app.py

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

# Voice input dependencies
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():
    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:
        waveform = torch.mean(waveform, dim=0, keepdim=True)
    if sample_rate != 16000:
        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):
    if state_key not in st.session_state:
        st.session_state[state_key] = ""
    audio_bytes = audio_recorder(
        key=state_key + "_audio",
        pause_threshold=0.8,
        text="🎙️ Speak your message",
        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}")
                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]

######################################
# Game Functions & Styling
######################################

@st.cache_resource
def get_help_agent():
    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');
        * { font-family: 'Inter', sans-serif; }
        .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; }
        .subtitle { font-size: 1.1rem; text-align: center; color: #64748B; margin-bottom: 2.5rem; }
        .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; color: black; }
        .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; }
        button { background: linear-gradient(45deg, #6C63FF, #3B82F6) !important;
                 color: white !important; border-radius: 10px !important;
                 padding: 12px 24px !important; font-weight: 600; }
        .final-reveal { 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; }
    </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": f"Bearer {os.getenv('GROQ_API_KEY')}",
        "Content-Type": "application/json"
    }
    system_prompt = f"""You're playing 20 questions to guess a {category}. Rules:
1. Ask strategic, non-repeating yes/no questions to narrow down.
2. Use all previous answers smartly.
3. If you're 80%+ sure, say: Final Guess: [your guess]
4. For places: ask about continent, country, landmarks, etc.
5. For people: ask if real, profession, gender, etc.
6. For objects: ask about use, size, material, etc."""
    
    prompt = f"""Based on these answers about a {category}, provide ONLY your final guess with no extra text:
{conversation_history}""" if is_final_guess else "Ask your next smart yes/no question."

    messages = [{"role": "system", "content": system_prompt}]
    messages += conversation_history
    messages.append({"role": "user", "content": prompt})

    data = {
        "model": "llama-3-70b-8192",
        "messages": messages,
        "temperature": 0.8,
        "max_tokens": 100
    }

    try:
        res = requests.post(api_url, headers=headers, json=data)
        res.raise_for_status()
        return res.json()["choices"][0]["message"]["content"]
    except Exception as e:
        st.error(f"❌ LLaMA API error: {e}")
        return "..."

######################################
# Main App Logic Here (UI, Game Loop)
######################################

def main():
    load_dotenv()
    inject_custom_css()

    st.title("🎮 Guess It! - 20 Questions Game")
    st.markdown("<div class='subtitle'>Think of a person, place, or object. LLaMA will try to guess it!</div>", unsafe_allow_html=True)

    category = st.selectbox("Category of your secret:", ["Person", "Place", "Object"])
    
    if "conversation" not in st.session_state:
        st.session_state.conversation = []
        st.session_state.last_bot_msg = ""

    if st.button("🔄 Restart Game"):
        st.session_state.conversation = []
        st.session_state.last_bot_msg = ""
        st.rerun()

    if not st.session_state.conversation:
        st.session_state.last_bot_msg = ask_llama([], category)
        st.session_state.conversation.append({"role": "assistant", "content": st.session_state.last_bot_msg})

    st.markdown(f"<div class='question-box'><strong>LLaMA:</strong> {st.session_state.last_bot_msg}</div>", unsafe_allow_html=True)

    user_input = get_voice_transcription("voice_input") or st.text_input("💬 Your answer (yes/no/sometimes):")

    if st.button("Submit Answer") and user_input:
        st.session_state.conversation.append({"role": "user", "content": user_input})
        with st.spinner("Thinking..."):
            response = ask_llama(st.session_state.conversation, category)
        st.session_state.last_bot_msg = response
        st.session_state.conversation.append({"role": "assistant", "content": response})
        st.rerun()

    if st.button("🤔 Make Final Guess"):
        with st.spinner("Making final guess..."):
            final_guess = ask_llama(st.session_state.conversation, category, is_final_guess=True)
            st.markdown(f"<div class='final-reveal'>🤯 Final Guess: <strong>{final_guess}</strong></div>", unsafe_allow_html=True)
            show_confetti()

if __name__ == "__main__":
    main()