Update app.py

app.py

@@ -116,14 +116,14 @@ KEYWORD_OVERRIDE = {
     "goodbye": ["مع السلامة","مع السلامه","باي","وداعا","bye","goodbye","تصبح على خير",
                 "في امان الله","الله يسلمك","سلامتك"],
     "greeting":["السلام عليكم","وعليكم السلام","اهلا","أهلا","هلا","هلو","مرحبا","مرحباً",
-                "صباح الخير","مساء الخير","هاي","hi","hello","صباح","مساء"],
+                "صباح الخير","مساء الخير","هاي","hi","hello","ازيك","اخبارك","عامل ايه","صباح","مساء"],
     "thanks":  ["شكرا","شكراً","تسلم","يسلمو","ممنون","مشكور","thanks","thank","الف شكر"],
 }
 CATEGORY_KEYWORDS = {
-    "restaurant":["مطعم","اكل","وجبات","مشويات","كباب","شاورما","كريب","سندوتش","سندوتشات","برجر","سمك","فرايد","مطاعم","فول","طعمية"],
-    "pharmacy":  ["صيدليه","صيدلية","دوا","ادويه","دواء","صيدله","دكتور","علاج"],
-    "cafe":      ["كافيه","كوفي","قهوه","قهوة","كافيتيريا","كافية","قهوي"],
-    "supermarket":["سوبرماركت","ماركت","بقاله","هايبر","بقالة","سوبر ماركت","سوبرماركات","جروسري"],
+    "restaurant":["مطعم","اكل","أكل","وجبات","مشويات","كباب","شاورما","كريب","سندوتش","سندوتشات","برجر","سمك","فرايد","مطاعم","فول","طعمية","restaurant","restaurants","food","meal","burger","pizza","shawarma","crepe","grill","grills"],
+    "pharmacy":  ["صيدليه","صيدلية","دوا","ادويه","دواء","صيدله","دكتور","علاج","pharmacy","drugstore","medicine"],
+    "cafe":      ["كافيه","كوفي","قهوه","قهوة","كافيتيريا","كافية","قهوي","cafe","coffee","coffee shop"],
+    "supermarket":["سوبرماركت","ماركت","بقاله","هايبر","بقالة","سوبر ماركت","سوبرماركات","جروسري","supermarket","market","grocery","hypermarket"],
     "housing":   ["شقه","شقة","ايجار","إيجار","فندق","هوستل","سكن","سكن"],
 }
 _CAT_MAP = {
@@ -420,15 +420,63 @@ def rebuild_embeddings(df: pd.DataFrame, force: bool = False):
 """HELPER FUNCTIONS"""
 
 def normalize_category(cat):
-    if not cat: return cat
+    """Normalize DB/API categories to the app canonical categories."""
+    if not cat:
+        return cat
     cat_s = str(cat).strip()
-    if cat_s in ("restaurant","pharmacy","cafe","supermarket","housing"):
-        return cat_s
     c = clean_text(cat_s)
-    if c in _CAT_MAP: return _CAT_MAP[c]
+
+    # exact canonical values
+    if c in ("restaurant", "pharmacy", "cafe", "supermarket", "housing"):
+        return c
+
+    # English / mixed API categories
+    if "pharmacy" in c or "صيدلي" in c or "صيدله" in c:
+        return "pharmacy"
+    if "supermarket" in c or "market" in c or "ماركت" in c or "بقال" in c or "هايبر" in c:
+        return "supermarket"
+    if "housing" in c or "hotel" in c or "hostel" in c or "شقه" in c or "سكن" in c or "فندق" in c:
+        return "housing"
+    # Restaurant & Café: لو مطلوب restaurant هيعدي بالـ contains في الفلتر، لكن normalized الأساسي restaurant
+    if "restaurant" in c or "مطعم" in c or "مطاعم" in c:
+        return "restaurant"
+    if "cafe" in c or "coffee" in c or "كافيه" in c or "كوفي" in c or "قهوه" in c or "قهوة" in c:
+        return "cafe"
+
+    if c in _CAT_MAP:
+        return _CAT_MAP[c]
     for ar, en in _CAT_MAP.items():
-        if ar in c or c in ar: return en
-    return cat_s
+        if ar in c or c in ar:
+            return en
+    return c
+
+
+def _category_match_value(raw_category: str, target_category: str) -> bool:
+    """Strict category match without allowing unrelated fallback."""
+    if not target_category:
+        return True
+    target = normalize_category(target_category)
+    raw = clean_text(raw_category)
+
+    if target == "restaurant":
+        return ("restaurant" in raw or "مطعم" in raw or "مطاعم" in raw) and "pharmacy" not in raw and "صيدلي" not in raw
+    if target == "cafe":
+        return "cafe" in raw or "coffee" in raw or "كافيه" in raw or "كوفي" in raw or "قهوه" in raw or "قهوة" in raw
+    if target == "pharmacy":
+        return "pharmacy" in raw or "صيدلي" in raw or "صيدله" in raw
+    if target == "supermarket":
+        return "supermarket" in raw or "market" in raw or "ماركت" in raw or "بقال" in raw or "هايبر" in raw
+    if target == "housing":
+        return "housing" in raw or "hotel" in raw or "hostel" in raw or "شقه" in raw or "سكن" in raw or "فندق" in raw
+    return target in raw
+
+
+def _is_proximity_query(text: str) -> bool:
+    t = norm(text)
+    return any(w in t for w in [
+        "اقرب", "أقرب", "الاقرب", "الأقرب", "قريب", "قريبه", "قريبة",
+        "جنبي", "حواليا", "حولي", "near", "nearest", "nearby", "close"
+    ])
 
 def apply_keyword_override(text):
     t = norm(text); tw = set(t.split())
@@ -621,11 +669,12 @@ def apply_filters(df, category=None, sub_category=None, location=None,
 
     # ── category ─────────────────────────────────────────────────
     if category:
-        cat_c = clean_text(category)
-        f_cat = f[f["category_clean"].astype(str).str.contains(re.escape(cat_c), na=False)]
-        if not f_cat.empty:
-            f = f_cat
-        # لو مفيش ونعم semantic ممكن تديه — الـ caller يقرر
+        target_cat = normalize_category(category)
+        f_cat = f[f["category"].astype(str).apply(lambda x: _category_match_value(x, target_cat))]
+        if f_cat.empty and "category_clean" in f.columns:
+            f_cat = f[f["category_clean"].astype(str).apply(lambda x: _category_match_value(x, target_cat))]
+        # strict: لو المستخدم طلب category معينة ممنوع نبدّلها بـ category تانية
+        f = f_cat
 
     # ── sub_category ─────────────────────────────────────────────
     if sub_category:
@@ -771,6 +820,15 @@ def rank(df, query, user_lat=None, user_lon=None, intent=None):
             axis=1
         )
 
+    if is_proximity and has_loc and "distance_km" in df.columns:
+        # في طلب "أقرب" ومعانا location: المسافة هي الحَكَم الأول
+        df = df.sort_values(
+            by=["distance_km", "rating_norm"],
+            ascending=[True, False],
+            na_position="last",
+        ).reset_index(drop=True)
+        return df
+
     return df.sort_values("final_score", ascending=False).reset_index(drop=True)
 
 def search_places(query, top_k_final=5, category=None, sub_category=None,
@@ -834,17 +892,8 @@ def search_places(query, top_k_final=5, category=None, sub_category=None,
         strict_location=bool(location),
     )
 
-    # لو مفيش location ومفيش نتائج بعد category، جرب من غير category فقط للأسئلة العامة
-    if filtered.empty and category and not location:
-        filtered = apply_filters(
-            all_cands,
-            sub_category=sub_category,
-            price_range=price_range,
-            open_now_only=open_now_only,
-            min_rating=min_rating,
-            strict_location=False,
-        )
-
+    # مهم: ممنوع نلغي category لو المستخدم طلب nearest_* أو category واضحة
+    # لأن ده كان بيرجع restaurant في سؤال nearest_pharmacy مثلاً.
     if filtered.empty:
         return pd.DataFrame()
 
@@ -1710,10 +1759,9 @@ def chat(text: str, session, user_lat=None, user_lon=None):
 
     # ── proximity query: لو مفيش lat/lon ────────────────────────
     t_low = norm(text)
-    PROX_REQ = ["قريب مني","جنبي","حواليا","أقرب ليا","الأقرب ليا"]
-    if any(p in t_low for p in PROX_REQ) and not (user_lat and user_lon):
-        reply = ("📍 عشان أجيب الأقرب ليك محتاج موقعك.\n"
-                 "فعّل الـ location من التطبيق أو قولي أنت في أنهي منطقة؟")
+    if _is_proximity_query(text) and not (user_lat and user_lon):
+        reply = ("📍 عشان أحدد الأقرب محتاج تسمحلي باللوكيشن.\n"
+                 "أو قولي المنطقة اللي أنت فيها.")
         result.update(reply=reply, intent="missing_info",
                       best_place=None, all_results=[])
         session.add(text, reply, "missing_info", ents, None, [])
@@ -1887,7 +1935,7 @@ SEARCH_INTENTS.add("show_menu")
 _MENU_TRIGGERS  = {"منيو","menu","القائمه","قايمة","قائمة","الاكلات","اكلات","عروضه","عرض المطعم"}
 _ITEM_TRIGGERS  = {
     # أكلات رئيسية
-    "بيتزا","برجر","كريب","شاورما","بروست","سوشي","رانش",
+    "بيتزا","برجر","كريب","شاورما","بروست","سوشي","رانش","pizza","burger","crepe","shawarma","broast","sushi","ranch",
     "ساندوتش","سندوتش","بسطرمه","كبده","كبدة","حواوشي",
     "طعميه","طعمية","فول","كشري","مكرونه","لازانيا",
     # لحوم ودجاج وسمك
@@ -2092,6 +2140,80 @@ def search_places_by_item(item_query: str) -> list[dict]:
     log.info(f"🔍 search_by_item fuzzy → {len(places)} results")
     return places
 
+
+def filter_item_results_strict(places: list, item_query: str) -> list:
+    """
+    فلتر قوي لنتائج search_by_item:
+    - لو السؤال عن أكل: restaurants فقط.
+    - لو السؤال عن مشروب: cafe أو restaurant فقط.
+    - لازم الصنف نفسه يكون ظاهر في matched_items / sub_category / description / name.
+    """
+    if not places:
+        return []
+
+    q = clean_text(item_query)
+
+    food_keywords = {
+        "شاورما", "برجر", "بيتزا", "كريب", "مشويات", "فراخ",
+        "دجاج", "كباب", "كفته", "كفتة", "سمك", "ساندوتش",
+        "سندوتش", "حواوشي", "كشري", "مكرونه", "مكرونة",
+        "وجبه", "وجبة", "اكل", "أكل", "burger", "pizza",
+        "shawarma", "crepe", "chicken", "grill", "grills"
+    }
+
+    drink_keywords = {
+        "قهوة", "قهوه", "كوفي", "عصير", "مشروب", "مشروبات",
+        "لاتيه", "اسبريسو", "نسكافيه", "سموذي",
+        "coffee", "juice", "latte", "espresso", "smoothie"
+    }
+
+    is_food = any(clean_text(w) in q for w in food_keywords)
+    is_drink = any(clean_text(w) in q for w in drink_keywords)
+
+    query_words = [w for w in q.split() if len(w) > 2]
+    filtered = []
+
+    for p in places:
+        raw_category = str(p.get("category", ""))
+        category = normalize_category(raw_category)
+
+        text_blob = clean_text(
+            f"{p.get('name','')} "
+            f"{p.get('name_ar','')} "
+            f"{p.get('name_en','')} "
+            f"{p.get('category','')} "
+            f"{p.get('sub_category','')} "
+            f"{p.get('description','')} "
+            f"{p.get('matched_item','')} "
+            f"{p.get('matched_items','')} "
+            f"{p.get('menu_items','')} "
+            f"{p.get('tags','')}"
+        )
+
+        if is_food and category != "restaurant" and not _category_match_value(raw_category, "restaurant"):
+            continue
+
+        if is_drink and not (
+            category in ("cafe", "restaurant") or
+            _category_match_value(raw_category, "cafe") or
+            _category_match_value(raw_category, "restaurant")
+        ):
+            continue
+
+        if query_words and not any(w in text_blob for w in query_words):
+            continue
+
+        filtered.append(p)
+
+    # dedup
+    seen, unique = set(), []
+    for p in filtered:
+        pid = str(p.get("place_id") or p.get("id") or p.get("name") or "")
+        if pid and pid not in seen:
+            seen.add(pid)
+            unique.append(p)
+    return unique
+
 def get_place_menu(place_id: str) -> list[dict]:
     """
     Calls GET /api/mobile/places/<place_id>/menu
@@ -2222,7 +2344,8 @@ def handle_search_by_item(text: str, session, user_lat=None, user_lon=None) -> d
     Handles search_by_item intent end-to-end.
     1. Extract item_query
     2. Call backend API
-    3. Return formatted response
+    3. Strictly filter wrong categories/items
+    4. Return formatted response
     """
     item_query = extract_item_query(text)
     item_query = clean_item_query(item_query)
@@ -2234,7 +2357,21 @@ def handle_search_by_item(text: str, session, user_lat=None, user_lon=None) -> d
             "place_cards": [],
             "best_place":  None,
         }
+
     places = search_places_by_item(item_query)
+    places = filter_item_results_strict(places, item_query)
+
+    if not places:
+        result = {
+            "reply":       f"مش لاقي أماكن بتقدم {item_query} حاليًا 😔",
+            "intent":      "search_by_item",
+            "ui_type":     CFG.UI_TEXT_ONLY,
+            "place_cards": [],
+            "best_place":  None,
+        }
+        session.add(text, result["reply"], "search_by_item", {"item": item_query}, None, [])
+        return result
+
     result = format_item_search_response(places, item_query, user_lat, user_lon)
     session.add(text, result["reply"], "search_by_item", {"item": item_query},
                 result.get("best_place"), places)
@@ -2598,4 +2735,4 @@ def reload_data(x_secret: Optional[str] = Header(default=None)):
 def reset_session(session_id: str):
     SESSIONS.pop(session_id, None)
     log.info(f"🗑️  Session {session_id} reset")
-    return {"status": "reset", "session_id": session_id}
+    return {"status": "reset", "session_id": session_id}