fix: citation author parsing bug + replace broken NHS K&L Hub link

Citation fix (utils/citations.py):
- Add _is_initials_token() helper — detects PubMed/Europe PMC "Surname JK"
format where the last token is run-together uppercase initials (1-4 chars)
- Add _initials_from_str() helper — expands "JK" → "J. K." for Harvard/APA
- Apply fix to all four formatters: Harvard, APA 7th, Vancouver, AMA
- Before: "Smith JK" → "JK, S." (last token wrongly treated as surname)
- After: "Smith JK" → "Smith, J. K." (Harvard) / "Smith JK" (Vancouver/AMA)
- Also handles multi-word surnames: "van der Berg JK" → "van der Berg, J. K."

Link fix (streamlit_app.py):
- Replace broken NHS K&L Hub (Scottish-only .nhs.scot URL) with TRIP Database
- TRIP is free, no-login EBP search covering RCTs, guidelines, reviews
- URL: https://www.tripdatabase.com/search?criteria={query}

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

streamlit_app.py

@@ -465,8 +465,8 @@ with tab_search:
             use_container_width=True,
         )
         link_cols[2].link_button(
-            "🟢 NHS K&L Hub",
-            f"https://www.knowledgeandevidence.nhs.scot/search?query={q_enc}",
+            "🟢 TRIP Database",
+            f"https://www.tripdatabase.com/search?criteria={q_enc}",
             use_container_width=True,
         )
         link_cols[3].link_button(

utils/citations.py

@@ -74,6 +74,11 @@ def _apa_single(name: str) -> str:
         return f"{last.strip()}, {initials}"
     parts = name.split()
     if len(parts) >= 2:
+        # Detect PubMed/Europe PMC "Surname JK" format
+        if _is_initials_token(parts[-1]):
+            last = " ".join(parts[:-1])   # handles "van der Berg JK"
+            initials = _initials_from_str(parts[-1])
+            return f"{last}, {initials}"
         last = parts[-1]
         initials = _initials(" ".join(parts[:-1]))
         return f"{last}, {initials}"
@@ -147,6 +152,11 @@ def _vancouver_single(name: str) -> str:
         return f"{last.strip()} {initials}"
     parts = name.split()
     if len(parts) >= 2:
+        # Detect PubMed/Europe PMC "Surname JK" format
+        if _is_initials_token(parts[-1]):
+            last = " ".join(parts[:-1])
+            initials = parts[-1].upper()   # already run-together: "JK" stays "JK"
+            return f"{last} {initials}"
         last = parts[-1]
         initials = "".join(p[0].upper() for p in parts[:-1] if p)
         return f"{last} {initials}"
@@ -222,6 +232,12 @@ def _harvard_single(name: str) -> str:
         return f"{last.strip()}, {initials}"
     parts = name.split()
     if len(parts) >= 2:
+        # Detect PubMed/Europe PMC "Surname JK" format (e.g. "Smith JK")
+        # Without this check, "Smith JK" → last="JK", initials="S." → "JK, S." (wrong)
+        if _is_initials_token(parts[-1]):
+            last = " ".join(parts[:-1])   # handles multi-word surnames too
+            initials = _initials_from_str(parts[-1])   # "JK" → "J. K."
+            return f"{last}, {initials}"
         last = parts[-1]
         initials = _initials(" ".join(parts[:-1]))
         return f"{last}, {initials}"
@@ -290,6 +306,11 @@ def _ama_single(name: str) -> str:
         return f"{last.strip()} {initials}"
     parts = name.split()
     if len(parts) >= 2:
+        # Detect PubMed/Europe PMC "Surname JK" format
+        if _is_initials_token(parts[-1]):
+            last = " ".join(parts[:-1])
+            initials = parts[-1].upper()   # AMA uses run-together: "Smith JK"
+            return f"{last} {initials}"
         last = parts[-1]
         initials = "".join(p[0].upper() for p in parts[:-1] if p)
         return f"{last} {initials}"
@@ -306,6 +327,34 @@ def _initials(first_middle: str) -> str:
     return " ".join(p[0].upper() + "." for p in parts if p)
 
 
+def _initials_from_str(token: str) -> str:
+    """
+    Expand a run-together initials token from PubMed/Europe PMC into dotted
+    initials: "JK" → "J. K."  |  "AB" → "A. B."  |  "J" → "J."
+    """
+    return " ".join(c.upper() + "." for c in token if c.isalpha())
+
+
+def _is_initials_token(token: str) -> bool:
+    """
+    Return True when a name token looks like run-together initials rather than
+    a proper word — i.e. it is 1–4 uppercase letters with no vowels OR is
+    entirely uppercase and short.
+
+    Examples that return True : "JK", "AB", "J", "JKL", "ABCD"
+    Examples that return False: "Smith", "John", "van", "De", "O'Brien"
+    """
+    if not token or not token.isalpha():
+        return False
+    if len(token) > 4:
+        return False
+    # Single letter is always an initial
+    if len(token) == 1:
+        return token.isupper()
+    # Multi-char: must be all-uppercase (PubMed stores "Smith JK", not "Smith jk")
+    return token.isupper()
+
+
 def _clean_title(title: str) -> str:
     return title.strip().rstrip(".")