Update app.py

app.py

@@ -304,6 +304,136 @@ def _load_staff_directory_from_kb():
 def _staff_lookup_candidates():
     return kb_staff_directory or STAFF_DIRECTORY
 
+def _role_key(name: str) -> str:
+    return re.sub(r"[^a-z0-9]+", " ", (name or "").lower()).strip()
+
+def _build_role_aliases(staff: dict):
+    role = (staff.get("role") or "").strip()
+    details = (staff.get("details") or "").strip()
+    full_name = staff.get("full_name", "")
+    key = _role_key(full_name)
+
+    aliases = []
+    if role:
+        aliases.append(role)
+        role_l = role.lower()
+        parts = [p.strip() for p in re.split(r"[\/,]|\band\b", role_l) if p.strip()]
+        aliases.extend(parts)
+        if "manager" in role_l:
+            mgr_tail = re.sub(r"^manager\s*,?\s*", "", role_l).strip()
+            if mgr_tail:
+                aliases.extend([
+                    mgr_tail,
+                    f"{mgr_tail} librarian",
+                    f"{mgr_tail.rstrip('s')} librarian",
+                    f"{mgr_tail} manager",
+                ])
+        if "librarian" in role_l:
+            base = role_l.replace("librarian", "").replace("/", " ").strip(" ,")
+            if base:
+                aliases.extend([
+                    base,
+                    f"{base} librarian",
+                    f"{base.rstrip('s')} librarian",
+                ])
+
+    best_for = ""
+    m = re.search(r"Best for:\s*([^|]+)", details, re.IGNORECASE)
+    if m:
+        best_for = m.group(1).strip()
+    if best_for:
+        aliases.extend([p.strip() for p in best_for.split(",") if p.strip()])
+
+    manual = {
+        "nikesh narayanan": [
+            "research librarian", "access services librarian", "research and access services librarian",
+            "research access librarian", "open access librarian", "orcid librarian",
+            "research impact librarian", "bibliometrics librarian", "scholarly communication librarian"
+        ],
+        "walter brian hall": [
+            "systems librarian", "system librarian", "library systems librarian",
+            "digital services librarian", "technology services librarian", "website librarian",
+            "technology librarian", "digital librarian"
+        ],
+        "rani anand": [
+            "e resources librarian", "e-resources librarian", "electronic resources librarian",
+            "database librarian", "database access librarian", "resource access librarian"
+        ],
+        "jason fetty": [
+            "medical librarian", "health sciences librarian", "clinical librarian",
+            "systematic review librarian", "pubmed librarian"
+        ],
+        "alia al harrasi": [
+            "acquisitions librarian", "acquisition librarian", "technical services librarian",
+            "technical service librarian", "collection development librarian", "cataloguing librarian",
+            "metadata librarian"
+        ],
+        "muna ahmad mohammad al blooshi": [
+            "public services librarian", "public service librarian", "circulation librarian",
+            "access services manager", "service desk librarian"
+        ],
+        "dr abdulla al hefeiti": [
+            "library director", "director of library", "director libraries", "assistant provost libraries"
+        ],
+        "abdulla al hefeiti": [
+            "library director", "director of library", "director libraries", "assistant provost libraries"
+        ],
+    }
+    aliases.extend(manual.get(key, []))
+
+    normalized = []
+    for a in aliases:
+        a = re.sub(r"[^a-z0-9&/ +()-]+", " ", (a or "").lower())
+        a = re.sub(r"\s+", " ", a).strip(" ,")
+        if a:
+            normalized.append(a)
+            if a.endswith(" services"):
+                normalized.append(a[:-1])
+            if a.endswith(" service"):
+                normalized.append(a + " librarian")
+            if a.endswith(" resources"):
+                normalized.append(a[:-1])
+    return _dedupe_keep_order(normalized)
+
+def _match_staff_role(question: str):
+    ql = re.sub(r"[^a-z0-9 ]+", " ", (question or "").lower())
+    ql = re.sub(r"\s+", " ", ql).strip()
+    if not ql:
+        return None
+
+    role_trigger_re = re.compile(
+        r"\b(librarian|library director|director|manager|services?|service|systems?|technology|website|research|access|open access|orcid|bibliometric|bibliometrics|public services?|technical services?|acquisitions?|collection development|catalogu(?:e|ing)|metadata|database|e resources|e-resources|electronic resources|medical|clinical|systematic review|circulation)\b"
+    )
+    if not role_trigger_re.search(ql):
+        return None
+
+    best = None
+    best_score = 0
+    for staff in _staff_lookup_candidates():
+        score = 0
+        for alias in _build_role_aliases(staff):
+            if not alias:
+                continue
+            alias_words = alias.split()
+            if alias in ql:
+                score = max(score, 100 + len(alias_words))
+            elif len(alias_words) >= 2 and all(w in ql.split() for w in alias_words):
+                score = max(score, 70 + len(alias_words))
+            else:
+                overlap = sum(1 for w in alias_words if len(w) > 2 and w in ql.split())
+                if overlap >= 2:
+                    score = max(score, 40 + overlap)
+        role_words = _normalize_name_query(staff.get("role", ""))
+        if role_words:
+            overlap = sum(1 for w in role_words if len(w) > 2 and w in ql.split())
+            if overlap >= 2:
+                score = max(score, 20 + overlap)
+        if score > best_score:
+            best_score = score
+            best = staff
+
+    return best if best_score >= 42 else None
+
 def _match_staff_name(question: str):
     tokens = _normalize_name_query(question)
     if not tokens or len(tokens) > 5:
@@ -327,6 +457,12 @@ def _staff_name_answer(staff: dict, partial: str) -> str:
         f"<strong>{staff['full_name']}</strong> is the <strong>{staff['role']}</strong>. {staff['details']}"
     )
 
+def _staff_role_answer(staff: dict, question: str) -> str:
+    return (
+        f"For <strong>{question}</strong>, the best match is <strong>{staff['full_name']}</strong> — "
+        f"<strong>{staff['role']}</strong>.<br><br>{staff['details']}"
+    )
+
 
 GROUNDED_LIBRARY_MAP = {
     "ill": "interlibrary loan ILL document delivery full text unavailable article not available borrow from another library",
@@ -1357,7 +1493,7 @@ async def agent_query(req: AgentRequest):
             "source_mode": "social",
         }
 
-    # ---- Partial staff-name handling ----
+    # ---- Staff direct matching (name first, then role semantics) ----
     staff_match = _match_staff_name(question)
     if staff_match:
         answer = _staff_name_answer(staff_match, question)
@@ -1380,6 +1516,28 @@ async def agent_query(req: AgentRequest):
             "source_mode": "staff_kb" if kb_staff_directory else "staff_directory",
         }
 
+    staff_role_match = _match_staff_role(question)
+    if staff_role_match:
+        answer = _staff_role_answer(staff_role_match, question)
+        elapsed = time.time() - start
+        source_title = staff_role_match.get("source_title", "")
+        source_url = staff_role_match.get("source", "")
+        return {
+            "answer": answer,
+            "intent": "library_info",
+            "tools_used": ["staff_role_match"],
+            "search_results": [],
+            "sources": ([{"title": source_title, "source": source_url}] if source_title or source_url else []),
+            "model_used": req.model,
+            "response_time": round(elapsed, 2),
+            "corrected_query": question,
+            "natural_query": question,
+            "database_query": question,
+            "original_question": question,
+            "is_follow_up": False,
+            "source_mode": "staff_kb" if kb_staff_directory else "staff_directory",
+        }
+
     # ---- Follow-up to the greeting menu ----
     if _is_greeting_menu_followup(question, history):
         answer = _greeting_menu_clarify_answer()