app.py

import os
import gradio as gr
import requests
import json
import base64
import logging
import io
import time
from typing import List, Dict, Any, Union, Tuple, Optional
from dotenv import load_dotenv

# Load environment variables from .env file
load_dotenv()

# Configure logging
logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(name)s - %(levelname)s - %(message)s')
logger = logging.getLogger(__name__)

# Gracefully import libraries with fallbacks
try:
    from PIL import Image
    HAS_PIL = True
except ImportError:
    logger.warning("PIL not installed. Image processing will be limited.")
    HAS_PIL = False

try:
    import PyPDF2
    HAS_PYPDF2 = True
except ImportError:
    logger.warning("PyPDF2 not installed. PDF processing will be limited.")
    HAS_PYPDF2 = False

try:
    import markdown
    HAS_MARKDOWN = True
except ImportError:
    logger.warning("Markdown not installed. Markdown processing will be limited.")
    HAS_MARKDOWN = False

try:
    import openai
    HAS_OPENAI = True
except ImportError:
    logger.warning("OpenAI package not installed. OpenAI models will be unavailable.")
    HAS_OPENAI = False

try:
    from groq import Groq
    HAS_GROQ = True
except ImportError:
    logger.warning("Groq client not installed. Groq API will be unavailable.")
    HAS_GROQ = False

try:
    import cohere
    HAS_COHERE = True
except ImportError:
    logger.warning("Cohere package not installed. Cohere models will be unavailable.")
    HAS_COHERE = False

try:
    from huggingface_hub import InferenceClient
    HAS_HF = True
except ImportError:
    logger.warning("HuggingFace hub not installed. HuggingFace models will be limited.")
    HAS_HF = False

# API keys from environment
OPENROUTER_API_KEY = os.environ.get("OPENROUTER_API_KEY", "")
OPENAI_API_KEY = os.environ.get("OPENAI_API_KEY", "")
GROQ_API_KEY = os.environ.get("GROQ_API_KEY", "")
COHERE_API_KEY = os.environ.get("COHERE_API_KEY", "")
GLHF_API_KEY = os.environ.get("GLHF_API_KEY", "")
HF_API_KEY = os.environ.get("HF_API_KEY", "")

# ==========================================================
# MODEL DEFINITIONS
# ==========================================================

# OPENROUTER MODELS
# These are the original models from the provided code
OPENROUTER_MODELS = [
    # 1M+ Context Models
    {"category": "1M+ Context", "models": [
        #("Google: Gemini Pro 2.0 Experimental", "google/gemini-2.0-pro-exp-02-05:free", 2000000),
        ("Google: Gemini 2.0 Flash Thinking Experimental 01-21", "google/gemini-2.0-flash-thinking-exp:free", 1048576),
        ("Google: Gemini Flash 2.0 Experimental", "google/gemini-2.0-flash-exp:free", 1048576),
        ("Google: Gemini Pro 2.5 Experimental", "google/gemini-2.5-pro-exp-03-25:free", 1000000),
        ("Google: Gemini Flash 1.5 8B Experimental", "google/gemini-flash-1.5-8b-exp", 1000000),
    ]},
    
    # 100K-1M Context Models
    {"category": "100K+ Context", "models": [
        ("DeepSeek: DeepSeek R1 Zero", "deepseek/deepseek-r1-zero:free", 163840),
        ("DeepSeek: R1", "deepseek/deepseek-r1:free", 163840),
        ("DeepSeek: DeepSeek V3 Base", "deepseek/deepseek-v3-base:free", 131072),
        ("DeepSeek: DeepSeek V3 0324", "deepseek/deepseek-chat-v3-0324:free", 131072),
        ("Google: Gemma 3 4B", "google/gemma-3-4b-it:free", 131072),
        ("Google: Gemma 3 12B", "google/gemma-3-12b-it:free", 131072),
        ("Nous: DeepHermes 3 Llama 3 8B Preview", "nousresearch/deephermes-3-llama-3-8b-preview:free", 131072),
        ("Qwen: Qwen2.5 VL 72B Instruct", "qwen/qwen2.5-vl-72b-instruct:free", 131072),
        ("DeepSeek: DeepSeek V3", "deepseek/deepseek-chat:free", 131072),
        ("NVIDIA: Llama 3.1 Nemotron 70B Instruct", "nvidia/llama-3.1-nemotron-70b-instruct:free", 131072),
        ("Meta: Llama 3.2 1B Instruct", "meta-llama/llama-3.2-1b-instruct:free", 131072),
        ("Meta: Llama 3.2 11B Vision Instruct", "meta-llama/llama-3.2-11b-vision-instruct:free", 131072),
        ("Meta: Llama 3.1 8B Instruct", "meta-llama/llama-3.1-8b-instruct:free", 131072),
        ("Mistral: Mistral Nemo", "mistralai/mistral-nemo:free", 128000),
    ]},
    
    # 64K-100K Context Models
    {"category": "64K-100K Context", "models": [
        ("Mistral: Mistral Small 3.1 24B", "mistralai/mistral-small-3.1-24b-instruct:free", 96000),
        ("Google: Gemma 3 27B", "google/gemma-3-27b-it:free", 96000),
        ("Qwen: Qwen2.5 VL 3B Instruct", "qwen/qwen2.5-vl-3b-instruct:free", 64000),
        ("DeepSeek: R1 Distill Qwen 14B", "deepseek/deepseek-r1-distill-qwen-14b:free", 64000),
        ("Qwen: Qwen2.5-VL 7B Instruct", "qwen/qwen-2.5-vl-7b-instruct:free", 64000),
    ]},
    
    # 32K-64K Context Models
    {"category": "32K-64K Context", "models": [
        ("Google: LearnLM 1.5 Pro Experimental", "google/learnlm-1.5-pro-experimental:free", 40960),
        ("Qwen: QwQ 32B", "qwen/qwq-32b:free", 40000),
        ("Google: Gemini 2.0 Flash Thinking Experimental", "google/gemini-2.0-flash-thinking-exp-1219:free", 40000),
        ("Bytedance: UI-TARS 72B", "bytedance-research/ui-tars-72b:free", 32768),
        ("Qwerky 72b", "featherless/qwerky-72b:free", 32768),
        ("OlympicCoder 7B", "open-r1/olympiccoder-7b:free", 32768),
        ("OlympicCoder 32B", "open-r1/olympiccoder-32b:free", 32768),
        ("Google: Gemma 3 1B", "google/gemma-3-1b-it:free", 32768),
        ("Reka: Flash 3", "rekaai/reka-flash-3:free", 32768),
        ("Dolphin3.0 R1 Mistral 24B", "cognitivecomputations/dolphin3.0-r1-mistral-24b:free", 32768),
        ("Dolphin3.0 Mistral 24B", "cognitivecomputations/dolphin3.0-mistral-24b:free", 32768),
        ("Mistral: Mistral Small 3", "mistralai/mistral-small-24b-instruct-2501:free", 32768),
        ("Qwen2.5 Coder 32B Instruct", "qwen/qwen-2.5-coder-32b-instruct:free", 32768),
        ("Qwen2.5 72B Instruct", "qwen/qwen-2.5-72b-instruct:free", 32768),
    ]},
    
    # 8K-32K Context Models
    {"category": "8K-32K Context", "models": [
        ("Meta: Llama 3.2 3B Instruct", "meta-llama/llama-3.2-3b-instruct:free", 20000),
        ("Qwen: QwQ 32B Preview", "qwen/qwq-32b-preview:free", 16384),
        ("DeepSeek: R1 Distill Qwen 32B", "deepseek/deepseek-r1-distill-qwen-32b:free", 16000),
        ("Qwen: Qwen2.5 VL 32B Instruct", "qwen/qwen2.5-vl-32b-instruct:free", 8192),
        ("Moonshot AI: Moonlight 16B A3B Instruct", "moonshotai/moonlight-16b-a3b-instruct:free", 8192),
        ("DeepSeek: R1 Distill Llama 70B", "deepseek/deepseek-r1-distill-llama-70b:free", 8192),
        ("Qwen 2 7B Instruct", "qwen/qwen-2-7b-instruct:free", 8192),
        ("Google: Gemma 2 9B", "google/gemma-2-9b-it:free", 8192),
        ("Mistral: Mistral 7B Instruct", "mistralai/mistral-7b-instruct:free", 8192),
        ("Microsoft: Phi-3 Mini 128K Instruct", "microsoft/phi-3-mini-128k-instruct:free", 8192),
        ("Microsoft: Phi-3 Medium 128K Instruct", "microsoft/phi-3-medium-128k-instruct:free", 8192),
        ("Meta: Llama 3 8B Instruct", "meta-llama/llama-3-8b-instruct:free", 8192),
        ("OpenChat 3.5 7B", "openchat/openchat-7b:free", 8192),
        ("Meta: Llama 3.3 70B Instruct", "meta-llama/llama-3.3-70b-instruct:free", 8000),
    ]},
    
    # <8K Context Models
    {"category": "4K Context", "models": [
        ("AllenAI: Molmo 7B D", "allenai/molmo-7b-d:free", 4096),
        ("Rogue Rose 103B v0.2", "sophosympatheia/rogue-rose-103b-v0.2:free", 4096),
        ("Toppy M 7B", "undi95/toppy-m-7b:free", 4096),
        ("Hugging Face: Zephyr 7B", "huggingfaceh4/zephyr-7b-beta:free", 4096),
        ("MythoMax 13B", "gryphe/mythomax-l2-13b:free", 4096),
    ]},

    # Vision-capable Models
    {"category": "Vision Models", "models": [
        #("Google: Gemini Pro 2.0 Experimental", "google/gemini-2.0-pro-exp-02-05:free", 2000000),
        ("Google: Gemini 2.0 Flash Thinking Experimental 01-21", "google/gemini-2.0-flash-thinking-exp:free", 1048576),
        ("Google: Gemini Flash 2.0 Experimental", "google/gemini-2.0-flash-exp:free", 1048576),
        ("Google: Gemini Pro 2.5 Experimental", "google/gemini-2.5-pro-exp-03-25:free", 1000000),
        ("Google: Gemini Flash 1.5 8B Experimental", "google/gemini-flash-1.5-8b-exp", 1000000),
        ("Google: Gemma 3 4B", "google/gemma-3-4b-it:free", 131072),
        ("Google: Gemma 3 12B", "google/gemma-3-12b-it:free", 131072),
        ("Qwen: Qwen2.5 VL 72B Instruct", "qwen/qwen2.5-vl-72b-instruct:free", 131072),
        ("Meta: Llama 3.2 11B Vision Instruct", "meta-llama/llama-3.2-11b-vision-instruct:free", 131072),
        ("Mistral: Mistral Small 3.1 24B", "mistralai/mistral-small-3.1-24b-instruct:free", 96000),
        ("Google: Gemma 3 27B", "google/gemma-3-27b-it:free", 96000),
        ("Qwen: Qwen2.5 VL 3B Instruct", "qwen/qwen2.5-vl-3b-instruct:free", 64000),
        ("Qwen: Qwen2.5-VL 7B Instruct", "qwen/qwen-2.5-vl-7b-instruct:free", 64000),
        ("Google: LearnLM 1.5 Pro Experimental", "google/learnlm-1.5-pro-experimental:free", 40960),
        ("Google: Gemini 2.0 Flash Thinking Experimental", "google/gemini-2.0-flash-thinking-exp-1219:free", 40000),
        ("Bytedance: UI-TARS 72B", "bytedance-research/ui-tars-72b:free", 32768),
        ("Google: Gemma 3 1B", "google/gemma-3-1b-it:free", 32768),
        ("Qwen: Qwen2.5 VL 32B Instruct", "qwen/qwen2.5-vl-32b-instruct:free", 8192),
        ("AllenAI: Molmo 7B D", "allenai/molmo-7b-d:free", 4096),
    ]},
]

# Flatten OpenRouter model list for easier access
OPENROUTER_ALL_MODELS = []
for category in OPENROUTER_MODELS:
    for model in category["models"]:
        if model not in OPENROUTER_ALL_MODELS:  # Avoid duplicates
            OPENROUTER_ALL_MODELS.append(model)

# OPENAI MODELS
OPENAI_MODELS = {
    "gpt-3.5-turbo": 16385,
    "gpt-3.5-turbo-0125": 16385,
    "gpt-3.5-turbo-1106": 16385,
    "gpt-3.5-turbo-instruct": 4096,
    "gpt-4": 8192,
    "gpt-4-0314": 8192,
    "gpt-4-0613": 8192,
    "gpt-4-turbo": 128000,
    "gpt-4-turbo-2024-04-09": 128000,
    "gpt-4-turbo-preview": 128000,
    "gpt-4-0125-preview": 128000,
    "gpt-4-1106-preview": 128000,
    "gpt-4o": 128000,
    "gpt-4o-2024-11-20": 128000,
    "gpt-4o-2024-08-06": 128000,
    "gpt-4o-2024-05-13": 128000,
    "chatgpt-4o-latest": 128000,
    "gpt-4o-mini": 128000,
    "gpt-4o-mini-2024-07-18": 128000,
    "gpt-4o-realtime-preview": 128000,
    "gpt-4o-realtime-preview-2024-10-01": 128000,
    "gpt-4o-audio-preview": 128000,
    "gpt-4o-audio-preview-2024-10-01": 128000,
    "o1-preview": 128000,
    "o1-preview-2024-09-12": 128000,
    "o1-mini": 128000,
    "o1-mini-2024-09-12": 128000,
}

# HUGGINGFACE MODELS 
HUGGINGFACE_MODELS = {
    "microsoft/phi-3-mini-4k-instruct": 4096,
    "microsoft/Phi-3-mini-128k-instruct": 131072,
    "HuggingFaceH4/zephyr-7b-beta": 8192,
    "deepseek-ai/DeepSeek-Coder-V2-Instruct": 8192,
    "mistralai/Mistral-7B-Instruct-v0.3": 32768,
    "NousResearch/Nous-Hermes-2-Mixtral-8x7B-DPO": 32768,
    "microsoft/Phi-3.5-mini-instruct": 4096,
    "HuggingFaceTB/SmolLM2-1.7B-Instruct": 2048,
    "google/gemma-2-2b-it": 2048,
    "openai-community/gpt2": 1024,
    "microsoft/phi-2": 2048,
    "TinyLlama/TinyLlama-1.1B-Chat-v1.0": 2048,
    "VAGOsolutions/Llama-3-SauerkrautLM-8b-Instruct": 2048,
    "VAGOsolutions/Llama-3.1-SauerkrautLM-8b-Instruct": 4096,
    "VAGOsolutions/SauerkrautLM-Nemo-12b-Instruct": 4096,
    "openGPT-X/Teuken-7B-instruct-research-v0.4": 4096,
    "Qwen/Qwen2.5-7B-Instruct": 131072,
    "tiiuae/falcon-7b-instruct": 8192,
    "Qwen/QwQ-32B-preview": 32768,
}

# GROQ MODELS - We'll populate this dynamically
DEFAULT_GROQ_MODELS = {
    "gemma2-9b-it": 8192,
    "gemma-7b-it": 8192,
    "llama-3.3-70b-versatile": 131072,
    "llama-3.1-70b-versatile": 131072,
    "llama-3.1-8b-instant": 131072,
    "llama-guard-3-8b": 8192,
    "llama3-70b-8192": 8192,
    "llama3-8b-8192": 8192,
    "mixtral-8x7b-32768": 32768,
    "llama3-groq-70b-8192-tool-use-preview": 8192,
    "llama3-groq-8b-8192-tool-use-preview": 8192,
    "llama-3.3-70b-specdec": 131072,
    "llama-3.1-70b-specdec": 131072,
    "llama-3.2-1b-preview": 131072,
    "llama-3.2-3b-preview": 131072,
}

# COHERE MODELS
COHERE_MODELS = {
    "command-r-plus-08-2024": 131072,
    "command-r-plus-04-2024": 131072,
    "command-r-plus": 131072,
    "command-r-08-2024": 131072,
    "command-r-03-2024": 131072,
    "command-r": 131072,
    "command": 4096,
    "command-nightly": 131072,
    "command-light": 4096,
    "command-light-nightly": 4096,
    "c4ai-aya-expanse-8b": 8192,
    "c4ai-aya-expanse-32b": 131072,
}

# GLHF MODELS
GLHF_MODELS = {
    "mistralai/Mixtral-8x7B-Instruct-v0.1": 32768,
    "01-ai/Yi-34B-Chat": 32768,
    "mistralai/Mistral-7B-Instruct-v0.3": 32768,
    "microsoft/phi-3-mini-4k-instruct": 4096,
    "microsoft/Phi-3.5-mini-instruct": 4096,
    "microsoft/Phi-3-mini-128k-instruct": 131072,
    "HuggingFaceH4/zephyr-7b-beta": 8192,
    "NousResearch/Nous-Hermes-2-Mixtral-8x7B-DPO": 32768,
    "google/gemma-2-2b-it": 2048,
    "microsoft/phi-2": 2048,
}

# ==========================================================
# HELPER FUNCTIONS
# ==========================================================

def fetch_groq_models():
    """Fetch available Groq models with proper error handling"""
    try:
        if not HAS_GROQ or not GROQ_API_KEY:
            logger.warning("Groq client not available or no API key. Using default model list.")
            return DEFAULT_GROQ_MODELS

        client = Groq(api_key=GROQ_API_KEY)
        models = client.models.list()
        
        # Create dictionary of model_id -> context size
        model_dict = {}
        for model in models.data:
            model_id = model.id
            # Map known context sizes or use a default
            if "llama-3" in model_id and "70b" in model_id:
                context_size = 131072
            elif "llama-3" in model_id and "8b" in model_id:
                context_size = 131072
            elif "mixtral" in model_id:
                context_size = 32768
            elif "gemma" in model_id:
                context_size = 8192
            else:
                context_size = 8192  # Default assumption
                
            model_dict[model_id] = context_size
            
        # Ensure we have models by combining with defaults
        if not model_dict:
            return DEFAULT_GROQ_MODELS
        return {**DEFAULT_GROQ_MODELS, **model_dict}
        
    except Exception as e:
        logger.error(f"Error fetching Groq models: {e}")
        return DEFAULT_GROQ_MODELS

# Initialize Groq models
GROQ_MODELS = fetch_groq_models()

def encode_image_to_base64(image_path):
    """Encode an image file to base64 string"""
    try:
        if isinstance(image_path, str):  # File path as string
            with open(image_path, "rb") as image_file:
                encoded_string = base64.b64encode(image_file.read()).decode('utf-8')
                file_extension = image_path.split('.')[-1].lower()
                mime_type = f"image/{file_extension}"
                if file_extension in ["jpg", "jpeg"]:
                    mime_type = "image/jpeg"
                elif file_extension == "png":
                    mime_type = "image/png"
                elif file_extension == "webp":
                    mime_type = "image/webp"
                return f"data:{mime_type};base64,{encoded_string}"
        elif hasattr(image_path, 'name'):  # Handle Gradio file objects directly
            with open(image_path.name, "rb") as image_file:
                encoded_string = base64.b64encode(image_file.read()).decode('utf-8')
                file_extension = image_path.name.split('.')[-1].lower()
                mime_type = f"image/{file_extension}"
                if file_extension in ["jpg", "jpeg"]:
                    mime_type = "image/jpeg"
                elif file_extension == "png":
                    mime_type = "image/png"
                elif file_extension == "webp":
                    mime_type = "image/webp"
                return f"data:{mime_type};base64,{encoded_string}"
        else:  # Handle file object or other types
            logger.error(f"Unsupported image type: {type(image_path)}")
            return None
    except Exception as e:
        logger.error(f"Error encoding image: {str(e)}")
        return None

def extract_text_from_file(file_path):
    """Extract text from various file types"""
    try:
        file_extension = file_path.split('.')[-1].lower()
        
        if file_extension == 'pdf':
            if HAS_PYPDF2:
                text = ""
                with open(file_path, 'rb') as file:
                    pdf_reader = PyPDF2.PdfReader(file)
                    for page_num in range(len(pdf_reader.pages)):
                        page = pdf_reader.pages[page_num]
                        text += page.extract_text() + "\n\n"
                return text
            else:
                return "PDF processing is not available (PyPDF2 not installed)"
        
        elif file_extension == 'md':
            with open(file_path, 'r', encoding='utf-8') as file:
                return file.read()
        
        elif file_extension == 'txt':
            with open(file_path, 'r', encoding='utf-8') as file:
                return file.read()
                
        else:
            return f"Unsupported file type: {file_extension}"
            
    except Exception as e:
        logger.error(f"Error extracting text from file: {str(e)}")
        return f"Error processing file: {str(e)}"

def prepare_message_with_media(text, images=None, documents=None):
    """Prepare a message with text, images, and document content"""
    # If no media, return text only
    if not images and not documents:
        return text
    
    # Start with text content
    if documents and len(documents) > 0:
        # If there are documents, append their content to the text
        document_texts = []
        for doc in documents:
            if doc is None:
                continue
            # Make sure to handle file objects properly
            doc_path = doc.name if hasattr(doc, 'name') else doc
            doc_text = extract_text_from_file(doc_path)
            if doc_text:
                document_texts.append(doc_text)
        
        # Add document content to text
        if document_texts:
            if not text:
                text = "Please analyze these documents:"
            else:
                text = f"{text}\n\nDocument content:\n\n"
            
            text += "\n\n".join(document_texts)
            
        # If no images, return text only
        if not images:
            return text
    
    # If we have images, create a multimodal content array
    content = [{"type": "text", "text": text}]
    
    # Add images if any
    if images:
        # Check if images is a list of image paths or file objects
        if isinstance(images, list):
            for img in images:
                if img is None:
                    continue
                
                encoded_image = encode_image_to_base64(img)
                if encoded_image:
                    content.append({
                        "type": "image_url",
                        "image_url": {"url": encoded_image}
                    })
        else:
            # For single image or Gallery component
            logger.warning(f"Images is not a list: {type(images)}")
            # Try to handle as single image
            encoded_image = encode_image_to_base64(images)
            if encoded_image:
                content.append({
                    "type": "image_url", 
                    "image_url": {"url": encoded_image}
                })
    
    return content

def format_to_message_dict(history):
    """Convert history to proper message format"""
    messages = []
    for pair in history:
        if len(pair) == 2:
            human, ai = pair
            if human:
                messages.append({"role": "user", "content": human})
            if ai:
                messages.append({"role": "assistant", "content": ai})
    return messages

def process_uploaded_images(files):
    """Process uploaded image files"""
    file_paths = []
    for file in files:
        if hasattr(file, 'name'):
            file_paths.append(file.name)
    return file_paths

def filter_models(provider, search_term):
    """Filter models based on search term and provider"""
    if provider == "OpenRouter":
        all_models = [model[0] for model in OPENROUTER_ALL_MODELS]
    elif provider == "OpenAI":
        all_models = list(OPENAI_MODELS.keys())
    elif provider == "HuggingFace":
        all_models = list(HUGGINGFACE_MODELS.keys())
    elif provider == "Groq":
        all_models = list(GROQ_MODELS.keys())
    elif provider == "Cohere":
        all_models = list(COHERE_MODELS.keys())
    elif provider == "GLHF":
        all_models = list(GLHF_MODELS.keys())
    else:
        return [], None
        
    if not search_term:
        return all_models, all_models[0] if all_models else None
        
    filtered_models = [model for model in all_models if search_term.lower() in model.lower()]
    
    if filtered_models:
        return filtered_models, filtered_models[0]
    else:
        return all_models, all_models[0] if all_models else None

def get_model_info(provider, model_choice):
    """Get model ID and context size based on provider and model name"""
    if provider == "OpenRouter":
        for name, model_id, ctx_size in OPENROUTER_ALL_MODELS:
            if name == model_choice:
                return model_id, ctx_size
    elif provider == "OpenAI":
        if model_choice in OPENAI_MODELS:
            return model_choice, OPENAI_MODELS[model_choice]
    elif provider == "HuggingFace":
        if model_choice in HUGGINGFACE_MODELS:
            return model_choice, HUGGINGFACE_MODELS[model_choice]
    elif provider == "Groq":
        if model_choice in GROQ_MODELS:
            return model_choice, GROQ_MODELS[model_choice]
    elif provider == "Cohere":
        if model_choice in COHERE_MODELS:
            return model_choice, COHERE_MODELS[model_choice]
    elif provider == "GLHF":
        if model_choice in GLHF_MODELS:
            return model_choice, GLHF_MODELS[model_choice]
    
    return None, 0

def update_context_display(provider, model_name):
    """Update context size display for the selected model"""
    _, ctx_size = get_model_info(provider, model_name)
    return f"{ctx_size:,}" if ctx_size else "Unknown"

def update_model_info(provider, model_name):
    """Generate HTML info display for the selected model"""
    model_id, ctx_size = get_model_info(provider, model_name)
    if not model_id:
        return "<p>Model information not available</p>"
        
    # Check if this is a vision model
    is_vision_model = False
    
    # For OpenRouter, check the vision models category
    if provider == "OpenRouter":
        for cat in OPENROUTER_MODELS:
            if cat["category"] == "Vision Models":
                if any(m[0] == model_name for m in cat["models"]):
                    is_vision_model = True
                    break
    # For other providers, use heuristics
    elif provider == "OpenAI" and any(x in model_name.lower() for x in ["gpt-4", "gpt-4o"]):
        is_vision_model = True
    elif provider == "HuggingFace" and any(x in model_name.lower() for x in ["vl", "vision"]):
        is_vision_model = True
    
    vision_badge = '<span style="background-color: #4CAF50; color: white; padding: 3px 6px; border-radius: 3px; font-size: 0.8em; margin-left: 5px;">Vision</span>' if is_vision_model else ''
    
    # For OpenRouter, show the model ID
    model_id_html = f"<p><strong>Model ID:</strong> {model_id}</p>" if provider == "OpenRouter" else ""
    
    # For others, the ID is the same as the name
    if provider != "OpenRouter":
        model_id_html = ""
    
    return f"""
    <div class="model-info">
        <h3>{model_name} {vision_badge}</h3>
        {model_id_html}
        <p><strong>Context Size:</strong> {ctx_size:,} tokens</p>
        <p><strong>Provider:</strong> {provider}</p>
        {f'<p><strong>Features:</strong> Supports image understanding</p>' if is_vision_model else ''}
    </div>
    """

# ==========================================================
# API HANDLERS
# ==========================================================

def call_openrouter_api(payload, api_key_override=None):
    """Make a call to OpenRouter API with error handling"""
    try:
        api_key = api_key_override if api_key_override else OPENROUTER_API_KEY
        if not api_key:
            raise ValueError("OpenRouter API key is required")
            
        response = requests.post(
            "https://openrouter.ai/api/v1/chat/completions",
            headers={
                "Content-Type": "application/json",
                "Authorization": f"Bearer {api_key}",
                "HTTP-Referer": "https://huggingface.co./spaces/user/MultiProviderCrispChat"
            },
            json=payload,
            timeout=180  # Longer timeout for document processing
        )
        return response
    except requests.RequestException as e:
        logger.error(f"OpenRouter API request error: {str(e)}")
        raise e

def call_openai_api(payload, api_key_override=None):
    """Make a call to OpenAI API with error handling"""
    try:
        if not HAS_OPENAI:
            raise ImportError("OpenAI package not installed")
            
        api_key = api_key_override if api_key_override else OPENAI_API_KEY
        if not api_key:
            raise ValueError("OpenAI API key is required")
            
        client = openai.OpenAI(api_key=api_key)
        
        # Extract parameters from payload
        model = payload.get("model", "gpt-3.5-turbo")
        messages = payload.get("messages", [])
        temperature = payload.get("temperature", 0.7)
        max_tokens = payload.get("max_tokens", 1000)
        stream = payload.get("stream", False)
        top_p = payload.get("top_p", 0.9)
        presence_penalty = payload.get("presence_penalty", 0)
        frequency_penalty = payload.get("frequency_penalty", 0)
        
        # Handle response format if specified
        response_format = None
        if payload.get("response_format") == "json_object":
            response_format = {"type": "json_object"}
            
        # Create completion
        response = client.chat.completions.create(
            model=model,
            messages=messages,
            temperature=temperature,
            max_tokens=max_tokens,
            stream=stream,
            top_p=top_p,
            presence_penalty=presence_penalty,
            frequency_penalty=frequency_penalty,
            response_format=response_format
        )
        
        return response
    except Exception as e:
        logger.error(f"OpenAI API error: {str(e)}")
        raise e

def call_huggingface_api(payload, api_key_override=None):
    """Make a call to HuggingFace API with error handling"""
    try:
        if not HAS_HF:
            raise ImportError("HuggingFace hub not installed")
            
        api_key = api_key_override if api_key_override else HF_API_KEY
        
        # Extract parameters from payload
        model_id = payload.get("model", "mistralai/Mistral-7B-Instruct-v0.3")
        messages = payload.get("messages", [])
        temperature = payload.get("temperature", 0.7)
        max_tokens = payload.get("max_tokens", 500)
        
        # Create a prompt from messages
        prompt = ""
        for msg in messages:
            role = msg["role"].upper()
            content = msg["content"]
            
            # Handle multimodal content
            if isinstance(content, list):
                text_parts = []
                for item in content:
                    if item["type"] == "text":
                        text_parts.append(item["text"])
                content = "\n".join(text_parts)
                
            prompt += f"{role}: {content}\n"
            
        prompt += "ASSISTANT: "
        
        # Create client with or without API key
        client = InferenceClient(token=api_key) if api_key else InferenceClient()
        
        # Generate response
        response = client.text_generation(
            prompt,
            model=model_id,
            max_new_tokens=max_tokens,
            temperature=temperature,
            repetition_penalty=1.1
        )
        
        return {"generated_text": str(response)}
    except Exception as e:
        logger.error(f"HuggingFace API error: {str(e)}")
        raise e

def call_groq_api(payload, api_key_override=None):
    """Make a call to Groq API with error handling"""
    try:
        if not HAS_GROQ:
            raise ImportError("Groq client not installed")
            
        api_key = api_key_override if api_key_override else GROQ_API_KEY
        if not api_key:
            raise ValueError("Groq API key is required")
            
        client = Groq(api_key=api_key)
        
        # Extract parameters from payload
        model = payload.get("model", "llama-3.1-8b-instant")
        messages = payload.get("messages", [])
        temperature = payload.get("temperature", 0.7)
        max_tokens = payload.get("max_tokens", 1000)
        stream = payload.get("stream", False)
        top_p = payload.get("top_p", 0.9)
        
        # Create completion
        response = client.chat.completions.create(
            model=model,
            messages=messages,
            temperature=temperature,
            max_tokens=max_tokens,
            stream=stream,
            top_p=top_p
        )
        
        return response
    except Exception as e:
        logger.error(f"Groq API error: {str(e)}")
        raise e

def call_cohere_api(payload, api_key_override=None):
    """Make a call to Cohere API with error handling"""
    try:
        if not HAS_COHERE:
            raise ImportError("Cohere package not installed")
            
        api_key = api_key_override if api_key_override else COHERE_API_KEY
        if not api_key:
            raise ValueError("Cohere API key is required")
            
        client = cohere.Client(api_key=api_key)
        
        # Extract parameters from payload
        model = payload.get("model", "command-r-plus")
        messages = payload.get("messages", [])
        temperature = payload.get("temperature", 0.7)
        max_tokens = payload.get("max_tokens", 1000)
        
        # Format messages for Cohere
        chat_history = []
        user_message = ""
        
        for msg in messages:
            if msg["role"] == "system":
                # For system message, we'll prepend to the user's first message
                system_content = msg["content"]
                if isinstance(system_content, list):  # Handle multimodal content
                    system_parts = []
                    for item in system_content:
                        if item["type"] == "text":
                            system_parts.append(item["text"])
                    system_content = "\n".join(system_parts)
                user_message = f"System: {system_content}\n\n" + user_message
            elif msg["role"] == "user":
                content = msg["content"]
                # Handle multimodal content
                if isinstance(content, list):
                    text_parts = []
                    for item in content:
                        if item["type"] == "text":
                            text_parts.append(item["text"])
                    content = "\n".join(text_parts)
                user_message = content
            elif msg["role"] == "assistant":
                content = msg["content"]
                if content:
                    chat_history.append({"role": "ASSISTANT", "message": content})
        
        # Create chat completion
        response = client.chat(
            message=user_message,
            chat_history=chat_history,
            model=model,
            temperature=temperature,
            max_tokens=max_tokens
        )
        
        return response
    except Exception as e:
        logger.error(f"Cohere API error: {str(e)}")
        raise e

def call_glhf_api(payload, api_key_override=None):
    """Make a call to GLHF API with error handling"""
    try:
        if not HAS_OPENAI:
            raise ImportError("OpenAI package not installed (required for GLHF API)")
            
        api_key = api_key_override if api_key_override else GLHF_API_KEY
        if not api_key:
            raise ValueError("GLHF API key is required")
            
        client = openai.OpenAI(
            api_key=api_key,
            base_url="https://glhf.chat/api/openai/v1"
        )
        
        # Extract parameters from payload
        model_name = payload.get("model", "mistralai/Mistral-7B-Instruct-v0.3")
        # Add "hf:" prefix if not already there
        if not model_name.startswith("hf:"):
            model = f"hf:{model_name}"
        else:
            model = model_name
            
        messages = payload.get("messages", [])
        temperature = payload.get("temperature", 0.7)
        max_tokens = payload.get("max_tokens", 1000)
        stream = payload.get("stream", False)
        
        # Create completion
        response = client.chat.completions.create(
            model=model,
            messages=messages,
            temperature=temperature,
            max_tokens=max_tokens,
            stream=stream
        )
        
        return response
    except Exception as e:
        logger.error(f"GLHF API error: {str(e)}")
        raise e

def extract_ai_response(result, provider):
    """Extract AI response based on provider format"""
    try:
        if provider == "OpenRouter":
            if isinstance(result, dict):
                if "choices" in result and len(result["choices"]) > 0:
                    if "message" in result["choices"][0]:
                        message = result["choices"][0]["message"]
                        if message.get("reasoning") and not message.get("content"):
                            reasoning = message.get("reasoning")
                            lines = reasoning.strip().split('\n')
                            for line in lines:
                                if line and not line.startswith('I should') and not line.startswith('Let me'):
                                    return line.strip()
                            for line in lines:
                                if line.strip():
                                    return line.strip()
                        return message.get("content", "")
                    elif "delta" in result["choices"][0]:
                        return result["choices"][0]["delta"].get("content", "")
                    
        elif provider == "OpenAI":
            if hasattr(result, "choices") and len(result.choices) > 0:
                return result.choices[0].message.content
                
        elif provider == "HuggingFace":
            return result.get("generated_text", "")
                
        elif provider == "Groq":
            if hasattr(result, "choices") and len(result.choices) > 0:
                return result.choices[0].message.content
                
        elif provider == "Cohere":
            if hasattr(result, "text"):
                return result.text
                
        elif provider == "GLHF":
            if hasattr(result, "choices") and len(result.choices) > 0:
                return result.choices[0].message.content
            
        logger.error(f"Unexpected response structure from {provider}: {result}")
        return f"Error: Could not extract response from {provider} API result"
    except Exception as e:
        logger.error(f"Error extracting AI response: {str(e)}")
        return f"Error: {str(e)}"

# ==========================================================
# STREAMING HANDLERS
# ==========================================================

def openrouter_streaming_handler(response, chatbot, message_idx, message):
    try:
        # First add the user message if needed
        if len(chatbot) == message_idx:
            chatbot.append([message, ""])
            
        for line in response.iter_lines():
            if not line:
                continue
                
            line = line.decode('utf-8')
            if not line.startswith('data: '):
                continue
                
            data = line[6:]
            if data.strip() == '[DONE]':
                break
                
            try:
                chunk = json.loads(data)
                if "choices" in chunk and len(chunk["choices"]) > 0:
                    delta = chunk["choices"][0].get("delta", {})
                    if "content" in delta and delta["content"]:
                        # Update the current response
                        chatbot[-1][1] += delta["content"]
                        yield chatbot
            except json.JSONDecodeError:
                logger.error(f"Failed to parse JSON from chunk: {data}")
    except Exception as e:
        logger.error(f"Error in streaming handler: {str(e)}")
        # Add error message to the current response
        if len(chatbot) > message_idx:
            chatbot[-1][1] += f"\n\nError during streaming: {str(e)}"
            yield chatbot

def openai_streaming_handler(response, chatbot, message_idx, message):
    try:
        # First add the user message if needed
        if len(chatbot) == message_idx:
            chatbot.append([message, ""])
            
        full_response = ""
        for chunk in response:
            if hasattr(chunk.choices[0].delta, "content") and chunk.choices[0].delta.content is not None:
                content = chunk.choices[0].delta.content
                full_response += content
                chatbot[-1][1] = full_response
                yield chatbot
                
    except Exception as e:
        logger.error(f"Error in OpenAI streaming handler: {str(e)}")
        # Add error message to the current response
        chatbot[-1][1] += f"\n\nError during streaming: {str(e)}"
        yield chatbot

def groq_streaming_handler(response, chatbot, message_idx, message):
    try:
        # First add the user message if needed
        if len(chatbot) == message_idx:
            chatbot.append([message, ""])
            
        full_response = ""
        for chunk in response:
            if hasattr(chunk.choices[0].delta, "content") and chunk.choices[0].delta.content is not None:
                content = chunk.choices[0].delta.content
                full_response += content
                chatbot[-1][1] = full_response
                yield chatbot
                
    except Exception as e:
        logger.error(f"Error in Groq streaming handler: {str(e)}")
        # Add error message to the current response
        chatbot[-1][1] += f"\n\nError during streaming: {str(e)}"
        yield chatbot

def glhf_streaming_handler(response, chatbot, message_idx, message):
    try:
        # First add the user message if needed
        if len(chatbot) == message_idx:
            chatbot.append([message, ""])
            
        full_response = ""
        for chunk in response:
            if hasattr(chunk.choices[0].delta, "content") and chunk.choices[0].delta.content is not None:
                content = chunk.choices[0].delta.content
                full_response += content
                chatbot[-1][1] = full_response
                yield chatbot
                
    except Exception as e:
        logger.error(f"Error in GLHF streaming handler: {str(e)}")
        # Add error message to the current response
        chatbot[-1][1] += f"\n\nError during streaming: {str(e)}"
        yield chatbot

# ==========================================================
# MAIN FUNCTION TO ASK AI
# ==========================================================

def ask_ai(message, history, provider, model_choice, temperature, max_tokens, top_p, 
           frequency_penalty, presence_penalty, repetition_penalty, top_k, min_p, 
           seed, top_a, stream_output, response_format, images, documents, 
           reasoning_effort, system_message, transforms, api_key_override=None):
    """Enhanced AI query function with support for multiple providers"""
    # Validate input
    if not message.strip() and not images and not documents:
        return history
    
    # Copy history to new list to avoid modifying the original
    chat_history = list(history)
    
    # Create messages from chat history
    messages = format_to_message_dict(chat_history)
    
    # Add system message if provided
    if system_message and system_message.strip():
        # Remove any existing system message
        messages = [msg for msg in messages if msg.get("role") != "system"]
        # Add new system message at the beginning
        messages.insert(0, {"role": "system", "content": system_message.strip()})
    
    # Prepare message with images and documents if any
    content = prepare_message_with_media(message, images, documents)
    
    # Add current message
    messages.append({"role": "user", "content": content})
    
    # Common parameters for all providers
    common_params = {
        "temperature": temperature,
        "max_tokens": max_tokens,
        "top_p": top_p,
        "frequency_penalty": frequency_penalty,
        "presence_penalty": presence_penalty,
        "stream": stream_output
    }
    
    try:
        # Process based on provider
        if provider == "OpenRouter":
            # Get model ID from registry
            model_id, _ = get_model_info(provider, model_choice)
            if not model_id:
                error_message = f"Error: Model '{model_choice}' not found in OpenRouter"
                chat_history.append([message, error_message])
                return chat_history
                
            # Build OpenRouter payload
            payload = {
                "model": model_id,
                "messages": messages,
                **common_params
            }
            
            # Add optional parameters if set
            if repetition_penalty != 1.0:
                payload["repetition_penalty"] = repetition_penalty
            
            if top_k > 0:
                payload["top_k"] = top_k
            
            if min_p > 0:
                payload["min_p"] = min_p
            
            if seed > 0:
                payload["seed"] = seed
            
            if top_a > 0:
                payload["top_a"] = top_a
            
            # Add response format if JSON is requested
            if response_format == "json_object":
                payload["response_format"] = {"type": "json_object"}
            
            # Add reasoning if selected
            if reasoning_effort != "none":
                payload["reasoning"] = {
                    "effort": reasoning_effort
                }
            
            # Add transforms if selected
            if transforms:
                payload["transforms"] = transforms
                
            # Call OpenRouter API
            logger.info(f"Sending request to OpenRouter model: {model_id}")
            
            response = call_openrouter_api(payload, api_key_override)
            
            # Handle streaming response
            if stream_output and response.status_code == 200:
                # Add empty response slot to history
                chat_history.append([message, ""])
                
                # Set up generator for streaming updates
                def streaming_generator():
                    for updated_history in openrouter_streaming_handler(response, chat_history, len(chat_history) - 1, message):
                        yield updated_history
                
                return streaming_generator()
            
            # Handle normal response
            elif response.status_code == 200:
                result = response.json()
                logger.info(f"Response content: {result}")
                
                # Extract AI response
                ai_response = extract_ai_response(result, provider)
                
                # Add response to history
                chat_history.append([message, ai_response])
                return chat_history
            
            # Handle error response
            else:
                error_message = f"Error: Status code {response.status_code}"
                try:
                    response_data = response.json()
                    error_message += f"\n\nDetails: {json.dumps(response_data, indent=2)}"
                except:
                    error_message += f"\n\nResponse: {response.text}"
                
                logger.error(error_message)
                chat_history.append([message, error_message])
                return chat_history
                
        elif provider == "OpenAI":
            # Get model ID from registry
            model_id, _ = get_model_info(provider, model_choice)
            if not model_id:
                error_message = f"Error: Model '{model_choice}' not found in OpenAI"
                chat_history.append([message, error_message])
                return chat_history
                
            # Build OpenAI payload
            payload = {
                "model": model_id,
                "messages": messages,
                **common_params
            }
            
            # Add response format if JSON is requested
            if response_format == "json_object":
                payload["response_format"] = {"type": "json_object"}
            
            # Call OpenAI API
            logger.info(f"Sending request to OpenAI model: {model_id}")
            
            try:
                response = call_openai_api(payload, api_key_override)
                
                # Handle streaming response
                if stream_output:
                    # Add empty response slot to history
                    chat_history.append([message, ""])
                    
                    # Set up generator for streaming updates
                    def streaming_generator():
                        for updated_history in openai_streaming_handler(response, chat_history, len(chat_history) - 1, message):
                            yield updated_history
                    
                    return streaming_generator()
                
                # Handle normal response
                else:
                    ai_response = extract_ai_response(response, provider)
                    chat_history.append([message, ai_response])
                    return chat_history
            except Exception as e:
                error_message = f"OpenAI API Error: {str(e)}"
                logger.error(error_message)
                chat_history.append([message, error_message])
                return chat_history
                
        elif provider == "HuggingFace":
            # Get model ID from registry
            model_id, _ = get_model_info(provider, model_choice)
            if not model_id:
                error_message = f"Error: Model '{model_choice}' not found in HuggingFace"
                chat_history.append([message, error_message])
                return chat_history
                
            # Build HuggingFace payload
            payload = {
                "model": model_id,
                "messages": messages,
                "temperature": temperature,
                "max_tokens": max_tokens
            }
            
            # Call HuggingFace API
            logger.info(f"Sending request to HuggingFace model: {model_id}")
            
            try:
                response = call_huggingface_api(payload, api_key_override)
                
                # Extract response
                ai_response = extract_ai_response(response, provider)
                chat_history.append([message, ai_response])
                return chat_history
            except Exception as e:
                error_message = f"HuggingFace API Error: {str(e)}"
                logger.error(error_message)
                chat_history.append([message, error_message])
                return chat_history
                
        elif provider == "Groq":
            # Get model ID from registry
            model_id, _ = get_model_info(provider, model_choice)
            if not model_id:
                error_message = f"Error: Model '{model_choice}' not found in Groq"
                chat_history.append([message, error_message])
                return chat_history
                
            # Build Groq payload
            payload = {
                "model": model_id,
                "messages": messages,
                "temperature": temperature,
                "max_tokens": max_tokens,
                "top_p": top_p,
                "stream": stream_output
            }
            
            # Call Groq API
            logger.info(f"Sending request to Groq model: {model_id}")
            
            try:
                response = call_groq_api(payload, api_key_override)
                
                # Handle streaming response
                if stream_output:
                    # Add empty response slot to history
                    chat_history.append([message, ""])
                    
                    # Set up generator for streaming updates
                    def streaming_generator():
                        for updated_history in groq_streaming_handler(response, chat_history, len(chat_history) - 1, message):
                            yield updated_history
                    
                    return streaming_generator()
                
                # Handle normal response
                else:
                    ai_response = extract_ai_response(response, provider)
                    chat_history.append([message, ai_response])
                    return chat_history
            except Exception as e:
                error_message = f"Groq API Error: {str(e)}"
                logger.error(error_message)
                chat_history.append([message, error_message])
                return chat_history
                
        elif provider == "Cohere":
            # Get model ID from registry
            model_id, _ = get_model_info(provider, model_choice)
            if not model_id:
                error_message = f"Error: Model '{model_choice}' not found in Cohere"
                chat_history.append([message, error_message])
                return chat_history
                
            # Build Cohere payload (doesn't support streaming the same way)
            payload = {
                "model": model_id,
                "messages": messages,
                "temperature": temperature,
                "max_tokens": max_tokens
            }
            
            # Call Cohere API
            logger.info(f"Sending request to Cohere model: {model_id}")
            
            try:
                response = call_cohere_api(payload, api_key_override)
                
                # Extract response
                ai_response = extract_ai_response(response, provider)
                chat_history.append([message, ai_response])
                return chat_history
            except Exception as e:
                error_message = f"Cohere API Error: {str(e)}"
                logger.error(error_message)
                chat_history.append([message, error_message])
                return chat_history
                
        elif provider == "GLHF":
            # Get model ID from registry
            model_id, _ = get_model_info(provider, model_choice)
            if not model_id:
                error_message = f"Error: Model '{model_choice}' not found in GLHF"
                chat_history.append([message, error_message])
                return chat_history
            
            # Build GLHF payload
            payload = {
                "model": model_id,  # The hf: prefix will be added in the API call
                "messages": messages,
                "temperature": temperature,
                "max_tokens": max_tokens,
                "stream": stream_output
            }
            
            # Call GLHF API
            logger.info(f"Sending request to GLHF model: {model_id}")
            
            try:
                response = call_glhf_api(payload, api_key_override)
                
                # Handle streaming response
                if stream_output:
                    # Add empty response slot to history
                    chat_history.append([message, ""])
                    
                    # Set up generator for streaming updates
                    def streaming_generator():
                        for updated_history in glhf_streaming_handler(response, chat_history, len(chat_history) - 1, message):
                            yield updated_history
                    
                    return streaming_generator()
                
                # Handle normal response
                else:
                    ai_response = extract_ai_response(response, provider)
                    chat_history.append([message, ai_response])
                    return chat_history
            except Exception as e:
                error_message = f"GLHF API Error: {str(e)}"
                logger.error(error_message)
                chat_history.append([message, error_message])
                return chat_history
            
        else:
            error_message = f"Error: Unsupported provider '{provider}'"
            chat_history.append([message, error_message])
            return chat_history
            
    except Exception as e:
        error_message = f"Error: {str(e)}"
        logger.error(f"Exception during API call: {error_message}")
        chat_history.append([message, error_message])
        return chat_history

def clear_chat():
    """Reset all inputs"""
    return [], "", [], [], 0.7, 1000, 0.8, 0.0, 0.0, 1.0, 40, 0.1, 0, 0.0, False, "default", "none", "", []

# ==========================================================
# UI CREATION
# ==========================================================

def create_app():
    """Create the Multi-Provider CrispChat Gradio application"""
    with gr.Blocks(
        title="Multi-Provider CrispChat",
        css="""
            .context-size { 
                font-size: 0.9em;
                color: #666;
                margin-left: 10px;
            }
            footer { display: none !important; }
            .model-selection-row {
                display: flex;
                align-items: center;
            }
            .parameter-grid {
                display: grid;
                grid-template-columns: 1fr 1fr;
                gap: 10px;
            }
            .vision-badge {
                background-color: #4CAF50;
                color: white;
                padding: 3px 6px;
                border-radius: 3px;
                font-size: 0.8em;
                margin-left: 5px;
            }
            .provider-selection {
                margin-bottom: 10px;
                padding: 10px;
                border-radius: 5px;
                background-color: #f5f5f5;
            }
        """
    ) as demo:
        gr.Markdown("""
        # 🤖 Multi-Provider CrispChat
        
        Chat with AI models from multiple providers: OpenRouter, OpenAI, HuggingFace, Groq, Cohere, and GLHF.
        """)
        
        with gr.Row():
            with gr.Column(scale=2):
                # Chatbot interface
                chatbot = gr.Chatbot(
                    height=500, 
                    show_copy_button=True, 
                    show_label=False,
                    avatar_images=(None, "https://upload.wikimedia.org/wikipedia/commons/0/04/ChatGPT_logo.svg"),
                    type="messages",
                    elem_id="chat-window"
                )
                
                with gr.Row():
                    message = gr.Textbox(
                        placeholder="Type your message here...",
                        label="Message",
                        lines=2,
                        elem_id="message-input",
                        scale=4
                    )
                
                with gr.Row():
                    with gr.Column(scale=3):
                        submit_btn = gr.Button("Send", variant="primary", elem_id="send-btn")
                    
                    with gr.Column(scale=1):
                        clear_btn = gr.Button("Clear Chat", variant="secondary")
                
                with gr.Row():
                    # Image upload
                    with gr.Accordion("Upload Images (for vision models)", open=False):
                        images = gr.File(
                            label="Uploaded Images",
                            file_types=["image"],
                            file_count="multiple"
                        )
                        
                        image_upload_btn = gr.UploadButton(
                            label="Upload Images",
                            file_types=["image"],
                            file_count="multiple"
                        )
                    
                    # Document upload
                    with gr.Accordion("Upload Documents (PDF, MD, TXT)", open=False):
                        documents = gr.File(
                            label="Uploaded Documents",
                            file_types=[".pdf", ".md", ".txt"], 
                            file_count="multiple"
                        )
            
            with gr.Column(scale=1):
                with gr.Group(elem_classes="provider-selection"):
                    gr.Markdown("### Provider Selection")
                    
                    # Provider selection
                    provider_choice = gr.Radio(
                        choices=["OpenRouter", "OpenAI", "HuggingFace", "Groq", "Cohere", "GLHF"],
                        value="OpenRouter",
                        label="AI Provider"
                    )
                    
                    # API key input
                    api_key_override = gr.Textbox(
                        placeholder="Override API key (leave empty to use environment variable)",
                        label="API Key Override",
                        type="password"
                    )
                
                with gr.Group():
                    gr.Markdown("### Model Selection")
                    
                    with gr.Row(elem_classes="model-selection-row"):
                        model_search = gr.Textbox(
                            placeholder="Search models...",
                            label="",
                            show_label=False
                        )
                    
                    # Provider-specific model dropdowns
                    openrouter_model = gr.Dropdown(
                        choices=[model[0] for model in OPENROUTER_ALL_MODELS],
                        value=OPENROUTER_ALL_MODELS[0][0] if OPENROUTER_ALL_MODELS else None,
                        label="OpenRouter Model",
                        elem_id="openrouter-model-choice",
                        visible=True
                    )
                    
                    openai_model = gr.Dropdown(
                        choices=list(OPENAI_MODELS.keys()),
                        value="gpt-3.5-turbo" if "gpt-3.5-turbo" in OPENAI_MODELS else None,
                        label="OpenAI Model",
                        elem_id="openai-model-choice",
                        visible=False
                    )
                    
                    hf_model = gr.Dropdown(
                        choices=list(HUGGINGFACE_MODELS.keys()),
                        value="mistralai/Mistral-7B-Instruct-v0.3" if "mistralai/Mistral-7B-Instruct-v0.3" in HUGGINGFACE_MODELS else None,
                        label="HuggingFace Model",
                        elem_id="hf-model-choice",
                        visible=False
                    )
                    
                    groq_model = gr.Dropdown(
                        choices=list(GROQ_MODELS.keys()),
                        value="llama-3.1-8b-instant" if "llama-3.1-8b-instant" in GROQ_MODELS else None,
                        label="Groq Model",
                        elem_id="groq-model-choice",
                        visible=False
                    )
                    
                    cohere_model = gr.Dropdown(
                        choices=list(COHERE_MODELS.keys()),
                        value="command-r-plus" if "command-r-plus" in COHERE_MODELS else None,
                        label="Cohere Model",
                        elem_id="cohere-model-choice",
                        visible=False
                    )
                    
                    glhf_model = gr.Dropdown(
                        choices=list(GLHF_MODELS.keys()),
                        value="mistralai/Mistral-7B-Instruct-v0.3" if "mistralai/Mistral-7B-Instruct-v0.3" in GLHF_MODELS else None,
                        label="GLHF Model",
                        elem_id="glhf-model-choice",
                        visible=False
                    )
                    
                    context_display = gr.Textbox(
                        value=update_context_display("OpenRouter", OPENROUTER_ALL_MODELS[0][0]),
                        label="Context Size",
                        interactive=False,
                        elem_classes="context-size"
                    )
                
                with gr.Accordion("Generation Parameters", open=False):
                    with gr.Group(elem_classes="parameter-grid"):
                        temperature = gr.Slider(
                            minimum=0.0, 
                            maximum=2.0, 
                            value=0.7, 
                            step=0.1,
                            label="Temperature"
                        )
                        
                        max_tokens = gr.Slider(
                            minimum=100, 
                            maximum=4000, 
                            value=1000, 
                            step=100,
                            label="Max Tokens"
                        )
                        
                        top_p = gr.Slider(
                            minimum=0.1, 
                            maximum=1.0, 
                            value=0.8, 
                            step=0.1,
                            label="Top P"
                        )
                        
                        frequency_penalty = gr.Slider(
                            minimum=-2.0, 
                            maximum=2.0, 
                            value=0.0, 
                            step=0.1,
                            label="Frequency Penalty"
                        )
                        
                        presence_penalty = gr.Slider(
                            minimum=-2.0, 
                            maximum=2.0, 
                            value=0.0, 
                            step=0.1,
                            label="Presence Penalty"
                        )
                        
                        reasoning_effort = gr.Radio(
                            ["none", "low", "medium", "high"],
                            value="none",
                            label="Reasoning Effort (OpenRouter)"
                        )
                
                with gr.Accordion("Advanced Options", open=False):
                    with gr.Row():
                        with gr.Column():
                            repetition_penalty = gr.Slider(
                                minimum=0.1, 
                                maximum=2.0, 
                                value=1.0, 
                                step=0.1,
                                label="Repetition Penalty"
                            )
                            
                            top_k = gr.Slider(
                                minimum=1, 
                                maximum=100, 
                                value=40, 
                                step=1,
                                label="Top K"
                            )
                            
                            min_p = gr.Slider(
                                minimum=0.0, 
                                maximum=1.0, 
                                value=0.1, 
                                step=0.05,
                                label="Min P"
                            )
                        
                        with gr.Column():
                            seed = gr.Number(
                                value=0,
                                label="Seed (0 for random)",
                                precision=0
                            )
                            
                            top_a = gr.Slider(
                                minimum=0.0, 
                                maximum=1.0, 
                                value=0.0, 
                                step=0.05,
                                label="Top A"
                            )
                            
                            stream_output = gr.Checkbox(
                                label="Stream Output",
                                value=False
                            )
                    
                    with gr.Row():
                        response_format = gr.Radio(
                            ["default", "json_object"],
                            value="default",
                            label="Response Format"
                        )
                        
                        gr.Markdown("""
                        * **json_object**: Forces the model to respond with valid JSON only.
                        * Only available on certain models - check model support.
                        """)
                
                # Custom instructing options
                with gr.Accordion("Custom Instructions", open=False):
                    system_message = gr.Textbox(
                        placeholder="Enter a system message to guide the model's behavior...",
                        label="System Message",
                        lines=3
                    )
                    
                    transforms = gr.CheckboxGroup(
                        ["prompt_optimize", "prompt_distill", "prompt_compress"],
                        label="Prompt Transforms (OpenRouter specific)"
                    )
                    
                    gr.Markdown("""
                    * **prompt_optimize**: Improve prompt for better responses.
                    * **prompt_distill**: Compress prompt to use fewer tokens without changing meaning.
                    * **prompt_compress**: Aggressively compress prompt to fit larger contexts.
                    """)
                
                # Add a model information section
                with gr.Accordion("About Selected Model", open=False):
                    model_info_display = gr.HTML(
                        value=update_model_info("OpenRouter", OPENROUTER_ALL_MODELS[0][0])
                    )
        
        # Add usage instructions
        with gr.Accordion("Usage Instructions", open=False):
            gr.Markdown("""
            ## Basic Usage
            1. Type your message in the input box
            2. Select a provider and model
            3. Click "Send" or press Enter
            
            ## Working with Files
            - **Images**: Upload images to use with vision-capable models
            - **Documents**: Upload PDF, Markdown, or text files to analyze their content
            
            ## Provider Information
            - **OpenRouter**: Free access to various models with context window sizes up to 2M tokens
            - **OpenAI**: Requires an API key, includes GPT-3.5 and GPT-4 models
            - **HuggingFace**: Direct access to open models, some models require API key
            - **Groq**: High-performance inference, requires API key
            - **Cohere**: Specialized in language understanding, requires API key
            - **GLHF**: Access to HuggingFace models, requires API key
            
            ## Advanced Parameters
            - **Temperature**: Controls randomness (higher = more creative, lower = more deterministic)
            - **Max Tokens**: Maximum length of the response
            - **Top P**: Nucleus sampling threshold (higher = consider more tokens)
            - **Reasoning Effort**: Some models can show their reasoning process (OpenRouter only)
            """)
        
        # Add a footer with version info
        footer_md = gr.Markdown("""
        ---
        ### Multi-Provider CrispChat v1.0
        Built with ❤️ using Gradio and multiple AI provider APIs | Context sizes shown next to model names
        """)
        
        # Define event handlers
        def toggle_model_dropdowns(provider):
            """Show/hide model dropdowns based on provider selection"""
            return [
                gr.update(visible=(provider == "OpenRouter")),
                gr.update(visible=(provider == "OpenAI")),
                gr.update(visible=(provider == "HuggingFace")),
                gr.update(visible=(provider == "Groq")),
                gr.update(visible=(provider == "Cohere")),
                gr.update(visible=(provider == "GLHF"))
            ]
            
        def update_context_for_provider(provider, openrouter_model, openai_model, hf_model, groq_model, cohere_model, glhf_model):
            """Update context display based on selected provider and model"""
            if provider == "OpenRouter":
                return update_context_display(provider, openrouter_model)
            elif provider == "OpenAI":
                return update_context_display(provider, openai_model)
            elif provider == "HuggingFace":
                return update_context_display(provider, hf_model)
            elif provider == "Groq":
                return update_context_display(provider, groq_model)
            elif provider == "Cohere":
                return update_context_display(provider, cohere_model)
            elif provider == "GLHF":
                return update_context_display(provider, glhf_model)
            return "Unknown"
            
        def update_model_info_for_provider(provider, openrouter_model, openai_model, hf_model, groq_model, cohere_model, glhf_model):
            """Update model info based on selected provider and model"""
            if provider == "OpenRouter":
                return update_model_info(provider, openrouter_model)
            elif provider == "OpenAI":
                return update_model_info(provider, openai_model)
            elif provider == "HuggingFace":
                return update_model_info(provider, hf_model)
            elif provider == "Groq":
                return update_model_info(provider, groq_model)
            elif provider == "Cohere":
                return update_model_info(provider, cohere_model)
            elif provider == "GLHF":
                return update_model_info(provider, glhf_model)
            return "<p>Model information not available</p>"
        
    # Handling model search function - Fixed compared to previous implementation
        def search_models(provider, search_term):
            """Filter models for the selected provider based on search term"""
            filtered_models = []
            
            if provider == "OpenRouter":
                all_models = [model[0] for model in OPENROUTER_ALL_MODELS]
                if search_term:
                    filtered_models = [model for model in all_models if search_term.lower() in model.lower()]
                else:
                    filtered_models = all_models
                
                return gr.update(choices=filtered_models, value=filtered_models[0] if filtered_models else None)
                
            elif provider == "OpenAI":
                all_models = list(OPENAI_MODELS.keys())
                if search_term:
                    filtered_models = [model for model in all_models if search_term.lower() in model.lower()]
                else:
                    filtered_models = all_models
                
                return gr.update(choices=filtered_models, value=filtered_models[0] if filtered_models else None)
                
            elif provider == "HuggingFace":
                all_models = list(HUGGINGFACE_MODELS.keys())
                if search_term:
                    filtered_models = [model for model in all_models if search_term.lower() in model.lower()]
                else:
                    filtered_models = all_models
                
                return gr.update(choices=filtered_models, value=filtered_models[0] if filtered_models else None)
                
            elif provider == "Groq":
                all_models = list(GROQ_MODELS.keys())
                if search_term:
                    filtered_models = [model for model in all_models if search_term.lower() in model.lower()]
                else:
                    filtered_models = all_models
                
                return gr.update(choices=filtered_models, value=filtered_models[0] if filtered_models else None)
                
            elif provider == "Cohere":
                all_models = list(COHERE_MODELS.keys())
                if search_term:
                    filtered_models = [model for model in all_models if search_term.lower() in model.lower()]
                else:
                    filtered_models = all_models
                
                return gr.update(choices=filtered_models, value=filtered_models[0] if filtered_models else None)
                
            elif provider == "GLHF":
                all_models = list(GLHF_MODELS.keys())
                if search_term:
                    filtered_models = [model for model in all_models if search_term.lower() in model.lower()]
                else:
                    filtered_models = all_models
                
                return gr.update(choices=filtered_models, value=filtered_models[0] if filtered_models else None)
                
            # Default return in case of unknown provider
            return gr.update(choices=[], value=None)
            
        def refresh_groq_models_list():
            """Refresh the list of Groq models"""
            global GROQ_MODELS
            GROQ_MODELS = fetch_groq_models()
            return gr.update(choices=list(GROQ_MODELS.keys()))
            
        def get_current_model(provider, openrouter_model, openai_model, hf_model, groq_model, cohere_model, glhf_model):
            """Get the currently selected model based on provider"""
            if provider == "OpenRouter":
                return openrouter_model
            elif provider == "OpenAI":
                return openai_model
            elif provider == "HuggingFace":
                return hf_model
            elif provider == "Groq":
                return groq_model
            elif provider == "Cohere":
                return cohere_model
            elif provider == "GLHF":
                return glhf_model
            return None
        
        # Process uploaded images
        image_upload_btn.upload(
            fn=lambda files: files,
            inputs=image_upload_btn,
            outputs=images
        )
        
        # Set up provider selection event
        provider_choice.change(
            fn=toggle_model_dropdowns,
            inputs=provider_choice,
            outputs=[openrouter_model, openai_model, hf_model, groq_model, cohere_model, glhf_model]
        ).then(
            fn=update_context_for_provider,
            inputs=[provider_choice, openrouter_model, openai_model, hf_model, groq_model, cohere_model, glhf_model],
            outputs=context_display
        ).then(
            fn=update_model_info_for_provider,
            inputs=[provider_choice, openrouter_model, openai_model, hf_model, groq_model, cohere_model, glhf_model],
            outputs=model_info_display
        )
        
        # Set up model search event - FIXED VERSION
        # Important: We need to return a proper Gradio component update for each dropdown
        model_search.change(
            fn=search_models,
            inputs=[provider_choice, model_search],
            outputs=[openrouter_model]  # This will be handled by the JS forwarding logic
        )
        
        # Set up model change events
        openrouter_model.change(
            fn=lambda model: update_context_display("OpenRouter", model),
            inputs=openrouter_model,
            outputs=context_display
        ).then(
            fn=lambda model: update_model_info("OpenRouter", model),
            inputs=openrouter_model,
            outputs=model_info_display
        )
        
        openai_model.change(
            fn=lambda model: update_context_display("OpenAI", model),
            inputs=openai_model,
            outputs=context_display
        ).then(
            fn=lambda model: update_model_info("OpenAI", model),
            inputs=openai_model,
            outputs=model_info_display
        )
        
        hf_model.change(
            fn=lambda model: update_context_display("HuggingFace", model),
            inputs=hf_model,
            outputs=context_display
        ).then(
            fn=lambda model: update_model_info("HuggingFace", model),
            inputs=hf_model,
            outputs=model_info_display
        )
        
        groq_model.change(
            fn=lambda model: update_context_display("Groq", model),
            inputs=groq_model,
            outputs=context_display
        ).then(
            fn=lambda model: update_model_info("Groq", model),
            inputs=groq_model,
            outputs=model_info_display
        )
        
        cohere_model.change(
            fn=lambda model: update_context_display("Cohere", model),
            inputs=cohere_model,
            outputs=context_display
        ).then(
            fn=lambda model: update_model_info("Cohere", model),
            inputs=cohere_model,
            outputs=model_info_display
        )
        
        glhf_model.change(
            fn=lambda model: update_context_display("GLHF", model),
            inputs=glhf_model,
            outputs=context_display
        ).then(
            fn=lambda model: update_model_info("GLHF", model),
            inputs=glhf_model,
            outputs=model_info_display
        )
        
        # Add custom JavaScript for routing model search to visible dropdown
        gr.HTML("""
        <script>
            // To be triggered after page load
            document.addEventListener('DOMContentLoaded', function() {
                // Find dropdowns
                const providerRadio = document.querySelector('input[name="provider_choice"]');
                const searchInput = document.getElementById('model_search');
                
                if (providerRadio && searchInput) {
                    // When provider changes, clear the search
                    providerRadio.addEventListener('change', function() {
                        searchInput.value = '';
                    });
                }
            });
        </script>
        """)
        
        # Set up submission event
        def submit_message(message, history, provider, openrouter_model, openai_model, hf_model, groq_model, cohere_model, glhf_model, 
                          temperature, max_tokens, top_p, frequency_penalty, presence_penalty, repetition_penalty, 
                          top_k, min_p, seed, top_a, stream_output, response_format,
                          images, documents, reasoning_effort, system_message, transforms, api_key_override):
            """Submit message to selected provider and model"""
            # Get the currently selected model
            model_choice = get_current_model(provider, openrouter_model, openai_model, hf_model, groq_model, cohere_model, glhf_model)
            
            # Check if model is selected
            if not model_choice:
                history.append([message, f"Error: No model selected for provider {provider}"])
                return history
                
            # Call the ask_ai function with the appropriate parameters
            return ask_ai(
                message=message,
                history=history,
                provider=provider,
                model_choice=model_choice,
                temperature=temperature,
                max_tokens=max_tokens,
                top_p=top_p,
                frequency_penalty=frequency_penalty,
                presence_penalty=presence_penalty,
                repetition_penalty=repetition_penalty,
                top_k=top_k,
                min_p=min_p,
                seed=seed,
                top_a=top_a,
                stream_output=stream_output,
                response_format=response_format,
                images=images,
                documents=documents,
                reasoning_effort=reasoning_effort,
                system_message=system_message,
                transforms=transforms,
                api_key_override=api_key_override
            )
        
        # Submit button click event
        submit_btn.click(
            fn=submit_message,
            inputs=[
                message, chatbot, provider_choice, 
                openrouter_model, openai_model, hf_model, groq_model, cohere_model, glhf_model,
                temperature, max_tokens, top_p, frequency_penalty, presence_penalty, repetition_penalty, 
                top_k, min_p, seed, top_a, stream_output, response_format,
                images, documents, reasoning_effort, system_message, transforms, api_key_override
            ],
            outputs=chatbot,
            show_progress="minimal",
        ).then(
            fn=lambda: "",  # Clear message box after sending
            inputs=None,
            outputs=message
        )
        
        # Also submit on Enter key
        message.submit(
            fn=submit_message,
            inputs=[
                message, chatbot, provider_choice, 
                openrouter_model, openai_model, hf_model, groq_model, cohere_model, glhf_model,
                temperature, max_tokens, top_p, frequency_penalty, presence_penalty, repetition_penalty, 
                top_k, min_p, seed, top_a, stream_output, response_format,
                images, documents, reasoning_effort, system_message, transforms, api_key_override
            ],
            outputs=chatbot,
            show_progress="minimal",
        ).then(
            fn=lambda: "",  # Clear message box after sending
            inputs=None,
            outputs=message
        )
        
        # Clear chat button
        clear_btn.click(
            fn=clear_chat,
            inputs=[],
            outputs=[
                chatbot, message, images, documents, temperature, 
                max_tokens, top_p, frequency_penalty, presence_penalty,
                repetition_penalty, top_k, min_p, seed, top_a, stream_output,
                response_format, reasoning_effort, system_message, transforms
            ]
        )
        
        return demo

# Launch the app
if __name__ == "__main__":
    # Check API keys and print status
    missing_keys = []
    
    if not OPENROUTER_API_KEY:
        logger.warning("WARNING: OPENROUTER_API_KEY environment variable is not set")
        missing_keys.append("OpenRouter")
    
    if not OPENAI_API_KEY:
        logger.warning("WARNING: OPENAI_API_KEY environment variable is not set")
        missing_keys.append("OpenAI")
        
    if not GROQ_API_KEY:
        logger.warning("WARNING: GROQ_API_KEY environment variable is not set")
        missing_keys.append("Groq")
        
    if not COHERE_API_KEY:
        logger.warning("WARNING: COHERE_API_KEY environment variable is not set")
        missing_keys.append("Cohere")
        
    if not GLHF_API_KEY:
        logger.warning("WARNING: GLHF_API_KEY environment variable is not set")
        missing_keys.append("GLHF")
        
    if missing_keys:
        print("Missing API keys for the following providers:")
        for key in missing_keys:
            print(f"- {key}")
        print("\nYou can still use the application, but some providers will require API keys.")
        print("You can provide API keys through environment variables or use the API Key Override field.")
        
        if "OpenRouter" in missing_keys:
            print("\nNote: OpenRouter offers free tier access to many models!")
            
    print("\nStarting Multi-Provider CrispChat application...")
    demo = create_app()
    demo.launch(
        server_name="0.0.0.0", 
        server_port=7860, 
        debug=True,
        show_error=True
    )