app.py

import gradio as gr
import requests
import os
from dotenv import load_dotenv
import pandas as pd
import numpy as np
from sklearn.ensemble import RandomForestClassifier
from sentence_transformers import SentenceTransformer, util

# Load Hugging Face token
load_dotenv()
HF_TOKEN = os.getenv("HF_TOKEN")
headers = {"Authorization": f"Bearer {HF_TOKEN}"}

# ----------------- FEATURE FUNCTIONS -----------------

def generate_quiz(text, num_questions):
    prompt = (
        f"Generate {num_questions} simple quiz questions from the following text:\n\n"
        f"{text}\n\n"
        f"Each question should be followed by options A, B, C, D and a correct answer."
    )

    API_URL = "https://api-inference.huggingface.co/models/google/flan-t5-small"
    payload = {
        "inputs": prompt,
        "parameters": {"max_new_tokens": 300}
    }

    response = requests.post(API_URL, headers=headers, json=payload)
    
    try:
        result = response.json()
        if isinstance(result, list) and "generated_text" in result[0]:
            output = result[0]["generated_text"].strip()
            return output if output else "⚠️ Model returned empty text. Try reducing question count or using a smaller paragraph."
        else:
            return f"⚠️ Model didn't return expected output. Try fewer questions or simpler content."
    except Exception as e:
        return f"⚠️ API Error: {e}"




def get_bot_response(query):
    API_URL = "https://api-inference.huggingface.co/models/HuggingFaceH4/zephyr-7b-beta"
    payload = {"inputs": f"Student: {query}\nAI:", "parameters": {"max_new_tokens": 150}}
    response = requests.post(API_URL, headers=headers, json=payload)
    try:
        return response.json()[0]["generated_text"].split("AI:")[-1].strip()
    except:
        return "⚠️ AI Assistant unavailable right now."

def summarize_text(text):
    API_URL = "https://api-inference.huggingface.co/models/facebook/bart-large-cnn"
    payload = {"inputs": text, "parameters": {"max_length": 120}}
    response = requests.post(API_URL, headers=headers, json=payload)
    try:
        return response.json()[0]["summary_text"]
    except:
        return "⚠️ Unable to summarize content."

def translate_text(text, target_lang="te"):
    API_URL = "https://libretranslate.de/translate"
    payload = {
        "q": text,
        "source": "en",
        "target": target_lang,
        "format": "text"
    }
    response = requests.post(API_URL, data=payload)
    try:
        return response.json()["translatedText"]
    except:
        return "⚠️ Translation failed."

def check_plagiarism(text1, text2):
    model = SentenceTransformer('all-MiniLM-L6-v2')
    embeddings = model.encode([text1, text2], convert_to_tensor=True)
    similarity = util.pytorch_cos_sim(embeddings[0], embeddings[1]).item()
    return f"Similarity Score: {similarity:.2f}\n{('⚠️ Possible Plagiarism' if similarity > 0.75 else '✅ No significant overlap')}"

def analyze_weakness(file):
    try:
        df = pd.read_csv(file.name)
        topic_scores = df.groupby('Topic')['Score'].mean().sort_values()
        weakest = topic_scores.head(3)
        return f"Weak Areas:\n{weakest.to_string()}"
    except:
        return "⚠️ Failed to analyze file. Ensure it contains 'Topic' and 'Score' columns."

def predict_engagement(attendance, login_freq, video_watch):
    X = np.array([[attendance, login_freq, video_watch]])
    y = [0, 1, 1, 0, 1]
    X_train = np.array([[90, 5, 80], [85, 4, 90], [95, 6, 85], [60, 2, 40], [88, 3, 75]])
    clf = RandomForestClassifier().fit(X_train, y)
    prediction = clf.predict(X)[0]
    return "✅ Likely to be Engaged" if prediction else "⚠️ At Risk of Disengagement"

def generate_badge(score, speed):
    if score >= 90 and speed <= 30:
        return "🏅 Gold Badge"
    elif score >= 75:
        return "🥈 Silver Badge"
    else:
        return "🥉 Bronze Badge"

# ----------------- UI -----------------
with gr.Blocks(title="Smart LMS AI Suite") as app:

    with gr.Tab("🧠 Quiz Generator"):
        quiz_text = gr.Textbox(label="Paste Topic Content")
        quiz_slider = gr.Slider(1, 10, label="Number of Questions", value=3)
        quiz_output = gr.Textbox(label="Generated Quiz")
        quiz_button = gr.Button("Generate Quiz")
        quiz_button.click(fn=generate_quiz, inputs=[quiz_text, quiz_slider], outputs=quiz_output)

    with gr.Tab("🤖 AI Teaching Assistant"):
        bot_input = gr.Textbox(label="Ask a Question")
        bot_output = gr.Textbox(label="AI Answer")
        bot_button = gr.Button("Get Answer")
        bot_button.click(fn=get_bot_response, inputs=bot_input, outputs=bot_output)

    with gr.Tab("📄 Summarizer"):
        sum_input = gr.Textbox(label="Paste Content")
        sum_output = gr.Textbox(label="Summary")
        sum_button = gr.Button("Summarize")
        sum_button.click(fn=summarize_text, inputs=sum_input, outputs=sum_output)

    with gr.Tab("🌍 Translator"):
        trans_input = gr.Textbox(label="Text in English")
        lang_dropdown = gr.Dropdown(["te", "hi", "ta", "fr"], value="te", label="Target Language Code")
        trans_output = gr.Textbox(label="Translated Text")
        trans_button = gr.Button("Translate")
        trans_button.click(fn=translate_text, inputs=[trans_input, lang_dropdown], outputs=trans_output)

    with gr.Tab("🧾 Plagiarism Checker"):
        plag_1 = gr.Textbox(label="Document 1")
        plag_2 = gr.Textbox(label="Document 2")
        plag_out = gr.Textbox(label="Result")
        plag_btn = gr.Button("Check Plagiarism")
        plag_btn.click(fn=check_plagiarism, inputs=[plag_1, plag_2], outputs=plag_out)

    with gr.Tab("📉 Weakness Analyzer"):
        csv_input = gr.File(label="Upload CSV with 'Topic' and 'Score' Columns")
        weak_out = gr.Textbox(label="Weak Topics")
        weak_btn = gr.Button("Analyze")
        weak_btn.click(fn=analyze_weakness, inputs=csv_input, outputs=weak_out)

    with gr.Tab("📊 Engagement Predictor"):
        att = gr.Slider(0, 100, value=85, label="Attendance %")
        login = gr.Slider(0, 10, value=5, label="Login Frequency")
        video = gr.Slider(0, 100, value=80, label="Video Watch %")
        engage_out = gr.Textbox(label="Prediction")
        engage_btn = gr.Button("Predict")
        engage_btn.click(fn=predict_engagement, inputs=[att, login, video], outputs=engage_out)

    with gr.Tab("🏅 Badge Generator"):
        score = gr.Slider(0, 100, label="Score")
        speed = gr.Slider(0, 60, label="Time Taken (mins)")
        badge_out = gr.Textbox(label="Badge Awarded")
        badge_btn = gr.Button("Generate Badge")
        badge_btn.click(fn=generate_badge, inputs=[score, speed], outputs=badge_out)

    gr.Markdown("🚀 Built using Hugging Face, Gradio, and Free APIs")

app.launch()