Upload 5 files

.gitignore

@@ -0,0 +1,8 @@
+__pycache__/
+*.pyc
+*.pyo
+*.pyd
+.env
+tmp/
+*.tmp
+*.log

README.md

@@ -0,0 +1,23 @@
+# Arabic Book Brain AI - Hugging Face Space
+
+## 📚 Description (English)
+
+Upload Arabic books (.pdf, .docx, .doc), train a system to understand the book's ideas, and answer Arabic questions based strictly on the uploaded content.
+
+## 📚 الوصف (بالعربية)
+
+رفع كتب باللغة العربية (بصيغة PDF أو DOCX أو DOC)، تدريب نظام ذكي لفهم أفكار الكتب، ثم الإجابة عن الأسئلة باللغة العربية بناءً فقط على المحتوى.
+
+---
+
+## 🛠 Installation
+
+```bash
+pip install -r requirements.txt
+```
+
+## 🧠 Start App
+
+```bash
+python app.py
+```

app.py

@@ -0,0 +1,122 @@
+import gradio as gr
+import os
+import tempfile
+import faiss
+import torch
+from langchain.embeddings import HuggingFaceEmbeddings
+from langchain.vectorstores import FAISS
+from langchain.text_splitter import RecursiveCharacterTextSplitter
+from langchain.prompts import PromptTemplate
+from langchain.chains import RetrievalQA
+from langchain.llms import HuggingFacePipeline
+from transformers import AutoTokenizer, AutoModelForSeq2SeqLM, pipeline
+from pdfminer.high_level import extract_text as extract_pdf_text
+import docx
+import nltk
+
+nltk.download('punkt')
+from nltk.tokenize import sent_tokenize
+
+uploaded_texts = []
+vector_store = None
+qa_chain = None
+
+embedding_model_name = "CAMeL-Lab/bert-base-arabic-camelbert-mix"
+embeddings = HuggingFaceEmbeddings(model_name=embedding_model_name)
+
+model_name = "csebuetnlp/mT5_small"
+tokenizer = AutoTokenizer.from_pretrained(model_name)
+model = AutoModelForSeq2SeqLM.from_pretrained(model_name)
+
+pipe = pipeline("text2text-generation", model=model, tokenizer=tokenizer, max_length=512)
+llm = HuggingFacePipeline(pipeline=pipe)
+
+ARABIC_PROMPT_TEMPLATE = """
+أنت نظام ذكي يجيب بناءً فقط على المعلومات المستخرجة من الكتب.
+لا تستخدم أي معلومات خارجية.
+السؤال: {question}
+الإجابة:
+"""
+
+def format_arabic_prompt(question):
+    return ARABIC_PROMPT_TEMPLATE.format(question=question)
+
+def extract_text_from_file(file_path):
+    if file_path.endswith(".pdf"):
+        return extract_pdf_text(file_path)
+    elif file_path.endswith(".docx") or file_path.endswith(".doc"):
+        doc = docx.Document(file_path)
+        return "\n".join([para.text for para in doc.paragraphs])
+    else:
+        raise ValueError("Unsupported file format")
+
+def arabic_split_text(text):
+    sentences = sent_tokenize(text, language='arabic')
+    chunks = []
+    chunk = ""
+    for sentence in sentences:
+        if len(chunk) + len(sentence) <= 500:
+            chunk += " " + sentence
+        else:
+            chunks.append(chunk.strip())
+            chunk = sentence
+    if chunk:
+        chunks.append(chunk.strip())
+    return chunks
+
+def train_from_texts(texts):
+    global vector_store, qa_chain
+
+    splitter = RecursiveCharacterTextSplitter(
+        chunk_size=500,
+        chunk_overlap=100,
+        length_function=len,
+    )
+    
+    all_chunks = []
+    for text in texts:
+        chunks = arabic_split_text(text)
+        all_chunks.extend(chunks)
+
+    vectors = embeddings.embed_documents(all_chunks)
+    dimension = len(vectors[0])
+    index = faiss.IndexFlatL2(dimension)
+    vector_store = FAISS(embedding_function=embeddings, index=index, documents=all_chunks)
+
+    retriever = vector_store.as_retriever(search_type="similarity", search_kwargs={"k": 10})
+    qa_chain = RetrievalQA.from_chain_type(llm=llm, retriever=retriever)
+
+def upload_book(file, progress=gr.Progress()):
+    with tempfile.NamedTemporaryFile(delete=False) as tmp:
+        tmp.write(file.read())
+        tmp_path = tmp.name
+
+    progress(0.2, desc="تحميل الملف...")
+    extracted_text = extract_text_from_file(tmp_path)
+    uploaded_texts.append(extracted_text)
+    progress(0.5, desc="معالجة النص...")
+
+    train_from_texts(uploaded_texts)
+    progress(1.0, desc="اكتمل التدريب!")
+    return "النظام جاهز للإجابة على أسئلتك"
+
+def answer_question(user_question):
+    if qa_chain is None:
+        return "الرجاء رفع كتاب أولاً."
+    prompt = format_arabic_prompt(user_question)
+    result = qa_chain.run(prompt)
+    return result
+
+with gr.Blocks() as demo:
+    with gr.Tab("تحميل الكتب"):
+        upload_button = gr.File(label="ارفع كتابك (.pdf .docx .doc)", file_types=[".pdf", ".docx", ".doc"])
+        upload_output = gr.Textbox(label="حالة النظام")
+        upload_button.upload(upload_book, inputs=upload_button, outputs=upload_output)
+
+    with gr.Tab("اسأل الكتاب"):
+        question = gr.Textbox(label="اكتب سؤالك بالعربية")
+        answer = gr.Textbox(label="الإجابة")
+        ask_button = gr.Button("إرسال السؤال")
+        ask_button.click(answer_question, inputs=question, outputs=answer)
+
+demo.launch(share=True)

requirements.txt

@@ -0,0 +1,9 @@
+gradio
+transformers
+sentence-transformers
+langchain
+faiss-cpu
+torch
+pdfminer.six
+python-docx
+nltk

space.yaml

@@ -0,0 +1,10 @@
+# Space metadata
+title: Arabic Book Brain AI
+emoji: 📚
+colorFrom: purple
+colorTo: indigo
+sdk: gradio
+sdk_version: 4.14.0
+app_file: app.py
+python_version: 3.10
+license: apache-2.0