app.py

import os
import gradio as gr
import fitz  # PyMuPDF
from datetime import date
from langchain.embeddings import HuggingFaceEmbeddings
from langchain.vectorstores import FAISS
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.chains import RetrievalQA
from langchain.llms import HuggingFaceHub

# ✅ 1. Extract text from PDF
def extract_text_from_pdf(file_path):
    doc = fitz.open(file_path)
    text = ""
    for page in doc:
        text += page.get_text()
    return text

# ✅ 2. Chunk text
def chunk_text(text):
    splitter = RecursiveCharacterTextSplitter(
        chunk_size=500,
        chunk_overlap=100
    )
    return splitter.split_text(text)

# ✅ 3. Create embeddings and retriever
def create_retriever(chunks):
    print("⚙️ Generating embeddings...")
    embeddings = HuggingFaceEmbeddings(model_name="sentence-transformers/all-MiniLM-L6-v2")
    vectorstore = FAISS.from_texts(chunks, embedding=embeddings)
    retriever = vectorstore.as_retriever()
    return retriever

# ✅ 4. Set up LLM chain (e.g., HuggingFaceHub, GROQ, etc.)
def create_chain(retriever):
    llm = HuggingFaceHub(repo_id="google/flan-t5-base", model_kwargs={"temperature": 0.1, "max_length": 256})
    chain = RetrievalQA.from_chain_type(llm=llm, retriever=retriever)
    return chain

# 🔄 5. Upload file → extract → chunk → embed
def upload_file(file):
    try:
        text = extract_text_from_pdf(file.name)
        print("📄 Extracted text:", text[:300])
        chunks = chunk_text(text)
        print("✅ Chunks created:", len(chunks))
        return chunks
    except Exception as e:
        print("❌ Upload error:", e)
        return f"Error while uploading file:\n{e}"

# 💬 6. Query & Generate Answer
def generate_response(name, today, query, chunks, model_choice):
    try:
        print("🧑‍🏫 Teacher:", name, "| 📅 Date:", today)
        print("🔍 Query:", query)
        print("📦 Chunks received:", len(chunks))
        print("🤖 Model selected:", model_choice)

        retriever = create_retriever(chunks)
        rag_chain = create_chain(retriever)
        answer = rag_chain.run(query)

        print("✅ Answer:", answer)
        formatted = f"👩‍🏫 Teacher: {name}\n📅 Date: {today}\n\n📌 Question: {query}\n\n📘 Answer:\n{answer}"
        return formatted
    except Exception as e:
        print("❌ Error in response generation:", e)
        return f"❌ Error:\n{e}"

# 🧱 7. Gradio UI
with gr.Blocks(title="📚 Curriculum Assistant") as demo:
    gr.Markdown("# 🧠 Curriculum Assistant\nUpload your curriculum PDF and ask questions from it!")

    with gr.Row():
        teacher_name = gr.Textbox(label="👩‍🏫 Teacher Name")
        today_date = gr.Textbox(value=str(date.today()), label="📅 Date")

    with gr.Row():
        file_input = gr.File(label="📎 Upload Curriculum PDF")
        upload_btn = gr.Button("📥 Extract")

    chunks_state = gr.State([])

    with gr.Row():
        query_input = gr.Textbox(label="💡 Enter your question")
        model_selector = gr.Dropdown(choices=["flan-t5-base"], value="flan-t5-base", label="🤖 Choose Model")
        submit_button = gr.Button("🚀 Generate Resource")

    answer_output = gr.Textbox(label="📘 Output", lines=12)

    # 🔁 Button connections
    upload_btn.click(fn=upload_file, inputs=[file_input], outputs=[chunks_state])
    submit_button.click(
        fn=generate_response,
        inputs=[teacher_name, today_date, query_input, chunks_state, model_selector],
        outputs=[answer_output]
    )

# 🚀 Launch the app
demo.launch()