app.py

import streamlit as st
import anthropic
import openai
import base64
from datetime import datetime
import plotly.graph_objects as go
import cv2
import glob
import json
import math
import os
import pytz
import random
import re
import requests
import streamlit.components.v1 as components
import textract
import time
import zipfile
from audio_recorder_streamlit import audio_recorder
from bs4 import BeautifulSoup
from collections import deque
from dotenv import load_dotenv
from gradio_client import Client, handle_file
from huggingface_hub import InferenceClient
from io import BytesIO
from moviepy.editor import VideoFileClip
from PIL import Image
from PyPDF2 import PdfReader
from urllib.parse import quote
from xml.etree import ElementTree as ET
from openai import OpenAI

# 1. Configuration and Setup
Site_Name = '🚲BikeAI🏆 Claude and GPT Multi-Agent Research AI'
title = "🚲BikeAI🏆 Claude and GPT Multi-Agent Research AI"
helpURL = 'https://huggingface.co/awacke1'
bugURL = 'https://huggingface.co/spaces/awacke1'
icons = '🚲🏆'

st.set_page_config(
    page_title=title,
    page_icon=icons,
    layout="wide",
    initial_sidebar_state="auto",
    menu_items={
        'Get Help': helpURL,
        'Report a bug': bugURL,
        'About': title
    }
)

# 2. Load environment variables and initialize clients
load_dotenv()

# OpenAI setup
openai.api_key = os.getenv('OPENAI_API_KEY')
if openai.api_key == None:
    openai.api_key = st.secrets['OPENAI_API_KEY']

openai_client = OpenAI(
    api_key=os.getenv('OPENAI_API_KEY'),
    organization=os.getenv('OPENAI_ORG_ID')
)

# 3. Claude setup
anthropic_key = os.getenv("ANTHROPIC_API_KEY_3")
if anthropic_key == None:
    anthropic_key = st.secrets["ANTHROPIC_API_KEY"]
claude_client = anthropic.Anthropic(api_key=anthropic_key)

# 4. Initialize session states
if 'transcript_history' not in st.session_state:
    st.session_state.transcript_history = []
if "chat_history" not in st.session_state:
    st.session_state.chat_history = []
if "openai_model" not in st.session_state:
    st.session_state["openai_model"] = "gpt-4o-2024-05-13"
if "messages" not in st.session_state:
    st.session_state.messages = []
if 'last_voice_input' not in st.session_state:
    st.session_state.last_voice_input = ""

# 5. # HuggingFace setup
API_URL = os.getenv('API_URL')
HF_KEY = os.getenv('HF_KEY')
MODEL1 = "meta-llama/Llama-2-7b-chat-hf"
MODEL2 = "openai/whisper-small.en"

headers = {
    "Authorization": f"Bearer {HF_KEY}",
    "Content-Type": "application/json"
}

# Initialize session states
if "chat_history" not in st.session_state:
    st.session_state.chat_history = []
if "openai_model" not in st.session_state:
    st.session_state["openai_model"] = "gpt-4o-2024-05-13"
if "messages" not in st.session_state:
    st.session_state.messages = []

# Custom CSS
st.markdown("""
<style>
    .main {
        background: linear-gradient(to right, #1a1a1a, #2d2d2d);
        color: #ffffff;
    }
    .stMarkdown {
        font-family: 'Helvetica Neue', sans-serif;
    }
    .category-header {
        background: linear-gradient(45deg, #2b5876, #4e4376);
        padding: 20px;
        border-radius: 10px;
        margin: 10px 0;
    }
    .scene-card {
        background: rgba(0,0,0,0.3);
        padding: 15px;
        border-radius: 8px;
        margin: 10px 0;
        border: 1px solid rgba(255,255,255,0.1);
    }
    .media-gallery {
        display: grid;
        gap: 1rem;
        padding: 1rem;
    }
    .bike-card {
        background: rgba(255,255,255,0.05);
        border-radius: 10px;
        padding: 15px;
        transition: transform 0.3s;
    }
    .bike-card:hover {
        transform: scale(1.02);
    }
</style>
""", unsafe_allow_html=True)


# Bike Collections
bike_collections = {
    "Celestial Collection 🌌": {
        "Eclipse Vaulter": {
            "prompt": """Cinematic shot of a sleek black mountain bike silhouetted against a total solar eclipse. 
                     The corona creates an ethereal halo effect, with lens flares accentuating key points of the frame.
                     Dynamic composition shows the bike mid-leap, with stardust particles trailing behind.
                     Camera angle: Low angle, wide shot
                     Lighting: Dramatic rim lighting from eclipse
                     Color palette: Deep purples, cosmic blues, corona gold""",
            "emoji": "🌑"
        },
        "Starlight Leaper": {
            "prompt": """A black bike performing an epic leap under a vast Milky Way galaxy.
                     Shimmering stars blanket the sky while the bike's wheels leave a trail of stardust.
                     Camera angle: Wide-angle upward shot
                     Lighting: Natural starlight with subtle rim lighting
                     Color palette: Deep blues, silver highlights, cosmic purples""",
            "emoji": "✨"
        },
        "Moonlit Hopper": {
            "prompt": """A sleek black bike mid-hop over a moonlit meadow,
                     the full moon illuminating the misty surroundings. Fireflies dance around the bike,
                     and soft shadows create a serene yet dynamic atmosphere.
                     Camera angle: Side profile with slight low angle
                     Lighting: Soft moonlight with atmospheric fog
                     Color palette: Silver blues, soft whites, deep shadows""",
            "emoji": "🌙"
        }
    },
    "Nature-Inspired Collection 🌲": {
        "Shadow Grasshopper": {
            "prompt": """A black bike jumping between forest paths,
                     with dappled sunlight streaming through the canopy. Shadows dance on the bike's frame
                     as it soars above mossy logs.
                     Camera angle: Through-the-trees tracking shot
                     Lighting: Natural forest lighting with sun rays
                     Color palette: Forest greens, golden sunlight, deep shadows""",
            "emoji": "🦗"
        },
        "Onyx Leapfrog": {
            "prompt": """A bike with obsidian-black finish jumping over a sparkling creek,
                     the reflection on the water broken into ripples by the leap. The surrounding forest
                     is vibrant with greens and browns.
                     Camera angle: Low angle from water level
                     Lighting: Golden hour side lighting
                     Color palette: Deep blacks, water blues, forest greens""",
            "emoji": "🐸"
        }
    }
}


# Helper Functions
def generate_filename(prompt, file_type):
    """Generate a safe filename using the prompt and file type."""
    central = pytz.timezone('US/Central')
    safe_date_time = datetime.now(central).strftime("%m%d_%H%M")
    replaced_prompt = re.sub(r'[<>:"/\\|?*\n]', ' ', prompt)
    safe_prompt = re.sub(r'\s+', ' ', replaced_prompt).strip()[:230]
    return f"{safe_date_time}_{safe_prompt}.{file_type}"




# Function to create and save a file (and avoid the black hole of lost data 🕳)
def create_file(filename, prompt, response, should_save=True):
    if not should_save:
        return
    with open(filename, 'w', encoding='utf-8') as file:
        file.write(prompt + "\n\n" + response)



def create_and_save_file(content, file_type="md", prompt=None, is_image=False, should_save=True):
    """Create and save file with proper handling of different types."""
    if not should_save:
        return None
    filename = generate_filename(prompt if prompt else content, file_type)
    with open(filename, "w", encoding="utf-8") as f:
        if is_image:
            f.write(content)
        else:
            f.write(prompt + "\n\n" + content if prompt else content)
    return filename

def get_download_link(file_path):
    """Create download link for file."""
    with open(file_path, "rb") as file:
        contents = file.read()
    b64 = base64.b64encode(contents).decode()
    return f'<a href="data:file/txt;base64,{b64}" download="{os.path.basename(file_path)}">Download {os.path.basename(file_path)}📂</a>'

@st.cache_resource
def SpeechSynthesis(result):
    """HTML5 Speech Synthesis."""
    documentHTML5 = f'''
    <!DOCTYPE html>
    <html>
    <head>
        <title>Read It Aloud</title>
        <script type="text/javascript">
            function readAloud() {{
                const text = document.getElementById("textArea").value;
                const speech = new SpeechSynthesisUtterance(text);
                window.speechSynthesis.speak(speech);
            }}
        </script>
    </head>
    <body>
        <h1>🔊 Read It Aloud</h1>
        <textarea id="textArea" rows="10" cols="80">{result}</textarea>
        <br>
        <button onclick="readAloud()">🔊 Read Aloud</button>
    </body>
    </html>
    '''
    components.html(documentHTML5, width=1280, height=300)

# Media Processing Functions
def process_image(image_input, user_prompt):
    """Process image with GPT-4o vision."""
    if isinstance(image_input, str):
        with open(image_input, "rb") as image_file:
            image_input = image_file.read()
            
    base64_image = base64.b64encode(image_input).decode("utf-8")
    
    response = openai_client.chat.completions.create(
        model=st.session_state["openai_model"],
        messages=[
            {"role": "system", "content": "You are a helpful assistant that responds in Markdown."},
            {"role": "user", "content": [
                {"type": "text", "text": user_prompt},
                {"type": "image_url", "image_url": {
                    "url": f"data:image/png;base64,{base64_image}"
                }}
            ]}
        ],
        temperature=0.0,
    )
    
    return response.choices[0].message.content

def process_audio(audio_input, text_input=''):
    """Process audio with Whisper and GPT."""
    if isinstance(audio_input, str):
        with open(audio_input, "rb") as file:
            audio_input = file.read()

    transcription = openai_client.audio.transcriptions.create(
        model="whisper-1",
        file=audio_input,
    )
    
    st.session_state.messages.append({"role": "user", "content": transcription.text})
    
    with st.chat_message("assistant"):
        st.markdown(transcription.text)
        SpeechSynthesis(transcription.text)
        
        filename = generate_filename(transcription.text, "wav")
        create_and_save_file(audio_input, "wav", transcription.text, True)

def process_video(video_path, seconds_per_frame=1):
    """Process video files for frame extraction and audio."""
    base64Frames = []
    video = cv2.VideoCapture(video_path)
    total_frames = int(video.get(cv2.CAP_PROP_FRAME_COUNT))
    fps = video.get(cv2.CAP_PROP_FPS)
    frames_to_skip = int(fps * seconds_per_frame)
    
    for frame_idx in range(0, total_frames, frames_to_skip):
        video.set(cv2.CAP_PROP_POS_FRAMES, frame_idx)
        success, frame = video.read()
        if not success:
            break
        _, buffer = cv2.imencode(".jpg", frame)
        base64Frames.append(base64.b64encode(buffer).decode("utf-8"))
    
    video.release()
    
    # Extract audio
    base_video_path = os.path.splitext(video_path)[0]
    audio_path = f"{base_video_path}.mp3"
    try:
        video_clip = VideoFileClip(video_path)
        video_clip.audio.write_audiofile(audio_path)
        video_clip.close()
    except:
        st.warning("No audio track found in video")
        audio_path = None
    
    return base64Frames, audio_path

def process_video_with_gpt(video_input, user_prompt):
    """Process video with GPT-4o vision."""
    base64Frames, audio_path = process_video(video_input)
    
    response = openai_client.chat.completions.create(
        model=st.session_state["openai_model"],
        messages=[
            {"role": "system", "content": "Analyze the video frames and provide a detailed description."},
            {"role": "user", "content": [
                {"type": "text", "text": user_prompt},
                *[{"type": "image_url", "image_url": {"url": f"data:image/jpeg;base64,{frame}"}}
                  for frame in base64Frames]
            ]}
        ]
    )
    
    return response.choices[0].message.content


def extract_urls(text):
    try:
        date_pattern = re.compile(r'### (\d{2} \w{3} \d{4})')
        abs_link_pattern = re.compile(r'\[(.*?)\]\((https://arxiv\.org/abs/\d+\.\d+)\)')
        pdf_link_pattern = re.compile(r'\[⬇️\]\((https://arxiv\.org/pdf/\d+\.\d+)\)')
        title_pattern = re.compile(r'### \d{2} \w{3} \d{4} \| \[(.*?)\]')
        date_matches = date_pattern.findall(text)
        abs_link_matches = abs_link_pattern.findall(text)
        pdf_link_matches = pdf_link_pattern.findall(text)
        title_matches = title_pattern.findall(text)

        # markdown with the extracted fields
        markdown_text = ""
        for i in range(len(date_matches)):
            date = date_matches[i]
            title = title_matches[i]
            abs_link = abs_link_matches[i][1]
            pdf_link = pdf_link_matches[i]
            markdown_text += f"**Date:** {date}\n\n"
            markdown_text += f"**Title:** {title}\n\n"
            markdown_text += f"**Abstract Link:** [{abs_link}]({abs_link})\n\n"
            markdown_text += f"**PDF Link:** [{pdf_link}]({pdf_link})\n\n"
            markdown_text += "---\n\n"
        return markdown_text
    
    except:
        st.write('.')
        return ''


def search_arxiv(query):
    
    st.write("Performing AI Lookup...")
    client = Client("awacke1/Arxiv-Paper-Search-And-QA-RAG-Pattern")
    
    result1 = client.predict(
        prompt=query,
        llm_model_picked="mistralai/Mixtral-8x7B-Instruct-v0.1",
        stream_outputs=True,
        api_name="/ask_llm"
    )
    st.markdown("### Mixtral-8x7B-Instruct-v0.1 Result")
    st.markdown(result1)

    result2 = client.predict(
        prompt=query,
        llm_model_picked="mistralai/Mistral-7B-Instruct-v0.2",
        stream_outputs=True,
        api_name="/ask_llm"
    )
    st.markdown("### Mistral-7B-Instruct-v0.2 Result")
    st.markdown(result2)
    combined_result = f"{result1}\n\n{result2}"
    return combined_result

    #return responseall


# Function to generate a filename based on prompt and time (because names matter 🕒)
def generate_filename(prompt, file_type):
    central = pytz.timezone('US/Central')
    safe_date_time = datetime.now(central).strftime("%m%d_%H%M")
    safe_prompt = re.sub(r'\W+', '_', prompt)[:90]
    return f"{safe_date_time}_{safe_prompt}.{file_type}"

# Function to create and save a file (and avoid the black hole of lost data 🕳)
def create_file(filename, prompt, response):
    with open(filename, 'w', encoding='utf-8') as file:
        file.write(prompt + "\n\n" + response)

    
def perform_ai_lookup(query):
    start_time = time.strftime("%Y-%m-%d %H:%M:%S")
    client = Client("awacke1/Arxiv-Paper-Search-And-QA-RAG-Pattern")
    response1 = client.predict(
        query,
        20,
        "Semantic Search",
        "mistralai/Mixtral-8x7B-Instruct-v0.1",
        api_name="/update_with_rag_md"
    )
    Question = '### 🔎 ' + query + '\r\n'  # Format for markdown display with links
    References =  response1[0]  
    ReferenceLinks = extract_urls(References)

    RunSecondQuery = True
    results=''
    if RunSecondQuery:
        # Search 2 - Retrieve the Summary with Papers Context and Original Query
        response2 = client.predict(
            query,
            "mistralai/Mixtral-8x7B-Instruct-v0.1",
            True,
            api_name="/ask_llm"
        )
        if len(response2) > 10:
            Answer = response2
            SpeechSynthesis(Answer)
            # Restructure results to follow format of Question, Answer, References, ReferenceLinks
            results = Question + '\r\n' + Answer + '\r\n' + References + '\r\n' + ReferenceLinks
            st.markdown(results)

    st.write('🔍Run of Multi-Agent System Paper Summary Spec is Complete')
    end_time = time.strftime("%Y-%m-%d %H:%M:%S")
    start_timestamp = time.mktime(time.strptime(start_time, "%Y-%m-%d %H:%M:%S"))
    end_timestamp = time.mktime(time.strptime(end_time, "%Y-%m-%d %H:%M:%S"))
    elapsed_seconds = end_timestamp - start_timestamp
    st.write(f"Start time: {start_time}")
    st.write(f"Finish time: {end_time}")
    st.write(f"Elapsed time: {elapsed_seconds:.2f} seconds")

    
    filename = generate_filename(query, "md")
    create_file(filename, query, results)
    return results

# Chat Processing Functions
def process_with_gpt(text_input):
    """Process text with GPT-4o."""
    if text_input:
        st.session_state.messages.append({"role": "user", "content": text_input})
        
        with st.chat_message("user"):
            st.markdown(text_input)
        
        with st.chat_message("assistant"):
            completion = openai_client.chat.completions.create(
                model=st.session_state["openai_model"],
                messages=[
                    {"role": m["role"], "content": m["content"]}
                    for m in st.session_state.messages
                ],
                stream=False
            )
            return_text = completion.choices[0].message.content
            st.write("GPT-4o: " + return_text)
            
            #filename = generate_filename(text_input, "md")
            filename = generate_filename("GPT-4o: " + return_text, "md")
            create_file(filename, text_input, return_text)
            st.session_state.messages.append({"role": "assistant", "content": return_text})
            return return_text

def process_with_claude(text_input):
    """Process text with Claude."""
    if text_input:

        with st.chat_message("user"):
            st.markdown(text_input)

        with st.chat_message("assistant"):
            response = claude_client.messages.create(
                model="claude-3-sonnet-20240229",
                max_tokens=1000,
                messages=[
                    {"role": "user", "content": text_input}
                ]
            )
            response_text = response.content[0].text
            st.write("Claude: " + response_text)
            
            #filename = generate_filename(text_input, "md")
            filename = generate_filename("Claude: " + response_text, "md")
            create_file(filename, text_input, response_text)
            
            st.session_state.chat_history.append({
                "user": text_input,
                "claude": response_text
            })
            return response_text

# File Management Functions
def load_file(file_name):
    """Load file content."""
    with open(file_name, "r", encoding='utf-8') as file:
        content = file.read()
    return content

def create_zip_of_files(files):
    """Create zip archive of files."""
    zip_name = "all_files.zip"
    with zipfile.ZipFile(zip_name, 'w') as zipf:
        for file in files:
            zipf.write(file)
    return zip_name



def get_media_html(media_path, media_type="video", width="100%"):
    """Generate HTML for media player."""
    media_data = base64.b64encode(open(media_path, 'rb').read()).decode()
    if media_type == "video":
        return f'''
        <video width="{width}" controls autoplay muted loop>
            <source src="data:video/mp4;base64,{media_data}" type="video/mp4">
            Your browser does not support the video tag.
        </video>
        '''
    else:  # audio
        return f'''
        <audio controls style="width: {width};">
            <source src="data:audio/mpeg;base64,{media_data}" type="audio/mpeg">
            Your browser does not support the audio element.
        </audio>
        '''

def create_media_gallery():
    """Create the media gallery interface."""
    st.header("🎬 Media Gallery")
    
    tabs = st.tabs(["🖼️ Images", "🎵 Audio", "🎥 Video", "🎨 Scene Generator"])
    
    with tabs[0]:
        image_files = glob.glob("*.png") + glob.glob("*.jpg")
        if image_files:
            num_cols = st.slider("Number of columns", 1, 5, 3)
            cols = st.columns(num_cols)
            for idx, image_file in enumerate(image_files):
                with cols[idx % num_cols]:
                    img = Image.open(image_file)
                    st.image(img, use_container_width=True)
                    
                    # Add GPT vision analysis option
                    if st.button(f"Analyze {os.path.basename(image_file)}"):
                        analysis = process_image(image_file, 
                                              "Describe this image in detail and identify key elements.")
                        st.markdown(analysis)
    
    with tabs[1]:
        audio_files = glob.glob("*.mp3") + glob.glob("*.wav")
        for audio_file in audio_files:
            with st.expander(f"🎵 {os.path.basename(audio_file)}"):
                st.markdown(get_media_html(audio_file, "audio"), unsafe_allow_html=True)
                if st.button(f"Transcribe {os.path.basename(audio_file)}"):
                    with open(audio_file, "rb") as f:
                        transcription = process_audio(f)
                        st.write(transcription)
    
    with tabs[2]:
        video_files = glob.glob("*.mp4")
        for video_file in video_files:
            with st.expander(f"🎥 {os.path.basename(video_file)}"):
                st.markdown(get_media_html(video_file, "video"), unsafe_allow_html=True)
                if st.button(f"Analyze {os.path.basename(video_file)}"):
                    analysis = process_video_with_gpt(video_file, 
                                                    "Describe what's happening in this video.")
                    st.markdown(analysis)
    
    with tabs[3]:
        for collection_name, bikes in bike_collections.items():
            st.subheader(collection_name)
            cols = st.columns(len(bikes))
            
            for idx, (bike_name, details) in enumerate(bikes.items()):
                with cols[idx]:
                    st.markdown(f"""
                    <div class='bike-card'>
                        <h3>{details['emoji']} {bike_name}</h3>
                        <p>{details['prompt']}</p>
                    </div>
                    """, unsafe_allow_html=True)
                    
                    if st.button(f"Generate {bike_name} Scene"):
                        prompt = details['prompt']
                        # Here you could integrate with image generation API
                        st.write(f"Generated scene description for {bike_name}:")
                        st.write(prompt)

def display_file_manager():
    """Display file management sidebar with guaranteed unique button keys."""
    st.sidebar.title("📁 File Management")
    
    all_files = glob.glob("*.md")
    all_files.sort(reverse=True)

    if st.sidebar.button("🗑 Delete All", key="delete_all_files_button"):
        for file in all_files:
            os.remove(file)
        st.rerun()

    if st.sidebar.button("⬇️ Download All", key="download_all_files_button"):
        zip_file = create_zip_of_files(all_files)
        st.sidebar.markdown(get_download_link(zip_file), unsafe_allow_html=True)

    # Create unique keys using file attributes
    for idx, file in enumerate(all_files):
        # Get file stats for unique identification
        file_stat = os.stat(file)
        unique_id = f"{idx}_{file_stat.st_size}_{file_stat.st_mtime}"
        
        col1, col2, col3, col4 = st.sidebar.columns([1,3,1,1])
        with col1:
            if st.button("🌐", key=f"view_{unique_id}"):
                st.session_state.current_file = file
                st.session_state.file_content = load_file(file)
        with col2:
            st.markdown(get_download_link(file), unsafe_allow_html=True)
        with col3:
            if st.button("📂", key=f"edit_{unique_id}"):
                st.session_state.current_file = file
                st.session_state.file_content = load_file(file)
        with col4:
            if st.button("🗑", key=f"delete_{unique_id}"):
                os.remove(file)
                st.rerun()


def main():
    st.sidebar.markdown("### 🚲BikeAI🏆 Claude and GPT Multi-Agent Research AI")
    
    # Main navigation
    tab_main = st.radio("Choose Action:", 
                        ["💬 Chat", "📸 Media Gallery", "🔍 Search ArXiv", "📝 File Editor"],
                        horizontal=True)
    
    if tab_main == "💬 Chat":
        # Model Selection
        model_choice = st.sidebar.radio(
            "Choose AI Model:",
            ["GPT-4o", "Claude-3", "GPT+Claude+Arxiv"]
        )
        
        # Chat Interface
        user_input = st.text_area("Message:", height=100)
        
        if st.button("Send 📨"):
            if user_input:
                if model_choice == "GPT-4o":
                    gpt_response = process_with_gpt(user_input)
                elif model_choice == "Claude-3":
                    claude_response = process_with_claude(user_input)
                else:  # Both
                    col1, col2, col3 = st.columns(3)
                    with col2:
                        st.subheader("Claude-3.5 Sonnet:")
                        try:
                            claude_response = process_with_claude(user_input)
                        except:
                            st.write('Claude 3.5 Sonnet out of tokens.')
                    with col1:
                        st.subheader("GPT-4o Omni:")
                        try:
                            gpt_response = process_with_gpt(user_input)  
                        except:
                            st.write('GPT 4o out of tokens')
                    with col3:
                        st.subheader("Arxiv and Mistral Research:")
                        with st.spinner("Searching ArXiv..."):
                            #results = search_arxiv(user_input)
                            results = perform_ai_lookup(user_input)
                            
                            st.markdown(results)
                            
        # Display Chat History
        st.subheader("Chat History 📜")
        tab1, tab2 = st.tabs(["Claude History", "GPT-4o History"])
        
        with tab1:
            for chat in st.session_state.chat_history:
                st.text_area("You:", chat["user"], height=100)
                st.text_area("Claude:", chat["claude"], height=200)
                st.markdown(chat["claude"])
        
        with tab2:
            for message in st.session_state.messages:
                with st.chat_message(message["role"]):
                    st.markdown(message["content"])
    
    elif tab_main == "📸 Media Gallery":
        create_media_gallery()
    
    elif tab_main == "🔍 Search ArXiv":
        query = st.text_input("Enter your research query:")
        if query:
            with st.spinner("Searching ArXiv..."):
                results = search_arxiv(query)
                st.markdown(results)
    
    elif tab_main == "📝 File Editor":
        if hasattr(st.session_state, 'current_file'):
            st.subheader(f"Editing: {st.session_state.current_file}")
            new_content = st.text_area("Content:", st.session_state.file_content, height=300)
            if st.button("Save Changes"):
                with open(st.session_state.current_file, 'w', encoding='utf-8') as file:
                    file.write(new_content)
                st.success("File updated successfully!")

    # Always show file manager in sidebar
    display_file_manager()

if __name__ == "__main__":
    main()

# Speech Recognition HTML Component
speech_recognition_html = """
<!DOCTYPE html>
<html>
<head>
    <title>Continuous Speech Demo</title>
    <style>
        body { 
            font-family: sans-serif; 
            padding: 20px; 
            max-width: 800px;
            margin: 0 auto;
        }
        button { 
            padding: 10px 20px; 
            margin: 10px 5px;
            font-size: 16px;
        }
        #status { 
            margin: 10px 0;
            padding: 10px;
            background: #e8f5e9;
            border-radius: 4px;
        }
        #output {
            white-space: pre-wrap;
            padding: 15px;
            background: #f5f5f5;
            border-radius: 4px;
            margin: 10px 0;
            min-height: 100px;
            max-height: 400px;
            overflow-y: auto;
        }
        .controls {
            margin: 10px 0;
        }
    </style>
</head>
<body>
    <div class="controls">
        <button id="start">Start Listening</button>
        <button id="stop" disabled>Stop Listening</button>
        <button id="clear">Clear Text</button>
    </div>
    <div id="status">Ready</div>
    <div id="output"></div>

    <script>
        if (!('webkitSpeechRecognition' in window)) {
            alert('Speech recognition not supported');
        } else {
            const recognition = new webkitSpeechRecognition();
            const startButton = document.getElementById('start');
            const stopButton = document.getElementById('stop');
            const clearButton = document.getElementById('clear');
            const status = document.getElementById('status');
            const output = document.getElementById('output');
            let fullTranscript = '';
            let lastUpdateTime = Date.now();

            // Configure recognition
            recognition.continuous = true;
            recognition.interimResults = true;

            // Function to start recognition
            const startRecognition = () => {
                try {
                    recognition.start();
                    status.textContent = 'Listening...';
                    startButton.disabled = true;
                    stopButton.disabled = false;
                } catch (e) {
                    console.error(e);
                    status.textContent = 'Error: ' + e.message;
                }
            };

            // Auto-start on load
            window.addEventListener('load', () => {
                setTimeout(startRecognition, 1000);
            });

            startButton.onclick = startRecognition;

            stopButton.onclick = () => {
                recognition.stop();
                status.textContent = 'Stopped';
                startButton.disabled = false;
                stopButton.disabled = true;
            };

            clearButton.onclick = () => {
                fullTranscript = '';
                output.textContent = '';
                window.parent.postMessage({
                    type: 'clear_transcript',
                }, '*');
            };

            recognition.onresult = (event) => {
                let interimTranscript = '';
                let finalTranscript = '';

                for (let i = event.resultIndex; i < event.results.length; i++) {
                    const transcript = event.results[i][0].transcript;
                    if (event.results[i].isFinal) {
                        finalTranscript += transcript + '\\n';
                    } else {
                        interimTranscript += transcript;
                    }
                }

                if (finalTranscript || (Date.now() - lastUpdateTime > 5000)) {
                    if (finalTranscript) {
                        fullTranscript += finalTranscript;
                        // Send to Streamlit
                        window.parent.postMessage({
                            type: 'final_transcript',
                            text: finalTranscript
                        }, '*');
                    }
                    lastUpdateTime = Date.now();
                }

                output.textContent = fullTranscript + (interimTranscript ? '... ' + interimTranscript : '');
                output.scrollTop = output.scrollHeight;
            };

            recognition.onend = () => {
                if (!stopButton.disabled) {
                    try {
                        recognition.start();
                        console.log('Restarted recognition');
                    } catch (e) {
                        console.error('Failed to restart recognition:', e);
                        status.textContent = 'Error restarting: ' + e.message;
                        startButton.disabled = false;
                        stopButton.disabled = true;
                    }
                }
            };

            recognition.onerror = (event) => {
                console.error('Recognition error:', event.error);
                status.textContent = 'Error: ' + event.error;
                
                if (event.error === 'not-allowed' || event.error === 'service-not-allowed') {
                    startButton.disabled = false;
                    stopButton.disabled = true;
                }
            };
        }
    </script>
</body>
</html>
"""

# Helper Functions
def generate_filename(prompt, file_type):
    central = pytz.timezone('US/Central')
    safe_date_time = datetime.now(central).strftime("%m%d_%H%M")
    replaced_prompt = re.sub(r'[<>:"/\\|?*\n]', ' ', prompt)
    safe_prompt = re.sub(r'\s+', ' ', replaced_prompt).strip()[:230]
    return f"{safe_date_time}_{safe_prompt}.{file_type}"

# File Management Functions
def load_file(file_name):
    """Load file content."""
    with open(file_name, "r", encoding='utf-8') as file:
        content = file.read()
    return content

def create_zip_of_files(files):
    """Create zip archive of files."""
    zip_name = "all_files.zip"
    with zipfile.ZipFile(zip_name, 'w') as zipf:
        for file in files:
            zipf.write(file)
    return zip_name

def get_download_link(file):
    """Create download link for file."""
    with open(file, "rb") as f:
        contents = f.read()
    b64 = base64.b64encode(contents).decode()
    return f'<a href="data:file/txt;base64,{b64}" download="{os.path.basename(file)}">Download {os.path.basename(file)}📂</a>'

def display_file_manager():
    """Display file management sidebar."""
    st.sidebar.title("📁 File Management")
    
    all_files = glob.glob("*.md")
    all_files.sort(reverse=True)

    if st.sidebar.button("🗑 Delete All"):
        for file in all_files:
            os.remove(file)
        st.rerun()

    if st.sidebar.button("⬇️ Download All"):
        zip_file = create_zip_of_files(all_files)
        st.sidebar.markdown(get_download_link(zip_file), unsafe_allow_html=True)

    for file in all_files:
        col1, col2, col3, col4 = st.sidebar.columns([1,3,1,1])
        with col1:
            if st.button("🌐", key="view_"+file):
                st.session_state.current_file = file
                st.session_state.file_content = load_file(file)
        with col2:
            st.markdown(get_download_link(file), unsafe_allow_html=True)
        with col3:
            if st.button("📂", key="edit_"+file):
                st.session_state.current_file = file
                st.session_state.file_content = load_file(file)
        with col4:
            if st.button("🗑", key="delete_"+file):
                os.remove(file)
                st.rerun()

def create_media_gallery():
    """Create the media gallery interface."""
    st.header("🎬 Media Gallery")
    
    tabs = st.tabs(["🖼️ Images", "🎵 Audio", "🎥 Video", "🎨 Scene Generator"])
    
    with tabs[0]:
        image_files = glob.glob("*.png") + glob.glob("*.jpg")
        if image_files:
            num_cols = st.slider("Number of columns", 1, 5, 3)
            cols = st.columns(num_cols)
            for idx, image_file in enumerate(image_files):
                with cols[idx % num_cols]:
                    img = Image.open(image_file)
                    st.image(img, use_container_width=True)
                    
                    # Add GPT vision analysis option
                    if st.button(f"Analyze {os.path.basename(image_file)}"):
                        analysis = process_image(image_file, 
                                              "Describe this image in detail and identify key elements.")
                        st.markdown(analysis)
    
    with tabs[1]:
        audio_files = glob.glob("*.mp3") + glob.glob("*.wav")
        for audio_file in audio_files:
            with st.expander(f"🎵 {os.path.basename(audio_file)}"):
                st.markdown(get_media_html(audio_file, "audio"), unsafe_allow_html=True)
                if st.button(f"Transcribe {os.path.basename(audio_file)}"):
                    with open(audio_file, "rb") as f:
                        transcription = process_audio(f)
                        st.write(transcription)
    
    with tabs[2]:
        video_files = glob.glob("*.mp4")
        for video_file in video_files:
            with st.expander(f"🎥 {os.path.basename(video_file)}"):
                st.markdown(get_media_html(video_file, "video"), unsafe_allow_html=True)
                if st.button(f"Analyze {os.path.basename(video_file)}"):
                    analysis = process_video_with_gpt(video_file, 
                                                    "Describe what's happening in this video.")
                    st.markdown(analysis)
    
    with tabs[3]:
        for collection_name, bikes in bike_collections.items():
            st.subheader(collection_name)
            cols = st.columns(len(bikes))
            
            for idx, (bike_name, details) in enumerate(bikes.items()):
                with cols[idx]:
                    st.markdown(f"""
                    <div class='bike-card'>
                        <h3>{details['emoji']} {bike_name}</h3>
                        <p>{details['prompt']}</p>
                    </div>
                    """, unsafe_allow_html=True)
                    
                    if st.button(f"Generate {bike_name} Scene"):
                        prompt = details['prompt']
                        # Here you could integrate with image generation API
                        st.write(f"Generated scene description for {bike_name}:")
                        st.write(prompt)

def get_media_html(media_path, media_type="video", width="100%"):
    """Generate HTML for media player."""
    media_data = base64.b64encode(open(media_path, 'rb').read()).decode()
    if media_type == "video":
        return f'''
        <video width="{width}" controls autoplay muted loop>
            <source src="data:video/mp4;base64,{media_data}" type="video/mp4">
            Your browser does not support the video tag.
        </video>
        '''
    else:  # audio
        return f'''
        <audio controls style="width: {width};">
            <source src="data:audio/mpeg;base64,{media_data}" type="audio/mpeg">
            Your browser does not support the audio element.
        </audio>
        '''


def process_transcription_with_ai(text):
    """Process transcribed text with all three AI models."""
    results = {
        "claude": None,
        "gpt": None,
        "arxiv": None
    }
    
    try:
        results["claude"] = process_with_claude(text)
    except Exception as e:
        st.error(f"Claude processing error: {e}")
        
    try:
        results["gpt"] = process_with_gpt(text)
    except Exception as e:
        st.error(f"GPT processing error: {e}")
        
    try:
        results["arxiv"] = perform_ai_lookup(text)
    except Exception as e:
        st.error(f"Arxiv processing error: {e}")
        
    return results

def handle_speech_recognition_component():
    """Handle the speech recognition component and AI processing."""
    st.subheader("Voice Recognition with Multi-Modal Output")
    
    # Initialize state for transcribed text
    if "transcribed_text" not in st.session_state:
        st.session_state.transcribed_text = ""
    
    # Render the React component
    component = components.declare_component(
        "speech_recognition",
        path="frontend/build"  # Update this path to match your React component location
    )
    
    # Handle component events
    component_result = component()
    if component_result:
        if component_result.get("type") == "process_ai":
            text = component_result.get("text", "").strip()
            if text:
                with st.spinner("Processing with AI models..."):
                    results = process_transcription_with_ai(text)
                    
                    # Display results in columns
                    col1, col2, col3 = st.columns(3)
                    with col1:
                        st.subheader("GPT-4o Results")
                        if results["gpt"]:
                            st.markdown(results["gpt"])
                    
                    with col2:
                        st.subheader("Claude Results")
                        if results["claude"]:
                            st.markdown(results["claude"])
                    
                    with col3:
                        st.subheader("Arxiv Results")
                        if results["arxiv"]:
                            st.markdown(results["arxiv"])
                                 

    
def main():
    st.sidebar.markdown("### 🚲BikeAI🏆 Claude and GPT Multi-Agent Research AI")
    
    # Main navigation
    tab_main = st.radio("Choose Action:", 
                        ["🎤 Voice Input", "💬 Chat", "📸 Media Gallery", "🔍 Search ArXiv", "📝 File Editor"],
                        horizontal=True)
    
    if tab_main == "🎤 Voice Input":
        handle_speech_recognition_component()
          
    if tab_main == "🎤 Voice Input":
        st.subheader("Voice Recognition")
        
        # Display speech recognition component
        speech_component = st.components.v1.html(speech_recognition_html, height=400)
        
        # Handle speech recognition output
        if speech_component:
            try:
                data = speech_component
                if isinstance(data, dict):
                    if data.get('type') == 'final_transcript':
                        text = data.get('text', '').strip()
                        if text:
                            st.session_state.last_voice_input = text
                            
                            # Process voice input with AI
                            st.subheader("AI Response to Voice Input:")
                            
                            col1, col2, col3 = st.columns(3)
                            with col2:
                                st.write("Claude-3.5 Sonnet:")
                                try:
                                    claude_response = process_with_claude(text)
                                except:
                                    st.write('Claude 3.5 Sonnet out of tokens.')
                            with col1:
                                st.write("GPT-4o Omni:")
                                try:
                                    gpt_response = process_with_gpt(text)
                                except:
                                    st.write('GPT 4o out of tokens')
                            with col3:
                                st.write("Arxiv and Mistral Research:")
                                with st.spinner("Searching ArXiv..."):
                                    results = perform_ai_lookup(text)
                                    st.markdown(results)
                    
                    elif data.get('type') == 'clear_transcript':
                        st.session_state.last_voice_input = ""
                        st.experimental_rerun()
                        
            except Exception as e:
                st.error(f"Error processing voice input: {e}")
        
        # Display last voice input
        if st.session_state.last_voice_input:
            st.text_area("Last Voice Input:", st.session_state.last_voice_input, height=100)
    
    # [Rest of the main function remains the same]
    elif tab_main == "💬 Chat":
        # [Previous chat interface code]
        pass
    
    elif tab_main == "📸 Media Gallery":
        create_media_gallery()
    
    elif tab_main == "🔍 Search ArXiv":
        query = st.text_input("Enter your research query:")
        if query:
            with st.spinner("Searching ArXiv..."):
                results = search_arxiv(query)
                st.markdown(results)
    
    elif tab_main == "📝 File Editor":
        if hasattr(st.session_state, 'current_file'):
            st.subheader(f"Editing: {st.session_state.current_file}")
            new_content = st.text_area("Content:", st.session_state.file_content, height=300)
            if st.button("Save Changes"):
                with open(st.session_state.current_file, 'w', encoding='utf-8') as file:
                    file.write(new_content)
                st.success("File updated successfully!")

    # Always show file manager in sidebar
    display_file_manager()

if __name__ == "__main__":
    main()