Update app.py

app.py

@@ -103,6 +103,8 @@ FOOD_DB = {
     "egg":   {"kcal":155,"carb_g":1.1,"protein_g":13, "fat_g":11, "sodium_mg":124,    "cat":"豆魚蛋肉類", "base_g":60,  "tip":"水煮/荷包少油，避免重鹹醬料"},
     "banana":{"kcal":89, "carb_g":23, "protein_g":1.1,"fat_g":0.3,"sodium_mg":1,      "cat":"水果類",     "base_g":100, "tip":"控制份量，避免一次過量"},
     "miso_soup":{"kcal":36,"carb_g":4.3,"protein_g":2.0,"fat_g":1.3,"sodium_mg":550,  "cat":"湯品/飲品",  "base_g":200, "tip":"味噌湯偏鹹，建議少量品嚐"},
+    "salad": {"kcal":30,"carb_g":5,"protein_g":1.5,"fat_g":0.5,"sodium_mg":40,"cat":"蔬菜類","base_g":100,"tip":"少醬少油，優先清爽調味"},
+    "fish":  {"kcal":170,"carb_g":0,"protein_g":22,"fat_g":8,"sodium_mg":70,"cat":"豆魚蛋肉類","base_g":120,"tip":"蒸/烤/煎少油，避免重鹹醬汁"},
 }
 
 ALIASES = {
@@ -137,22 +139,69 @@ def detect_foods_from_text(text: str) -> List[str]:
 import re
 DEFAULT_BASE_G = 100
 STOPWORDS = {
+    # 英文
     "a","an","the","with","and","of","on","in","to","served","over","side","sides",
-    "plate","bento","box","set","dish","meal","mixed","assorted","fresh","hot","cold",
-    "grilled","roasted","fried","deep","steamed","boiled","braised","stir","stirred","sautéed",
-    "sauce","soup","salad","topped","seasoned","style","japanese","taiwanese","korean","chinese",
+    "set","dish","meal","mixed","assorted","fresh","hot","cold","topped","style","seasoned",
+    # 中文
     "便當","套餐","一盤","一碗","配菜","附餐","湯","沙拉","醬","佐","搭配","附","拌","炒","滷","炸","烤","蒸","煮"
 }
+COLOR_WORDS = {"white","black","red","green","yellow","orange","brown","purple","pink","golden"}
+UTENSILS = {"plate","bowl","tray","box","cup","glass","plateful","bento"}
+ADJ_MISC = {"piece","slice","fillet","serving","topped","mixed","assorted"}
+
+# 常見食物名詞（沒有就先列為候選）
+FOOD_LIKE = {
+    "salad","fish","chicken","beef","pork","shrimp","tofu","egg",
+    "rice","noodles","bread","soup","vegetables","veggies","fruit"
+}
+
+import re
+
 def extract_food_terms_free(text: str):
-    parts = re.split(r"(?:,|\.|;|\band\b|\bwith\b|\bserved with\b|\baccompanied by\b|\n)+", text, flags=re.I)
+    """
+    從 caption 中抽食物詞（允許未知）：
+    - 解析片語：piece/slice/fillet/serving of X → X
+    - 切片（逗號/and/with），去掉顏色、器皿、形容詞停用詞
+    - 取片尾名詞；若無，掃描整句抓常見食物名詞
+    - Alias → 主鍵；沒對到就保留原字（當未知）
+    """
+    t = text.strip().lower()
     hits = set()
+
+    # 1) 特例：「X of Y」→ 直接抓 Y
+    for pat in [r"(?:piece|slice|fillet|serving)\s+of\s+([a-z\u4e00-\u9fff]+)"]:
+        for m in re.findall(pat, t, flags=re.I):
+            y = m.strip()
+            if y in COLOR_WORDS or y in UTENSILS or y in ADJ_MISC or y in STOPWORDS:
+                continue
+            hits.add(ALIASES.get(y, y))
+
+    # 2) 片段切分（逗號、分號、and、with、換行）
+    parts = re.split(r"(?:,|;|\.|\band\b|\bwith\b|\n)+", t, flags=re.I)
     for p in parts:
-        if not p: continue
-        toks = re.findall(r"[A-Za-z\u4e00-\u9fff]+", p.lower())
-        toks = [w for w in toks if len(w) >= 2 and w not in STOPWORDS]
-        if not toks: continue
-        head = toks[-1]
+        if not p:
+            continue
+        # 擷取英/中文字
+        toks = re.findall(r"[a-z\u4e00-\u9fff]+", p)
+        # 過濾顏色/器皿/形容詞/停用詞
+        toks = [
+            w for w in toks
+            if w not in COLOR_WORDS
+            and w not in UTENSILS
+            and w not in ADJ_MISC
+            and w not in STOPWORDS
+            and len(w) >= 2
+        ]
+        if not toks:
+            continue
+        head = toks[-1]  # 片尾通常是名詞，如 "salad"/"fish"
         hits.add(ALIASES.get(head, head))
+
+    # 3) 萬一片段沒抓到，再從整句補常見食物名詞
+    for w in FOOD_LIKE:
+        if re.search(rf"\b{re.escape(w)}\b", t):
+            hits.add(ALIASES.get(w, w))
+
     return list(hits)
 
 def estimate_weight(name: str, plate_cm: int, portion: str) -> int: