Update app.py

app.py

@@ -1,18 +1,38 @@
-from fastapi import FastAPI
 import gradio as gr
 from sentence_transformers import SentenceTransformer, util
 import torch
+import logging
 
-# Load model (auto-downloads if not cached)
-model = SentenceTransformer("CAMeL-Lab/bert-base-arabic-camelbert-ca")
+# Set up logging
+logging.basicConfig(level=logging.INFO)
+logger = logging.getLogger(__name__)
 
-# Load knowledge file and generate embeddings
-with open("knowledge.txt", "r", encoding="utf-8") as f:
-    knowledge_text = f.read()
+# Load model
+logger.info("Loading model...")
+try:
+    model = SentenceTransformer("CAMeL-Lab/bert-base-arabic-camelbert-ca")
+except Exception as e:
+    logger.error(f"Failed to load model: {e}")
+    raise
 
-# Split the text into chunks
-def split_text(text, chunk_size=400):
-    sentences = text.split("،")  # Split on Arabic comma
+# Load knowledge file
+logger.info("Loading knowledge file...")
+try:
+    with open("knowledge.txt", "r", encoding="utf-8") as f:
+        knowledge_text = f.read()
+    if not knowledge_text.strip():
+        raise ValueError("knowledge.txt is empty")
+except FileNotFoundError:
+    logger.error("knowledge.txt not found")
+    raise
+except Exception as e:
+    logger.error(f"Error reading knowledge.txt: {e}")
+    raise
+
+# Split text into chunks
+def split_text(text, chunk_size=200):
+    logger.info("Splitting text into chunks...")
+    sentences = text.split("،")
     chunks, chunk = [], ""
     for sentence in sentences:
         if len(chunk) + len(sentence) < chunk_size:
@@ -24,29 +44,39 @@ def split_text(text, chunk_size=400):
         chunks.append(chunk.strip())
     return chunks
 
-chunks = split_text(knowledge_text)
-corpus_embeddings = model.encode(chunks, convert_to_tensor=True)
+# Generate embeddings
+logger.info("Generating embeddings...")
+try:
+    chunks = split_text(knowledge_text)
+    corpus_embeddings = model.encode(chunks, convert_to_tensor=True, batch_size=16)
+except Exception as e:
+    logger.error(f"Error generating embeddings: {e}")
+    raise
 
 # Search function
 def answer_question(question):
-    question_embedding = model.encode(question, convert_to_tensor=True)
-    scores = util.cos_sim(question_embedding, corpus_embeddings)[0]
-    best_idx = torch.argmax(scores).item()
-    return chunks[best_idx]
+    try:
+        if not question.strip():
+            return "الرجاء إدخال سؤال."
+        question_embedding = model.encode(question, convert_to_tensor=True)
+        scores = util.cos_sim(question_embedding, corpus_embeddings)[0]
+        best_idx = torch.argmax(scores).item()
+        score = scores[best_idx].item()
+        logger.info(f"Question: {question}, Best chunk: {chunks[best_idx][:50]}..., Score: {score:.4f}")
+        return chunks[best_idx] if score > 0.3 else "عذرًا، لم أجد إجابة مناسبة. حاول صياغة السؤال بطريقة أخرى."
+    except Exception as e:
+        logger.error(f"Error answering question: {e}")
+        return "حدث خطأ أثناء معالجة السؤال. حاول مرة أخرى."
 
 # Gradio UI
+logger.info("Setting up Gradio UI...")
 with gr.Blocks() as demo:
     gr.Markdown("### 🤖 اسأل عن التنمية المستدامة أو الموازنة التشاركية")
     with gr.Row():
-        inp = gr.Textbox(label="اكتب سؤالك هنا", placeholder="مثال: ما هي أهداف التنمية المستدامة؟")
-    out = gr.Textbox(label="الإجابة")
+        inp = gr.Textbox(label="اكتب سؤالك هنا", placeholder="مثال: ما هي أهداف التنمية المستدامة؟", dir="rtl")
+    out = gr.Textbox(label="الإجابة", dir="rtl")
     btn = gr.Button("إجابة")
     btn.click(fn=answer_question, inputs=inp, outputs=out)
 
-# FastAPI + Gradio mount
-app = FastAPI()
-@app.get("/")
-def read_root():
-    return {"message": "Arabic Q&A Chatbot running."}
-
-app = gr.mount_gradio_app(app, demo, path="/")
+# Launch Gradio app (Hugging Face Spaces will handle this automatically)
+demo.launch(server_name="0.0.0.0", server_port=7860)