app.py

import gradio as gr
import numpy as np
import pandas as pd
import plotly.express as px
from sklearn.decomposition import PCA
from sentence_transformers import SentenceTransformer

# Load pre-trained sentence transformer model
model = SentenceTransformer('all-MiniLM-L6-v2')

# Function to compute document embeddings and apply PCA
# Modify the Gradio interface to accept a list of identifiers and texts
def compute_pca(data):
    # data is expected to be a list of dictionaries with 'Identifier' and 'Text' keys
    df = pd.DataFrame(data)

    # Remove rows where 'Identifier' or 'Text' is empty or contains only whitespace
    valid_entries = df[
        (df['Identifier'].str.strip() != '') & 
        (df['Text'].str.strip() != '')
    ]
    
    if valid_entries.empty:
        return gr.Plot.update(value=None, label="No data to process. Please fill in the boxes.")
    
    # Generate embeddings
    embeddings = model.encode(valid_entries['Text'].tolist())
    
    # Perform PCA to reduce to 2 dimensions
    pca = PCA(n_components=2)
    pca_result = pca.fit_transform(embeddings)
    
    # Add PCA results to the DataFrame
    valid_entries = valid_entries.reset_index(drop=True)
    valid_entries['PC1'] = pca_result[:, 0]
    valid_entries['PC2'] = pca_result[:, 1]
    
    # Plot the PCA result with identifiers as labels
    fig = px.scatter(valid_entries, x='PC1', y='PC2', text='Identifier', title='PCA of Text Embeddings')
    return fig

def text_editor_app():
    with gr.Blocks() as demo:
        # Use a DataFrame component for inputs with fixed columns
        data_input = gr.Dataframe(
            headers=["Identifier", "Text"],
            datatype=["str", "str"],
            row_count=(4, "dynamic"),
            col_count=(2, "fixed"),
            label="Input Data (Enter at least two texts)",
            wrap=True,
            placeholder="Please enter at least two identifier-text pairs."
        )

        # Button to run the analysis
        analyze_button = gr.Button("Run Analysis")

        # Output plot
        output_plot = gr.Plot(label="PCA Visualization")

        # Run analysis when the button is clicked
        analyze_button.click(
            fn=compute_pca, inputs=data_input, outputs=output_plot
        )

    return demo


# Launch the app
text_editor_app().launch()