_archive/app copy.py

import gradio as gr
from ui.dataset_input import create_dataset_input, load_example_dataset
from ui.analysis_screen import create_analysis_screen, process_analysis_request
from visualization.bow_visualizer import process_and_visualize_analysis
import nltk
import os
import json

# Download necessary NLTK resources function remains unchanged
def download_nltk_resources():
    """Download required NLTK resources if not already downloaded"""
    try:
        # Create nltk_data directory in the user's home directory if it doesn't exist
        nltk_data_path = os.path.expanduser("~/nltk_data")
        os.makedirs(nltk_data_path, exist_ok=True)
        
        # Add this path to NLTK's data path
        nltk.data.path.append(nltk_data_path)
        
        # Download required resources
        resources = ['punkt', 'wordnet', 'stopwords', 'punkt_tab']
        for resource in resources:
            try:
                # Different resources can be in different directories in NLTK
                locations = [
                    f'tokenizers/{resource}',
                    f'corpora/{resource}',
                    f'taggers/{resource}',
                    f'{resource}'
                ]
                
                found = False
                for location in locations:
                    try:
                        nltk.data.find(location)
                        print(f"Resource {resource} already downloaded")
                        found = True
                        break
                    except LookupError:
                        continue
                
                if not found:
                    print(f"Downloading {resource}...")
                    nltk.download(resource, quiet=True)
            except Exception as e:
                print(f"Error with resource {resource}: {e}")
        
        print("NLTK resources check completed")
    except Exception as e:
        print(f"Error downloading NLTK resources: {e}")

def create_app():
    """
    Create a streamlined Gradio app for dataset input and Bag of Words analysis. 
    
    Returns:
        gr.Blocks: The Gradio application
    """
    with gr.Blocks(title="LLM Response Comparator") as app:
        # Application state to share data between tabs
        dataset_state = gr.State({})
        analysis_results_state = gr.State({})
        
        # Dataset Input Tab
        with gr.Tab("Dataset Input"):
            dataset_inputs, example_dropdown, load_example_btn, create_btn, prompt, response1, model1, response2, model2 = create_dataset_input()
            
            # Add status indicator to show when dataset is created
            dataset_status = gr.Markdown("*No dataset loaded*")
            
            # Load example dataset
            load_example_btn.click(
                fn=load_example_dataset,
                inputs=[example_dropdown],
                outputs=[prompt, response1, model1, response2, model2]  # Update all field values
            )

            # Save dataset to state and update status
            def create_dataset(p, r1, m1, r2, m2):
                if not p or not r1 or not r2:
                    return {}, "❌ **Error:** Please fill in at least the prompt and both responses"
                
                dataset = {
                    "entries": [
                        {"prompt": p, "response": r1, "model": m1 or "Model 1"},
                        {"prompt": p, "response": r2, "model": m2 or "Model 2"}
                    ]
                }
                return dataset, "✅ **Dataset created successfully!** You can now go to the Analysis tab"
                
            create_btn.click(
                fn=create_dataset,
                inputs=[prompt, response1, model1, response2, model2],
                outputs=[dataset_state, dataset_status]
            )
        
        # Analysis Tab
        with gr.Tab("Analysis"):
            # Use create_analysis_screen to get UI components including visualization container
            analysis_options, analysis_params, run_analysis_btn, analysis_output, bow_top_slider = create_analysis_screen()
            
            # Pre-create visualization components (initially hidden)
            visualization_area_visible = gr.Checkbox(value=False, visible=False, label="Visualization Visible")
            analysis_title = gr.Markdown("## Analysis Results", visible=False)
            prompt_title = gr.Markdown(visible=False)
            models_compared = gr.Markdown(visible=False)
            
            # Container for model 1 words
            model1_title = gr.Markdown(visible=False)
            model1_words = gr.Markdown(visible=False)
            
            # Container for model 2 words
            model2_title = gr.Markdown(visible=False)
            model2_words = gr.Markdown(visible=False)
            
            # Similarity metrics
            similarity_metrics_title = gr.Markdown("### Similarity Metrics", visible=False)
            similarity_metrics = gr.Markdown(visible=False)
            
            # Status or error message area
            status_message_visible = gr.Checkbox(value=False, visible=False, label="Status Message Visible")
            status_message = gr.Markdown(visible=False)
            
            # Define a helper function to extract parameter values and run the analysis
            def run_analysis(dataset, selected_analyses, bow_top, ngram_n, ngram_top):
                try:
                    if not dataset or "entries" not in dataset or not dataset["entries"]:
                        return (
                            {},  # analysis_results_state
                            False,  # analysis_output visibility
                            False,  # visualization_area_visible
                            gr.update(visible=False),  # analysis_title
                            gr.update(visible=False),  # prompt_title
                            gr.update(visible=False),  # models_compared
                            gr.update(visible=False),  # model1_title
                            gr.update(visible=False),  # model1_words
                            gr.update(visible=False),  # model2_title
                            gr.update(visible=False),  # model2_words
                            gr.update(visible=False),  # similarity_metrics_title
                            gr.update(visible=False),  # similarity_metrics
                            True,  # status_message_visible
                            gr.update(visible=True, value="❌ **Error:** No dataset loaded. Please create or load a dataset first.")  # status_message
                        )
                    
                    parameters = {
                        "bow_top": bow_top,
                        "ngram_n": ngram_n,
                        "ngram_top": ngram_top
                    }
                    print("Running analysis with parameters:", parameters)
                    
                    # Process the analysis request
                    analysis_results, _ = process_analysis_request(dataset, selected_analyses, parameters)
                    
                    # If there's an error or no results
                    if not analysis_results or "analyses" not in analysis_results or not analysis_results["analyses"]:
                        return (
                            analysis_results,
                            False,
                            False,
                            gr.update(visible=False),
                            gr.update(visible=False),
                            gr.update(visible=False),
                            gr.update(visible=False),
                            gr.update(visible=False),
                            gr.update(visible=False),
                            gr.update(visible=False),
                            gr.update(visible=False),
                            gr.update(visible=False),
                            True,
                            gr.update(visible=True, value="❌ **No results found.** Try different analysis options.")
                        )
                    
                    # Extract information to display in components
                    prompt = list(analysis_results["analyses"].keys())[0]
                    analyses = analysis_results["analyses"][prompt]
                    
                    if "bag_of_words" not in analyses:
                        return (
                            analysis_results,
                            False,
                            False,
                            gr.update(visible=False),
                            gr.update(visible=False),
                            gr.update(visible=False),
                            gr.update(visible=False),
                            gr.update(visible=False),
                            gr.update(visible=False),
                            gr.update(visible=False),
                            gr.update(visible=False),
                            gr.update(visible=False),
                            True,
                            gr.update(visible=True, value="❌ **No Bag of Words analysis found.** Make sure to select it in the options.")
                        )
                    
                    bow_results = analyses["bag_of_words"]
                    models = bow_results.get("models", [])
                    
                    if len(models) < 2:
                        return (
                            analysis_results,
                            False,
                            False,
                            gr.update(visible=False),
                            gr.update(visible=False),
                            gr.update(visible=False),
                            gr.update(visible=False),
                            gr.update(visible=False),
                            gr.update(visible=False),
                            gr.update(visible=False),
                            gr.update(visible=False),
                            gr.update(visible=False),
                            True,
                            gr.update(visible=True, value="❌ **Not enough models to compare.** Please ensure you have two model responses.")
                        )
                    
                    # Extract and format information for display
                    model1_name = models[0]
                    model2_name = models[1]
                    
                    # Format important words for each model
                    important_words = bow_results.get("important_words", {})
                    model1_words_text = "No important words found"
                    model2_words_text = "No important words found"
                    
                    if model1_name in important_words:
                        word_list = [f"**{item['word']}** ({item['count']})" for item in important_words[model1_name][:10]]
                        model1_words_text = ", ".join(word_list)
                    
                    if model2_name in important_words:
                        word_list = [f"**{item['word']}** ({item['count']})" for item in important_words[model2_name][:10]]
                        model2_words_text = ", ".join(word_list)
                    
                    # Format similarity metrics
                    similarity_text = "No similarity metrics found"
                    comparisons = bow_results.get("comparisons", {})
                    comparison_key = f"{model1_name} vs {model2_name}"

                    if comparison_key in comparisons:
                        metrics = comparisons[comparison_key]
                        cosine = metrics.get("cosine_similarity", 0)
                        jaccard = metrics.get("jaccard_similarity", 0)
                        semantic = metrics.get("semantic_similarity", 0)  # Add semantic similarity
                        common_words = metrics.get("common_word_count", 0)
                        
                        similarity_text = f"""
                        - **Cosine Similarity**: {cosine:.2f} (higher means more similar word frequency patterns)
                        - **Jaccard Similarity**: {jaccard:.2f} (higher means more word overlap)
                        - **Semantic Similarity**: {semantic:.2f} (higher means more similar meaning)
                        - **Common Words**: {common_words} words appear in both responses
                        """
                        
                    # Return all updated component values
                    return (
                        analysis_results,  # analysis_results_state
                        False,  # analysis_output visibility
                        True,   # visualization_area_visible
                        gr.update(visible=True),  # analysis_title
                        gr.update(visible=True, value=f"## Analysis of Prompt: \"{prompt[:100]}...\""),  # prompt_title
                        gr.update(visible=True, value=f"### Comparing responses from {model1_name} and {model2_name}"),  # models_compared
                        gr.update(visible=True, value=f"#### Top Words Used by {model1_name}"),  # model1_title
                        gr.update(visible=True, value=model1_words_text),  # model1_words
                        gr.update(visible=True, value=f"#### Top Words Used by {model2_name}"),  # model2_title
                        gr.update(visible=True, value=model2_words_text),  # model2_words
                        gr.update(visible=True),  # similarity_metrics_title
                        gr.update(visible=True, value=similarity_text),  # similarity_metrics
                        False,  # status_message_visible
                        gr.update(visible=False)  # status_message
                    )
                
                except Exception as e:
                    import traceback
                    error_msg = f"Error in analysis: {str(e)}\n{traceback.format_exc()}"
                    print(error_msg)
                    
                    return (
                        {"error": error_msg},  # analysis_results_state
                        True,  # analysis_output visibility (show raw JSON for debugging)
                        False,  # visualization_area_visible
                        gr.update(visible=False),
                        gr.update(visible=False),
                        gr.update(visible=False),
                        gr.update(visible=False),
                        gr.update(visible=False),
                        gr.update(visible=False),
                        gr.update(visible=False),
                        gr.update(visible=False),
                        gr.update(visible=False),
                        True,  # status_message_visible
                        gr.update(visible=True, value=f"❌ **Error during analysis:**\n\n```\n{str(e)}\n```")  # status_message
                    )

            # Function to update visibility based on checkbox state
            def update_visibility(viz_visible, status_visible):
                return [
                    gr.update(visible=viz_visible),  # analysis_title
                    gr.update(visible=viz_visible),  # prompt_title
                    gr.update(visible=viz_visible),  # models_compared
                    gr.update(visible=viz_visible),  # model1_title
                    gr.update(visible=viz_visible),  # model1_words
                    gr.update(visible=viz_visible),  # model2_title
                    gr.update(visible=viz_visible),  # model2_words
                    gr.update(visible=viz_visible),  # similarity_metrics_title
                    gr.update(visible=viz_visible),  # similarity_metrics
                    gr.update(visible=status_visible)  # status_message
                ]

            # Connect visibility checkboxes to update function
            visualization_area_visible.change(
                fn=update_visibility,
                inputs=[visualization_area_visible, status_message_visible],
                outputs=[
                    analysis_title,
                    prompt_title,
                    models_compared,
                    model1_title,
                    model1_words,
                    model2_title,
                    model2_words,
                    similarity_metrics_title,
                    similarity_metrics,
                    status_message
                ]
            )

            # Run analysis with proper parameters
            run_analysis_btn.click(
                fn=run_analysis,
                inputs=[dataset_state, analysis_options, bow_top_slider, ngram_n, ngram_top],
                outputs=[
                    analysis_results_state,
                    analysis_output,
                    visualization_area_visible,
                    analysis_title,
                    prompt_title,
                    models_compared,
                    model1_title,
                    model1_words,
                    model2_title,
                    model2_words,
                    similarity_metrics_title,
                    similarity_metrics,
                    status_message_visible,
                    status_message
                ]
            )
    
    return app

if __name__ == "__main__":
    # Download required NLTK resources before launching the app
    download_nltk_resources()
    
    app = create_app()
    app.launch()