app.py

import gradio as gr
import pandas as pd
import tensorflow as tf
import nltk
import spacy
import re
from nltk.corpus import stopwords
from nltk.tokenize import word_tokenize
from tensorflow.keras.preprocessing.sequence import pad_sequences
import requests
import pickle

# Download necessary resources
import spacy.cli
spacy.cli.download("en_core_web_sm")
nltk.download('punkt_tab')
nltk.download('stopwords')
stop_words = set(stopwords.words('english'))
nlp = spacy.load('en_core_web_sm')

# Download the model file from Hugging Face
model_url = "https://huggingface.co/Zmorell/HIPA_2/resolve/main/saved_keras_model.keras"
local_model_path = "saved_keras_model.keras"

response = requests.get(model_url)
with open(local_model_path, 'wb') as f:
    f.write(response.content)

print(f"Model downloaded to {local_model_path}")

# Load the downloaded model
model = tf.keras.models.load_model(local_model_path)
print(f"Model loaded from {local_model_path}")

# Load the tokenizer
tokenizer_file_path = "tokenizer.pickle"
with open(tokenizer_file_path, 'rb') as handle:
    tokenizer = pickle.load(handle)

print("Tokenizer loaded from tokenizer.pickle")

def preprocess_text(text):
    text = re.sub(r'[^a-zA-Z0-9\s]', '', text)
    tokens = word_tokenize(text.lower())
    tokens = [word for word in tokens if word not in stop_words]
    doc = nlp(' '.join(tokens))
    lemmas = [token.lemma_ for token in doc]
    return ' '.join(lemmas)

def predict(text):
    try:
        print(f"Input text: {text}")
        inputs = preprocess_text(text)
        print(f"Preprocessed text: {inputs}")
        
        inputs = tokenizer.texts_to_sequences([inputs])
        print(f"Tokenized text: {inputs}")
        
        inputs = pad_sequences(inputs, maxlen=750, padding='post')
        print(f"Padded text: {inputs}")

        outputs = model.predict(inputs)
        print(f"Model outputs: {outputs}")
        
        # Interpret the output as a prediction
        prediction = outputs[0][0]
        if prediction >= 0.5:
            result = f"True = {prediction:.2f}"
        else:
            result = f"False = {prediction:.2f}"
        
        return result
    except Exception as e:
        print(f"Error during prediction: {e}")
        return f"Error during prediction: {e}"
ui_css = """
#body {
    height: 700px;
    width: 500px;
    background-color: rgb(108, 207, 239);
    border-radius: 15px;
}
#hipaa-image {
    width: 75px;
}
#input-box {
    width: 480px;
    border: 2px solid black;
    margin-left: 8px;
    margin-right: 8px;
    overflow-y: scroll;
    height: 150px;
    max-height: 150px;
}
#output-elems {
    width: 480px;
    border: 2px solid black;
    margin-left: 8px;
    margin-right: 8px;
    padding: 1em;
}
#submit-button, #clear-button {
    color: white;
    height: 45px;
    width: 60px;
    margin: 10px;
    border-radius: 5px;
    border: 5px solid black;
}
#submit-button {
    background-color: red;
}
#clear-button {
    background-color: grey;
}
#addinfo {
font-size: 16;
justify-self: center;
}
"""
# Set up the Gradio interface
with gr.Blocks(css=ui_css) as demo:
    with gr.Column(elem_id="body"):
        with gr.Row(elem_id="header"):
            with gr.Row(elem_id="hipaa-image"):
                gr.Image(value="hipaa-e1638383751916.png")
            with gr.Row():
                gr.Markdown("Enter text below to determine if it is a HIPAA violation. Smaller inputs may be less accurate.", elem_id="addinfo")
            
        with gr.Row(elem_id="interactives"):
            inputs=gr.Textbox(label="Enter Input Text Here", elem_id="input-box", lines=5)
            
        with gr.Row(elem_id="output-elems"):
            gr.Markdown("This text is a violation: ")
            outputs=gr.Textbox(label="", elem_id="output-box", interactive=False)
                
        with gr.Row():
            submit_button = gr.Button("Submit", elem_id="submit-button")
            clear_button = gr.Button("Clear", elem_id="clear-button")
        
    submit_button.click(predict, inputs=inputs, outputs=outputs)
    clear_button.click(lambda: ("", ""), inputs=None, outputs=[inputs, outputs])
demo.launch()