app.py

#!/usr/bin/env python3
import streamlit as st
from gradio_client import Client
from groq import Groq
from PIL import Image
import moviepy.editor as mp
from natsort import natsorted
import os
from dotenv import load_dotenv
import json

# Load environment variables
load_dotenv()

# Constants
HF_TOKEN = os.getenv("HF_TOKEN")
GROQ_API_KEY = os.getenv("GROQ_API_KEY")
IMAGE_GENERATION_SPACE_NAME = "stabilityai/stable-diffusion-3.5-large-turbo"

# Initialize Groq client
groq_client = Groq(api_key=GROQ_API_KEY)

# LLM Models (free options)
LLM_MODELS = {
    "Mixtral 8x7B (Groq)": "mixtral-8x7b-32768",
    "Mistral 7B (HF)": "mistralai/Mixtral-7B-Instruct-v0.1",
    "LLaMA 13B (HF)": "meta-llama/Llama-13b-hf"  # Note: May require approval; replace if needed
}

# Utility Functions
def generate_tutor_output(subject, difficulty, student_input, model):
    prompt = f"""
    You are an expert tutor in {subject} at the {difficulty} level. 
    The student has provided the following input: "{student_input}"
    
    Please generate:
    1. A brief, engaging lesson on the topic (2-3 paragraphs)
    2. A thought-provoking question to check understanding
    3. Constructive feedback on the student's input
    
    Format your response as a JSON object with keys: "lesson", "question", "feedback"
    """
    
    if model.startswith("mixtral"):  # Groq model
        completion = groq_client.chat.completions.create(
            messages=[{
                "role": "system",
                "content": f"You are the world's best AI tutor for {subject}, renowned for clear, engaging explanations."
            }, {
                "role": "user",
                "content": prompt
            }],
            model=model,
            max_tokens=1000
        )
        return json.loads(completion.choices[0].message.content)
    else:  # Hugging Face models
        try:
            client = Client("https://api-inference.huggingface.co/models/" + model, hf_token=HF_TOKEN)
            response = client.predict(prompt, api_name="/generate")
            return json.loads(response)
        except:
            st.warning(f"HF model {model} failed, falling back to Mixtral.")
            return generate_tutor_output(subject, difficulty, student_input, "mixtral-8x7b-32768")

def generate_image(prompt, path='temp_image.png'):
    try:
        client = Client(IMAGE_GENERATION_SPACE_NAME, hf_token=HF_TOKEN)
        result = client.predict(
            prompt=prompt,
            width=512,  # Reduced for speed
            height=512,
            api_name="/predict"  # Correct endpoint
        )
        image = Image.open(result)
        image.save(path)
        return path
    except Exception as e:
        st.error(f"Error generating image: {e}")
        return None

def generate_video(images, audio_text, language, speaker, path='temp_video.mp4'):
    try:
        audio_client = Client("habib926653/Multilingual-TTS")
        audio_result = audio_client.predict(
            text=audio_text,
            language_code=language,
            speaker=speaker,
            api_name="/text_to_speech_edge"
        )
        audio_file = audio_result[1]
        with open(audio_file, 'rb') as f:
            audio_bytes = f.read()
        audio_path = "temp_audio.mp3"
        with open(audio_path, 'wb') as f:
            f.write(audio_bytes)
        
        audio_clip = mp.AudioFileClip(audio_path)
        duration_per_image = audio_clip.duration / len(images)
        image_clips = [mp.ImageClip(img).set_duration(duration_per_image) for img in images if img]
        video = mp.concatenate_videoclips(image_clips, method="compose").set_audio(audio_clip)
        video.write_videofile(path, fps=24, codec='libx264')
        return path
    except Exception as e:
        st.error(f"Error generating video: {e}")
        return None

# Streamlit App
def main():
    st.markdown("<h1 style='text-align: center;'>EduAI: Your Interactive Tutor</h1>", unsafe_allow_html=True)
    st.markdown("<p style='text-align: center;'>Learn, Ask, Visualize! ❤️</p>", unsafe_allow_html=True)

    subject = st.selectbox("Choose Subject:", ["Math", "Science", "History", "Literature", "Code", "AI"])
    difficulty = st.selectbox("Difficulty Level:", ["Beginner", "Intermediate", "Advanced"])
    model = st.selectbox("Choose LLM Model:", list(LLM_MODELS.keys()))
    student_input = st.text_area("Your Question/Input (max 1500 chars):", max_chars=1500)

    if 'tutor_response' not in st.session_state:
        st.session_state.tutor_response = None

    if st.button("Generate Answer & Question"):
        if student_input:
            with st.spinner("Generating your lesson..."):
                response = generate_tutor_output(subject, difficulty, student_input, LLM_MODELS[model])
                st.session_state.tutor_response = response
        else:
            st.warning("Please provide an input!")

    if st.session_state.tutor_response:
        st.markdown("### Lesson")
        st.write(st.session_state.tutor_response["lesson"])
        st.markdown("### Comprehension Question")
        st.write(st.session_state.tutor_response["question"])
        st.markdown("### Feedback")
        st.write(st.session_state.tutor_response["feedback"])

        col1, col2 = st.columns(2)
        with col1:
            if st.button("Generate Image"):
                with st.spinner("Creating image..."):
                    image_path = generate_image(st.session_state.tutor_response["lesson"])
                    if image_path:
                        st.image(image_path, caption="Visual of your lesson")
        with col2:
            if st.button("Generate Video"):
                with st.spinner("Creating video..."):
                    audio_client = Client("habib926653/Multilingual-TTS")
                    speakers_response = audio_client.predict(language="English", api_name="/get_speakers")
                    speaker = speakers_response["choices"][0][0]
                    images = [generate_image(st.session_state.tutor_response["lesson"])]
                    video_path = generate_video(images, st.session_state.tutor_response["lesson"], "English", speaker)
                    if video_path:
                        st.video(video_path)

    st.markdown("---")
    st.markdown("<p style='text-align: center;'>Built for learning, powered by AI!</p>", unsafe_allow_html=True)

if __name__ == "__main__":
    main()