fix(references): keep non-Latin reference labels

clean_label rejected any label without [A-Za-z0-9], so people whose
names render entirely in Cyrillic (or another non-Latin script) were
silently dropped from search-result references. Require a Unicode
letter or digit instead; pure punctuation is still rejected.

(cherry picked from commit f2d7d073cb5342d33eb315f2f5acda2b24d655e3)

Author: Fazliev

linkedin_mcp_server/scraping/link_metadata.py

@@ -358,7 +358,10 @@ def clean_label(value: str, kind: ReferenceKind) -> str | None:
         return None
     if len(value) > 80:
         return None
-    if not re.search(r"[A-Za-z0-9]", value):
+    # Require at least one letter or digit in ANY script — Cyrillic (and
+    # other non-Latin) profile names are valid labels. [^\W_] matches a
+    # Unicode word character that is not an underscore.
+    if not re.search(r"[^\W_]", value):
         return None
 
     return value

tests/test_link_metadata.py

@@ -257,6 +257,53 @@ def test_rejects_single_character_labels(self):
             }
         ]
 
+    def test_keeps_cyrillic_only_names(self):
+        # Regression: a non-Latin (Cyrillic) name must survive label
+        # cleaning. The alphanumeric guard uses a Unicode word check, not
+        # [A-Za-z0-9], so search results from RU/BY/other locales keep their
+        # person references instead of being dropped.
+        references = build_references(
+            [
+                {
+                    "href": "https://www.linkedin.com/in/margo-yunanova/",
+                    "text": "Маргарита Юнанова",
+                }
+            ],
+            "search_results",
+        )
+
+        assert references == [
+            {
+                "kind": "person",
+                "url": "/in/margo-yunanova/",
+                "text": "Маргарита Юнанова",
+                "context": "search result",
+            }
+        ]
+
+    def test_rejects_punctuation_only_labels_across_scripts(self):
+        # The Unicode word guard must still reject pure punctuation/symbols
+        # (no letter or digit in any script).
+        references = build_references(
+            [
+                {
+                    "href": "https://www.linkedin.com/in/williamhgates/",
+                    "text": "—·—",
+                    "aria_label": "Bill Gates",
+                }
+            ],
+            "main_profile",
+        )
+
+        assert references == [
+            {
+                "kind": "person",
+                "url": "/in/williamhgates/",
+                "text": "Bill Gates",
+                "context": "top card",
+            }
+        ]
+
     def test_preserves_words_starting_with_view(self):
         references = build_references(
             [