app.py

import os
import random
import csv
import requests
from sentence_transformers import SentenceTransformer
import gradio as gr

with open("secrets.csv", 'w') as file:
    file.write(os.getenv('secrets'))

# Load the CSV file
csv_file = "secrets.csv"
questions = []
answers = []

with open(csv_file, "r") as file:
    reader = csv.DictReader(file)
    for row in reader:
        questions.append(row["questions"])
        answers.append(row["answers"])

# Initialize the SentenceTransformer model
model = SentenceTransformer("BAAI/bge-base-en-v1.5")

def predict_prompt(name, email):
    # Select 10 random answers from the CSV file
    selected_answers = random.sample(answers, 10)
    
    total_score = 0
    results = []
    
    # Iterate over the selected answers
    for i, answer in enumerate(selected_answers, 1):
        actual_question = questions[answers.index(answer)]
        
        # Prompt the user to enter their predicted question
        predicted_question = gr.components.Textbox(label=f"Answer {i}: {answer}")
        
        # Encode the predicted and actual questions
        predicted = model.encode(predicted_question, normalize_embeddings=True)
        actual = model.encode(actual_question, normalize_embeddings=True)
        
        # Calculate the similarity score
        similarity = predicted @ actual.T
        score = float(similarity)
        
        results.append(f"Answer {i}: {answer}\nPredicted Question: {predicted_question}\nScore: {score}\n")
        total_score += score
    
    results.append(f"\nTotal Score: {total_score}")
    
    # Make an API call to submit the user's name, email, and score
    url = "https://api.example.com/submit"  # Replace with the actual API endpoint URL
    data = {
        "name": name,
        "email": email,
        "total_score": total_score
    }
    
    response = requests.post(url, json=data)
    
    if response.status_code == 200:
        results.append("Score submitted successfully!")
    else:
        results.append("Failed to submit the score.")
    
    return "\n".join(results)

# Create the Gradio interface
iface = gr.Interface(
    fn=predict_prompt,
    inputs=[
        gr.components.Textbox(label="Name"),
        gr.components.Textbox(label="Email")
    ],
    outputs="text",
    title="Predict the Prompt",
    description="Enter your name and email to start the competition. You will be presented with 10 random answers, and your task is to predict the question for each answer. Your score will be calculated based on the similarity between your predicted question and the actual question.",
)

# Launch the Gradio app
iface.launch()