app.py

import fitz
import faiss
import torch
import requests
import json
import time
import gradio as gr
import datetime
import os
from sentence_transformers import SentenceTransformer

embed_model = SentenceTransformer("intfloat/multilingual-e5-large-instruct")

chunks = []
index = None
qa_history = []
uploaded_filename = ""

def split_into_chunks(text, chunk_size=512, overlap=64):
    words = text.split()
    return [" ".join(words[i:i+chunk_size]) for i in range(0, len(words), chunk_size - overlap)]

def get_embeddings(texts):
    prompts = [f"query: {t}" for t in texts]
    return embed_model.encode(prompts, normalize_embeddings=True)

def ask_question_stream(query, history):
    if index is None:
        yield "❌ Please upload and process a PDF first."
        return

    query_vec = get_embeddings([query])[0].reshape(1, -1)
    _, I = index.search(query_vec, 4)
    context = "\n".join([chunks[i] for i in I[0]])

    prompt = f"""Answer the question using only the below context.

Context:
{context}

Question: {query}
Answer:"""

    headers = {
        "Authorization": f"Bearer {os.getenv('OPENROUTER_API_KEY')}",
        "Content-Type": "application/json"
    }

    payload = {
        "model": "deepseek/deepseek-chat-v3-0324:free",
        "messages": [{"role": "user", "content": prompt}]
    }

    try:
        res = requests.post("https://openrouter.ai/api/v1/chat/completions", headers=headers, data=json.dumps(payload))
        res_json = res.json()
        response = res_json["choices"][0]["message"]["content"]
        qa_history.append((query, response))

        words = response.strip().split()
        for i in range(len(words)):
            yield " ".join(words[:i+1])
            time.sleep(0.02)
    except Exception as e:
        yield f"❌ Error: {str(e)}"

def process_pdf(pdf_file):
    global chunks, index, uploaded_filename
    if pdf_file is None:
        return "❌ No file selected."
    uploaded_filename = pdf_file.name.split("/")[-1]
    doc = fitz.open(pdf_file.name)
    full_text = "\n".join([page.get_text() for page in doc])
    chunks = split_into_chunks(full_text)
    embeddings = get_embeddings(chunks)
    if not embeddings.any():
        return "❌ No text extracted."
    dim = embeddings[0].shape[0]
    index = faiss.IndexFlatIP(dim)
    index.add(embeddings)
    return "✅ Processed. Ready for Q&A."

def clear_cache():
    global chunks, index, qa_history, uploaded_filename
    chunks, index, qa_history, uploaded_filename = [], None, [], ""
    return "🗑️ Cache cleared."

def export_history():
    if not qa_history:
        return None
    timestamp = datetime.datetime.now().strftime("%Y-%m-%d_%H-%M-%S")
    filename = f"qa_history_{timestamp}.txt"
    with open(filename, "w", encoding="utf-8") as f:
        for q, a in qa_history:
            f.write(f"Q: {q}\nA: {a}\n\n")
    return filename

custom_css = """
#popup-alert {
  background-color: #fef3c7;
  color: #92400e;
  padding: 12px 20px;
  border-radius: 8px;
  border: 1px solid #fcd34d;
  font-weight: bold;
  position: relative;
  margin-bottom: 12px;
}
#popup-alert button {
  position: absolute;
  top: 4px;
  right: 8px;
  background: none;
  color: #92400e;
  border: none;
  font-size: 18px;
  cursor: pointer;
}
"""

with gr.Blocks(theme=gr.themes.Soft(), css=custom_css) as app:
    with gr.Row():
        gr.HTML(
            """<div style='text-align:center'>
                <h2>🤖 Chat with Your Research Paper</h2>
                <div id='popup-alert' style="display: inline-block;">
                    ⚠️ Please click ‘Clear Cache’ before uploading a new PDF.
                    <button onclick="this.parentElement.style.display='none';">&times;</button>
                </div>
            </div>"""
        )

    with gr.Row(equal_height=False):
        with gr.Column(scale=1, min_width=250):
            pdf_upload = gr.File(label="📁 Upload PDF", file_types=[".pdf"])
            upload_status = gr.Textbox(label="Status", interactive=False)
            clear_button = gr.Button("🧹 Clear Cache")
            export_button = gr.Button("📤 Export Q&A History")
            download_box = gr.File(visible=False)

            pdf_upload.change(fn=process_pdf, inputs=pdf_upload, outputs=upload_status)
            clear_button.click(fn=clear_cache, outputs=upload_status)
            export_button.click(fn=export_history, inputs=[], outputs=download_box)
            download_box.change(lambda x: gr.update(visible=True) if x else gr.update(visible=False), inputs=download_box, outputs=download_box)

        with gr.Column(scale=4, min_width=600):
            gr.ChatInterface(
                fn=ask_question_stream,
                chatbot=gr.Chatbot(label="📄 PDF Chatbot", show_copy_button=True),
                textbox=gr.Textbox(placeholder="Ask about the uploaded paper...", container=False, scale=7),
                examples=["What is the conclusion?", "Who are the authors?", "What are the key findings?"]
            )

    app.launch()