Delete app.py

app.py

@@ -1,117 +0,0 @@
-# app.py
-
-import gradio as gr
-import os
-import pdfminer.high_level
-import docx
-from sentence_transformers import SentenceTransformer
-from transformers import AutoModelForCausalLM, AutoTokenizer
-import torch
-import faiss
-import tempfile
-
-# ====== Settings ======
-EMBEDDING_MODEL_NAME = "sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2"
-GENERATION_MODEL_NAME = "aubmindlab/aragpt2-small"
-CHUNK_SIZE = 500
-CHUNK_OVERLAP = 50
-TOP_K = 5
-
-# ====== Load Models ======
-embedder = SentenceTransformer(EMBEDDING_MODEL_NAME)
-gen_tokenizer = AutoTokenizer.from_pretrained(GENERATION_MODEL_NAME)
-gen_model = AutoModelForCausalLM.from_pretrained(GENERATION_MODEL_NAME)
-
-# ====== Globals ======
-index = None
-chunks = []
-
-# ====== Helpers ======
-def extract_text_from_pdf(file_path):
-    with open(file_path, 'rb') as f:
-        return pdfminer.high_level.extract_text(f)
-
-def extract_text_from_docx(file_path):
-    doc = docx.Document(file_path)
-    return "\n".join([para.text for para in doc.paragraphs])
-
-def chunk_text(text):
-    words = text.split()
-    chunks = []
-    for i in range(0, len(words), CHUNK_SIZE - CHUNK_OVERLAP):
-        chunk = " ".join(words[i:i+CHUNK_SIZE])
-        chunks.append(chunk)
-    return chunks
-
-def build_vector_store(chunks):
-    vectors = embedder.encode(chunks)
-    dim = vectors.shape[1]
-    idx = faiss.IndexFlatL2(dim)
-    idx.add(vectors)
-    return idx, vectors
-
-def retrieve_relevant_chunks(question, idx, chunks, vectors):
-    q_vec = embedder.encode([question])
-    D, I = idx.search(q_vec, TOP_K)
-    return [chunks[i] for i in I[0] if i < len(chunks)]
-
-def generate_answer(context_chunks, question):
-    context = " \n".join(context_chunks)
-    prompt = f"سؤال: {question}\nمحتوى ذو صلة: {context}\nجواب:"
-    inputs = gen_tokenizer(prompt, return_tensors="pt", max_length=1024, truncation=True)
-    outputs = gen_model.generate(**inputs, max_new_tokens=100)
-    answer = gen_tokenizer.decode(outputs[0], skip_special_tokens=True)
-    answer = answer.split("جواب:")[-1].strip()
-    return answer
-
-# ====== Gradio Functions ======
-def upload_and_train(files):
-    global index, chunks
-
-    all_text = ""
-    for file in files:
-        suffix = os.path.splitext(file.name)[-1].lower()
-        with tempfile.NamedTemporaryFile(delete=False, suffix=suffix) as tmp:
-            tmp.write(file.read())
-            tmp_path = tmp.name
-        if suffix == ".pdf":
-            all_text += extract_text_from_pdf(tmp_path) + "\n"
-        elif suffix in [".docx", ".doc"]:
-            all_text += extract_text_from_docx(tmp_path) + "\n"
-        os.unlink(tmp_path)
-
-    chunks = chunk_text(all_text)
-    index, vectors = build_vector_store(chunks)
-
-    return "✅ النظام جاهز للإجابة على أسئلتك"
-
-def ask_question(user_question):
-    if index is None:
-        return "الرجاء رفع الكتب أولاً وتدريب النظام."
-    rel_chunks = retrieve_relevant_chunks(user_question, index, chunks, None)
-    answer = generate_answer(rel_chunks, user_question)
-    return answer
-
-# ====== Gradio Interface ======
-upload = gr.File(file_types=[".pdf", ".doc", ".docx"], file_count="multiple")
-train_btn = gr.Button("ابدأ التدريب")
-train_output = gr.Textbox()
-question_input = gr.Textbox(placeholder="اكتب سؤالك هنا باللغة العربية")
-answer_output = gr.Textbox()
-ask_btn = gr.Button("أرسل السؤال")
-
-with gr.Blocks() as demo:
-    gr.Markdown("# 🧠 محاكاة دماغ المؤلف - نظام ذكي للإجابة على الأسئلة من كتبك بالعربية")
-    upload.render()
-    train_btn.render()
-    train_output.render()
-    question_input.render()
-    ask_btn.render()
-    answer_output.render()
-
-    train_btn.click(upload_and_train, inputs=[upload], outputs=[train_output])
-    ask_btn.click(ask_question, inputs=[question_input], outputs=[answer_output])
-
-# Launch
-if __name__ == "__main__":
-    demo.launch()