app.py

import streamlit as st
from transformers import pipeline
from fpdf import FPDF  # For PDF generation

# Initialize the pipeline for text generation using Llama model
llama_model = "meta-llama/Llama-2-7b-hf"  # Update with the specific Llama model you want to use

# Function to get code from Hugging Face Llama model
def generate_code(summary, language):
    try:
        generator = pipeline('text-generation', model=llama_model, tokenizer=llama_model)
        generated_code = generator(f"Generate {language} code: {summary}", max_length=150)[0]['generated_text']
        return generated_code
    except Exception as e:
        st.error(f"Error generating code: {e}")
        return ""

# Function to explain the generated code using Llama (or a suitable summarization model)
def explain_code(code):
    try:
        explainer = pipeline('summarization', model=llama_model, tokenizer=llama_model)  # You can use a summarization model or a specific Llama-based explanation model
        explanation = explainer(f"Explain the following code:\n\n{code}", max_length=250, min_length=50)[0]['summary_text']
        return explanation
    except Exception as e:
        st.error(f"Error explaining code: {e}")
        return ""

# Function to save code as a PDF
def save_code_as_pdf(code, file_name="generated_code.pdf"):
    pdf = FPDF()
    pdf.add_page()
    pdf.set_font("Arial", size=12)
    pdf.multi_cell(0, 10, code)
    pdf.output(file_name)
    return file_name

# --- Streamlit Interface ---
st.set_page_config(page_title="Generative AI Code Generator", page_icon="🧑‍💻", layout="wide")

# Page Title
st.title("🚀 Generative AI Code Generator Using Hugging Face Llama")

# Input Fields
summary = st.text_area("📝 Enter the Task Summary", "For example: Create a function to add two numbers.")
language = st.selectbox("🌐 Select Programming Language", ["Python", "Java", "JavaScript", "C++"])

# Generate Code Button
if st.button("Generate Code"):
    if summary:
        generated_code = generate_code(summary, language)

        if generated_code:
            st.subheader(f"✨ Generated {language} Code:")
            st.code(generated_code, language=language.lower())

            # Code Modification Section
            modified_code = st.text_area("✏️ Modify the Code (Optional):", value=generated_code, height=200)

            # Explanation Button
            if st.button("Explain Code"):
                explanation = explain_code(generated_code)
                st.subheader("📖 Code Explanation:")
                st.write(explanation)

            # Download Code as PDF
            if st.button("Download Code as PDF"):
                pdf_path = save_code_as_pdf(modified_code)  # Use modified code if edited
                with open(pdf_path, "rb") as pdf_file:
                    st.download_button(
                        label="📥 Download PDF",
                        data=pdf_file,
                        file_name="generated_code.pdf",
                        mime="application/pdf",
                    )

# New Code Button
if st.button("Generate New Code"):
    st.rerun()  # Refresh the page to clear inputs

# Footer Information
st.markdown("---")
st.write("🌟 Powered by **Streamlit**, **Hugging Face**, and **Transformers** | Deployed on Hugging Face")