import os
import gradio as gr
import asyncio
import tempfile
import subprocess
import shutil
from dotenv import find_dotenv, load_dotenv
from langchain.chat_models import init_chat_model

# Auto-install Deno if not found (for Hugging Face Spaces)
def ensure_deno_installed():
    """Install Deno if not already installed (for Hugging Face Spaces compatibility)"""
    try:
        # Check if Deno is already installed
        result = subprocess.run(['deno', '--version'], capture_output=True, text=True)
        if result.returncode == 0:
            print(f"✅ Deno already installed: {result.stdout.split()[1]}")
            return True
    except FileNotFoundError:
        pass
    
    print("🔧 Deno not found. Installing Deno for PyodideSandbox...")
    
    try:
        # Install Deno using the official installer
        install_cmd = "curl -fsSL https://deno.land/install.sh | sh"
        result = subprocess.run(install_cmd, shell=True, capture_output=True, text=True)
        
        if result.returncode == 0:
            # Add Deno to PATH
            deno_path = os.path.expanduser("~/.deno/bin")
            if deno_path not in os.environ.get("PATH", ""):
                os.environ["PATH"] = f"{deno_path}:{os.environ.get('PATH', '')}"
            
            print("✅ Deno installed successfully!")
            return True
        else:
            print(f"❌ Deno installation failed: {result.stderr}")
            return False
            
    except Exception as e:
        print(f"❌ Error installing Deno: {e}")
        return False

# Install Deno before importing sandbox dependencies
print("🔍 Checking Deno installation...")
deno_available = ensure_deno_installed()

# Import the CodeAct agent functionality
from agent import FileInjectedPyodideSandbox, create_pyodide_eval_fn, create_codeact

# Import the new guided analysis functionality
from graph import analyze_file_with_guidance_sync, guided_analysis_graph
from graph_streaming import streaming_analyze_file_with_guidance

# Load environment variables
load_dotenv(find_dotenv())

# Initialize model for file analysis
codeact_model = init_chat_model("o3-2025-04-16", model_provider="openai")

# Store uploaded file path globally
uploaded_file_path = None

def handle_file_upload(file):
    """Handle file upload and store the path globally"""
    global uploaded_file_path
    try:
        if file is not None:
            # With type="filepath", Gradio returns the file path as a string
            uploaded_file_path = file
            filename = os.path.basename(file)
            return f"✅ File uploaded successfully: {filename}"
        else:
            uploaded_file_path = None
            return "❌ No file uploaded"
    except Exception as e:
        uploaded_file_path = None
        return f"❌ Upload error: {str(e)}"

def streaming_analyze_file_with_question(user_question):
    """
    Streaming version that yields progress updates in real-time
    """
    global uploaded_file_path, deno_available
    
    try:
        if not uploaded_file_path or not os.path.exists(uploaded_file_path):
            yield "❌ No file uploaded or file not found. Please upload a file first."
            return
        
        if not user_question or user_question.strip() == "":
            user_question = "Provide a comprehensive analysis of this file including security, performance, and data insights."
        
        # Check if Deno is available for sandbox operations
        if not deno_available:
            yield """❌ Deno runtime not available. This is required for code execution in the sandbox.
            
📋 Troubleshooting:
1. This usually happens on deployment platforms that don't have Deno pre-installed
2. The app attempted to install Deno automatically but failed
3. Try restarting the space or contact support

🔄 Alternative: You can still upload files, but advanced code analysis may be limited."""
            return
        
        # Use the streaming guided analysis approach
        for chunk in streaming_analyze_file_with_guidance(uploaded_file_path, user_question):
            yield chunk
        
    except Exception as e:
        error_msg = str(e)
        if "Deno" in error_msg or "deno" in error_msg:
            yield f"""❌ Deno-related error in analysis: {error_msg}

🔧 This appears to be a Deno runtime issue. The sandbox requires Deno for code execution.
Try restarting the application or contact support if this persists."""
        else:
            yield f"❌ Error in guided analysis: {error_msg}"

def analyze_file_with_question(user_question):
    """
    Non-streaming version for backward compatibility
    """
    global uploaded_file_path, deno_available
    
    try:
        if not uploaded_file_path or not os.path.exists(uploaded_file_path):
            return "❌ No file uploaded or file not found. Please upload a file first."
        
        if not user_question or user_question.strip() == "":
            user_question = "Provide a comprehensive analysis of this file including security, performance, and data insights."
        
        # Check if Deno is available for sandbox operations
        if not deno_available:
            return """❌ Deno runtime not available. This is required for code execution in the sandbox.
            
📋 Troubleshooting:
1. This usually happens on deployment platforms that don't have Deno pre-installed
2. The app attempted to install Deno automatically but failed
3. Try restarting the space or contact support

🔄 Alternative: You can still upload files, but advanced code analysis may be limited."""
        
        # Use the new guided analysis approach
        result = analyze_file_with_guidance_sync(uploaded_file_path, user_question)
        return result
        
    except Exception as e:
        error_msg = str(e)
        if "Deno" in error_msg or "deno" in error_msg:
            return f"""❌ Deno-related error in analysis: {error_msg}

🔧 This appears to be a Deno runtime issue. The sandbox requires Deno for code execution.
Try restarting the application or contact support if this persists."""
        else:
            return f"❌ Error in guided analysis: {error_msg}"

async def analyze_uploaded_file():
    """Legacy function - kept for backward compatibility"""
    return analyze_file_with_question("Provide a comprehensive analysis of this file.")

def run_file_analysis():
    """Wrapper to run async file analysis in sync context"""
    return asyncio.run(analyze_uploaded_file())

# Create the Gradio interface
with gr.Blocks(title="DataForge - AI CodeAct Agent") as demo:
    gr.Markdown("# 🤖 DataForge - AI CodeAct Agent")
    
    # Demo Video Section
    gr.Markdown("""
    ## 🎥 **Demo Video - See DataForge in Action!**
    
    **📺 [Watch the full demo on YouTube](https://www.youtube.com/watch?v=f5jp2i3engM)** - Learn how to use DataForge in just a few minutes!
    
    ---
    """)
    
    gr.Markdown("""
    ## 🔑 **AI Writes Code to Analyze Your Data Locally**
    
    **Why DataForge handles massive files when other AI tools fail:**
    
    ❌ **Other AI Tools**: Upload data to LLM → Hit limits → Fail on large files  
    ✅ **DataForge**: AI writes code → Code processes data locally → No limits!
    
    ### 💪 **Key Benefits:**
    - **♾️ No Size Limits** - Process GB+ files locally
    - **🛡️ Complete Privacy** - Data never leaves your machine  
    - **⚡ Lightning Fast** - No uploads, pure local processing
    - **🎯 Custom Analysis** - Code written for your specific question
    """)
    
    # Supported File Types - Simple Version
    gr.Markdown("## 📋 **Supported Files**")
    gr.Markdown("""
    **📊 Data:** CSV, JSON, XML, TSV  
    **📝 Logs:** Application, access, error, audit logs  
    **🗂️ Text:** Any text file, code files, configs  
    **💾 Size:** No limits - handles multi-GB files locally
    """)
    
    with gr.Row():
        with gr.Column(scale=1):
            # File Upload Section
            gr.Markdown("### 📤 File Upload")
            file_upload = gr.File(
                label="Upload File for Analysis",
                type="filepath"
            )
            upload_status = gr.Textbox(
                label="Upload Status",
                value="No file uploaded",
                interactive=False
            )
            
            
            # Question Section
            gr.Markdown("### ❓ Ask Your Question")
            user_question = gr.Textbox(
                label="Your Question about the File",
                placeholder="What would you like to know about this file? (e.g., 'Find security threats', 'Show performance issues', 'What errors are present?')",
                lines=4,
                value=""
            )
            
            analyze_btn = gr.Button("🤖 Activate CodeAct Agent", variant="primary", size="lg")
            
            # How it works
            gr.Markdown("### 🔬 **How It Works**")
            gr.Markdown("""
            1. **🔍 AI samples** your file structure  
            2. **⚡ AI writes** custom analysis code  
            3. **🚀 Code processes** your entire file locally  
            4. **📊 Results** delivered to you
            
            **Your data never leaves your machine!**
            """)
        
        with gr.Column(scale=2):
            analysis_output = gr.Textbox(
                label="🤖 CodeAct Agent Analysis Results",
                lines=25,
                max_lines=35,
                placeholder="Upload a file, ask your question, and click 'Activate CodeAct Agent' to watch the AI write and execute custom analysis code in real-time...",
                interactive=False
            )
    
    # Event handlers
    file_upload.change(
        fn=handle_file_upload,
        inputs=[file_upload],
        outputs=[upload_status]
    )
    
    analyze_btn.click(
        fn=streaming_analyze_file_with_question,
        inputs=[user_question],
        outputs=[analysis_output]
    )
    
    gr.Markdown("---")
    gr.Markdown("## 💡 **Real CodeAct Use Cases - When LLMs Fail, Code Succeeds**")
    gr.Markdown("""
    **🔥 Real Problem:** You have a 500MB server log file and want to ask: *"Which IP addresses made the most requests yesterday?"*
    
    ❌ **Traditional LLM:** "File too large, please upload smaller chunks"  
    ✅ **DataForge CodeAct:** AI writes Python code with regex + Counter to process entire file → Gets exact answer
    
    **🔥 Real Problem:** You have a 2GB CSV with sales data and ask: *"How many orders were placed in each month of 2024?"*
    
    ❌ **Traditional LLM:** Crashes on upload or hits token limits  
    ✅ **DataForge CodeAct:** AI writes pandas code to parse dates and count by month → Simple monthly breakdown
    
    **🔥 Real Problem:** You have 1GB of JSON API logs and ask: *"Find all 500 errors and group them by endpoint and time of day"*
    
    ❌ **Traditional LLM:** "Cannot process this much data"  
    ✅ **DataForge CodeAct:** AI writes JSON parsing + datetime analysis code → Complete error breakdown
    
    **🎯 The Key:** Instead of sending your data to an LLM, the LLM writes code that processes your data locally!
    """)

if __name__ == "__main__":
    print("🤖 Starting DataForge CodeAct Agent Application...")
    print("🚀 Initializing advanced AI-powered file analysis capabilities...")
    demo.launch()