Update app.py

app.py

@@ -1,606 +1,866 @@
-from fastapi import FastAPI, HTTPException
-from fastapi.responses import HTMLResponse
-from transformers import AutoModelForCausalLM, AutoTokenizer
+from fastapi import FastAPI, UploadFile, File, Form, HTTPException, Request
+from fastapi.staticfiles import StaticFiles
+from fastapi.responses import RedirectResponse, JSONResponse, HTMLResponse
+from transformers import pipeline, ViltProcessor, ViltForQuestionAnswering, M2M100ForConditionalGeneration, M2M100Tokenizer
+from typing import Optional, Dict, Any, List
+import logging
+import time
+import os
+import io
+import json
+import re
+from PIL import Image
+from docx import Document
+import fitz  # PyMuPDF
+import pandas as pd
+from functools import lru_cache
+import torch
+import numpy as np
+from pydantic import BaseModel
+import asyncio
+import google.generativeai as genai
+from spellchecker import SpellChecker
+import nltk
+from nltk.tokenize import sent_tokenize
+
+# Configure logging
+logging.basicConfig(
+    level=logging.INFO,
+    format='%(asctime)s - %(name)s - %(levelname)s - %(message)s'
+)
+logger = logging.getLogger("cosmic_ai")
+
+# Set a custom NLTK data directory
+nltk_data_dir = os.getenv('NLTK_DATA_DIR', '/tmp/nltk_data')
+os.makedirs(nltk_data_dir, exist_ok=True)
+nltk.data.path.append(nltk_data_dir)
+
+# Download punkt_tab data if not already present
+try:
+    nltk.download('punkt_tab', download_dir=nltk_data_dir, quiet=True, raise_on_error=True)
+    logger.info(f"NLTK punkt_tab verified in {nltk_data_dir}")
+except Exception as e:
+    logger.error(f"Error verifying NLTK punkt_tab: {str(e)}")
+    raise Exception(f"Failed to verify NLTK punkt_tab: {str(e)}")
+
+# Create app directory if it doesn't exist
+upload_dir = os.getenv('UPLOAD_DIR', '/tmp/uploads')
+os.makedirs(upload_dir, exist_ok=True)
+
+app = FastAPI(
+    title="Cosmic AI Assistant",
+    description="An advanced AI assistant with space-themed interface, translation, and file question-answering features",
+    version="2.0.0"
+)
+
+# Mount static files
+app.mount("/static", StaticFiles(directory="static"), name="static")
+
+# Mount images directory
+app.mount("/images", StaticFiles(directory="images"), name="images")
+
+# Gemini API Configuration
+API_KEY = "AIzaSyDtLhhmXpy8ubSGb84ImaxM_ywlL0l_8bo"  # Replace with your actual API key
+genai.configure(api_key=API_KEY)
+
+# Model configurations
+MODELS = {
+    "summarization": "sshleifer/distilbart-cnn-12-6",
+    "image-to-text": "Salesforce/blip-image-captioning-large",
+    "visual-qa": "dandelin/vilt-b32-finetuned-vqa",
+    "chatbot": "gemini-1.5-pro",
+    "translation": "facebook/m2m100_418M",
+    "file-qa": "distilbert-base-cased-distilled-squad"
+}
+
+# Supported languages for translation
+SUPPORTED_LANGUAGES = {
+    "english": "en",
+    "french": "fr",
+    "german": "de",
+    "spanish": "es",
+    "italian": "it",
+    "russian": "ru",
+    "chinese": "zh",
+    "japanese": "ja",
+    "arabic": "ar",
+    "hindi": "hi",
+    "portuguese": "pt",
+    "korean": "ko"
+}
+
+# Global variables for pre-loaded translation model
+translation_model = None
+translation_tokenizer = None
+
+# Initialize spell checker
+spell = SpellChecker()
+
+# Cache for model loading (excluding translation)
+@lru_cache(maxsize=8)
+def load_model(task: str, model_name: str = None):
+    """Cached model loader with proper task names and error handling"""
+    try:
+        logger.info(f"Loading model for task: {task}, model: {model_name or MODELS.get(task)}")
+        start_time = time.time()
+        
+        model_to_load = model_name or MODELS.get(task)
+        
+        if task == "chatbot":
+            return genai.GenerativeModel(model_to_load)
+            
+        if task == "visual-qa":
+            processor = ViltProcessor.from_pretrained(model_to_load)
+            model = ViltForQuestionAnswering.from_pretrained(model_to_load)
+            device = "cuda" if torch.cuda.is_available() else "cpu"
+            model.to(device)
+            
+            def vqa_function(image, question, **generate_kwargs):
+                if image.mode != "RGB":
+                    image = image.convert("RGB")
+                inputs = processor(image, question, return_tensors="pt").to(device)
+                logger.info(f"VQA inputs - question: {question}, image size: {image.size}")
+                with torch.no_grad():
+                    outputs = model(**inputs)
+                logits = outputs.logits
+                idx = logits.argmax(-1).item()
+                answer = model.config.id2label[idx]
+                logger.info(f"VQA raw output: {answer}")
+                return answer
+            
+            return vqa_function
+        
+        # Use pipeline for summarization, image-to-text, and file-qa
+        return pipeline(task if task != "file-qa" else "question-answering", model=model_to_load)
+        
+    except Exception as e:
+        logger.error(f"Model load failed: {str(e)}")
+        raise HTTPException(status_code=500, detail=f"Model loading failed: {task} - {str(e)}")
 
-app = FastAPI(title="Chatbot")
+def get_gemini_response(user_input: str, is_generation: bool = False):
+    """Function to generate response with Gemini for both chat and text generation"""
+    if not user_input:
+        return "Please provide some input."
+    try:
+        chatbot = load_model("chatbot")
+        if is_generation:
+            prompt = f"Generate creative text based on this prompt: {user_input}"
+        else:
+            prompt = user_input
+        response = chatbot.generate_content(prompt)
+        return response.text.strip()
+    except Exception as e:
+        return f"Error: {str(e)}"
 
-# Load the chatbot model (DialoGPT-medium) at startup
-model_name = "microsoft/DialoGPT-medium"
-tokenizer = AutoTokenizer.from_pretrained(model_name)
-model = AutoModelForCausalLM.from_pretrained(model_name)
+def translate_text(text: str, target_language: str):
+    """Translate text to any target language using pre-loaded M2M100 model"""
+    if not text:
+        return "Please provide text to translate."
+    
+    try:
+        global translation_model, translation_tokenizer
+        
+        target_lang = target_language.lower()
+        if target_lang not in SUPPORTED_LANGUAGES:
+            similar = [lang for lang in SUPPORTED_LANGUAGES if target_lang in lang or lang in target_lang]
+            if similar:
+                target_lang = similar[0]
+            else:
+                return f"Language '{target_language}' not supported. Available languages: {', '.join(SUPPORTED_LANGUAGES.keys())}"
+        
+        lang_code = SUPPORTED_LANGUAGES[target_lang]
+        
+        if translation_model is None or translation_tokenizer is None:
+            raise Exception("Translation model not initialized")
+        
+        match = re.search(r'how to say\s+(.+?)\s+in\s+(\w+)', text.lower())
+        if match:
+            text_to_translate = match.group(1)
+        else:
+            content_match = re.search(r'(?:translate|convert).*to\s+[a-zA-Z]+\s*[:\s]*(.+)', text, re.IGNORECASE)
+            text_to_translate = content_match.group(1) if content_match else text
+        
+        translation_tokenizer.src_lang = "en"
+        encoded = translation_tokenizer(text_to_translate, return_tensors="pt", padding=True, truncation=True).to(translation_model.device)
+        
+        start_time = time.time()
+        generated_tokens = translation_model.generate(
+            **encoded,
+            forced_bos_token_id=translation_tokenizer.get_lang_id(lang_code),
+            max_length=512,
+            num_beams=1,
+            early_stopping=True
+        )
+        translated_text = translation_tokenizer.batch_decode(generated_tokens, skip_special_tokens=True)[0]
+        logger.info(f"Translation took {time.time() - start_time:.2f} seconds")
+        
+        return translated_text
+        
+    except Exception as e:
+        logger.error(f"Translation error: {str(e)}", exc_info=True)
+        return f"Translation error: {str(e)}"
+
+def detect_intent(text: str = None, file: UploadFile = None, intent: str = None) -> tuple[str, str]:
+    """Enhanced intent detection with explicit intent parameter support"""
+    target_language = "English"  # Default
+    valid_intents = [
+        "chatbot", "translate", "file-translate", "summarize", "image-to-text",
+        "visual-qa", "visualize", "text-generation", "file-qa"
+    ]
+
+    # Check if an explicit intent is provided and valid
+    if intent and intent in valid_intents:
+        logger.info(f"Using explicit intent: {intent}")
+        # For translation intents, check if target language is specified in text
+        if intent in ["translate", "file-translate"] and text:
+            translate_patterns = [
+                r'translate.*to\s+\[?([a-zA-Z]+)\]?:?\s*(.*)',
+                r'convert.*to\s+\[?([a-zA-Z]+)\]?:?\s*(.*)',
+                r'how to say.*in\s+\[?([a-zA-Z]+)\]?:?\s*(.*)'
+            ]
+            for pattern in translate_patterns:
+                translate_match = re.search(pattern, text.lower())
+                if translate_match:
+                    potential_lang = translate_match.group(1).lower()
+                    if potential_lang in SUPPORTED_LANGUAGES:
+                        target_language = potential_lang.capitalize()
+                        break
+        return intent, target_language
+
+    # Existing intent detection logic for cases where intent is not provided
+    if file and text:
+        text_lower = text.lower()
+        filename = file.filename.lower() if file.filename else ""
+        
+        # Check for file translation intent
+        translate_patterns = [
+            r'translate.*to\s+\[?([a-zA-Z]+)\]?:?\s*(.*)',
+            r'convert.*to\s+\[?([a-zA-Z]+)\]?:?\s*(.*)',
+            r'how to say.*in\s+\[?([a-zA-Z]+)\]?:?\s*(.*)'
+        ]
+        for pattern in translate_patterns:
+            translate_match = re.search(pattern, text_lower)
+            if translate_match and filename.endswith(('.pdf', '.docx', '.txt', '.rtf')):
+                potential_lang = translate_match.group(1).lower()
+                if potential_lang in SUPPORTED_LANGUAGES:
+                    target_language = potential_lang.capitalize()
+                    return "file-translate", target_language
+        
+        # Image-related intents
+        content_type = file.content_type.lower() if file.content_type else ""
+        if content_type.startswith('image/') and text:
+            if "what’s this" in text_lower or "does this fly" in text_lower or ("fly" in text_lower and any(q in text_lower for q in ['does', 'can', 'will'])):
+                return "visual-qa", target_language
+            if any(q in text_lower for q in ['what is', 'what\'s', 'describe', 'tell me about', 'explain', 'how many', 'what color', 'is there', 'are they', 'does the']):
+                return "visual-qa", target_language
+            if "generate a caption" in text_lower or "caption" in text_lower:
+                return "image-to-text", target_language
+        
+        # File-related intents
+        if filename.endswith(('.xlsx', '.xls', '.csv')):
+            return "visualize", target_language
+        elif filename.endswith(('.pdf', '.docx', '.doc', '.txt', '.rtf')):
+            if any(q in text_lower for q in ['what is', 'who is', 'where', 'when', 'why', 'how', 'what are', 'who are']):
+                return "file-qa", target_language
+            return "summarize", target_language
+    
+    if not text:
+        # If only a file is provided, infer intent based on file type
+        if file:
+            filename = file.filename.lower() if file.filename else ""
+            content_type = file.content_type.lower() if file.content_type else ""
+            if content_type.startswith('image/'):
+                return "image-to-text", target_language  # Default to image-to-text for images
+            elif filename.endswith(('.pdf', '.docx', '.doc', '.txt', '.rtf')):
+                return "summarize", target_language  # Default to summarize for text files
+            elif filename.endswith(('.xlsx', '.xls', '.csv')):
+                return "visualize", target_language
+        return "chatbot", target_language
+    
+    text_lower = text.lower()
+    
+    if any(keyword in text_lower for keyword in ['chat', 'talk', 'converse', 'ask gemini']):
+        return "chatbot", target_language
+
+    # Text translation intent
+    translate_patterns = [
+        r'translate.*to\s+\[?([a-zA-Z]+)\]?:?\s*(.*)',
+        r'convert.*to\s+\[?([a-zA-Z]+)\]?:?\s*(.*)',
+        r'how to say.*in\s+\[?([a-zA-Z]+)\]?:?\s*(.*)'
+    ]
+    
+    for pattern in translate_patterns:
+        translate_match = re.search(pattern, text_lower)
+        if translate_match:
+            potential_lang = translate_match.group(1).lower()
+            if potential_lang in SUPPORTED_LANGUAGES:
+                target_language = potential_lang.capitalize()
+                return "translate", target_language
+            else:
+                logger.warning(f"Invalid language detected: {potential_lang}")
+                return "chatbot", target_language
+
+    vqa_patterns = [
+        r'how (many|much)',
+        r'what (color|size|position|shape)',
+        r'is (there|that|this) (a|an)',
+        r'are (they|there) (any|some)',
+        r'does (the|this) (image|picture) (show|contain)'
+    ]
+    
+    if any(re.search(pattern, text_lower) for pattern in vqa_patterns):
+        return "visual-qa", target_language
+
+    summarization_patterns = [
+        r'\b(summar(y|ize|ise)|brief( overview)?)\b',
+        r'\b(long article|text|document)\b',
+        r'\bcan you (summar|brief|condense)\b',
+        r'\b(short summary|brief explanation)\b',
+        r'\b(overview|main points|key ideas)\b',
+        r'\b(tl;?dr|too long didn\'?t read)\b'
+    ]
+    
+    if any(re.search(pattern, text_lower) for pattern in summarization_patterns):
+        return "summarize", target_language
+    
+    generation_patterns = [
+        r'\b(write|generate|create|compose)\b',
+        r'\b(story|poem|essay|text|content)\b'
+    ]
+    
+    if any(re.search(pattern, text_lower) for pattern in generation_patterns):
+        return "text-generation", target_language
+    
+    if len(text) > 100:
+        return "summarize", target_language
+    
+    return "chatbot", target_language
+
+def preprocess_text(text: str) -> str:
+    """Correct spelling errors and improve text readability."""
+    words = text.split()
+    corrected_words = [spell.correction(word) if spell.correction(word) else word for word in words]
+    corrected_text = " ".join(corrected_words)
+    sentences = sent_tokenize(corrected_text)
+    return ". ".join(sentence.capitalize() for sentence in sentences) + (". " if sentences else "")
+
+class ProcessResponse(BaseModel):
+    response: str
+    type: str
+    additional_data: Optional[Dict[str, Any]] = None
+
+@app.get("/chatbot")
+async def chatbot_interface():
+    """Redirect to the static index.html file for the chatbot interface"""
+    return RedirectResponse(url="/static/index.html")
 
-# Function to generate chatbot response
-def get_chatbot_response(user_input: str, max_length=1000):
-    if not user_input:
-        return "Please say something!"
+@app.post("/chat")
+async def chat_endpoint(data: dict):
+    """Endpoint for chatbot interactions"""
+    message = data.get("message", "")
+    if not message:
+        raise HTTPException(status_code=400, detail="No message provided")
+    try:
+        response = get_gemini_response(message)
+        return {"response": response}
+    except Exception as e:
+        raise HTTPException(status_code=500, detail=f"Chat error: {str(e)}")
+
+@app.post("/process", response_model=ProcessResponse)
+async def process_input(
+    request: Request,
+    text: str = Form(None),
+    file: UploadFile = File(None),
+    intent: str = Form(None)
+):
+    """Enhanced unified endpoint with dynamic translation and file translation"""
+    start_time = time.time()
+    client_ip = request.client.host
+    logger.info(f"Request from {client_ip}: text={text[:50] + '...' if text and len(text) > 50 else text}, file={file.filename if file else None}, intent={intent}")
     
-    input_ids = tokenizer.encode(user_input + tokenizer.eos_token, return_tensors="pt")
-    chat_history_ids = model.generate(
-        input_ids,
-        max_length=max_length,
-        pad_token_id=tokenizer.eos_token_id,
-        no_repeat_ngram_size=3,
-        do_sample=True,
-        top_k=50,
-        top_p=0.95,
-        temperature=0.8
-    )
-    response = tokenizer.decode(chat_history_ids[:, input_ids.shape[-1]:][0], skip_special_tokens=True)
-    return response.strip()
-
-# HTML, CSS, and JS with two modes (dark and soft)
-HTML_CONTENT = """
-<!DOCTYPE html>
-<html lang="en">
-<head>
-    <meta charset="UTF-8">
-    <meta name="viewport" content="width=device-width, initial-scale=1.0">
-    <meta name="description" content="Vion IA: Your friendly AI chatbot powered by xAI">
-    <link rel="stylesheet" href="https://cdnjs.cloudflare.com/ajax/libs/highlight.js/11.9.0/styles/github-dark.min.css">
-    <link href='https://unpkg.com/boxicons@2.1.4/css/boxicons.min.css' rel='stylesheet'>
-    <title>Vion IA | Powered by xAI</title>
-    <style>
-        @import url('https://fonts.googleapis.com/css2?family=Open+Sans:wght@300;400;600;800&display=swap');
-        * { margin: 0; padding: 0; outline: none; box-sizing: border-box; font-family: "Open Sans", sans-serif; }
-        :root {
-            --primary-color: #0A0A0B; --secondary-color: #1A1B1E; --secondary-hover-color: #2A2B2E;
-            --focus-color: #242528; --focus-hover-color: #343538; --button-hover-color: #252627;
-            --text-color: #FFFFFF; --text-secondary-color: #A0A1A3; --heading-secondary-color: #606162;
-            --placeholder-color: #78797A; --accent-color: #00A3FF; --star-color: #FFFFFF;
-            --background-color: #000; --error-color: #FF4D4D; --success-color: #4CAF50;
-        }
-        .soft_mode {
-            --primary-color: rgb(220, 230, 240); /* Soft pastel blue */
-            --secondary-color: rgb(235, 240, 245); /* Lighter pastel */
-            --secondary-hover-color: rgb(200, 210, 225);
-            --focus-color: rgb(245, 245, 250); --focus-hover-color: rgb(230, 230, 235);
-            --button-hover-color: rgb(210, 220, 230);
-            --text-color: rgb(40, 50, 60); --text-secondary-color: rgb(90, 100, 110);
-            --heading-secondary-color: rgb(150, 160, 170);
-            --placeholder-color: rgb(120, 130, 140); --accent-color: rgb(100, 150, 255); /* Softer blue */
-            --star-color: rgb(180, 190, 200); --background-color: rgb(240, 245, 250); /* Very light pastel */
-            --error-color: rgb(255, 100, 100); --success-color: rgb(100, 200, 100);
-        }
-        body {
-            min-height: 100vh; display: flex; flex-direction: row; justify-content: space-between;
-            background: var(--background-color); position: relative; overflow-y: hidden; transition: background 0.3s ease;
-        }
-        .stars {
-            position: fixed; width: 100%; height: 100%;
-            background: url('https://www.transparenttextures.com/patterns/stardust.png');
-            animation: starsAnim 100s linear infinite; z-index: -1; transition: opacity 0.3s ease; opacity: 0.8;
-        }
-        .soft_mode .stars {
-            background: url('https://www.transparenttextures.com/patterns/stardust.png') rgba(240, 245, 250, 0.4);
-            opacity: 0.3;
-        }
-        @keyframes starsAnim { from { background-position: 0 0; } to { background-position: 10000px 10000px; } }
-        @keyframes pulse { 0% { transform: scale(1); } 50% { transform: scale(1.1); } 100% { transform: scale(1); } }
-        @keyframes fadeIn { from { opacity: 0; transform: translateY(10px); } to { opacity: 1; transform: translateY(0); } }
-        .sidebar {
-            width: 70px; height: 100vh; background: rgba(26, 27, 30, 0.95); display: flex; flex-direction: column;
-            align-items: center; padding: 1rem 0; position: fixed; left: 0; top: 0; transition: width 0.3s ease; z-index: 1001;
-        }
-        .sidebar:hover { width: 200px; }
-        .soft_mode .sidebar { background: rgba(235, 240, 245, 0.95); }
-        .sidebar__item {
-            width: 100%; padding: 1rem; color: var(--text-secondary-color); text-decoration: none;
-            display: flex; align-items: center; gap: 1rem; transition: all 0.3s ease; position: relative;
-        }
-        .sidebar__item:hover, .sidebar__item:focus {
-            background: var(--secondary-hover-color); color: var(--accent-color); padding-left: 1.5rem; transform: scale(1.05);
-        }
-        .sidebar__item i { font-size: 1.5rem; }
-        .sidebar__item span { display: none; font-size: 1rem; }
-        .sidebar:hover .sidebar__item span { display: inline; }
-        .tooltip {
-            visibility: hidden; background: var(--secondary-color); color: var(--text-color); font-size: 0.8rem;
-            padding: 0.5rem; border-radius: 0.3rem; position: absolute; top: -30px; left: 50%; transform: translateX(-50%);
-            white-space: nowrap; z-index: 1002; transition: visibility 0.2s, opacity 0.2s; opacity: 0;
-        }
-        .sidebar__item:hover .tooltip, .sidebar__item:focus .tooltip { visibility: visible; opacity: 1; }
-        .main-content {
-            flex: 1; display: flex; flex-direction: column; padding-bottom: 100px; padding-top: 2rem; margin-left: 70px;
-            height: 50vh; overflow: hidden;
-        }
-        .header { max-width: 900px; text-align: center; padding: 0 2rem; margin: 0 auto; }
-        .header__title h1 {
-            color: var(--text-color); font-size: 3.5rem; font-weight: 800; margin-bottom: 1rem;
-            text-shadow: 0 0 10px rgba(0, 163, 255, 0.5); animation: fadeIn 1s ease-in;
-        }
-        .header__title h2 {
-            color: var(--text-secondary-color); font-size: 1.5rem; font-weight: 400;
-            max-width: 600px; margin: 0 auto; text-shadow: 0 0 5px rgba(0, 0, 0, 0.3);
-            transition: opacity 0.3s ease, height 0.3s ease;
-        }
-        .suggests {
-            display: grid; grid-template-columns: repeat(auto-fit, minmax(200px, 1fr));
-            margin: 2rem auto; max-width: 900px; gap: 1rem; padding: 0 2rem; transition: opacity 0.3s ease, height 0.3s ease;
-            animation: fadeIn 0.5s ease-in;
-        }
-        .suggests.hidden, .header__title h2.hidden {
-            opacity: 0; height: 0; margin: 0; overflow: hidden;
-        }
-        .suggests__item {
-            background: rgba(26, 27, 30, 0.9); color: var(--text-secondary-color); padding: 1.5rem;
-            border-radius: 0.5rem; cursor: pointer; transition: all 0.3s ease; border: 1px solid var(--focus-color);
-            position: relative;
-        }
-        .soft_mode .suggests__item { background: rgba(235, 240, 245, 0.9); }
-        .suggests__item:hover, .suggests__item:focus {
-            background: var(--secondary-hover-color); border-color: var(--accent-color); color: var(--text-color);
-            transform: translateY(-3px);
-        }
-        .suggests__item-icon { margin-top: 1rem; color: var(--accent-color); transition: transform 0.2s ease; }
-        .suggests__item:hover .suggests__item-icon, .suggests__item:focus .suggests__item-icon { transform: scale(1.2); }
-        .suggests__item .tooltip { top: -40px; left: 50%; transform: translateX(-50%); }
-        .suggests__item:hover .tooltip, .suggests__item:focus .tooltip { visibility: visible; opacity: 1; }
-        .prompt {
-            position: fixed; background: rgba(10, 10, 11, 0.9); z-index: 1000; width: calc(100% - 70px);
-            left: 70px; bottom: 0; padding: 1rem; border-top: 1px solid var(--secondary-color); transition: all 0.3s ease;
-        }
-        .soft_mode .prompt { background: rgba(235, 240, 245, 0.9); border-top: 1px solid var(--focus-color); }
-        .prompt__input-wrapper {
-            max-width: 900px; margin: 0 auto; position: relative; display: flex; align-items: center;
-            background: var(--secondary-color); border: 1px solid var(--focus-color); border-radius: 0.5rem;
-            padding: 0.2rem; transition: all 0.3s ease; animation: fadeIn 0.5s ease-in;
-        }
-        .prompt__input-wrapper:focus-within {
-            border-color: var(--accent-color); background: var(--focus-color); transform: scale(1.02);
-        }
-        .prompt__input-wrapper.dragover {
-            border: 2px dashed var(--accent-color); background: var(--focus-hover-color);
-        }
-        .prompt__form-input {
-            flex-grow: 1; border: none; resize: none; font-size: 1.1rem; color: var(--text-color);
-            padding: 0.3rem 0.5rem; background: transparent; outline: none; transition: all 0.3s ease;
-        }
-        .prompt__form-input::placeholder { color: var(--placeholder-color); transition: opacity 0.3s ease; }
-        .prompt__form-input:focus::placeholder { opacity: 0.5; }
-        .prompt__action-buttons {
-            display: flex; align-items: center; gap: 0.3rem; padding-right: 0.3rem; position: relative;
-        }
-        .prompt__action-buttons.advanced { display: none; }
-        .prompt__action-buttons.advanced.active { display: flex; }
-        .prompt__form-button {
-            background: none; border: none; color: var(--text-secondary-color); font-size: 1.3rem;
-            cursor: pointer; padding: 0.3rem; transition: all 0.3s ease; position: relative;
-        }
-        .prompt__form-button:hover, .prompt__form-button:focus { color: var(--accent-color); transform: scale(1.1); }
-        .prompt__form-button.send { font-size: 1.5rem; }
-        .prompt__form-button .tooltip { top: -35px; left: 50%; transform: translateX(-50%); }
-        .prompt__form-button:hover .tooltip, .prompt__form-button:focus .tooltip { visibility: visible; opacity: 1; }
-        .prompt__disclaim {
-            text-align: center; color: var(--placeholder-color); font-size: 0.8rem; margin-top: 1rem;
-            max-width: 900px; margin-left: auto; margin-right: auto; transition: opacity 0.3s ease;
-        }
-        .chat-bar {
-            max-width: 900px; margin: 2rem auto; padding: 0 2rem; display: flex; flex-direction: column;
-            overflow-y: auto; max-height: calc(100vh - 180px); -ms-overflow-style: none; scrollbar-width: none;
-        }
-        .chat-bar::-webkit-scrollbar { display: none; }
-        .chat-message {
-            margin-bottom: 1rem; padding: 1rem; border-radius: 0.5rem; background: rgba(26, 27, 30, 0.9);
-            color: var(--text-color); word-wrap: break-word; animation: fadeIn 0.3s ease-in; position: relative;
-        }
-        .soft_mode .chat-message { background: rgba(235, 240, 245, 0.9); }
-        .chat-message.user {
-            background: rgba(122, 92, 250, 0.2); border: 1px solid var(--accent-color); border-radius: 0.5rem;
-        }
-        .chat-message.bot { background: rgba(36, 37, 40, 0.9); }
-        .soft_mode .chat-message.bot { background: rgba(245, 245, 250, 0.9); }
-        .chat-message.user.bubble-rounded { border-radius: 1rem; }
-        .chat-message.user.bubble-sharp { border-radius: 0; border: 2px solid var(--accent-color); }
-        .chat-message.user.bubble-starry {
-            border-radius: 0.5rem; border: 1px dashed var(--accent-color);
-            background: rgba(122, 92, 250, 0.2) url('https://www.transparenttextures.com/patterns/stardust.png') repeat;
-            background-size: 100px 100px;
-        }
-        .chat-message.feedback::after {
-            content: 'Was this helpful?'; color: var(--text-secondary-color); font-size: 0.8rem; display: block;
-            margin-top: 0.5rem; cursor: pointer; text-decoration: underline;
-        }
-        .chat-message.feedback .feedback-options {
-            display: none; position: absolute; bottom: -30px; left: 1rem; gap: 0.5rem;
-        }
-        .chat-message.feedback:hover .feedback-options { display: flex; }
-        .feedback-options button {
-            background: none; border: none; color: var(--text-secondary-color); font-size: 1rem; cursor: pointer;
-            transition: color 0.2s ease;
-        }
-        .feedback-options button:hover, .feedback-options button:focus { color: var(--accent-color); }
-        .error-message {
-            background: rgba(255, 77, 77, 0.2); border: 1px solid var(--error-color); color: var(--text-color);
-            padding: 1rem; border-radius: 0.5rem; margin-bottom: 1rem; animation: fadeIn 0.3s ease-in;
-            display: flex; justify-content: space-between; align-items: center;
-        }
-        .error-message button {
-            background: var(--error-color); color: var(--text-color); border: none; padding: 0.3rem 0.6rem;
-            border-radius: 0.3rem; cursor: pointer; transition: background 0.2s ease;
-        }
-        .error-message button:hover, .error-message button:focus { background: var(--button-hover-color); }
-        .back-to-latest {
-            display: none; position: fixed; bottom: 100px; right: 2rem; background: var(--secondary-color);
-            color: var(--text-color); padding: 0.5rem 1rem; border-radius: 0.5rem; cursor: pointer;
-            border: 1px solid var(--accent-color); transition: all 0.3s ease; z-index: 1000;
-        }
-        .back-to-latest.visible { display: block; }
-        .back-to-latest:hover, .back-to-latest:focus { background: var(--secondary-hover-color); transform: scale(1.05); }
-        .processing-dots {
-            display: none; position: absolute; right: 60px; color: var(--accent-color); font-size: 1.2rem;
-        }
-        .processing-dots.active { display: inline; animation: pulse 1.5s infinite; }
-        @keyframes blink {
-            0% { opacity: 1; } 50% { opacity: 0.3; } 100% { opacity: 1; }
-        }
-        .processing-dots span {
-            animation: blink 1s infinite; animation-delay: calc(0.2s * var(--i));
-        }
-    </style>
-</head>
-<body>
-    <div class="stars"></div>
-    <nav class="sidebar" aria-label="Main navigation">
-        <a href="#" class="sidebar__item" tabindex="0" aria-label="Home"><i class='bx bx-home'></i><span>Home</span><div class="tooltip">Go to Home</div></a>
-        <a href="#" class="sidebar__item" tabindex="0" aria-label="Profile"><i class='bx bx-user'></i><span>Profile</span><div class="tooltip">View Profile</div></a>
-        <a href="#" class="sidebar__item" tabindex="0" aria-label="Settings"><i class='bx bx-cog'></i><span>Settings</span><div class="tooltip">Adjust Settings</div></a>
-        <a href="#" class="sidebar__item" tabindex="0" aria-label="Help"><i class='bx bx-help-circle'></i><span>Help</span><div class="tooltip">Get Help</div></a>
-    </nav>
-    <div class="main-content">
-        <header class="header">
-            <div class="header__title">
-                <h1>Vion IA</h1>
-                <h2 id="welcome-text">Ask me anything and I'll provide helpful and truthful answers from an outside perspective on humanity.</h2>
-            </div>
-        </header>
-        <div class="suggests">
-            <div class="suggests__item" tabindex="0"><p class="suggests__item-text">What is the meaning of life?</p><div class="suggests__item-icon"><i class='bx bx-bulb'></i></div><div class="tooltip">Explore life's purpose</div></div>
-            <div class="suggests__item" tabindex="0"><p class="suggests__item-text">Explain quantum physics simply</p><div class="suggests__item-icon"><i class='bx bx-atom'></i></div><div class="tooltip">Learn about quantum physics</div></div>
-            <div class="suggests__item" tabindex="0"><p class="suggests__item-text">How does the universe work?</p><div class="suggests__item-icon"><i class='bx bx-planet'></i></div><div class="tooltip">Discover the universe</div></div>
-        </div>
-        <div class="chat-bar" id="chatBar" aria-live="polite"></div>
-        <button class="back-to-latest" id="backToLatest" tabindex="0">Back to Latest</button>
-    </div>
-    <section class="prompt">
-        <form action="#" class="prompt__form" novalidate>
-            <div class="prompt__input-wrapper">
-                <input type="text" placeholder="Ask me anything..." class="prompt__form-input" id="chatInput" required aria-label="Chat input">
-                <div class="prompt__action-buttons basic">
-                    <button type="button" class="prompt__form-button send" id="sendButton" tabindex="0" aria-label="Send message"><i class='bx bx-send'></i><div class="tooltip">Send Message (Ctrl+S)</div></button>
-                    <button type="button" class="prompt__form-button" id="moreOptions" tabindex="0" aria-label="Show more options"><i class='bx bx-dots-horizontal-rounded'></i><div class="tooltip">More Options</div></button>
-                </div>
-                <div class="prompt__action-buttons advanced" id="advancedOptions">
-                    <label for="fileInput" class="prompt__form-button" tabindex="0" aria-label="Upload file"><i class='bx bx-upload'></i><div class="tooltip">Upload File</div></label>
-                    <input type="file" id="fileInput" style="display: none;" accept=".txt,.pdf,.jpg,.png">
-                    <button type="button" class="prompt__form-button" id="deepSearchButton" tabindex="0" aria-label="Deep search"><i class='bx bx-search'></i><div class="tooltip">Deep Search</div></button>
-                    <button type="button" class="prompt__form-button" id="thinkButton" tabindex="0" aria-label="Think mode"><i class='bx bx-brain'></i><div class="tooltip">Think Mode</div></button>
-                    <button type="button" class="prompt__form-button" id="bubbleToggle" tabindex="0" aria-label="Toggle bubble style"><i class='bx bx-chat'></i><div class="tooltip">Change Bubble Style</div></button>
-                    <button type="button" class="prompt__form-button" id="soundToggle" tabindex="0" aria-label="Toggle sound"><i class='bx bx-volume-full'></i><div class="tooltip">Toggle Sound</div></button>
-                    <button type="button" class="prompt__form-button" id="themeToggler" tabindex="0" aria-label="Toggle theme"><i class='bx bx-adjust'></i><div class="tooltip">Toggle Dark/Soft (Ctrl+T)</div></button>
-                    <span class="processing-dots" id="processingDots"><span style="--i:1">.</span><span style="--i:2">.</span><span style="--i:3">.</span></span>
-                </div>
-            </div>
-        </form>
-        <p class="prompt__disclaim">Vion IA provides answers based on its training and design. It may make mistakes.</p>
-    </section>
-    <audio id="sendSound" src="https://www.soundjay.com/buttons/beep-01a.mp3"></audio>
-    <audio id="responseSound" src="https://www.soundjay.com/buttons/beep-02.mp3"></audio>
-    <script src="https://cdn.jsdelivr.net/npm/marked/marked.min.js"></script>
-    <script src="https://cdnjs.cloudflare.com/ajax/libs/highlight.js/11.9.0/highlight.min.js"></script>
-    <script>
-        // Initialize state
-        let soundEnabled = localStorage.getItem('soundEnabled') !== 'false';
-        let messageHistory = [];
-        let currentBubbleIndex = 0;
-        const bubbleStyles = ['bubble-rounded', 'bubble-sharp', 'bubble-starry'];
-        const chatBar = document.getElementById('chatBar');
-        const inputField = document.getElementById('chatInput');
-        const suggests = document.querySelector('.suggests');
-        const welcomeText = document.getElementById('welcome-text');
-        const processingDots = document.getElementById('processingDots');
-        const sendButton = document.getElementById('sendButton');
-        const backToLatest = document.getElementById('backToLatest');
-        const themeToggler = document.getElementById('themeToggler');
-        const soundToggle = document.getElementById('soundToggle');
-        const advancedOptions = document.getElementById('advancedOptions');
-        const moreOptions = document.getElementById('moreOptions');
-
-        // Personalized greeting
-        function setGreeting() {
-            const hour = new Date().getHours();
-            const greeting = hour < 12 ? 'Good morning!' : hour < 18 ? 'Good afternoon!' : 'Good evening!';
-            const returning = localStorage.getItem('visited');
-            welcomeText.textContent = returning ? `Welcome back! ${greeting}` : `${greeting} Ask me anything!`;
-            localStorage.setItem('visited', 'true');
-        }
-        setGreeting();
-
-        // Theme management (dark or soft)
-        function setTheme(theme) {
-            document.body.classList.remove('soft_mode');
-            if (theme === 'soft') document.body.classList.add('soft_mode');
-            themeToggler.innerHTML = theme === 'soft' ? "<i class='bx bx-moon'></i>" : "<i class='bx bx-adjust'></i>";
-            localStorage.setItem('theme', theme);
-        }
-        const savedTheme = localStorage.getItem('theme') || 'dark';
-        setTheme(savedTheme);
-
-        // Toggle theme with button or Ctrl+T
-        themeToggler.addEventListener('click', () => {
-            const newTheme = document.body.classList.contains('soft_mode') ? 'dark' : 'soft';
-            setTheme(newTheme);
-        });
-        document.addEventListener('keydown', (e) => {
-            if (e.ctrlKey && e.key === 't') {
-                e.preventDefault();
-                const newTheme = document.body.classList.contains('soft_mode') ? 'dark' : 'soft';
-                setTheme(newTheme);
-            }
-        });
-
-        // Sound toggle
-        soundToggle.innerHTML = soundEnabled ? "<i class='bx bx-volume-full'></i>" : "<i class='bx bx-volume-mute'></i>";
-        soundToggle.addEventListener('click', () => {
-            soundEnabled = !soundEnabled;
-            localStorage.setItem('soundEnabled', soundEnabled);
-            soundToggle.innerHTML = soundEnabled ? "<i class='bx bx-volume-full'></i>" : "<i class='bx bx-volume-mute'></i>";
-        });
-
-        // Progressive disclosure
-        moreOptions.addEventListener('click', () => {
-            advancedOptions.classList.toggle('active');
-            moreOptions.style.transform = advancedOptions.classList.contains('active') ? 'rotate(90deg)' : 'rotate(0)';
-        });
-
-        // Bubble customization
-        document.getElementById('bubbleToggle').addEventListener('click', () => {
-            currentBubbleIndex = (currentBubbleIndex + 1) % bubbleStyles.length;
-            applyBubbleStyle();
-        });
-
-        function applyBubbleStyle() {
-            const userMessages = document.querySelectorAll('.chat-message.user');
-            userMessages.forEach(msg => {
-                bubbleStyles.forEach(style => msg.classList.remove(style));
-                msg.classList.add(bubbleStyles[currentBubbleIndex]);
-            });
-        }
-
-        // Add message with feedback option
-        function addMessage(content, isUser = false, isError = false) {
-            if (isError) {
-                const errorDiv = document.createElement('div');
-                errorDiv.classList.add('error-message');
-                errorDiv.innerHTML = `${content} <button onclick="retryLastMessage()" tabindex="0">Retry</button>`;
-                chatBar.appendChild(errorDiv);
-            } else {
-                const messageDiv = document.createElement('div');
-                messageDiv.classList.add('chat-message', isUser ? 'user' : 'bot');
-                if (isUser) messageDiv.classList.add(bubbleStyles[currentBubbleIndex]);
-                if (!isUser) messageDiv.classList.add('feedback');
-                messageDiv.textContent = content;
-                if (!isUser) {
-                    const feedbackDiv = document.createElement('div');
-                    feedbackDiv.classList.add('feedback-options');
-                    feedbackDiv.innerHTML = `
-                        <button onclick="handleFeedback('up')" tabindex="0" aria-label="Thumbs up"><i class='bx bx-thumbs-up'></i></button>
-                        <button onclick="handleFeedback('down')" tabindex="0" aria-label="Thumbs down"><i class='bx bx-thumbs-down'></i></button>
-                    `;
-                    messageDiv.appendChild(feedbackDiv);
+    detected_intent, target_language = detect_intent(text, file, intent)
+    logger.info(f"Detected intent: {detected_intent}, target_language: {target_language}")
+    
+    try:
+        if detected_intent == "chatbot":
+            response = get_gemini_response(text)
+            return {"response": response, "type": "chat"}
+        elif detected_intent == "translate":
+            content = await extract_text_from_file(file) if file else text
+            if "all languages" in text.lower():
+                translations = {}
+                phrase_to_translate = "I want to explore the stars" if "I want to explore the stars" in text else content
+                for lang, code in SUPPORTED_LANGUAGES.items():
+                    translation_tokenizer.src_lang = "en"
+                    encoded = translation_tokenizer(phrase_to_translate, return_tensors="pt").to(translation_model.device)
+                    generated_tokens = translation_model.generate(
+                        **encoded,
+                        forced_bos_token_id=translation_tokenizer.get_lang_id(code),
+                        max_length=512,
+                        num_beams=1
+                    )
+                    translations[lang] = translation_tokenizer.batch_decode(generated_tokens, skip_special_tokens=True)[0]
+                response = "\n".join(f"{lang.capitalize()}: {translations[lang]}" for lang in translations)
+                logger.info(f"Translated to all supported languages: {', '.join(translations.keys())}")
+                return {"response": response, "type": "translation"}
+            else:
+                translated_text = translate_text(content, target_language)
+                return {"response": translated_text, "type": "translation"}
+            
+        elif detected_intent == "file-translate":
+            if not file or not file.filename.lower().endswith(('.pdf', '.docx', '.txt', '.rtf')):
+                raise HTTPException(status_code=400, detail="A text-based file (PDF, DOCX, TXT, RTF) is required")
+            if not text:
+                raise HTTPException(status_code=400, detail="Please specify a target language for translation")
+            
+            content = await extract_text_from_file(file)
+            if not content.strip():
+                raise HTTPException(status_code=400, detail="No text could be extracted from the file")
+            
+            # Split content into chunks to handle large files
+            max_chunk_size = 512
+            chunks = [content[i:i+max_chunk_size] for i in range(0, len(content), max_chunk_size)]
+            translated_chunks = []
+            
+            for chunk in chunks:
+                translated_chunk = translate_text(chunk, target_language)
+                translated_chunks.append(translated_chunk)
+            
+            translated_text = " ".join(translated_chunks)
+            translated_text = translated_text.strip().capitalize()
+            if not translated_text.endswith(('.', '!', '?')):
+                translated_text += '.'
+            
+            logger.info(f"File translated to {target_language}: {translated_text[:100]}...")
+            
+            return {
+                "response": translated_text,
+                "type": "file_translation",
+                "additional_data": {
+                    "file_name": file.filename,
+                    "target_language": target_language
                 }
-                chatBar.appendChild(messageDiv);
-                messageHistory.push({ content, isUser });
             }
-            if (chatBar.scrollTop + chatBar.clientHeight >= chatBar.scrollHeight - 100) {
-                chatBar.scrollTop = chatBar.scrollHeight;
-            }
-        }
-
-        // Handle feedback (placeholder)
-        function handleFeedback(type) {
-            console.log(`Feedback: ${type}`);
-            alert(`Thanks for your feedback! (${type === 'up' ? 'Thumbs up' : 'Thumbs down'})`);
-        }
-
-        // Retry last message
-        function retryLastMessage() {
-            const lastUserMessage = messageHistory.filter(m => m.isUser).slice(-1)[0];
-            if (lastUserMessage) sendMessage(lastUserMessage.content);
-        }
-
-        // Input events
-        inputField.addEventListener('input', () => {
-            const input = inputField.value.trim();
-            suggests.classList.toggle('hidden', input.length > 0);
-            welcomeText.classList.toggle('hidden', input.length > 0);
-            inputField.setCustomValidity(input.length < 3 && input.length > 0 ? 'Please enter at least 3 characters.' : '');
-        });
-
-        inputField.addEventListener('focus', () => {
-            inputField.parentElement.style.boxShadow = `0 0 5px var(--accent-color)`;
-        });
-
-        inputField.addEventListener('blur', () => {
-            inputField.parentElement.style.boxShadow = 'none';
-        });
-
-        inputField.addEventListener('keydown', (e) => {
-            if (e.key === 'Enter' && !e.shiftKey) {
-                e.preventDefault();
-                sendMessage();
-            } else if (e.ctrlKey && e.key === 's') {
-                e.preventDefault();
-                sendMessage();
-            } else if (e.key === 'Escape') {
-                e.preventDefault();
-                inputField.value = '';
-                suggests.classList.remove('hidden');
-                welcomeText.classList.remove('hidden');
-            }
-        });
-
-        // Drag and drop
-        const inputWrapper = document.querySelector('.prompt__input-wrapper');
-        inputWrapper.addEventListener('dragover', (e) => {
-            e.preventDefault();
-            inputWrapper.classList.add('dragover');
-        });
-        inputWrapper.addEventListener('dragleave', () => {
-            inputWrapper.classList.remove('dragover');
-        });
-        inputWrapper.addEventListener('drop', (e) => {
-            e.preventDefault();
-            inputWrapper.classList.remove('dragover');
-            const file = e.dataTransfer.files[0];
-            if (file) alert(`File dropped: ${file.name} (Processing to be implemented)`);
-        });
-
-        // Suggestion interactions
-        document.querySelectorAll('.suggests__item').forEach(item => {
-            item.addEventListener('click', () => {
-                inputField.value = item.querySelector('.suggests__item-text').textContent;
-                suggests.classList.add('hidden');
-                welcomeText.classList.add('hidden');
-                inputField.focus();
-            });
-            item.addEventListener('dblclick', () => {
-                alert(`Pinned suggestion: ${item.querySelector('.suggests__item-text').textContent} (Placeholder)`);
-            });
-            item.addEventListener('keydown', (e) => {
-                if (e.key === 'Enter' || e.key === ' ') {
-                    e.preventDefault();
-                    item.click();
+            
+        elif detected_intent == "summarize":
+            content = await extract_text_from_file(file) if file else text
+            if not content.strip():
+                raise HTTPException(status_code=400, detail="No content to summarize")
+
+            content = preprocess_text(content)
+            logger.info(f"Preprocessed content: {content[:100]}...")
+
+            summarizer = load_model("summarization")
+
+            content_length = len(content.split())
+            max_len = max(50, min(200, content_length))
+            min_len = max(20, min(50, content_length // 3))
+
+            try:
+                if len(content) > 1024:
+                    chunks = [content[i:i+1024] for i in range(0, len(content), 1024)]
+                    summaries = []
+                    
+                    for chunk in chunks[:3]:
+                        summary = summarizer(
+                            chunk,
+                            max_length=max_len,
+                            min_length=min_len,
+                            do_sample=False,
+                            truncation=True
+                        )
+                        summaries.append(summary[0]['summary_text'])
+                    
+                    final_summary = " ".join(summaries)
+                else:
+                    summary = summarizer(
+                        content,
+                        max_length=max_len,
+                        min_length=min_len,
+                        do_sample=False,
+                        truncation=True
+                    )
+                    final_summary = summary[0]['summary_text']
+
+                final_summary = re.sub(r'\s+', ' ', final_summary).strip()
+                if not final_summary or final_summary.lower().startswith(content.lower()[:30]):
+                    logger.warning("Summarizer produced inadequate output, falling back to Gemini")
+                    final_summary = get_gemini_response(
+                        f"Summarize this text in a concise and meaningful way: {content}"
+                    )
+
+                if not final_summary.endswith(('.', '!', '?')):
+                    final_summary += '.'
+
+                logger.info(f"Generated summary: {final_summary}")
+                return {"response": final_summary, "type": "summary", "message": "Text was preprocessed to correct spelling errors"}
+
+            except Exception as e:
+                logger.error(f"Summarization error: {str(e)}")
+                final_summary = get_gemini_response(
+                    f"Summarize this text in a concise and meaningful way: {content}"
+                )
+                return {"response": final_summary, "type": "summary", "message": "Text was preprocessed to correct spelling errors"}
+            
+        elif detected_intent == "image-to-text":
+            if not file or not file.content_type.startswith('image/'):
+                raise HTTPException(status_code=400, detail="An image file is required")
+            
+            image = Image.open(io.BytesIO(await file.read()))
+            captioner = load_model("image-to-text")
+            
+            caption = captioner(image, max_new_tokens=50)
+            
+            return {
+                "response": caption[0]['generated_text'],
+                "type": "caption",
+                "additional_data": {
+                    "image_size": f"{image.width}x{image.height}"
                 }
-            });
-        });
-
-        // Scroll handling
-        chatBar.addEventListener('scroll', () => {
-            const isScrolledUp = chatBar.scrollTop < chatBar.scrollHeight - chatBar.clientHeight - 100;
-            backToLatest.classList.toggle('visible', isScrolledUp);
-            if (chatBar.scrollTop < 100 && messageHistory.length > 10) {
-                loadMoreMessages();
-            }
-        });
-
-        backToLatest.addEventListener('click', () => {
-            chatBar.scrollTop = chatBar.scrollHeight;
-        });
-        backToLatest.addEventListener('keydown', (e) => {
-            if (e.key === 'Enter' || e.key === ' ') {
-                e.preventDefault();
-                chatBar.scrollTop = chatBar.scrollHeight;
-            }
-        });
-
-        // Placeholder for loading more messages
-        function loadMoreMessages() {
-            console.log('Loading more messages (placeholder)');
-        }
-
-        // Send message
-        function sendMessage(input = inputField.value.trim()) {
-            if (!input) {
-                addMessage('Oops, please type something to ask!', false, true);
-                return;
-            }
-            if (input.length < 3) {
-                addMessage('Your query is too short—try adding more details!', false, true);
-                return;
             }
 
-            sendButton.disabled = true;
-            processingDots.classList.add('active');
-            addMessage(input, true);
-            inputField.value = '';
-            if (soundEnabled) document.getElementById('sendSound').play();
-
-            fetch('/chat', {
-                method: 'POST',
-                headers: { 'Content-Type': 'application/json' },
-                body: JSON.stringify({ message: input })
-            })
-            .then(res => res.json())
-            .then(data => {
-                if (data.response) {
-                    addMessage(data.response);
-                    suggests.classList.add('hidden');
-                    if (soundEnabled) document.getElementById('responseSound').play();
-                } else {
-                    addMessage(data.detail || 'Something went wrong! Please try again.', false, true);
+        elif detected_intent == "visual-qa":
+            if not file or not file.content_type.startswith('image/'):
+                raise HTTPException(status_code=400, detail="An image file is required")
+            if not text:
+                raise HTTPException(status_code=400, detail="A question is required for VQA")
+            
+            image = Image.open(io.BytesIO(await file.read())).convert("RGB")
+            vqa_pipeline = load_model("visual-qa")
+            
+            question = text.strip()
+            if not question.endswith('?'):
+                question += '?'
+            
+            answer = vqa_pipeline(
+                image=image,
+                question=question
+            )
+            
+            answer = answer.strip()
+            if not answer or answer.lower() == question.lower():
+                logger.warning(f"VQA failed to generate a meaningful answer: {answer}")
+                answer = "I couldn't determine the answer from the image."
+            else:
+                answer = answer.capitalize()
+                if not answer.endswith(('.', '!', '?')):
+                    answer += '.'
+            
+            # Check if the question asks for a specific, factual detail like color
+            factual_questions = ['color', 'size', 'number', 'how many', 'what is the']
+            is_factual = any(keyword in question.lower() for keyword in factual_questions)
+            
+            if is_factual:
+                # Return the raw VQA answer for factual questions
+                final_answer = answer
+            else:
+                # Apply cosmic tone for non-factual, open-ended questions
+                chatbot = load_model("chatbot")
+                if "fly" in question.lower():
+                    final_answer = chatbot.generate_content(f"Make this fun and spacey: {answer}").text.strip()
+                else:
+                    final_answer = chatbot.generate_content(f"Make this cosmic and poetic: {answer}").text.strip()
+            
+            logger.info(f"Final VQA answer: {final_answer}")
+            
+            return {
+                "response": final_answer,
+                "type": "visual_qa",
+                "additional_data": {
+                    "question": text,
+                    "image_size": f"{image.width}x{image.height}"
                 }
-            })
-            .catch(error => {
-                addMessage('Failed to process the query! Please check your connection.', false, true);
-            })
-            .finally(() => {
-                sendButton.disabled = false;
-                processingDots.classList.remove('active');
-            });
-        }
-
-        sendButton.addEventListener('click', () => sendMessage());
-        sendButton.addEventListener('keydown', (e) => {
-            if (e.key === 'Enter' || e.key === ' ') {
-                e.preventDefault();
-                sendMessage();
             }
-        });
-
-        // Placeholder button actions
-        document.getElementById('deepSearchButton').addEventListener('click', () => {
-            alert('Initiating DeepSearch... (Functionality to be implemented)');
-        });
-        document.getElementById('thinkButton').addEventListener('click', () => {
-            alert('Activating Think mode... (Functionality to be implemented)');
-        });
-        document.getElementById('fileInput').addEventListener('change', (e) => {
-            const file = e.target.files[0];
-            if (file) alert(`File selected: ${file.name} (Processing to be implemented)`);
-        });
-
-        // Responsive adjustments
-        window.addEventListener('resize', () => {
-            const width = window.innerWidth;
-            chatBar.style.maxHeight = width < 768 ? 'calc(100vh - 200px)' : 'calc(100vh - 180px)';
-        });
-
-        // Starry background effects
-        document.addEventListener('mousemove', (e) => {
-            let x = e.clientX / window.innerWidth;
-            let y = e.clientY / window.innerHeight;
-            document.querySelector('.stars').style.transform = `translate(${x * 50}px, ${y * 50}px)`;
-        });
-
-        // Enhanced keyboard navigation
-        document.querySelectorAll('.sidebar__item, .suggests__item, .prompt__form-button, .back-to-latest, .feedback-options button').forEach(el => {
-            el.addEventListener('keydown', (e) => {
-                if (e.key === 'Enter' || e.key === ' ') {
-                    e.preventDefault();
-                    el.click();
+      
+        elif detected_intent == "visualize":
+            if not file:
+                raise HTTPException(status_code=400, detail="An Excel file is required")
+            
+            file_content = await file.read()
+            
+            if file.filename.endswith('.csv'):
+                df = pd.read_csv(io.BytesIO(file_content))
+            else:
+                df = pd.read_excel(io.BytesIO(file_content))
+            
+            code = generate_visualization_code(df, text)
+            stats = df.describe().to_string()
+            response = f"Stats:\n{stats}\n\nChart Code:\n{code}"
+            
+            return {"response": response, "type": "visualization_code"}
+            
+        elif detected_intent == "text-generation":
+            response = get_gemini_response(text, is_generation=True)
+            lines = response.split(". ")
+            formatted_poem = "\n".join(line.strip() + ("." if not line.endswith(".") else "") for line in lines if line)
+            return {"response": formatted_poem, "type": "generated_text"}
+            
+        elif detected_intent == "file-qa":
+            if not file or not file.filename.lower().endswith(('.pdf', '.docx', '.doc', '.txt', '.rtf')):
+                raise HTTPException(status_code=400, detail="A text-based file (PDF, DOCX, TXT, RTF) is required")
+            if not text:
+                raise HTTPException(status_code=400, detail="A question about the file is required")
+            
+            content = await extract_text_from_file(file)
+            if not content.strip():
+                raise HTTPException(status_code=400, detail="No text could be extracted from the file")
+            
+            qa_pipeline = load_model("file-qa")
+            
+            question = text.strip()
+            if not question.endswith('?'):
+                question += '?'
+            
+            if len(content) > 512:
+                chunks = [content[i:i+512] for i in range(0, len(content), 512)]
+                answers = []
+                for chunk in chunks[:3]:
+                    result = qa_pipeline(question=question, context=chunk)
+                    if result['score'] > 0.1:
+                        answers.append((result['answer'], result['score']))
+                if answers:
+                    best_answer = max(answers, key=lambda x: x[1])[0]
+                else:
+                    best_answer = "I couldn't find a clear answer in the document."
+            else:
+                result = qa_pipeline(question=question, context=content)
+                best_answer = result['answer'] if result['score'] > 0.1 else "I couldn't find a clear answer in the document."
+            
+            best_answer = best_answer.strip().capitalize()
+            if not best_answer.endswith(('.', '!', '?')):
+                best_answer += '.'
+            
+            try:
+                chatbot = load_model("chatbot")
+                final_answer = chatbot.generate_content(f"Make this cosmic and poetic: {best_answer}").text.strip()
+            except Exception as e:
+                logger.warning(f"Failed to add cosmic tone: {str(e)}. Using raw answer.")
+                final_answer = best_answer
+            
+            logger.info(f"File QA answer: {final_answer}")
+            
+            return {
+                "response": final_answer,
+                "type": "file_qa",
+                "additional_data": {
+                    "question": text,
+                    "file_name": file.filename
                 }
-            });
-        });
-    </script>
-</body>
-</html>
-"""
+            }
+            
+        else:
+            response = get_gemini_response(text or "Hello! How can I assist you?")
+            return {"response": response, "type": "chat"}
+            
+    except Exception as e:
+        logger.error(f"Processing error: {str(e)}", exc_info=True)
+        raise HTTPException(status_code=500, detail=str(e))
+    finally:
+        process_time = time.time() - start_time
+        logger.info(f"Request processed in {process_time:.2f} seconds")
 
-@app.get("/", response_class=HTMLResponse)
-async def read_root():
-    return HTML_CONTENT
+async def extract_text_from_file(file: UploadFile) -> str:
+    """Enhanced text extraction with multiple fallbacks"""
+    if not file:
+        return ""
+
+    content = await file.read()
+    filename = file.filename.lower()
 
-@app.post("/chat")
-async def chat_endpoint(data: dict):
-    message = data.get("message", "")
-    if not message:
-        raise HTTPException(status_code=400, detail="No message provided")
     try:
-        response = get_chatbot_response(message)
-        return {"response": response}
+        if filename.endswith('.pdf'):
+            try:
+                doc = fitz.open(stream=content, filetype="pdf")
+                if doc.is_encrypted:
+                    return "PDF is encrypted and cannot be read"
+                text = ""
+                for page in doc:
+                    text += page.get_text()
+                return text
+            except Exception as pdf_error:
+                logger.warning(f"PyMuPDF failed: {str(pdf_error)}. Trying pdfminer.six...")
+                from pdfminer.high_level import extract_text
+                from io import BytesIO
+                return extract_text(BytesIO(content))
+
+        elif filename.endswith(('.docx', '.doc')):
+            doc = Document(io.BytesIO(content))
+            return "\n".join(para.text for para in doc.paragraphs)
+
+        elif filename.endswith('.txt'):
+            return content.decode('utf-8', errors='replace')
+
+        elif filename.endswith('.rtf'):
+            text = content.decode('utf-8', errors='replace')
+            text = re.sub(r'\\[a-z]+', ' ', text)
+            text = re.sub(r'\{|\}|\\', '', text)
+            return text
+
+        else:
+            raise HTTPException(status_code=400, detail=f"Unsupported file format: {filename}")
+
     except Exception as e:
-        raise HTTPException(status_code=500, detail=f"Chat error: {str(e)}")
+        logger.error(f"File extraction error: {str(e)}", exc_info=True)
+        raise HTTPException(
+            status_code=500,
+            detail=f"Error extracting text: {str(e)}. Supported formats: PDF, DOCX, TXT, RTF"
+        )
+
+def generate_visualization_code(df: pd.DataFrame, request: str = None) -> str:
+    """Generate visualization code based on data analysis"""
+    num_rows, num_cols = df.shape
+    numeric_cols = df.select_dtypes(include=[np.number]).columns.tolist()
+    categorical_cols = df.select_dtypes(include=['object']).columns.tolist()
+    date_cols = [col for col in df.columns if df[col].dtype == 'datetime64[ns]' or 
+                (isinstance(df[col].dtype, np.dtype) and pd.to_datetime(df[col], errors='coerce').notna().all())]
+    
+    if request:
+        request_lower = request.lower()
+    else:
+        request_lower = ""
+    
+    if len(numeric_cols) >= 2 and ("scatter" in request_lower or "correlation" in request_lower):
+        x_col = numeric_cols[0]
+        y_col = numeric_cols[1]
+        return f"""import pandas as pd
+import matplotlib.pyplot as plt
+import seaborn as sns
+df = pd.read_excel('data.xlsx')
+plt.figure(figsize=(10, 6))
+sns.regplot(x='{x_col}', y='{y_col}', data=df, scatter_kws={{'alpha': 0.6}})
+plt.title('Correlation between {x_col} and {y_col}')
+plt.grid(True, alpha=0.3)
+plt.tight_layout()
+plt.savefig('correlation_plot.png')
+plt.show()
+correlation = df['{x_col}'].corr(df['{y_col}'])
+print(f"Correlation coefficient: {{correlation:.4f}}")"""
+
+    elif len(numeric_cols) >= 1 and len(categorical_cols) >= 1 and ("bar" in request_lower or "comparison" in request_lower):
+        cat_col = categorical_cols[0]
+        num_col = numeric_cols[0]
+        return f"""import pandas as pd
+import matplotlib.pyplot as plt
+import seaborn as sns
+df = pd.read_excel('data.xlsx')
+plt.figure(figsize=(12, 7))
+ax = sns.barplot(x='{cat_col}', y='{num_col}', data=df, palette='viridis')
+for p in ax.patches:
+    ax.annotate(f'{{p.get_height():.1f}}', 
+                (p.get_x() + p.get_width() / 2., p.get_height()), 
+                ha='center', va='bottom', fontsize=10, color='black', xytext=(0, 5),
+                textcoords='offset points')
+plt.title('Comparison of {num_col} by {cat_col}', fontsize=15)
+plt.xlabel('{cat_col}', fontsize=12)
+plt.ylabel('{num_col}', fontsize=12)
+plt.xticks(rotation=45, ha='right')
+plt.grid(axis='y', alpha=0.3)
+plt.tight_layout()
+plt.savefig('comparison_chart.png')
+plt.show()"""
+
+    elif len(numeric_cols) >= 1 and ("distribution" in request_lower or "histogram" in request_lower):
+        num_col = numeric_cols[0]
+        return f"""import pandas as pd
+import matplotlib.pyplot as plt
+import seaborn as sns
+df = pd.read_excel('data.xlsx')
+plt.figure(figsize=(10, 6))
+sns.histplot(df['{num_col}'], kde=True, bins=20, color='purple')
+plt.title('Distribution of {num_col}', fontsize=15)
+plt.xlabel('{num_col}', fontsize=12)
+plt.ylabel('Frequency', fontsize=12)
+plt.grid(True, alpha=0.3)
+plt.tight_layout()
+plt.savefig('distribution_plot.png')
+plt.show()
+print(df['{num_col}'].describe())"""
+
+    else:
+        return f"""import pandas as pd
+import matplotlib.pyplot as plt
+import seaborn as sns
+import numpy as np
+df = pd.read_excel('data.xlsx')
+print("Descriptive statistics:")
+print(df.describe())
+fig, axes = plt.subplots(2, 2, figsize=(15, 12))
+numeric_df = df.select_dtypes(include=[np.number])
+if not numeric_df.empty and numeric_df.shape[1] > 1:
+    sns.heatmap(numeric_df.corr(), annot=True, cmap='coolwarm', fmt='.2f', ax=axes[0, 0])
+    axes[0, 0].set_title('Correlation Matrix')
+if not numeric_df.empty:
+    for i, col in enumerate(numeric_df.columns[:1]):
+        sns.histplot(df[col], kde=True, ax=axes[0, 1], color='purple')
+        axes[0, 1].set_title(f'Distribution of {col}')
+        axes[0, 1].set_xlabel(col)
+        axes[0, 1].set_ylabel('Frequency')
+categorical_cols = df.select_dtypes(include=['object']).columns
+if len(categorical_cols) > 0 and not numeric_df.empty:
+    cat_col = categorical_cols[0]
+    num_col = numeric_df.columns[0]
+    sns.barplot(x=cat_col, y=num_col, data=df, ax=axes[1, 0], palette='viridis')
+    axes[1, 0].set_title(f'{num_col} by {cat_col}')
+    axes[1, 0].set_xticklabels(axes[1, 0].get_xticklabels(), rotation=45, ha='right')
+if not numeric_df.empty and len(categorical_cols) > 0:
+    cat_col = categorical_cols[0]
+    num_col = numeric_df.columns[0]
+    sns.boxplot(x=cat_col, y=num_col, data=df, ax=axes[1, 1], palette='Set3')
+    axes[1, 1].set_title(f'Distribution of {num_col} by {cat_col}')
+    axes[1, 1].set_xticklabels(axes[1, 1].get_xticklabels(), rotation=45, ha='right')
+plt.tight_layout()
+plt.savefig('dashboard.png')
+plt.show()"""
+
+@app.get("/", include_in_schema=False)
+async def home():
+    """Redirect to the static index.html file"""
+    return RedirectResponse(url="/static/index.html")
+
+@app.get("/health", include_in_schema=True)
+async def health_check():
+    """Health check endpoint"""
+    return {"status": "healthy", "version": "2.0.0"}
+
+@app.get("/models", include_in_schema=True)
+async def list_models():
+    """List available models"""
+    return {"models": MODELS}
+
+@app.on_event("startup")
+async def startup_event():
+    """Pre-load models at startup with timeout"""
+    global translation_model, translation_tokenizer
+    logger.info("Starting model pre-loading...")
+    
+    async def load_model_with_timeout(task):
+        try:
+            await asyncio.wait_for(asyncio.to_thread(load_model, task), timeout=60.0)
+            logger.info(f"Successfully loaded {task} model")
+        except asyncio.TimeoutError:
+            logger.warning(f"Timeout loading {task} model - will load on demand")
+        except Exception as e:
+            logger.error(f"Error pre-loading {task}: {str(e)}")
+    
+    try:
+        model_name = MODELS["translation"]
+        translation_model = M2M100ForConditionalGeneration.from_pretrained(model_name)
+        translation_tokenizer = M2M100Tokenizer.from_pretrained(model_name)
+        device = "cuda" if torch.cuda.is_available() else "cpu"
+        translation_model.to(device)
+        logger.info("Translation model pre-loaded successfully")
+    except Exception as e:
+        logger.error(f"Error pre-loading translation model: {str(e)}")
+    
+    await asyncio.gather(
+        load_model_with_timeout("summarization"),
+        load_model_with_timeout("image-to-text"),
+        load_model_with_timeout("visual-qa"),
+        load_model_with_timeout("chatbot"),
+        load_model_with_timeout("file-qa")
+    )
 
 if __name__ == "__main__":
     import uvicorn
-    uvicorn.run(app, host="0.0.0.0", port=7860)
+    uvicorn.run("app:app", host="0.0.0.0", port=7860, reload=True)
+