app.py

import os
import gradio as gr
import requests

API_TOKEN = os.getenv("HF_API_TOKEN")
MODEL = "meta-llama/Llama-2-7b-chat-hf"

def query_model(prompt):
    headers = {"Authorization": f"Bearer {API_TOKEN}"}
    payload = {"inputs": prompt, "parameters": {"max_new_tokens": 100}}
    response = requests.post(
        f"https://api-inference.huggingface.co/models/{MODEL}",
        headers=headers,
        json=payload
    )
    if response.status_code == 200:
        output = response.json()
        try:
            return output[0]['generated_text']
        except:
            return "Error: Unexpected output format"
    else:
        return f"Error: {response.status_code} - {response.text}"

def evaluate_response(prompt, response):
    prompt_keywords = set(prompt.lower().split())
    response_words = set(response.lower().split())
    common_words = prompt_keywords.intersection(response_words)
    return "✔✔ Understood" if len(common_words) >= 3 else "❌ Not Understood"

def process_prompt(prompt):
    response = query_model(prompt)
    evaluation = evaluate_response(prompt, response)
    return response, evaluation

gr.Interface(
    fn=process_prompt,
    inputs=gr.Textbox(lines=3, placeholder="Enter your prompt here...", label="Prompt"),
    outputs=[gr.Textbox(label="Model Response"), gr.Textbox(label="Evaluation Result")],
    title="Prompt Understanding Tester",
    description="Test if the Hugging Face model understands your prompt."
).launch()