app.py

import pandas as pd
from datasets import Dataset
from transformers import DistilBertForSequenceClassification, DistilBertTokenizer
import torch
import gradio as gr
from huggingface_hub import HfApi, Repository
import os

# Step 1: Set your Hugging Face access token
hf_token = ""  # Replace with your actual token

# Initialize the Hugging Face API client
api = HfApi()

# Step 2: Create a sample dataset
data = {
    "Name": ["John Doe", "Jane Smith", "Mike Johnson"],
    "Email": ["johndoe@example.com", "janesmith@example.com", "mikej@example.com"],
    "Case Problem": ["Login Issues", "Payment Failure", "UI Bug"],
    "Feedback": ["Negative", "Positive", "Neutral"],
    "Details": [
        "Unable to login after password reset.",
        "Payment went through after retrying.",
        "The interface is a bit confusing at times."
    ]
}

# Create a DataFrame from the sample data
df = pd.DataFrame(data)

# Convert the DataFrame into a Hugging Face Dataset
dataset = Dataset.from_pandas(df)

# Step 3: Upload the dataset to Hugging Face
repo_id = "SailajaS/case-feedback-dataset"  # Replace with your actual Hugging Face repo name
try:
    # Create a repository on Hugging Face to store the dataset
    api.create_repo(repo_id=repo_id, repo_type="dataset", token=hf_token)
    print(f"Successfully created repository: {repo_id}")
except Exception as e:
    print(f"Error creating repository: {e}")

# Initialize the repository and push the dataset to Hugging Face
repo = Repository(local_dir="./dataset_repo", clone_from=repo_id)
dataset.to_csv("./dataset_repo/dataset.csv")
repo.push_to_hub()

# Step 4: Load pre-trained DistilBERT model and tokenizer
model_name = "distilbert-base-uncased"
model = DistilBertForSequenceClassification.from_pretrained(model_name)
tokenizer = DistilBertTokenizer.from_pretrained(model_name)

# Step 5: Define a function to predict feedback based on case details
def predict_case_feedback(details):
    inputs = tokenizer(details, return_tensors="pt", padding=True, truncation=True, max_length=512)
    with torch.no_grad():
        outputs = model(**inputs)
    logits = outputs.logits
    predicted_class = logits.argmax(dim=-1).item()
    
    feedback_labels = ["Negative", "Positive", "Neutral"]
    feedback = feedback_labels[predicted_class]
    
    return f"Predicted Feedback: {feedback}"

# Step 6: Create Gradio interface for the prediction function
interface = gr.Interface(
    fn=predict_case_feedback,
    inputs="text",              # Input: Text
    outputs="text",             # Output: Text
    title="Case Feedback Prediction",
    description="Enter the case details to predict feedback (Positive, Negative, Neutral)"
)

# Step 7: Launch the Gradio app
interface.launch()