app.py

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

DB_DIR = "chroma_db"
embedding_model = HuggingFaceEmbeddings(model_name="sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2")
llm = HuggingFaceHub(repo_id="mistralai/Mistral-7B-Instruct-v0.2", model_kwargs={"temperature":0.3, "max_new_tokens":500})

def load_and_index(files):
    all_texts = []
    for file in files:
        loader = PyPDFLoader(file.name)
        docs = loader.load()
        splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=200)
        texts = splitter.split_documents(docs)
        all_texts.extend(texts)
    vectordb = Chroma.from_documents(all_texts, embedding=embedding_model, persist_directory=DB_DIR)
    vectordb.persist()
    return "✅ تم تحميل وفهرسة الملفات."

def answer_question(query):
    vectordb = Chroma(persist_directory=DB_DIR, embedding_function=embedding_model)
    qa_chain = RetrievalQA.from_chain_type(llm=llm, retriever=vectordb.as_retriever())
    answer = qa_chain.run(query)
    return answer

with gr.Blocks(title="Smart PDF Assistant") as demo:
    gr.Markdown("# 🤖 Smart PDF Assistant\nحمّل ملفات PDF واسأل أي سؤال 📚")
    with gr.Row():
        uploader = gr.File(file_types=[".pdf"], file_count="multiple", label="تحميل ملفات PDF")
        index_btn = gr.Button("فهرسة الملفات")
    index_output = gr.Textbox(label="حالة الفهرسة")
    index_btn.click(load_and_index, inputs=[uploader], outputs=[index_output])

    query = gr.Textbox(label="اكتب سؤالك")
    answer_btn = gr.Button("أجب")
    answer_output = gr.Textbox(label="الإجابة")
    answer_btn.click(answer_question, inputs=[query], outputs=[answer_output])

demo.launch()