Update app.py

app.py

@@ -58,7 +58,7 @@ MEDICAL_KEYWORDS = [
 ]
 
 CURRENT_INFO_RE = re.compile(r"\b(current|currently|latest|today|recent|recently|now|this week|this month|news|update|updated|president|prime minister|ceo|minister|king|ruler|who won)\b", re.IGNORECASE)
-RESEARCH_CUE_RE = re.compile(r"\b(find|search|look for|show me|get me|give me|locate|recommend|suggest|articles?|papers?|books?|journals?|studies|literature|research|systematic review|evidence|sources|database|databases|peer reviewed|open access)\b", re.IGNORECASE)
+RESEARCH_CUE_RE = re.compile(r"\b(find|search|look for|show me|get me|give me|locate|recommend|suggest|articles?|papers?|books?|journals?|studies|literature|research|systematic review|evidence|sources|database|databases|peer reviewed|open access|advances?|approach(es)?|method(s|ology)?|technique[s]?|application[s]?|framework[s]?|model[s]?|algorithm[s]?|publication[s]?|recent work|current state|state of the art)\b", re.IGNORECASE)
 LIBRARY_CUE_RE = re.compile(r"\b(library|librarian|borrow|loan|renew|fine|study room|room booking|reserve a room|account|my library|interlibrary|ill|khazna|orcid|open access|apc|refworks|libkey|hours|location|contact|visitor|alumni|database access|off campus|remote access)\b", re.IGNORECASE)
 MEDICAL_SEARCH_RE = re.compile(r"\b(pubmed|embase|cinahl|clinicalkey|cochrane|uptodate|medline|systematic review|clinical trial|biomedical literature|medical literature)\b", re.IGNORECASE)
 SOCIAL_RE = re.compile(r"^(hi|hello|hey|good morning|good afternoon|good evening|how are you|how old are you|who are you|what are you|are you a bot|are you a robot|what can you do|thanks|thank you|ok|okay|bye|goodbye|lol|haha|hehe|you are silly|are you silly|are you dumb|stupid bot|dumb bot|idiot bot|joke|tell me a joke)\b", re.IGNORECASE)
@@ -876,28 +876,26 @@ def build_vectorstore(docs, force_rebuild=False):
 # ===== TOOL: SEARCH PRIMO =====
 def _boolean_to_primo_params(boolean_query: str) -> str:
     """
-    Convert Boolean string to PRIMO multi-query parameter format.
-    ("artificial intelligence" OR "machine learning") AND ("cancer diagnosis")
-    → query=any,contains,artificial intelligence OR machine learning,AND
-      &query=any,contains,cancer diagnosis
-    Each top-level AND group becomes a separate query= parameter.
+    Convert Boolean string to PRIMO REST API multi-query format.
+    PRIMO API uses q= (not query=) and requires %20 encoded spaces.
+
+    ("AI" OR "ML") AND ("cancer") AND ("deep learning")
+    → q=any,contains,AI%20OR%20ML,AND&q=any,contains,cancer,AND&q=any,contains,deep%20learning
+
+    Single concept: q=any,contains,machine%20learning
     """
     from urllib.parse import quote
 
-    # Strip outer parens groups split by top-level AND
-    # First flatten quotes and clean
-    cleaned = boolean_query.strip()
+    # If no Boolean operators — simple single query
+    if not (re.search(r'\bAND\b', boolean_query) and '(' in boolean_query):
+        clean = boolean_query.replace('"', '').strip()
+        return f"q=any,contains,{quote(clean)}"
 
-    # Split on top-level AND (not inside parentheses)
+    # Split on top-level AND
     groups = []
     depth = 0
-    current = []
-    i = 0
-    tokens = re.split(r'(\(|\)|\bAND\b|\bOR\b)', cleaned)
-    # Simpler approach: split on AND at depth 0
     chunk = ""
-    depth = 0
-    for char in cleaned:
+    for char in boolean_query:
         if char == '(':
             depth += 1
             chunk += char
@@ -906,34 +904,33 @@ def _boolean_to_primo_params(boolean_query: str) -> str:
             chunk += char
         else:
             chunk += char
-        # Check for AND at depth 0
         if depth == 0 and chunk.upper().endswith(' AND '):
             groups.append(chunk[:-5].strip())
             chunk = ""
     if chunk.strip():
         groups.append(chunk.strip())
-
     if not groups:
-        groups = [cleaned]
+        clean = boolean_query.replace('"', '').strip()
+        return f"q=any,contains,{quote(clean)}"
 
-    # Clean each group: remove outer parens, strip quotes, normalise OR
     primo_params = []
     for i, group in enumerate(groups):
-        # Remove outer parentheses
         g = group.strip()
+        # Remove outer parentheses
         if g.startswith('(') and g.endswith(')'):
             g = g[1:-1].strip()
-        # Remove double quotes (PRIMO doesn't need them in query= param)
+        # Remove double quotes
         g = g.replace('"', '')
-        # Normalise spacing around OR
+        # Normalise OR spacing
         g = re.sub(r'\s+OR\s+', ' OR ', g).strip()
         if not g:
             continue
-        # All except the last get ,AND suffix
+        # URL encode properly — spaces become %20
+        encoded = quote(g)
         suffix = ',AND' if i < len(groups) - 1 else ''
-        primo_params.append(f"query=any,contains,{quote(g, safe=' OR')}{suffix}")
+        primo_params.append(f"q=any,contains,{encoded}{suffix}")
 
-    return '&'.join(primo_params) if primo_params else f"query=any,contains,{quote(cleaned)}"
+    return '&'.join(primo_params) if primo_params else f"q=any,contains,{quote(boolean_query.replace(chr(34), '').strip())}"
 
 
 async def tool_search_primo(query, limit=5, peer_reviewed=False, open_access=False, year_from=None, year_to=None):
@@ -1577,7 +1574,22 @@ def _looks_research_question(question: str) -> bool:
     q = question.lower()
     if RESEARCH_CUE_RE.search(q):
         return True
-    return bool(re.search(r'\bimpact of\b|\beffects? of\b|\bcauses? of\b|\brelationship between\b|\bliterature on\b', q))
+    # Natural academic questions that don't use explicit search trigger words
+    # e.g. "What are the recent advances in AI for cancer diagnosis?"
+    # e.g. "How effective are mRNA vaccines in treating lung cancer?"
+    # e.g. "What is the impact of climate change on water security in the UAE?"
+    academic_patterns = re.compile(
+        r"\b(impact of|effect[s]? of|role of|use of|application of|advance[s] in|"
+        r"development[s] in|progress in|challenge[s] of|benefit[s] of|"
+        r"causes? of|relationship between|comparison of|effectiveness of|"
+        r"how (effective|efficient|useful|does|do|can|is|are)|"
+        r"what (are the|is the) (recent|latest|current|new|emerging)|"
+        r"published in the last|in the last \d+ years?|"
+        r"systematic review|meta.?analysis|clinical trial[s]?|"
+        r"literature on|survey of|overview of|state of the art)\b",
+        re.IGNORECASE
+    )
+    return bool(academic_patterns.search(q))
 
 def _looks_medical_search(question: str) -> bool:
     q = question.lower()
@@ -1904,6 +1916,11 @@ RULES:
   3. "I am fed up with searching" alone (no specific topic/database) = social (expressing frustration).
   4. "I need help finding articles on diabetes" = search_medical.
   5. KU = Khalifa University. Do NOT confuse with Kuwait University.
+  6. ANY question asking about advances, approaches, methods, effectiveness, impact, or relationship
+     on a specific academic/scientific topic = search_academic or search_medical. NOT library_info.
+     e.g. "What are the recent advances in AI for predictive maintenance?" = search_academic
+     e.g. "How effective are mRNA vaccines for cancer treatment?" = search_medical
+     e.g. "What is the impact of climate change on food security?" = search_academic
 
 For "social" intent ONLY: also include "casual_answer" — a warm 1-3 sentence LibBee response
 (friendly librarian tone, no markdown, no bullet points). Offer to help with library services.