app.py

import gradio as gr
from llama_cpp import Llama
from huggingface_hub import hf_hub_download
import json
import re

def load_model():
    repo_id = "forestav/gguf_lora_model"
    model_file = "unsloth.F16.gguf"  
    
    local_path = hf_hub_download(
        repo_id=repo_id,
        filename=model_file
    )
    
    print(f"Loading model from: {local_path}")
    
    model = Llama(
        model_path=local_path,
        n_ctx=2048,
        n_threads=8
    )
    
    return model

# Enhanced generation with multiple modes
def generate_response(message, history, mode='chat'):
    # Preprocessing based on mode
    if mode == 'code':
        system_prompt = "You are an expert coding assistant. Provide clean, efficient code solutions."
    elif mode == 'creative':
        system_prompt = "You are a creative writing assistant. Generate imaginative and engaging content."
    elif mode == 'analytical':
        system_prompt = "You are an analytical assistant. Provide deep, structured insights and reasoning."
    else:
        system_prompt = "You are a helpful AI assistant."
    
    # Prepare messages with system context
    messages = [
        {"role": "system", "content": system_prompt},
        *[{"role": "user" if i % 2 == 0 else "assistant", "content": msg} 
          for i, msg in enumerate(sum(history, []))],
        {"role": "user", "content": message}
    ]
    
    # Generate response
    response = model.create_chat_completion(
        messages=messages,
        max_tokens=512,
        temperature=0.7,
        top_p=0.95,
    )
    
    return response['choices'][0]['message']['content']

# Extract structured data from text
def extract_structured_data(text):
    try:
        # Try to extract JSON-like structures
        json_match = re.search(r'\{.*\}', text, re.DOTALL)
        if json_match:
            try:
                return json.loads(json_match.group(0))
            except json.JSONDecodeError:
                pass
        
        # Fall back to custom parsing for key-value pairs
        data = {}
        for line in text.split('\n'):
            if ':' in line:
                key, value = line.split(':', 1)
                data[key.strip()] = value.strip()
        
        return data
    except Exception as e:
        return {"error": str(e)}

# Create Gradio interface with multiple tabs
def create_interface():
    with gr.Blocks() as demo:
        gr.Markdown("# Multi-Mode AI Assistant")
        
        with gr.Tabs():
            # Chat Interface
            with gr.TabItem("Conversational Chat"):
                chat_interface = gr.ChatInterface(
                    fn=lambda message, history: generate_response(message, history, 'chat'),
                    title="Conversational AI",
                    description="General-purpose conversation mode"
                )
            
            # Code Generation Tab
            with gr.TabItem("Code Assistant"):
                code_interface = gr.ChatInterface(
                    fn=lambda message, history: generate_response(message, history, 'code'),
                    title="AI Code Generator",
                    description="Generate code snippets and solve programming challenges"
                )
            
            # Creative Writing Tab
            with gr.TabItem("Creative Writing"):
                creative_interface = gr.ChatInterface(
                    fn=lambda message, history: generate_response(message, history, 'creative'),
                    title="Creative Writing Assistant",
                    description="Generate stories, poems, and creative content"
                )
            
            # Data Extraction Tab
            with gr.TabItem("Data Extractor"):
                with gr.Row():
                    text_input = gr.Textbox(label="Input Text")
                    extract_btn = gr.Button("Extract Structured Data")
                    json_output = gr.JSON(label="Extracted Data")
                
                extract_btn.click(
                    fn=extract_structured_data, 
                    inputs=text_input, 
                    outputs=json_output
                )
    
    return demo

# Load model globally
print("Starting model loading...")
model = load_model()
print("Model loaded successfully!")

# Create and launch the interface
demo = create_interface()
demo.launch(
    server_name="0.0.0.0",  # Necessary for Spaces
    server_port=7860,       # Standard port for Spaces
    share=True            # Don't need share link in Spaces
)