app.py

import gradio as gr
import fitz  # PyMuPDF
import faiss
import numpy as np
from sentence_transformers import SentenceTransformer
import os
import requests
import json
from typing import List

# Load the sentence transformer model
embedder = SentenceTransformer("all-MiniLM-L6-v2")

# Initialize FAISS index
dimension = 384  # vector size for MiniLM-L6-v2
index = faiss.IndexFlatL2(dimension)
stored_chunks = []
stored_embeddings = []

# Set your Groq API key here (safe in Colab if you use secrets or input())
GROQ_API_KEY = "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"
LLM_MODEL = "llama3-8b-8192"

def extract_text_from_pdf(pdf_file):
    doc = fitz.open(stream=pdf_file.read(), filetype="pdf")
    text = ""
    for page in doc:
        text += page.get_text()
    return text

def chunk_text(text: str, chunk_size: int = 500, overlap: int = 50) -> List[str]:
    words = text.split()
    chunks = []
    for i in range(0, len(words), chunk_size - overlap):
        chunk = words[i:i + chunk_size]
        chunks.append(" ".join(chunk))
    return chunks

def embed_and_store(chunks):
    global stored_chunks, stored_embeddings
    embeddings = embedder.encode(chunks)
    index.add(np.array(embeddings, dtype=np.float32))
    stored_chunks.extend(chunks)
    stored_embeddings.extend(embeddings)

def query_groq(prompt):
    url = "https://api.groq.com/openai/v1/chat/completions"
    headers = {
        "Authorization": f"Bearer {GROQ_API_KEY}",
        "Content-Type": "application/json"
    }

    payload = {
        "model": LLM_MODEL,
        "messages": [
            {"role": "system", "content": "You are a helpful academic supervisor helping students study uploaded research papers."},
            {"role": "user", "content": prompt}
        ],
        "temperature": 0.2
    }

    response = requests.post(url, headers=headers, json=payload)
    return response.json()["choices"][0]["message"]["content"]

def retrieve_answer(user_query):
    embedded_query = embedder.encode([user_query])
    D, I = index.search(np.array(embedded_query, dtype=np.float32), k=3)
    context = "\n\n".join([stored_chunks[i] for i in I[0]])
    prompt = f"Based on the following context:\n\n{context}\n\nAnswer this question:\n{user_query}"
    return query_groq(prompt)

def handle_upload(file):
    text = extract_text_from_pdf(file)
    chunks = chunk_text(text)
    embed_and_store(chunks)
    return "PDF processed and indexed. You can now ask questions."

def handle_question(question):
    if not stored_chunks:
        return "Please upload a PDF first."
    return retrieve_answer(question)

with gr.Blocks() as demo:
    with gr.Row():
        file_input = gr.File(label="Upload your PDF")
        upload_btn = gr.Button("Process PDF")
    output_text = gr.Textbox(label="Status / Answer")

    upload_btn.click(fn=handle_upload, inputs=file_input, outputs=output_text)

    with gr.Row():
        query_input = gr.Textbox(label="Ask a Question")
        query_btn = gr.Button("Submit")
    query_btn.click(fn=handle_question, inputs=query_input, outputs=output_text)

demo.launch()