { transform: translateY(-2px); box-shadow: 0 5px 15px rgba(108, 99, 255, 0.3) !important; }
        .final-reveal { animation: fadeInUp 1s ease; font-size: 2.8rem;
                        background: linear-gradient(45deg, #6C63FF, #3B82F6);
                        -webkit-background-clip: text; -webkit-text-fill-color: transparent;
                        text-align: center; margin: 2rem 0; font-weight: 800; }
        .help-chat { background: rgba(255,255,255,0.9); backdrop-filter: blur(10px);
                     border-radius: 15px; padding: 1rem; margin: 1rem 0;
                     box-shadow: 0 8px 30px rgba(0,0,0,0.12); }
        @keyframes fadeInSlide { 0% { opacity: 0; transform: translateY(20px); }
                                 100% { opacity: 1; transform: translateY(0); } }
        @keyframes fadeInUp { 0% { opacity: 0; transform: translateY(30px); }
                              100% { opacity: 1; transform: translateY(0); } }
        .progress-bar { height: 6px; background: #e2e8f0; border-radius: 3px;
                        margin: 1.5rem 0; overflow: hidden; }
        .progress-fill { height: 100%; background: linear-gradient(90deg, #6C63FF, #3B82F6);
                         transition: width 0.5s ease; }
        .question-count { color: #6C63FF; font-weight: 600; font-size: 0.9rem; margin-bottom: 0.5rem; }


        /* Add these new styles for the info modal */
        .info-modal {
            position: fixed;
            top: 50%;
            left: 50%;
            transform: translate(-50%, -50%);
            background: white;
            padding: 2rem;
            border-radius: 20px;
            box-shadow: 0 10px 30px rgba(0,0,0,0.2);
            z-index: 1000;
            max-width: 600px;
            width: 90%;
            max-height: 80vh;
            overflow-y: auto;
        }
        .info-modal-backdrop {
            position: fixed;
            top: 0;
            left: 0;
            right: 0;
            bottom: 0;
            background: rgba(0,0,0,0.5);
            z-index: 999;
        }
        .tech-item {
            margin-bottom: 1.5rem;
            padding-bottom: 1.5rem;
            border-bottom: 1px solid #e2e8f0;
        }
        .tech-item:last-child {
            border-bottom: none;
            margin-bottom: 0;
            padding-bottom: 0;
        }
        .tech-title {
            color: #6C63FF;
            font-weight: 600;
            margin-bottom: 0.5rem;
            display: flex;
            align-items: center;
            gap: 0.5rem;
        }
        .close-modal-btn {
            position: absolute;
            top: 1rem;
            right: 1rem;
            background: none;
            border: none;
            font-size: 1.5rem;
            cursor: pointer;
            color: #64748B;
        }
        
    </style>
    """, unsafe_allow_html=True)

def show_confetti():
    html("""
    <canvas id="confetti-canvas" class="confetti"></canvas>
    <script src="https://cdn.jsdelivr.net/npm/[email protected]/dist/confetti.browser.min.js"></script>
    <script>
    const count = 200;
    const defaults = {
        origin: { y: 0.7 },
        zIndex: 1050
    };
    function fire(particleRatio, opts) {
        confetti(Object.assign({}, defaults, opts, {
            particleCount: Math.floor(count * particleRatio)
        }));
    }
    fire(0.25, { spread: 26, startVelocity: 55 });
    fire(0.2, { spread: 60 });
    fire(0.35, { spread: 100, decay: 0.91, scalar: 0.8 });
    fire(0.1, { spread: 120, startVelocity: 25, decay: 0.92, scalar: 1.2 });
    fire(0.1, { spread: 120, startVelocity: 45 });
    </script>
    """)

def ask_llama(conversation_history, category, is_final_guess=False):
    api_url = "https://api.groq.com/openai/v1/chat/completions"
    headers = {
        "Authorization": "Bearer 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",
        "Content-Type": "application/json"
    }
    system_prompt = f"""You're playing 20 questions to guess a {category}. Follow these rules:
1. Ask strategic, non-repeating yes/no questions that narrow down possibilities
2. Consider all previous answers carefully before asking next question
3. If you're very confident (80%+ sure), respond with "Final Guess: [your guess]"
4. For places: ask about continent, climate, famous landmarks, country, city or population
5. For people: ask about fictional or real, profession, gender, alive/dead, nationality, or fame
6. For objects: ask about size, color, usage, material, or where it's found
7. Never repeat questions and always make progress toward guessing"""
    if is_final_guess:
        prompt = f"""Based on these answers about a {category}, provide ONLY your final guess with no extra text:
{conversation_history}"""
    else:
        prompt = "Ask your next strategic yes/no question that will best narrow down the possibilities."
    messages = [
        {"role": "system", "content": system_prompt},
        *conversation_history,
        {"role": "user", "content": prompt}
    ]
    data = {
        "model": "llama-3.3-70b-versatile",
        "messages": messages,
        "temperature": 0.7 if is_final_guess else 0.8,
        "max_tokens": 100
    }
    try:
        response = requests.post(api_url, headers=headers, json=data)
        response.raise_for_status()
        return response.json()["choices"][0]["message"]["content"]
    except Exception as e:
        st.error(f"Error calling Llama API: {str(e)}")
        return "Could not generate question"

MISTRAL_API_KEY = "wm5eLl09b9I9cOxR3E9n5rrRr1CRQQjn"
def ask_help_agent(query):
    try:
        url = "https://api.mistral.ai/v1/chat/completions"
        headers = {
            "Authorization": f"Bearer {MISTRAL_API_KEY}",
            "Content-Type": "application/json"
        }
        system_message = "You are a friendly Chatbot."
        messages = [{"role": "system", "content": system_message}]
        if "help_conversation" in st.session_state:
            for msg in st.session_state.help_conversation:
                if msg.get("query"):
                    messages.append({"role": "user", "content": msg["query"]})
                if msg.get("response"):
                    messages.append({"role": "assistant", "content": msg["response"]})
        messages.append({"role": "user", "content": query})
        payload = {
            "model": "mistral-tiny",
            "messages": messages,
            "temperature": 0.7,
            "top_p": 0.95
        }
        response = requests.post(url, headers=headers, json=payload)
        if response.status_code == 200:
            result = response.json()
            return result["choices"][0]["message"]["content"]
        else:
            return f"API Error {response.status_code}: {response.text}"
    except Exception as e:
        return f"Error in help agent: {str(e)}"


def show_techniques_modal():
    # Use Streamlit's expander with markdown
    with st.expander("ℹ️ Project Techniques & Limitations", expanded=True):
        st.markdown("""
        **AI Models Used:**
        1. Groq Llama 3.3-70B - For generating strategic questions and final guesses
        2. Mistral Tiny - Powers the help chat assistant  
        3. OpenAI Whisper - Converts speech to text in real-time
        4. Hard Prompt Tuning - Carefully engineered prompts to optimize model performance

        **Known Limitations:**
        1. Voice input may take 5-10 seconds to process sentences, which is fine. (Whisper model trained on long sentences, so it's very accurate, but good for a sentence, but not for a word)
        2. Single words (like "yes", "object") may take 10-20 minutes, which is irritating.
        3. Language Support - While Whisper understands and writes Urdu, but the game only supports English responses
        """)
        
        if st.button("Close", key="modal_close_btn"):
            pass  # The expander will automatically close
    
######################################
# Main Game Logic with Voice Integration
######################################

def main():
    inject_custom_css()
    st.markdown('<div class="title">KASOTI</div>', unsafe_allow_html=True)
    st.markdown('<div class="subtitle">AI-Powered Guessing Game Challenge</div>', unsafe_allow_html=True)

    if st.button("ℹ️ Project Techniques & Limitations", key="info_btn"):
        show_techniques_modal()
    
        
    if 'game_state' not in st.session_state:
        st.session_state.game_state = "start"
        st.session_state.questions = []
        st.session_state.current_q = 0
        st.session_state.answers = []
        st.session_state.conversation_history = []
        st.session_state.category = None
        st.session_state.final_guess = None
        st.session_state.help_conversation = []  # separate history for help agent

    # Start screen with enhanced layout
    if st.session_state.game_state == "start":
        with st.container():
            st.markdown("""
            <div class="question-box">
                <h3 style="color: #6C63FF; margin-bottom: 1.5rem;">🎮 Welcome to KASOTI</h3>
                <p style="line-height: 1.6; color: #64748B;">
                    Think of something and I'll try to guess it in 20 questions or less!<br>
                    Choose from these categories:
                </p>
                <div style="display: grid; gap: 1rem; margin: 2rem 0;">
                    <div style="padding: 1.5rem; background: #f8f9fa; border-radius: 12px;">
                        <h4 style="margin: 0; color: #6C63FF;">🧑 Person</h4>
                        <p style="margin: 0.5rem 0 0; color: #64748B;">Celebrity, fictional character, historical figure</p>
                    </div>
                    <div style="padding: 1.5rem; background: #f8f9fa; border-radius: 12px;">
                        <h4 style="margin: 0; color: #6C63FF;">🌍 Place</h4>
                        <p style="margin: 0.5rem 0 0; color: #64748B;">City, country, landmark, geographical location</p>
                    </div>
                    <div style="padding: 1.5rem; background: #f8f9fa; border-radius: 12px;">
                        <h4 style="margin: 0; color: #6C63FF;">🎯 Object</h4>
                        <p style="margin: 0.5rem 0 0; color: #64748B;">Everyday item, tool, vehicle</p>
                    </div>
                </div>
            </div>
            """, unsafe_allow_html=True)

        with st.form("start_form"):
            # --- Voice Input for Category ---
            st.markdown("#### Use Voice (English/Urdu) for Category Input")
            voice_category = get_voice_transcription("voice_category")
            # The text input now defaults to any spoken words
            category_input = st.text_input("Enter category (person/place/object):", 
                                           value=voice_category.strip(), 
                                           key="category_input").strip().lower()
            if st.form_submit_button("Start Game"):
                if not category_input:
                    st.error("Please enter a category!")
                elif category_input not in ["person", "place", "object"]:
                    st.error("Please enter either 'person', 'place', or 'object'!")
                else:
                    st.session_state.category = category_input
                    first_question = ask_llama([
                        {"role": "user", "content": "Ask your first strategic yes/no question."}
                    ], category_input)
                    st.session_state.questions = [first_question]
                    st.session_state.conversation_history = [
                        {"role": "assistant", "content": first_question}
                    ]
                    st.session_state.game_state = "gameplay"
                    st.experimental_rerun()

    # Gameplay screen with progress bar
    elif st.session_state.game_state == "gameplay":
        with st.container():
            progress = (st.session_state.current_q + 1) / 20
            st.markdown(f"""
            <div class="question-count">QUESTION {st.session_state.current_q + 1} OF 20</div>
            <div class="progress-bar">
                <div class="progress-fill" style="width: {progress * 100}%"></div>
            </div>
            """, unsafe_allow_html=True)
            current_question = st.session_state.questions[st.session_state.current_q]
            st.markdown(f'''
            <div class="question-box">
                <div style="display: flex; align-items: center; gap: 1rem; margin-bottom: 1.5rem;">
                    <div style="background: #6C63FF; width: 40px; height: 40px; border-radius: 50%; 
                            display: flex; align-items: center; justify-content: center; color: white;">
                        <i class="fas fa-robot"></i>
                    </div>
                    <h3 style="margin: 0; color: #1E293B;">AI Question</h3>
                </div>
                <p style="font-size: 1.1rem; line-height: 1.6; color: #1E293B;">{current_question}</p>
            </div>
            ''', unsafe_allow_html=True)
        if "Final Guess:" in current_question:
            st.session_state.final_guess = current_question.split("Final Guess:")[1].strip()
            st.session_state.game_state = "confirm_guess"
            st.experimental_rerun()
        with st.form("answer_form"):
            # --- Voice Input for Answer ---
            st.markdown("#### Use Voice (English/Urdu) for Your Answer")
            voice_answer = get_voice_transcription("voice_answer")
            answer_input = st.text_input("Your answer (yes/no/both):",
                                          value=voice_answer.strip(),
                                          key=f"answer_{st.session_state.current_q}").strip().lower()
            if st.form_submit_button("Submit"):
                if answer_input not in ["yes", "no", "both"]:
                    st.error("Please answer with 'yes', 'no', or 'both'!")
                else:
                    st.session_state.answers.append(answer_input)
                    st.session_state.conversation_history.append(
                        {"role": "user", "content": answer_input}
                    )
                    next_response = ask_llama(
                        st.session_state.conversation_history,
                        st.session_state.category
                    )
                    if "Final Guess:" in next_response:
                        st.session_state.final_guess = next_response.split("Final Guess:")[1].strip()
                        st.session_state.game_state = "confirm_guess"
                    else:
                        st.session_state.questions.append(next_response)
                        st.session_state.conversation_history.append(
                            {"role": "assistant", "content": next_response}
                        )
                        st.session_state.current_q += 1
                        if st.session_state.current_q >= 20:
                            st.session_state.game_state = "result"
                    st.experimental_rerun()
        with st.expander("Need Help? Chat with AI Assistant"):
            # --- Voice Input for Help Query ---
            st.markdown("#### Use Voice (English/Urdu) for Help Query")
            voice_help = get_voice_transcription("voice_help")
            help_query = st.text_input("Enter your help query:", 
                                       value=voice_help.strip(), 
                                       key="help_query")
            if st.button("Send", key="send_help"):
                if help_query:
                    help_response = ask_help_agent(help_query)
                    st.session_state.help_conversation.append({"query": help_query, "response": help_response})
                else:
                    st.error("Please enter a query!")
            if st.session_state.help_conversation:
                for msg in st.session_state.help_conversation:
                    st.markdown(f"**You:** {msg['query']}")
                    st.markdown(f"**Help Assistant:** {msg['response']}")

    elif st.session_state.game_state == "confirm_guess":
        st.markdown(f'''
        <div class="question-box">
            <div style="display: flex; align-items: center; gap: 1rem; margin-bottom: 1.5rem;">
                <div style="background: #6C63FF; width: 40px; height: 40px; border-radius: 50%; 
                        display: flex; align-items: center; justify-content: center; color: white;">
                    <i class="fas fa-lightbulb"></i>
                </div>
                <h3 style="margin: 0; color: #1E293B;">AI's Final Guess</h3>
            </div>
            <p style="font-size: 1.2rem; line-height: 1.6; color: #1E293B;">
                Is it <strong style="color: #6C63FF;">{st.session_state.final_guess}</strong>?
            </p>
        </div>
        ''', unsafe_allow_html=True)
        with st.form("confirm_form"):
            confirm_input = st.text_input("Type your answer (yes/no/both):", key="confirm_input").strip().lower()
            if st.form_submit_button("Submit"):
                if confirm_input not in ["yes", "no", "both"]:
                    st.error("Please answer with 'yes', 'no', or 'both'!")
                else:
                    if confirm_input == "yes":
                        st.session_state.game_state = "result"
                        st.experimental_rerun()
                        st.stop()
                    else:
                        st.session_state.conversation_history.append(
                            {"role": "user", "content": "no"}
                        )
                        st.session_state.game_state = "gameplay"
                        next_response = ask_llama(
                            st.session_state.conversation_history,
                            st.session_state.category
                        )
                        st.session_state.questions.append(next_response)
                        st.session_state.conversation_history.append(
                            {"role": "assistant", "content": next_response}
                        )
                        st.session_state.current_q += 1
                        st.experimental_rerun()

    elif st.session_state.game_state == "result":
        if not st.session_state.final_guess:
            qa_history = "\n".join(
                [f"Q{i+1}: {q}\nA: {a}"
                 for i, (q, a) in enumerate(zip(st.session_state.questions, st.session_state.answers))]
            )
            final_guess = ask_llama(
                [{"role": "user", "content": qa_history}],
                st.session_state.category,
                is_final_guess=True
            )
            st.session_state.final_guess = final_guess.split("Final Guess:")[-1].strip()
        show_confetti()
        st.markdown(f'<div class="final-reveal">🎉 It\'s...</div>', unsafe_allow_html=True)
        time.sleep(1)
        st.markdown(f'<div class="final-reveal" style="font-size:3.5rem;color:#6C63FF;">{st.session_state.final_guess}</div>',
                    unsafe_allow_html=True)
        st.markdown(f"<p style='text-align:center; color:#64748B;'>Guessed in {len(st.session_state.questions)} questions</p>",
                    unsafe_allow_html=True)
        if st.button("Play Again", key="play_again"):
            st.session_state.clear()
            st.experimental_rerun()

if __name__ == "__main__":
    main()