fix(dictionary): strip Latin romanization debris from Russian translations and sanitize dictionary cache

app/dictionary.py

@@ -223,6 +223,60 @@ async def fetch_free_dictionary_api(word: str, client: Optional[httpx.AsyncClien
         pass
     return {"phonetic": None, "audio_url": None}
 
+def clean_russian_translation(text: str) -> str:
+    """
+    Cleans up Russian translations by stripping Latin romanization / phonetic debris
+    (e.g. 'át’sjaпорхать át’/', 'krýl’jami/', 'ževáto-koríčnevyj/', '] ]/]')
+    and restoring clean, deduplicated Cyrillic phrases, while preserving grammatical notes.
+    """
+    if not text:
+        return ""
+
+    # Protect grammatical notes like (прош. вр. от go), (3-е л. ед. ч. от flutter)
+    notes = []
+    def save_note(m):
+        notes.append(m.group(0))
+        return f"«ПРИМЕЧАНИЕ_{len(notes)-1}»"
+
+    s = re.sub(r"\([а-яА-ЯёЁ0-9\s.,-]*\bот\s+[a-zA-Z\x27-]+\)", save_note, text)
+
+    # 1. Clean out brackets junk and stray backslashes
+    s = re.sub(r"\[+[\s\]/\\,;]*\]*", " ", s)
+    s = re.sub(r"[\]\[\\]+", " ", s)
+
+    # 2. Match Latin romanization fragments with accents/slashes/apostrophes
+    latin_rom = r"[a-zA-Z\u00C0-\u024F\u1E00-\u1EFF\x27\u2019/]+"
+    s = re.sub(rf"\s*{latin_rom}\s*", ", ", s)
+
+    # 3. Clean up multiple commas/semicolons
+    s = re.sub(r"[,;\s]*,[,;\s]*", ", ", s)
+
+    # 4. Clean and deduplicate phrases
+    items = []
+    seen = set()
+    for part in s.split(","):
+        p = part.strip(" /\\,;-\x27\u2019")
+        if p.count("(") > p.count(")"):
+            p += ")"
+        elif p.count(")") > p.count("("):
+            p = re.sub(r"\)+$", "", p).strip()
+        p = re.sub(r"\s+", " ", p)
+        for idx, n in enumerate(notes):
+            token = f"«ПРИМЕЧАНИЕ_{idx}»"
+            if token in p:
+                p = p.replace(token, n).strip()
+        if p and re.search(r"[\u0400-\u04FF0-9]", p):
+            p_lower = p.lower()
+            if p_lower not in seen:
+                seen.add(p_lower)
+                items.append(p)
+
+    res = ", ".join(items) if items else text.strip()
+    for idx, n in enumerate(notes):
+        res = res.replace(f"«ПРИМЕЧАНИЕ_{idx}»", n)
+    return res
+
+
 async def lookup_term(query: str) -> Dict[str, Any]:
     query_clean = query.strip()
     if not query_clean:
@@ -251,13 +305,14 @@ async def lookup_term(query: str) -> Dict[str, Any]:
                 ).fetchone()
                 if off_row:
                     raw_trans = off_row["translation"] or ""
-                    tr_list = [t.strip() for t in raw_trans.split(",") if t.strip()]
+                    clean_trans = clean_russian_translation(raw_trans)
+                    tr_list = [t.strip() for t in clean_trans.split(",") if t.strip()]
                     res = {
                         "query": query_clean,
                         "is_sentence": False,
-                        "translation": raw_trans,
+                        "translation": clean_trans,
                         "translations_list": tr_list,
-                        "translation_ru": raw_trans,
+                        "translation_ru": clean_trans,
                         "phonetic": off_row["phonetic"] or "",
                         "audio_file": "",
                         "audio_url": "",