Update chatbot_updated.py

chatbot_updated.py

@@ -1,162 +1,61 @@
-# ==============================
-# ✅ chatbot_gradio_fixed.py
-# ==============================
-import os
-import gradio as gr
-import json
-import numpy as np
-from deep_translator import GoogleTranslator
-from ultralytics import YOLO
-from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
-
-# ==============================
-# 🔹 Paths (نسبيّة مع مكان المشروع)
-# ==============================
-BASE_DIR = os.path.dirname(__file__)  # مكان هذا الملف
-yolo_model_path = os.path.join(BASE_DIR, "best_egypt.pt")  # ضع موديل YOLO هنا
-artifacts_json_path = os.path.join(BASE_DIR, "artifacts.json")  # ملف JSON للأثار
-
-# ==============================
-# ✅ Load YOLO Model
-# ==============================
-yolo_model = YOLO(yolo_model_path)
-
-# ==============================
-# ✅ Load FLAN-T5 Model
-# ==============================
-model_name = "google/flan-t5-base"
-tokenizer = AutoTokenizer.from_pretrained(model_name)
-model = AutoModelForSeq2SeqLM.from_pretrained(model_name)
-
-# ==============================
-# ✅ Load Artifacts JSON
-# ==============================
-with open(artifacts_json_path, "r", encoding="utf-8") as f:
-    data = json.load(f)
-
-# ==============================
-# ✅ Helper Functions
-# ==============================
-def is_arabic(text):
-    return any('\u0600' <= ch <= '\u06FF' for ch in text)
-
-def translate_to_en(text):
-    return GoogleTranslator(source='auto', target='en').translate(text)
-
-def translate_to_ar(text):
-    return GoogleTranslator(source='auto', target='ar').translate(text)
-
-def detect_artifact(image_path):
-    results = yolo_model(image_path)
-    result = results[0]
-
-    if result.boxes is None or len(result.boxes) == 0:
-        return None
-
-    scores = result.boxes.conf.cpu().numpy()
-    best_index = np.argmax(scores)
-
-    class_id = int(result.boxes.cls[best_index])
-    return result.names[class_id]
-
-def get_artifact(artifact_name):
-    artifact_name = artifact_name.lower().strip()
-
-    # 1️⃣ Exact match
-    for doc in data:
-        if artifact_name == doc['name'].lower().strip():
-            return doc
-
-    # 2️⃣ Partial match
-    for doc in data:
-        if artifact_name in doc['name'].lower():
-            return doc
-
-    # 3️⃣ Match keywords
-    for doc in data:
-        if 'keywords' in doc:
-            for k in doc['keywords']:
-                if k.lower() in artifact_name:
-                    return doc
-    return None
-
-# ==============================
-# ✅ Chatbot Function
-# ==============================
-def chatbot_gradio(image, question):
-    user_lang = "ar" if is_arabic(question) else "en"
-    question_en = translate_to_en(question) if user_lang == "ar" else question
-
-    artifact_name = detect_artifact(image.name)
-    if artifact_name is None:
-        return "لم يتم التعرف على الأثر في الصورة" if user_lang == "ar" else "Artifact not detected."
-
-    artifact = get_artifact(artifact_name)
-    if artifact is None:
-        return "لا توجد بيانات عن هذا الأثر" if user_lang == "ar" else "No data found for this artifact."
-
-    q = question_en.lower()
-    if any(word in q for word in ["who built", "creator", "made by"]):
-        answer = artifact['creator']
-    elif any(word in q for word in ["where", "location", "found"]):
-        answer = artifact['location_found']
-    elif any(word in q for word in ["material", "made of"]):
-        answer = artifact['material']
-    elif any(word in q for word in ["describe", "appearance", "look like", "shape", "what do you know about it", "tell me about it"]):
-        answer = artifact['description'] + " " + artifact['importance']
-    else:
-        answer = artifact['description'] + " " + artifact['importance']
-
-    # إعداد الـ Prompt للموديل
-    instruction = "Provide a detailed and well-explained answer using ONLY the artifact information."
-    prompt = f"""
-You are an expert assistant specialized in Egyptian artifacts.
-IMPORTANT RULES:
-- Use ONLY the artifact data provided below.
-- Do NOT use any external knowledge.
-- Do NOT guess or hallucinate.
-- If the answer cannot be found in the data, respond exactly:
-"I don't know based on the given artifact data."
-- Always provide detailed and well-explained answers.
-Instruction:
-{instruction}
-Artifact Information:
-Name: {artifact['name']}
-Creator: {artifact['creator']}
-Built Year: {artifact['built_year']}
-Type: {artifact['type']}
-Era: {artifact['era']}
-Material: {artifact['material']}
-Description: {artifact['description']}
-Importance: {artifact['importance']}
-Location Found: {artifact['location_found']}
-Current Location: {artifact['current_location']}
-Question:
-{question_en}
-Answer:
-"""
-
-    inputs = tokenizer(prompt, return_tensors="pt")
-    outputs = model.generate(**inputs, max_new_tokens=400, do_sample=False)
-    response_en = tokenizer.decode(outputs[0], skip_special_tokens=True)
-
-    if "Answer:" in response_en:
-        response_en = response_en.split("Answer:")[-1].strip()
-
-    return translate_to_ar(response_en) if user_lang == "ar" else response_en
-
-# ==============================
-# ✅ Gradio Interface
-# ==============================
-iface = gr.Interface(
-    fn=chatbot_gradio,
-    inputs=[
-        gr.Image(type="file", label="Upload Artifact Image"),
-        gr.Textbox(lines=2, placeholder="Ask a question about the artifact...", label="Question")
-    ],
-    outputs=gr.Textbox(label="Answer"),
-    title="Egyptian Artifacts Chatbot",
-    description="Upload an image of an Egyptian artifact and ask any question about it."
-)
-
-iface.launch()
+# ==============================
+# 🔹 Main Entry Point
+# ==============================
+def chatbot_updated(question, image=None):
+    global LAST_ARTIFACT
+
+    if not question or question.strip() == "":
+        return "أهلاً بك! كيف يمكنني مساعدتك في التعرف على الآثار؟"
+
+    user_lang = "ar" if is_arabic(question) else "en"
+    question_en = translate_to_en(question) if user_lang == "ar" else question
+
+    artifact_name = None
+
+    # 1️⃣ التعرف من الصورة
+    if image is not None:
+        detected, _ = detect_artifact(image)
+        if detected and get_artifact(detected):
+            artifact_name = detected
+            LAST_ARTIFACT = detected
+
+    # 2️⃣ التعرف من النص
+    if artifact_name is None:
+        for doc in data:
+            if doc['name'].lower() in question_en.lower():
+                artifact_name = doc['name']
+                LAST_ARTIFACT = doc['name']
+                break
+
+    # 3️⃣ لو ما لقيناش أثر نهائي
+    if artifact_name is None or not get_artifact(artifact_name):
+        return "لا توجد معلومات عن هذا الأثر." if user_lang == "ar" else "No data found for this artifact."
+
+    artifact = get_artifact(artifact_name)
+
+    # معالجة الأسئلة المتعددة
+    raw_parts = split_questions(question_en)
+    if not raw_parts:
+        raw_parts = [question_en]
+
+    all_answers = []
+    for part in raw_parts:
+        # حل الضمائر it → اسم الأثر
+        if " it " in f" {part.lower()} ":
+            part = part.replace(" it ", f" {artifact_name} ")
+        answer = get_best_response(part, artifact)
+        if answer:
+            all_answers.append(answer)
+
+    # دمج الإجابات بجملة واحدة سلسة
+    if not all_answers:
+        return "لا يمكنني العثور على إجابة محددة."
+
+    final_answer_en = " ".join(all_answers)
+
+    # الترجمة النهائية
+    if user_lang == "ar":
+        ar_answer = translate_to_ar(final_answer_en)
+        return ar_answer.replace("  ", " ").strip()
+    
+    return final_answer_en