Create app.py

app.py

@@ -0,0 +1,974 @@
+import streamlit as st
+import asyncio
+import websockets
+import uuid
+from datetime import datetime
+import os
+import random
+import time
+import hashlib
+from PIL import Image
+import glob
+import base64
+import io
+import streamlit.components.v1 as components
+import edge_tts
+from audio_recorder_streamlit import audio_recorder
+import nest_asyncio
+import re
+import pytz
+import shutil
+import anthropic
+import openai
+from PyPDF2 import PdfReader
+import threading
+import json
+import zipfile
+from gradio_client import Client
+from dotenv import load_dotenv
+from streamlit_marquee import streamlit_marquee
+from collections import defaultdict, Counter
+import pandas as pd
+
+# 🛠️ Patch asyncio for nesting
+nest_asyncio.apply()
+
+# 🎨 Page Config
+st.set_page_config(
+    page_title="🚲TalkingAIResearcher🏆",
+    page_icon="🚲🏆",
+    layout="wide",
+    initial_sidebar_state="auto"
+)
+
+# 🌟 Static Config
+icons = '🤖🧠🔬📝'
+Site_Name = '🤖🧠Chat & Quote Node📝🔬'
+START_ROOM = "Sector 🌌"
+FUN_USERNAMES = {
+    "CosmicJester 🌌": "en-US-AriaNeural",
+    "PixelPanda 🐼": "en-US-JennyNeural",
+    "QuantumQuack 🦆": "en-GB-SoniaNeural",
+    "StellarSquirrel 🐿️": "en-AU-NatashaNeural",
+    "GizmoGuru ⚙️": "en-CA-ClaraNeural",
+    "NebulaNinja 🌠": "en-US-GuyNeural",
+    "ByteBuster 💾": "en-GB-RyanNeural",
+    "GalacticGopher 🌍": "en-AU-WilliamNeural",
+    "RocketRaccoon 🚀": "en-CA-LiamNeural",
+    "EchoElf 🧝": "en-US-AnaNeural",
+    "PhantomFox 🦊": "en-US-BrandonNeural",
+    "WittyWizard 🧙": "en-GB-ThomasNeural",
+    "LunarLlama 🌙": "en-AU-FreyaNeural",
+    "SolarSloth ☀️": "en-CA-LindaNeural",
+    "AstroAlpaca 🦙": "en-US-ChristopherNeural",
+    "CyberCoyote 🐺": "en-GB-ElliotNeural",
+    "MysticMoose 🦌": "en-AU-JamesNeural",
+    "GlitchGnome 🧚": "en-CA-EthanNeural",
+    "VortexViper 🐍": "en-US-AmberNeural",
+    "ChronoChimp 🐒": "en-GB-LibbyNeural"
+}
+EDGE_TTS_VOICES = list(set(FUN_USERNAMES.values()))
+FILE_EMOJIS = {"md": "📝", "mp3": "🎵", "png": "🖼️", "mp4": "🎥", "zip": "📦"}
+
+# 📁 Directories
+for d in ["chat_logs", "vote_logs", "audio_logs", "history_logs", "audio_cache", "paper_metadata"]:
+    os.makedirs(d, exist_ok=True)
+
+CHAT_DIR = "chat_logs"
+VOTE_DIR = "vote_logs"
+MEDIA_DIR = "."
+AUDIO_CACHE_DIR = "audio_cache"
+AUDIO_DIR = "audio_logs"
+PAPER_DIR = "paper_metadata"
+STATE_FILE = "user_state.txt"
+
+CHAT_FILE = os.path.join(CHAT_DIR, "global_chat.md")
+QUOTE_VOTES_FILE = os.path.join(VOTE_DIR, "quote_votes.md")
+IMAGE_VOTES_FILE = os.path.join(VOTE_DIR, "image_votes.md")
+HISTORY_FILE = os.path.join(VOTE_DIR, "vote_history.md")
+
+# 🔑 API Keys
+load_dotenv()
+anthropic_key = os.getenv('ANTHROPIC_API_KEY', st.secrets.get('ANTHROPIC_API_KEY', ""))
+openai_api_key = os.getenv('OPENAI_API_KEY', st.secrets.get('OPENAI_API_KEY', ""))
+openai_client = openai.OpenAI(api_key=openai_api_key)
+
+# 🕒 Timestamp Helper
+def format_timestamp_prefix(username=""):
+    central = pytz.timezone('US/Central')
+    now = datetime.now(central)
+    return f"{now.strftime('%Y%m%d_%H%M%S')}-by-{username}"
+
+# 📈 Performance Timer
+class PerformanceTimer:
+    def __init__(self, name):
+        self.name, self.start = name, None
+    def __enter__(self):
+        self.start = time.time()
+        return self
+    def __exit__(self, *args):
+        duration = time.time() - self.start
+        st.session_state['operation_timings'][self.name] = duration
+        st.session_state['performance_metrics'][self.name].append(duration)
+
+# 🎛️ Session State Init
+def init_session_state():
+    defaults = {
+        'server_running': False, 'server_task': None, 'active_connections': {},
+        'media_notifications': [], 'last_chat_update': 0, 'displayed_chat_lines': [],
+        'message_text': "", 'audio_cache': {}, 'pasted_image_data': None,
+        'quote_line': None, 'refresh_rate': 10, 'base64_cache': {},
+        'transcript_history': [], 'last_transcript': "", 'image_hashes': set(),
+        'tts_voice': "en-US-AriaNeural", 'chat_history': [], 'marquee_settings': {
+            "background": "#1E1E1E", "color": "#FFFFFF", "font-size": "14px",
+            "animationDuration": "20s", "width": "100%", "lineHeight": "35px"
+        }, 'operation_timings': {}, 'performance_metrics': defaultdict(list),
+        'enable_audio': True, 'download_link_cache': {}, 'username': None,
+        'autosend': True, 'autosearch': True, 'last_message': "", 'last_query': "",
+        'mp3_files': {}, 'timer_start': time.time(), 'quote_index': 0,
+        'quote_source': "famous", 'last_sent_transcript': "", 'old_val': None,
+        'last_refresh': time.time(), 'paper_metadata': {}, 'paste_image_base64': "",
+        'use_arxiv': True, 'use_arxiv_audio': False, 'speech_processed': False,
+        'auto_refresh': True
+    }
+    for k, v in defaults.items():
+        if k not in st.session_state:
+            st.session_state[k] = v
+
+# 🖌️ Marquee Helpers
+def update_marquee_settings_ui():
+    st.sidebar.markdown("### 🎯 Marquee Settings")
+    cols = st.sidebar.columns(2)
+    with cols[0]:
+        st.session_state['marquee_settings']['background'] = st.color_picker("🎨 Background", "#1E1E1E")
+        st.session_state['marquee_settings']['color'] = st.color_picker("✍️ Text", "#FFFFFF")
+    with cols[1]:
+        st.session_state['marquee_settings']['font-size'] = f"{st.slider('📏 Size', 10, 24, 14)}px"
+        st.session_state['marquee_settings']['animationDuration'] = f"{st.slider('⏱️ Speed', 1, 20, 20)}s"
+
+def display_marquee(text, settings, key_suffix=""):
+    truncated = text[:280] + "..." if len(text) > 280 else text
+    streamlit_marquee(content=truncated, **settings, key=f"marquee_{key_suffix}")
+    st.write("")
+
+# 📝 Text & File Helpers
+def clean_text_for_tts(text):
+    return re.sub(r'[#*!\[\]]+', '', ' '.join(text.split()))[:200] or "No text"
+
+def clean_text_for_filename(text):
+    return '_'.join(re.sub(r'[^\w\s-]', '', text.lower()).split())[:50]
+
+def get_high_info_terms(text, top_n=10):
+    stop_words = {'the', 'a', 'an', 'and', 'or', 'but', 'in', 'on', 'at', 'to', 'for', 'of', 'with'}
+    words = re.findall(r'\b\w+(?:-\w+)*\b', text.lower())
+    bi_grams = [' '.join(pair) for pair in zip(words, words[1:])]
+    filtered = [t for t in words + bi_grams if t not in stop_words and len(t.split()) <= 2]
+    return [t for t, _ in Counter(filtered).most_common(top_n)]
+
+def generate_filename(prompt, username, file_type="md", title=None):
+    timestamp = format_timestamp_prefix(username)
+    if title:
+        high_info = '-'.join(get_high_info_terms(title, 5))
+        return f"{timestamp}-{clean_text_for_filename(prompt[:20])}-{high_info}.{file_type}"
+    hash_val = hashlib.md5(prompt.encode()).hexdigest()[:8]
+    return f"{timestamp}-{hash_val}.{file_type}"
+
+def create_file(prompt, username, file_type="md", title=None):
+    filename = generate_filename(prompt, username, file_type, title)
+    with open(filename, 'w', encoding='utf-8') as f:
+        f.write(prompt)
+    return filename
+
+def get_download_link(file, file_type="mp3"):
+    cache_key = f"dl_{file}"
+    if cache_key not in st.session_state['download_link_cache']:
+        with open(file, "rb") as f:
+            b64 = base64.b64encode(f.read()).decode()
+        mime_types = {"mp3": "audio/mpeg", "png": "image/png", "mp4": "video/mp4", "md": "text/markdown", "zip": "application/zip"}
+        st.session_state['download_link_cache'][cache_key] = f'<a href="data:{mime_types.get(file_type, "application/octet-stream")};base64,{b64}" download="{os.path.basename(file)}">{FILE_EMOJIS.get(file_type, "Download")} Download {os.path.basename(file)}</a>'
+    return st.session_state['download_link_cache'][cache_key]
+
+def save_username(username):
+    try:
+        with open(STATE_FILE, 'w') as f:
+            f.write(username)
+    except Exception as e:
+        print(f"Failed to save username: {e}")
+
+def load_username():
+    if os.path.exists(STATE_FILE):
+        try:
+            with open(STATE_FILE, 'r') as f:
+                return f.read().strip()
+        except Exception as e:
+            print(f"Failed to load username: {e}")
+    return None
+
+def concatenate_markdown_files(exclude_files=["README.md"]):
+    md_files = sorted([f for f in glob.glob("*.md") if os.path.basename(f) not in exclude_files], key=os.path.getmtime)
+    all_md_content = ""
+    for i, md_file in enumerate(md_files, 1):
+        with open(md_file, 'r', encoding='utf-8') as f:
+            content = f.read().strip()
+            all_md_content += f"{i}. {content}\n"
+    return all_md_content.rstrip()
+
+def get_chat_text_only():
+    md_files = sorted(glob.glob("*.md"), key=os.path.getmtime)
+    chat_text = ""
+    for i, md_file in enumerate(md_files, 1):
+        with open(md_file, 'r', encoding='utf-8') as f:
+            content = f.read().strip()
+            # Extract text after colon, removing timestamp and speaker
+            lines = content.split('\n')
+            for line in lines:
+                if ':' in line and not line.startswith('#'):
+                    text = line.split(':', 1)[1].strip()
+                    if text.startswith('```markdown'):
+                        text = text.replace('```markdown', '').replace('```', '').strip()
+                    chat_text += f"{text}\n"
+    return chat_text.rstrip()
+
+# 🎶 Audio Processing
+async def async_edge_tts_generate(text, voice, username, rate=0, pitch=0, file_format="mp3"):
+    cache_key = f"{text[:100]}_{voice}_{rate}_{pitch}_{file_format}"
+    if cache_key in st.session_state['audio_cache']:
+        return st.session_state['audio_cache'][cache_key], 0
+    start_time = time.time()
+    text = clean_text_for_tts(text)
+    if not text or text == "No text":
+        print(f"Skipping audio generation for empty/invalid text: '{text}'")
+        return None, 0
+    filename = f"{format_timestamp_prefix(username)}-{hashlib.md5(text.encode()).hexdigest()[:8]}.{file_format}"
+    try:
+        communicate = edge_tts.Communicate(text, voice, rate=f"{rate:+d}%", pitch=f"{pitch:+d}Hz")
+        await communicate.save(filename)
+        if os.path.exists(filename) and os.path.getsize(filename) > 0:
+            st.session_state['audio_cache'][cache_key] = filename
+            return filename, time.time() - start_time
+        else:
+            print(f"Audio file {filename} was not created or is empty.")
+            return None, 0
+    except edge_tts.exceptions.NoAudioReceived as e:
+        print(f"No audio received for text: '{text}' with voice: {voice}. Error: {e}")
+        return None, 0
+    except Exception as e:
+        print(f"Error generating audio for text: '{text}' with voice: {voice}. Error: {e}")
+        return None, 0
+
+def play_and_download_audio(file_path):
+    if file_path and os.path.exists(file_path):
+        st.audio(file_path)
+        st.markdown(get_download_link(file_path), unsafe_allow_html=True)
+    else:
+        st.warning(f"Audio file not found: {file_path}")
+
+def load_mp3_viewer():
+    mp3_files = sorted(glob.glob("*.mp3"), key=os.path.getmtime)
+    for i, mp3 in enumerate(mp3_files, 1):
+        filename = os.path.basename(mp3)
+        if filename not in st.session_state['mp3_files']:
+            st.session_state['mp3_files'][filename] = (i, mp3)
+
+async def save_chat_entry(username, message, voice, is_markdown=False):
+    if not message.strip() or message == st.session_state.last_transcript:
+        return None, None
+    central = pytz.timezone('US/Central')
+    timestamp = datetime.now(central).strftime("%Y-%m-%d %H:%M:%S")
+    entry = f"[{timestamp}] {username} ({voice}): {message}" if not is_markdown else f"[{timestamp}] {username} ({voice}):\n```markdown\n{message}\n```"
+    md_file = create_file(entry, username, "md")
+    with open(CHAT_FILE, 'a') as f:
+        f.write(f"{entry}\n")
+    audio_file, _ = await async_edge_tts_generate(message, voice, username)
+    if audio_file:
+        with open(HISTORY_FILE, 'a') as f:
+            f.write(f"[{timestamp}] {username}: Audio - {audio_file}\n")
+        st.session_state['mp3_files'][os.path.basename(audio_file)] = (len(st.session_state['chat_history']) + 1, audio_file)
+        if st.session_state.get('speech_processed', False) and st.session_state.get('message_input', '') == message:
+            st.session_state['message_input'] = ""
+            st.session_state['speech_processed'] = False
+    else:
+        st.warning(f"Failed to generate audio for: {message}")
+    await broadcast_message(f"{username}|{message}", "chat")
+    st.session_state.last_chat_update = time.time()
+    st.session_state.chat_history.append(entry)
+    st.session_state.last_transcript = message
+    return md_file, audio_file
+
+async def load_chat():
+    if not os.path.exists(CHAT_FILE):
+        with open(CHAT_FILE, 'a') as f:
+            f.write(f"# {START_ROOM} Chat\n\nWelcome to the cosmic hub! 🎤\n")
+    with open(CHAT_FILE, 'r') as f:
+        content = f.read().strip()
+    lines = content.split('\n')
+    unique_lines = list(dict.fromkeys(line for line in lines if line.strip()))
+    return unique_lines
+
+# Claude Search Function with Image Support
+async def perform_claude_search(query, username, image=None):
+    if not query.strip() or query == st.session_state.last_transcript:
+        return None, None, None
+    client = anthropic.Anthropic(api_key=anthropic_key)
+    message_content = [{"type": "text", "text": query}]
+    if image:
+        buffered = io.BytesIO()
+        image.save(buffered, format="PNG")
+        img_base64 = base64.b64encode(buffered.getvalue()).decode('utf-8')
+        message_content.append({
+            "type": "image",
+            "source": {
+                "type": "base64",
+                "media_type": "image/png",
+                "data": img_base64
+            }
+        })
+    try:
+        response = client.messages.create(
+            model="claude-3-sonnet-20240229",
+            max_tokens=1000,
+            messages=[{"role": "user", "content": message_content}]
+        )
+        result = response.content[0].text
+        st.markdown(f"### Claude's Reply 🧠\n{result}")
+    except Exception as e:
+        st.error(f"Claude processing failed: {e}")
+        return None, None, None
+    
+    voice = FUN_USERNAMES.get(username, "en-US-AriaNeural")
+    md_file, audio_file = await save_chat_entry(username, f"Claude Search: {query}\nResponse: {result}", voice, True)
+    return md_file, audio_file, result
+
+# ArXiv Search Function
+async def perform_arxiv_search(query, username, claude_result=None):
+    if not query.strip() or query == st.session_state.last_transcript:
+        return None, None
+    if claude_result is None:
+        client = anthropic.Anthropic(api_key=anthropic_key)
+        claude_response = client.messages.create(
+            model="claude-3-sonnet-20240229",
+            max_tokens=1000,
+            messages=[{"role": "user", "content": query}]
+        )
+        claude_result = claude_response.content[0].text
+        st.markdown(f"### Claude's Reply 🧠\n{claude_result}")
+    
+    enhanced_query = f"{query}\n\n{claude_result}"
+    gradio_client = Client("awacke1/Arxiv-Paper-Search-And-QA-RAG-Pattern")
+    refs = gradio_client.predict(
+        enhanced_query, 10, "Semantic Search", "mistralai/Mixtral-8x7B-Instruct-v0.1", api_name="/update_with_rag_md"
+    )[0]
+    result = f"🔎 {enhanced_query}\n\n{refs}"
+    voice = FUN_USERNAMES.get(username, "en-US-AriaNeural")
+    md_file, audio_file = await save_chat_entry(username, f"ArXiv Search: {query}\nClaude Response: {claude_result}\nArXiv Results: {refs}", voice, True)
+    return md_file, audio_file
+
+async def perform_ai_lookup(q, vocal_summary=True, extended_refs=False, titles_summary=True, full_audio=False, useArxiv=True, useArxivAudio=False):
+    start = time.time()
+    client = anthropic.Anthropic(api_key=anthropic_key)
+    response = client.messages.create(
+        model="claude-3-sonnet-20240229",
+        max_tokens=1000,
+        messages=[{"role": "user", "content": q}]
+    )
+    st.write("Claude's reply 🧠:")
+    st.markdown(response.content[0].text)
+
+    result = response.content[0].text
+    md_file = create_file(result, "System", "md")
+    audio_file, _ = await async_edge_tts_generate(result, st.session_state['tts_voice'], "System")
+    st.subheader("📝 Main Response Audio")
+    play_and_download_audio(audio_file)
+
+    papers = []
+    if useArxiv:
+        q = q + result
+        st.write('Running Arxiv RAG with Claude inputs.')
+        gradio_client = Client("awacke1/Arxiv-Paper-Search-And-QA-RAG-Pattern")
+        refs = gradio_client.predict(
+            q, 20, "Semantic Search", "mistralai/Mixtral-8x7B-Instruct-v0.1", api_name="/update_with_rag_md"
+        )[0]
+        papers = parse_arxiv_refs(refs, q)
+        for paper in papers:
+            filename = create_file(generate_5min_feature_markdown(paper), "System", "md", paper['title'])
+            paper['md_file'] = filename
+            st.session_state['paper_metadata'][paper['title']] = filename
+        if papers and useArxivAudio:
+            await create_paper_audio_files(papers, q)
+    elapsed = time.time() - start
+    st.write(f"**Total Elapsed:** {elapsed:.2f} s")
+    return result, papers
+
+# 🌐 WebSocket Handling
+async def websocket_handler(websocket, path):
+    client_id = str(uuid.uuid4())
+    room_id = "chat"
+    if room_id not in st.session_state.active_connections:
+        st.session_state.active_connections[room_id] = {}
+    st.session_state.active_connections[room_id][client_id] = websocket
+    username = st.session_state.get('username', random.choice(list(FUN_USERNAMES.keys())))
+    chat_content = await load_chat()
+    if not any(f"Client-{client_id}" in line for line in chat_content):
+        await save_chat_entry("System 🌟", f"{username} has joined {START_ROOM}!", "en-US-AriaNeural")
+    try:
+        async for message in websocket:
+            if '|' in message:
+                username, content = message.split('|', 1)
+                voice = FUN_USERNAMES.get(username, "en-US-AriaNeural")
+                await save_chat_entry(username, content, voice)
+            else:
+                await websocket.send("ERROR|Message format: username|content")
+    except websockets.ConnectionClosed:
+        await save_chat_entry("System 🌟", f"{username} has left {START_ROOM}!", "en-US-AriaNeural")
+    finally:
+        if room_id in st.session_state.active_connections and client_id in st.session_state.active_connections[room_id]:
+            del st.session_state.active_connections[room_id][client_id]
+
+async def broadcast_message(message, room_id):
+    if room_id in st.session_state.active_connections:
+        disconnected = []
+        for client_id, ws in st.session_state.active_connections[room_id].items():
+            try:
+                await ws.send(message)
+            except websockets.ConnectionClosed:
+                disconnected.append(client_id)
+        for client_id in disconnected:
+            if client_id in st.session_state.active_connections[room_id]:
+                del st.session_state.active_connections[room_id][client_id]
+
+async def run_websocket_server():
+    if not st.session_state.get('server_running', False):
+        server = await websockets.serve(websocket_handler, '0.0.0.0', 8765)
+        st.session_state['server_running'] = True
+        await server.wait_closed()
+
+def start_websocket_server():
+    loop = asyncio.new_event_loop()
+    asyncio.set_event_loop(loop)
+    loop.run_until_complete(run_websocket_server())
+
+# 📚 PDF to Audio
+class AudioProcessor:
+    def __init__(self):
+        self.cache_dir = AUDIO_CACHE_DIR
+        os.makedirs(self.cache_dir, exist_ok=True)
+        self.metadata = json.load(open(f"{self.cache_dir}/metadata.json")) if os.path.exists(f"{self.cache_dir}/metadata.json") else {}
+
+    def _save_metadata(self):
+        with open(f"{self.cache_dir}/metadata.json", 'w') as f:
+            json.dump(self.metadata, f)
+
+    async def create_audio(self, text, voice='en-US-AriaNeural'):
+        cache_key = hashlib.md5(f"{text}:{voice}".encode()).hexdigest()
+        cache_path = f"{self.cache_dir}/{cache_key}.mp3"
+        if cache_key in self.metadata and os.path.exists(cache_path):
+            return cache_path
+        text = clean_text_for_tts(text)
+        if not text:
+            return None
+        communicate = edge_tts.Communicate(text, voice)
+        await communicate.save(cache_path)
+        self.metadata[cache_key] = {'timestamp': datetime.now().isoformat(), 'text_length': len(text), 'voice': voice}
+        self._save_metadata()
+        return cache_path
+
+def process_pdf(pdf_file, max_pages, voice, audio_processor):
+    reader = PdfReader(pdf_file)
+    total_pages = min(len(reader.pages), max_pages)
+    texts, audios = [], {}
+    async def process_page(i, text):
+        audio_path = await audio_processor.create_audio(text, voice)
+        if audio_path:
+            audios[i] = audio_path
+    for i in range(total_pages):
+        text = reader.pages[i].extract_text()
+        texts.append(text)
+        threading.Thread(target=lambda: asyncio.run(process_page(i, text))).start()
+    return texts, audios, total_pages
+
+# 🔍 ArXiv & AI Lookup
+def parse_arxiv_refs(ref_text, query):
+    if not ref_text:
+        return []
+    papers = []
+    current = {}
+    for line in ref_text.split('\n'):
+        if line.count('|') == 2:
+            if current:
+                papers.append(current)
+            date, title, *_ = line.strip('* ').split('|')
+            url = re.search(r'(https://arxiv.org/\S+)', line).group(1) if re.search(r'(https://arxiv.org/\S+)', line) else f"paper_{len(papers)}"
+            current = {'date': date, 'title': title, 'url': url, 'authors': '', 'summary': '', 'full_audio': None, 'download_base64': '', 'query': query}
+        elif current:
+            if not current['authors']:
+                current['authors'] = line.strip('* ')
+            else:
+                current['summary'] += ' ' + line.strip() if current['summary'] else line.strip()
+    if current:
+        papers.append(current)
+    return papers[:20]
+
+def generate_5min_feature_markdown(paper):
+    title, summary, authors, date, url = paper['title'], paper['summary'], paper['authors'], paper['date'], paper['url']
+    pdf_url = url.replace("abs", "pdf") + (".pdf" if not url.endswith(".pdf") else "")
+    wct, sw = len(title.split()), len(summary.split())
+    terms = get_high_info_terms(summary, 15)
+    rouge = round((len(terms) / max(sw, 1)) * 100, 2)
+    mermaid = "```mermaid\nflowchart TD\n" + "\n".join(f'    T{i+1}["{t}"] --> T{i+2}["{terms[i+1]}"]' for i in range(len(terms)-1)) + "\n```"
+    return f"""
+## 📄 {title}
+**Authors:** {authors}  
+**Date:** {date}  
+**Words:** Title: {wct}, Summary: {sw}  
+**Links:** [Abstract]({url}) | [PDF]({pdf_url})  
+**Terms:** {', '.join(terms)}  
+**ROUGE:** {rouge}%  
+### 🎤 TTF Read Aloud
+- **Title:** {title}  
+- **Terms:** {', '.join(terms)}  
+- **ROUGE:** {rouge}%  
+#### Concepts Graph
+{mermaid}
+---
+"""
+
+async def create_paper_audio_files(papers, query):
+    for p in papers:
+        audio_text = clean_text_for_tts(f"{p['title']} by {p['authors']}. {p['summary']}")
+        p['full_audio'], _ = await async_edge_tts_generate(audio_text, st.session_state['tts_voice'], p['authors'])
+        if p['full_audio']:
+            p['download_base64'] = get_download_link(p['full_audio'])
+
+def save_vote(file, item, user_hash):
+    timestamp = datetime.now().strftime("%Y-%m-%d %H:%M:%S")
+    entry = f"[{timestamp}] {user_hash} voted for {item}"
+    try:
+        with open(file, 'a') as f:
+            f.write(f"{entry}\n")
+        with open(HISTORY_FILE, 'a') as f:
+            f.write(f"- {timestamp} - User {user_hash} voted for {item}\n")
+        return True
+    except Exception as e:
+        print(f"Vote save flop: {e}")
+        return False
+
+def load_votes(file):
+    if not os.path.exists(file):
+        with open(file, 'w') as f:
+            f.write("# Vote Tally\n\nNo votes yet - get clicking! 🖱️\n")
+    try:
+        with open(file, 'r') as f:
+            lines = f.read().strip().split('\n')
+        votes = {}
+        for line in lines[2:]:
+            if line.strip() and 'voted for' in line:
+                item = line.split('voted for ')[1]
+                votes[item] = votes.get(item, 0) + 1
+        return votes
+    except Exception as e:
+        print(f"Vote load oopsie: {e}")
+        return {}
+
+def generate_user_hash():
+    if 'user_hash' not in st.session_state:
+        session_id = str(random.getrandbits(128))
+        hash_object = hashlib.md5(session_id.encode())
+        st.session_state['user_hash'] = hash_object.hexdigest()[:8]
+    return st.session_state['user_hash']
+
+async def save_pasted_image(image, username, prompt=""):
+    img_hash = hashlib.md5(image.tobytes()).hexdigest()[:8]
+    if img_hash in st.session_state.image_hashes:
+        return None
+    context = prompt if prompt else st.session_state.get('last_message', "pasted_image")
+    timestamp = format_timestamp_prefix(username)
+    filename = f"{timestamp}-{clean_text_for_filename(context)}-{img_hash}.png"
+    filepath = filename
+    try:
+        image.save(filepath, "PNG")
+        st.session_state.image_hashes.add(img_hash)
+        await save_chat_entry(username, f"Pasted image saved: {filepath}", FUN_USERNAMES.get(username, "en-US-AriaNeural"))
+        return filepath
+    except Exception as e:
+        st.error(f"Failed to save image: {e}")
+        return None
+
+# 📦 Zip and Delete Files
+def create_zip_of_files(files, prefix="All", query="latest"):
+    if not files:
+        return None
+    terms = get_high_info_terms(" ".join([open(f, 'r', encoding='utf-8').read() if f.endswith('.md') else os.path.splitext(os.path.basename(f))[0].replace('_', ' ') for f in files] + [query]), 5)
+    zip_name = f"{prefix}_{format_timestamp_prefix()}_{'-'.join(terms)[:20]}.zip"
+    with zipfile.ZipFile(zip_name, 'w') as z:
+        [z.write(f) for f in files]
+    return zip_name
+
+def delete_files(file_type, exclude_files=["README.md"]):
+    files = glob.glob(f"*.{file_type}")
+    if file_type == "md":
+        files = [f for f in files if os.path.basename(f) not in exclude_files]
+    for f in files:
+        try:
+            os.remove(f)
+            st.session_state['mp3_files'] = {k: v for k, v in st.session_state['mp3_files'].items() if not k.endswith(f".{file_type}")}
+        except Exception as e:
+            st.error(f"Failed to delete {f}: {e}")
+    if file_type in ["md", "mp3", "png", "mp4"]:
+        st.session_state['download_link_cache'] = {}
+
+# Custom Paste Image Component
+def paste_image_component():
+    st.markdown("""
+    <input type="text" id="pasteImageInput" placeholder="Paste image URL or base64 here" style="width: 100%;">
+    <button onclick="document.getElementById('pasteImageInput').value = ''; document.getElementById('pasteImageForm').submit();">Paste Image 📋</button>
+    <form id="pasteImageForm" style="display: none;">
+        <input type="hidden" name="pasted_image" id="hiddenPasteInput">
+    </form>
+    <script>
+    document.querySelector('button').addEventListener('click', function() {
+        const input = document.getElementById('pasteImageInput').value;
+        document.getElementById('hiddenPasteInput').value = input;
+        document.getElementById('pasteImageForm').submit();
+    });
+    </script>
+    """, unsafe_allow_html=True)
+    
+    pasted_image_data = st.experimental_get_query_params().get("pasted_image", [""])[0]
+    if pasted_image_data:
+        try:
+            if pasted_image_data.startswith('data:image'):
+                base64_str = pasted_image_data.split(',')[1]
+                img_bytes = base64.b64decode(base64_str)
+                img = Image.open(io.BytesIO(img_bytes))
+                st.image(img, caption="Image Pasted Successfully", use_column_width=True)
+                return img
+            else:
+                st.warning("Pasted data is not a valid base64 image.")
+                return None
+        except Exception as e:
+            st.error(f"Error decoding pasted image: {e}")
+            return None
+    return None
+
+# 🎮 Main Interface
+def main():
+    init_session_state()
+    load_mp3_viewer()
+    saved_username = load_username()
+    if saved_username and saved_username in FUN_USERNAMES:
+        st.session_state.username = saved_username
+    if not st.session_state.username:
+        available = [n for n in FUN_USERNAMES if not any(f"{n} has joined" in l for l in asyncio.run(load_chat()))]
+        st.session_state.username = random.choice(available or list(FUN_USERNAMES.keys()))
+        st.session_state.tts_voice = FUN_USERNAMES[st.session_state.username]
+        asyncio.run(save_chat_entry("System 🌟", f"{st.session_state.username} has joined {START_ROOM}!", "en-US-AriaNeural"))
+    save_username(st.session_state.username)
+
+    st.title(f"{Site_Name} for {st.session_state.username}")
+    update_marquee_settings_ui()
+    chat_text = get_chat_text_only()
+    display_marquee(f"🚀 Welcome to {START_ROOM} | 🤖 {st.session_state.username} | Chat: {chat_text}", st.session_state['marquee_settings'], "welcome")
+
+    mycomponent = components.declare_component("mycomponent", path="mycomponent")
+    val = mycomponent(my_input_value="", key=f"speech_{st.session_state.get('speech_processed', False)}")
+    if val and val != st.session_state.last_transcript:
+        val_stripped = val.strip().replace('\n', ' ')
+        if val_stripped:
+            voice = FUN_USERNAMES.get(st.session_state.username, "en-US-AriaNeural")
+            st.session_state['speech_processed'] = True
+            md_file, audio_file = asyncio.run(save_chat_entry(st.session_state.username, val_stripped, voice))
+            if audio_file:
+                play_and_download_audio(audio_file)
+            st.rerun()
+
+    tab_main = st.radio("Action:", ["🎤 Chat & Voice", "🔍 ArXiv", "📚 PDF to Audio"], horizontal=True, key="tab_main")
+    st.checkbox("Search ArXiv", key="use_arxiv")
+    st.checkbox("ArXiv Audio", key="use_arxiv_audio")
+    st.checkbox("Autosend Chat", key="autosend")
+    st.checkbox("Autosearch ArXiv", key="autosearch")
+
+    if tab_main == "🎤 Chat & Voice":
+        st.subheader(f"{START_ROOM} Chat 💬")
+        chat_content = asyncio.run(load_chat())
+        chat_container = st.container()
+        with chat_container:
+            numbered_content = "\n".join(f"{i+1}. {line}" for i, line in enumerate(chat_content))
+            st.code(numbered_content, language="python")
+
+        message = st.text_input(f"Message as {st.session_state.username}", key="message_input")
+        
+        pasted_image = paste_image_component()
+        if pasted_image is not None:
+            if st.session_state['paste_image_base64'] != base64.b64encode(pasted_image.tobytes()).decode('utf-8'):
+                st.session_state['paste_image_base64'] = base64.b64encode(pasted_image.tobytes()).decode('utf-8')
+                voice = FUN_USERNAMES.get(st.session_state.username, "en-US-AriaNeural")
+                image_prompt = st.text_input("Add a prompt for Claude (e.g., 'OCR this image')", key="image_prompt", value="")
+                with st.spinner("Saving image..."):
+                    filename = asyncio.run(save_pasted_image(pasted_image, st.session_state.username, image_prompt))
+                if filename:
+                    st.success(f"Image saved as: {filename}")
+                    if image_prompt:
+                        with st.spinner("Processing with Claude..."):
+                            md_file_claude, audio_file_claude, claude_result = asyncio.run(
+                                perform_claude_search(image_prompt, st.session_state.username, pasted_image)
+                            )
+                            if audio_file_claude:
+                                play_and_download_audio(audio_file_claude)
+                            if claude_result:
+                                with st.spinner("Searching ArXiv..."):
+                                    md_file_arxiv, audio_file_arxiv = asyncio.run(
+                                        perform_arxiv_search(image_prompt, st.session_state.username, claude_result)
+                                    )
+                                    if audio_file_arxiv:
+                                        play_and_download_audio(audio_file_arxiv)
+                    st.session_state.pasted_image_data = None
+                    st.session_state['paste_image_base64'] = ""
+                    st.session_state.timer_start = time.time()
+                    save_username(st.session_state.username)
+                    st.rerun()
+
+        if (message and message != st.session_state.last_message) or (st.session_state.pasted_image_data and not st.session_state['paste_image_base64']):
+            st.session_state.last_message = message
+            col_send, col_claude, col_arxiv = st.columns([1, 1, 1])
+            
+            with col_send:
+                if st.session_state.autosend or st.button("Send 🚀", key="send_button"):
+                    voice = FUN_USERNAMES.get(st.session_state.username, "en-US-AriaNeural")
+                    if message.strip():
+                        md_file, audio_file = asyncio.run(save_chat_entry(st.session_state.username, message, voice, True))
+                        if audio_file:
+                            play_and_download_audio(audio_file)
+                    if st.session_state.pasted_image_data:
+                        asyncio.run(save_chat_entry(st.session_state.username, f"Pasted image: {st.session_state.pasted_image_data}", voice))
+                        st.session_state.pasted_image_data = None
+                    st.session_state.timer_start = time.time()
+                    save_username(st.session_state.username)
+                    st.rerun()
+
+            with col_claude:
+                if st.button("🧠 Claude", key="claude_button"):
+                    voice = FUN_USERNAMES.get(st.session_state.username, "en-US-AriaNeural")
+                    if message.strip():
+                        md_file, audio_file, _ = asyncio.run(perform_claude_search(message, st.session_state.username))
+                        if audio_file:
+                            play_and_download_audio(audio_file)
+                        st.session_state.timer_start = time.time()
+                        save_username(st.session_state.username)
+                        st.rerun()
+
+            with col_arxiv:
+                if st.button("🔍 ArXiv", key="arxiv_button"):
+                    voice = FUN_USERNAMES.get(st.session_state.username, "en-US-AriaNeural")
+                    if message.strip():
+                        md_file, audio_file = asyncio.run(perform_arxiv_search(message, st.session_state.username))
+                        if audio_file:
+                            play_and_download_audio(audio_file)
+                        st.session_state.timer_start = time.time()
+                        save_username(st.session_state.username)
+                        st.rerun()
+
+    elif tab_main == "🔍 ArXiv":
+        st.subheader("🔍 Query ArXiv")
+        q = st.text_input("🔍 Query:", key="arxiv_query")
+        if q and q != st.session_state.last_query:
+            st.session_state.last_query = q
+            if st.session_state.autosearch or st.button("🔍 Run", key="arxiv_run"):
+                result, papers = asyncio.run(perform_ai_lookup(q, useArxiv=st.session_state['use_arxiv'], useArxivAudio=st.session_state['use_arxiv_audio']))
+                st.markdown(f"### Query: {q}")
+                for i, p in enumerate(papers, 1):
+                    expander_label = f"{p['title']} | [arXiv Link]({p['url']})"
+                    with st.expander(expander_label):
+                        with open(p['md_file'], 'r', encoding='utf-8') as f:
+                            content = f.read()
+                        numbered_content = "\n".join(f"{j+1}. {line}" for j, line in enumerate(content.split('\n')))
+                        st.code(numbered_content, language="python")
+
+    elif tab_main == "📚 PDF to Audio":
+        audio_processor = AudioProcessor()
+        pdf_file = st.file_uploader("Choose PDF", "pdf", key="pdf_upload")
+        max_pages = st.slider('Pages', 1, 100, 10, key="pdf_pages")
+        if pdf_file:
+            with st.spinner('Processing...'):
+                texts, audios, total = process_pdf(pdf_file, max_pages, st.session_state['tts_voice'], audio_processor)
+                for i, text in enumerate(texts):
+                    with st.expander(f"Page {i+1}"):
+                        st.markdown(text)
+                        while i not in audios:
+                            time.sleep(0.1)
+                        if audios.get(i):
+                            st.audio(audios[i])
+                            st.markdown(get_download_link(audios[i], "mp3"), unsafe_allow_html=True)
+                            voice = FUN_USERNAMES.get(st.session_state.username, "en-US-AriaNeural")
+                            asyncio.run(save_chat_entry(st.session_state.username, f"PDF Page {i+1} converted to audio: {audios[i]}", voice))
+
+    # Always Visible Media Gallery
+    st.header("📸 Media Gallery")
+    all_files = sorted(glob.glob("*.md") + glob.glob("*.mp3") + glob.glob("*.png") + glob.glob("*.mp4"), key=os.path.getmtime)
+    md_files = [f for f in all_files if f.endswith('.md') and os.path.basename(f) != "README.md"]
+    mp3_files = [f for f in all_files if f.endswith('.mp3')]
+    png_files = [f for f in all_files if f.endswith('.png')]
+    mp4_files = [f for f in all_files if f.endswith('.mp4')]
+
+    st.subheader("All Submitted Text")
+    all_md_content = concatenate_markdown_files()
+    with st.expander("View All Markdown Content"):
+        st.markdown(all_md_content)
+
+    st.subheader("🎵 Audio (MP3)")
+    for filename, (num, mp3) in sorted(st.session_state['mp3_files'].items(), key=lambda x: x[1][0]):
+        with st.expander(f"{num}. {os.path.basename(mp3)}"):
+            st.audio(mp3)
+            st.markdown(get_download_link(mp3, "mp3"), unsafe_allow_html=True)
+
+    st.subheader("🖼️ Images (PNG)")
+    for png in sorted(png_files, key=os.path.getmtime):
+        with st.expander(os.path.basename(png)):
+            st.image(png, use_container_width=True)
+            st.markdown(get_download_link(png, "png"), unsafe_allow_html=True)
+
+    st.subheader("🎥 Videos (MP4)")
+    for mp4 in sorted(mp4_files, key=os.path.getmtime):
+        with st.expander(os.path.basename(mp4)):
+            st.video(mp4)
+            st.markdown(get_download_link(mp4, "mp4"), unsafe_allow_html=True)
+
+    # 🗂️ Sidebar with Dialog and Audio
+    st.sidebar.subheader("Voice Settings")
+    new_username = st.sidebar.selectbox("Change Name/Voice", list(FUN_USERNAMES.keys()), index=list(FUN_USERNAMES.keys()).index(st.session_state.username), key="username_select")
+    if new_username != st.session_state.username:
+        asyncio.run(save_chat_entry("System 🌟", f"{st.session_state.username} changed to {new_username}", "en-US-AriaNeural"))
+        st.session_state.username, st.session_state.tts_voice = new_username, FUN_USERNAMES[new_username]
+        st.session_state.timer_start = time.time()
+        save_username(st.session_state.username)
+        st.rerun()
+
+    st.sidebar.markdown("### 💬 Chat Dialog")
+    chat_content = asyncio.run(load_chat())
+    with st.sidebar.expander("Chat History"):
+        numbered_content = "\n".join(f"{i+1}. {line}" for i, line in enumerate(chat_content))
+        st.code(numbered_content, language="python")
+
+    st.sidebar.markdown("### 💬 Chat Text Only")
+    chat_text_only = get_chat_text_only()
+    with st.sidebar.expander("Text Only History"):
+        numbered_text = "\n".join(f"{i+1}. {line}" for i, line in enumerate(chat_text_only.split('\n')))
+        st.code(numbered_text, language="python")
+
+    st.sidebar.subheader("Vote Totals")
+    chat_votes = load_votes(QUOTE_VOTES_FILE)
+    image_votes = load_votes(IMAGE_VOTES_FILE)
+    for item, count in chat_votes.items():
+        st.sidebar.write(f"{item}: {count} votes")
+    for image, count in image_votes.items():
+        st.sidebar.write(f"{image}: {count} votes")
+
+    st.sidebar.markdown("### 📂 File History")
+    for f in all_files[:10]:
+        st.sidebar.write(f"{FILE_EMOJIS.get(f.split('.')[-1], '📄')} {os.path.basename(f)}")
+
+    st.sidebar.subheader("📦 Zip & Delete")
+    col_zip, col_del = st.sidebar.columns(2)
+    with col_zip:
+        if st.button("⬇️ Zip All", key="zip_all"):
+            zip_name = create_zip_of_files(all_files, "All")
+            if zip_name:
+                st.session_state['download_link_cache'] = {}
+        if st.button("⬇️ Zip All MD", key="zip_md"):
+            zip_name = create_zip_of_files(md_files, "MD")
+            if zip_name:
+                st.session_state['download_link_cache'] = {}
+        if st.button("⬇️ Zip All MP3", key="zip_mp3"):
+            zip_name = create_zip_of_files(mp3_files, "MP3")
+            if zip_name:
+                st.session_state['download_link_cache'] = {}
+        if st.button("⬇️ Zip All PNG", key="zip_png"):
+            zip_name = create_zip_of_files(png_files, "PNG")
+            if zip_name:
+                st.session_state['download_link_cache'] = {}
+        if st.button("⬇️ Zip All MP4", key="zip_mp4"):
+            zip_name = create_zip_of_files(mp4_files, "MP4")
+            if zip_name:
+                st.session_state['download_link_cache'] = {}
+    with col_del:
+        if st.button("🗑️ Del All", key="del_all"):
+            for ft in ["md", "mp3", "png", "mp4"]:
+                delete_files(ft)
+            st.rerun()
+        if st.button("🗑️ Del All MD", key="del_md"):
+            delete_files("md")
+            st.rerun()
+        if st.button("🗑️ Del All MP3", key="del_mp3"):
+            delete_files("mp3")
+            st.rerun()
+        if st.button("🗑️ Del All PNG", key="del_png"):
+            delete_files("png")
+            st.rerun()
+        if st.button("🗑️ Del All MP4", key="del_mp4"):
+            delete_files("mp4")
+            st.rerun()
+        if st.button("🗑️ Del All Zip", key="del_zip"):
+            delete_files("zip", exclude_files=[])
+            st.rerun()
+
+    zip_files = sorted(glob.glob("*.zip"), key=os.path.getmtime, reverse=True)
+    for zip_file in zip_files:
+        st.sidebar.markdown(get_download_link(zip_file, "zip"), unsafe_allow_html=True)
+
+    # Refresh Timer in Sidebar
+    st.sidebar.subheader("Set Refresh Rate ⏳")
+    st.session_state['auto_refresh'] = st.sidebar.radio("Auto Refresh", ["On", "Off"], index=0 if st.session_state['auto_refresh'] else 1) == "On"
+    st.markdown("""
+        <style>
+        .timer {
+            font-size: 24px;
+            color: #ffcc00;
+            text-align: center;
+            animation: pulse 1s infinite;
+        }
+        @keyframes pulse {
+            0% { transform: scale(1); }
+            50% { transform: scale(1.1); }
+            100% { transform: scale(1); }
+        }
+        </style>
+    """, unsafe_allow_html=True)
+
+    refresh_rate = st.sidebar.slider("Refresh Rate (seconds)", min_value=1, max_value=300, value=st.session_state.refresh_rate, step=1)
+    if refresh_rate != st.session_state.refresh_rate:
+        st.session_state.refresh_rate = refresh_rate
+        st.session_state.timer_start = time.time()
+        save_username(st.session_state.username)
+
+    col1, col2, col3 = st.sidebar.columns(3)
+    with col1:
+        if st.button("🐇 Small (1s)"):
+            st.session_state.refresh_rate = 1
+            st.session_state.timer_start = time.time()
+            save_username(st.session_state.username)
+    with col2:
+        if st.button("🐢 Medium (10s)"):
+            st.session_state.refresh_rate = 10
+            st.session_state.timer_start = time.time()
+            save_username(st.session_state.username)
+    with col3:
+        if st.button("🐘 Large (5m)"):
+            st.session_state.refresh_rate = 300
+            st.session_state.timer_start = time.time()
+            save_username(st.session_state.username)
+
+    timer_placeholder = st.sidebar.empty()
+    start_time = st.session_state.timer_start
+    remaining_time = int(st.session_state.refresh_rate - (time.time() - start_time))
+    if st.session_state['auto_refresh'] and remaining_time <= 0:
+        st.session_state.timer_start = time.time()
+        st.session_state.last_refresh = time.time()
+        st.rerun()
+    else:
+        timer_placeholder.markdown(f"<p class='timer'>⏳ Next refresh in: {remaining_time} seconds</p>", unsafe_allow_html=True)
+
+    if not st.session_state.get('server_running', False) and not st.session_state.get('server_task', None):
+        st.session_state.server_task = threading.Thread(target=start_websocket_server, daemon=True)
+        st.session_state.server_task.start()
+
+if __name__ == "__main__":
+    main()