app.py

import gradio as gr
from langchain_community.vectorstores import Chroma
from langchain_community.embeddings import HuggingFaceEmbeddings
from langchain.chains import RetrievalQA
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.document_loaders import PyPDFLoader
import os
import shutil

CHROMA_PATH = "chroma_db"
EMBED_MODEL = "sentence-transformers/all-MiniLM-L6-v2"

def load_and_prepare_file(file_path):
    # تنظيف المجلد القديم
    if os.path.exists(CHROMA_PATH):
        shutil.rmtree(CHROMA_PATH)

    # تحميل وتقطيع النص
    loader = PyPDFLoader(file_path)
    pages = loader.load_and_split()
    text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=100)
    chunks = text_splitter.split_documents(pages)

    # إنشاء قاعدة بيانات المتجهات
    embedding_function = HuggingFaceEmbeddings(model_name=EMBED_MODEL)
    vectordb = Chroma.from_documents(chunks, embedding_function, persist_directory=CHROMA_PATH)
    vectordb.persist()
    return "✅ تم تجهيز الملف بنجاح، يمكنك الآن طرح الأسئلة."

def answer_question(question):
    embedding_function = HuggingFaceEmbeddings(model_name=EMBED_MODEL)
    vectordb = Chroma(persist_directory=CHROMA_PATH, embedding_function=embedding_function)
    retriever = vectordb.as_retriever()
    qa = RetrievalQA.from_chain_type(llm="gpt2", retriever=retriever)
    result = qa.run(question)
    return result

with gr.Blocks() as demo:
    gr.Markdown("### 📚 Smart PDF Assistant - مساعد PDF الذكي")

    file_input = gr.File(label="📄 ارفع ملف PDF", type="filepath")
    upload_output = gr.Textbox(label="نتيجة الرفع")
    upload_button = gr.Button("تحميل ومعالجة الملف")

    question_input = gr.Textbox(label="✍️ اكتب سؤالك هنا")
    answer_output = gr.Textbox(label="🔎 الإجابة")

    upload_button.click(load_and_prepare_file, inputs=file_input, outputs=upload_output)
    question_input.submit(answer_question, inputs=question_input, outputs=answer_output)

demo.launch()