app.py

import threading
import time
import gradio as gr
import logging
import json
import re
import torch
import tempfile
import subprocess
import ast
import os
import dataclasses
from pathlib import Path
from typing import Dict, List, Tuple, Optional, Any, Union
from dataclasses import dataclass, field
from enum import Enum
from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline
from sentence_transformers import SentenceTransformer
import faiss
import numpy as np
from PIL import Image

# Configure logging
logging.basicConfig(
    level=logging.INFO,
    format='%(asctime)s - %(name)s - %(levelname)s - %(message)s',
    handlers=[
        logging.StreamHandler(),
        logging.FileHandler('gradio_builder.log')
    ]
)
logger = logging.getLogger(__name__)

# Constants
DEFAULT_PORT = 7860
MODEL_CACHE_DIR = Path("model_cache")
TEMPLATE_DIR = Path("templates")
TEMP_DIR = Path("temp")
DATABASE_PATH = Path("code_database.json") #Path for our simple database


# Ensure directories exist
for directory in [MODEL_CACHE_DIR, TEMPLATE_DIR, TEMP_DIR]:
    directory.mkdir(exist_ok=True, parents=True)


@dataclass
class Template:
    code: str
    description: str
    components: List[str]
    metadata: Dict[str, Any] = field(default_factory=dict)
    version: str = "1.0"

class TemplateManager:
    # ... (TemplateManager remains the same) ...


class RAGSystem:
    def __init__(self, model_name: str = "gpt2", device: str = "cuda" if torch.cuda.is_available() else "cpu", embedding_model="all-mpnet-base-v2"):
        try:
            self.tokenizer = AutoTokenizer.from_pretrained(model_name)
            self.model = AutoModelForCausalLM.from_pretrained(model_name).to(device)
            self.device = device
            self.pipe = pipeline("text-generation", model=self.model, tokenizer=self.tokenizer, device=self.device)
            self.embedding_model = SentenceTransformer(embedding_model)
            self.load_database()
        except Exception as e:
            logger.error(f"Error loading language model or embedding model: {e}.  Falling back to placeholder generation.")
            self.pipe = None
            self.embedding_model = None
            self.code_embeddings = None


    def load_database(self):
      """Loads or creates the code database"""
      if DATABASE_PATH.exists():
        try:
          with open(DATABASE_PATH, 'r', encoding='utf-8') as f:
            self.database = json.load(f)
            self.code_embeddings = np.array(self.database['embeddings'])
          logger.info("Loaded code database from file")
        except (json.JSONDecodeError, KeyError) as e:
          logger.error(f"Error loading code database: {e}.  Creating new database.")
          self.database = {'codes': [], 'embeddings': []}
          self.code_embeddings = np.array([])

      else:
        logger.info("Code database does not exist. Creating new database.")
        self.database = {'codes': [], 'embeddings': []}
        self.code_embeddings = np.array([])

      if self.embedding_model and len(self.database['codes']) != len(self.database['embeddings']):
        logger.warning("Mismatch between number of codes and embeddings, rebuilding embeddings")
        self.rebuild_embeddings()
      elif self.embedding_model is None:
        logger.warning("Embeddings are not supported in this context. ")
      #Index the embeddings for efficient searching
      if len(self.code_embeddings) > 0 and self.embedding_model:
        self.index = faiss.IndexFlatL2(self.code_embeddings.shape[1]) #L2 distance
        self.index.add(self.code_embeddings)

    def add_to_database(self, code: str):
        """Adds a code snippet to the database"""
        try:
          embedding = self.embedding_model.encode(code)
          self.database['codes'].append(code)
          self.database['embeddings'].append(embedding.tolist())
          self.code_embeddings = np.vstack((self.code_embeddings, embedding))
          self.index.add(np.array([embedding])) # update FAISS index
          self.save_database()
          logger.info(f"Added code snippet to database. Total size:{len(self.database['codes'])}")
        except Exception as e:
          logger.error(f"Error adding to database: {e}")


    def save_database(self):
      """Saves the database to a file"""
      try:
        with open(DATABASE_PATH, 'w', encoding='utf-8') as f:
          json.dump(self.database, f, indent=2)
        logger.info(f"Saved database to {DATABASE_PATH}")
      except Exception as e:
        logger.error(f"Error saving database: {e}")

    def rebuild_embeddings(self):
      """rebuilds embeddings from the codes"""
      try:
        embeddings = self.embedding_model.encode(self.database['codes'])
        self.code_embeddings = embeddings
        self.database['embeddings'] = embeddings.tolist()
        self.index = faiss.IndexFlatL2(embeddings.shape[1]) #L2 distance
        self.index.add(embeddings)
        self.save_database()
        logger.info("Rebuilt and saved embeddings to the database")
      except Exception as e:
        logger.error(f"Error rebuilding embeddings: {e}")


    def retrieve_similar_code(self, description: str, top_k: int = 3) -> List[str]:
      """Retrieves similar code snippets from the database"""
      if self.embedding_model is None:
        return []
      try:
        embedding = self.embedding_model.encode(description)
        D, I = self.index.search(np.array([embedding]), top_k)
        return [self.database['codes'][i] for i in I[0]]
      except Exception as e:
        logger.error(f"Error retrieving similar code: {e}")
        return []

    def generate_code(self, description: str, template_code: str) -> str:
        retrieved_codes = self.retrieve_similar_code(description)
        prompt = f"Description: {description}\nRetrieved Code Snippets:\n{''.join([f'```python\n{code}\n```\n' for code in retrieved_codes])}\nTemplate:\n```python\n{template_code}\n```\nGenerated Code:\n```python\n"
        if self.pipe:
            try:
                generated_text = self.pipe(prompt, max_length=500, num_return_sequences=1)[0]['generated_text']
                generated_code = generated_text.split("Generated Code:")[1].strip().split('```')[0]
                return generated_code
            except Exception as e:
                logger.error(f"Error generating code with language model: {e}. Returning template code.")
                return template_code
        else:
            return f"# Placeholder code generation. Description: {description}\n{template_code}"

    def generate_interface(self, screenshot: Optional[Image.Image], description: str) -> str:
        retrieved_codes = self.retrieve_similar_code(description)
        prompt = f"Create a Gradio interface based on this description: {description}\nRetrieved Code Snippets:\n{''.join([f'```python\n{code}\n```\n' for code in retrieved_codes])}"
        if screenshot:
            prompt += "\nThe interface should resemble the provided screenshot."
        prompt += "\n```python\n"
        if self.pipe:
            try:
                generated_text = self.pipe(prompt, max_length=500, num_return_sequences=1)[0]['generated_text']
                generated_code = generated_text.split("```")[1].strip()
                return generated_code
            except Exception as e:
                logger.error(f"Error generating interface with language model: {e}. Returning placeholder.")
                return "import gradio as gr\n\ndemo = gr.Interface(fn=lambda x:x, inputs='text', outputs='text')\ndemo.launch()"
        else:
            return "import gradio as gr\n\ndemo = gr.Interface(fn=lambda x:x, inputs='text', outputs='text')\ndemo.launch()"

class PreviewManager:
    # ... (PreviewManager remains largely the same) ...


class GradioInterface:
    def __init__(self):
        self.template_manager = TemplateManager(TEMPLATE_DIR)
        self.template_manager.load_templates()
        self.current_code = ""
        self.rag_system = RAGSystem()
        self.preview_manager = PreviewManager()

    def _extract_components(self, code: str) -> List[str]:
        """Extract components from the code."""
        # This logic should analyze the code and extract components.
        # For example, you might look for function definitions, classes, etc.
        components = []
        # Simple regex to find function definitions
        function_matches = re.findall(r'def (\w+)', code)
        components.extend(function_matches)

        # Simple regex to find class definitions
        class_matches = re.findall(r'class (\w+)', code)
        components.extend(class_matches)

        # You can add more sophisticated logic here as needed
        return components

    def _get_template_choices(self) -> List[str]:
        """Get available template choices."""
        return list(self.template_manager.templates.keys())

    def launch(self, **kwargs):
        with gr.Blocks() as interface:
            gr.Markdown("## Code Generation Interface")
            description_input = gr.Textbox(label="Description", placeholder="Enter a description for the code you want to generate.")
            code_output = gr.Textbox(label="Generated Code", interactive=False)
            generate_button = gr.Button("Generate Code")
            template_choice = gr.Dropdown(label="Select Template", choices=self._get_template_choices(), value=None)
            save_button = gr.Button("Save as Template")

            # Generate code button action
            generate_button.click(
                fn=self.generate_code,
                inputs=description_input,
                outputs=code_output
            )

            # Save template button action
            save_button.click(
                fn=self.save_template,
                inputs=[code_output, template_choice, description_input],
                outputs=code_output
            )

            # Additional UI elements can be added here
            gr.Markdown("### Preview")
            preview_output = gr.Textbox(label="Preview", interactive=False)
            self.preview_manager.update_preview(code_output)  # Update preview with generated code

            # Update preview when code is generated
            generate_button.click(
                fn=lambda code: self.preview_manager.update_preview(code),
                inputs=code_output,
                outputs=preview_output
            )

        interface.launch(**kwargs)

    def generate_code(self, description: str) -> str:
        """Generate code based on the description."""
        template_code = ""  # Placeholder for template code
        return self.rag_system.generate_code(description, template_code)

    def save_template(self, code: str, name: str, description: str) -> str:
        """Save the generated code as a template."""
        try:
            components = self._extract_components(code)
            template = Template(code=code, description=description, components=components)
            if self.template_manager.save_template(name, template):
                self.rag_system.add_to_database(code)  # Add code to the database
                return f"✅ Template '{name}' saved successfully."
            else:
                return "❌ Failed to save template."
        except Exception as e:
            logger.error(f"Error saving template: {e}")
            return f"❌ Error saving template: {str(e)}"


def main():
    # Configure logging
    logging.basicConfig(
        level=logging.INFO,
        format='%(asctime)s - %(name)s - %(levelname)s - %(message)s',
        handlers=[
            logging.StreamHandler(),
            logging.FileHandler('gradio_builder.log')
        ]
    )
    logger = logging.getLogger(__name__)
    logger.info("=== Application Startup ===")
    
    try:
        # Initialize and launch interface
        interface = GradioInterface()
        interface.launch(
            server_port=DEFAULT_PORT,
            share=False,
            debug=True
        )
    except Exception as e:
        logger.error(f"Application error: {e}")
        raise
    finally:
        logger.info("=== Application Shutdown ===")

if __name__ == "__main__":
    main()