app.py

import fitz  # PyMuPDF
import numpy as np
from sentence_transformers import SentenceTransformer
import faiss
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig
import gradio as gr
import os

# Hugging Face token from environment variable
hf_token = os.getenv("HF_TOKEN", "your-token-here")

# Quantization config for 4-bit loading
quant_config = BitsAndBytesConfig(
    load_in_4bit=True,
    bnb_4bit_quant_type="nf4",
    bnb_4bit_compute_dtype=torch.float16,
    bnb_4bit_use_double_quant=True
)

# Load models with authentication and quantization
embedder = SentenceTransformer('all-MiniLM-L6-v2')
model_name = "mistralai/Mistral-7B-Instruct-v0.3"
tokenizer = AutoTokenizer.from_pretrained(model_name, token=hf_token)
model = AutoModelForCausalLM.from_pretrained(
    model_name,
    token=hf_token,
    quantization_config=quant_config,
    device_map="auto",
    low_cpu_mem_usage=True
)

# Text extraction function for PDFs
def extract_text_from_pdf(pdf_path):
    doc = fitz.open(pdf_path)
    text = ""
    for page in doc:
        text += page.get_text()
    return text

# RAG implementation
def create_vector_store(text):
    sentences = text.split(". ")
    embeddings = embedder.encode(sentences, convert_to_tensor=False)
    index = faiss.IndexFlatL2(embeddings.shape[1])
    index.add(embeddings)
    return index, sentences, embeddings

def retrieve_context(query, index, sentences, embeddings, k=3):
    query_embedding = embedder.encode([query], convert_to_tensor=False)
    distances, indices = index.search(query_embedding, k)
    return [sentences[i] for i in indices[0]]

def generate_explanation(query, context):
    prompt = f"As a teacher, explain this concept: {query}\nContext: {' '.join(context)}"
    inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
    outputs = model.generate(**inputs, max_new_tokens=50)
    return tokenizer.decode(outputs[0], skip_special_tokens=True)

# Process input and explain (PDF only)
def process_input(file, query):
    if not file.name.endswith('.pdf'):
        return "Please upload a PDF file."
    text = extract_text_from_pdf(file.name)
    index, sentences, embeddings = create_vector_store(text)
    context = retrieve_context(query, index, sentences, embeddings)
    explanation = generate_explanation(query, context)
    return explanation

# Gradio interface
interface = gr.Interface(
    fn=process_input,
    inputs=[
        gr.File(label="Upload a PDF file", file_types=[".pdf"]),
        gr.Textbox(label="Ask a question about the content")
    ],
    outputs=gr.Textbox(label="Explanation"),
    title="AI Tutor",
    description="Upload a PDF file and ask questions about its content!"
)

print("App initialized successfully!")
interface.launch()