Update src/api/agent.py

src/api/agent.py

@@ -1219,6 +1219,9 @@ BOOLEAN RULES — follow strictly:
 - Max 3 AND groups. Each group max 4 OR terms.
 - NEVER put years as Boolean terms — put them in year_from/year_to
 - Keep broad enough to return results
+- NAMED ENTITIES: Any specific keyword that is a country, island, city, person,
+  organisation, or proper noun MUST be kept as its own AND group. Never drop it.
+  e.g. "Maldives" → AND ("Maldives" OR "Maldivian") ; "UAE" → AND ("UAE" OR "United Arab Emirates")
 
 EXAMPLES:
 Input: "find articles on global warming and renewable energy"
@@ -1229,6 +1232,14 @@ Input: "peer reviewed articles last 5 years on machine learning healthcare"
 → boolean:("machine learning" OR "deep learning" OR "AI") AND ("healthcare" OR "medicine" OR "clinical")
 → year_from:"{CURRENT_YEAR - 5}", peer_reviewed:true
 
+Input: "impact of global warming in Maldives"
+→ boolean:("global warming" OR "climate change") AND ("Maldives" OR "sea level rise" OR "small island")
+→ year_from:"", peer_reviewed:false
+
+Input: "water scarcity solutions in UAE"
+→ boolean:("water scarcity" OR "water stress") AND ("UAE" OR "United Arab Emirates" OR "Gulf region")
+→ year_from:"", peer_reviewed:false
+
 Input: "summarize shift in personalized genomic medicine from N-of-1 trials last 24 months"
 → boolean:("personalized genomic medicine" OR "precision medicine") AND ("N-of-1 trials" OR "scalable precision")
 → year_from:"{CURRENT_YEAR - 2}", peer_reviewed:true
@@ -1399,31 +1410,27 @@ async def _semantic_scholar_search(topic: str, limit: int = 8) -> List[dict]:
 
 async def _openalex_search(topic: str, limit: int = 8) -> List[dict]:
     """
-    Search OpenAlex using the ?search= relevance engine — same approach as the
-    clickable OpenAlex link shown to users, which consistently returns relevant results.
-
-    Previous approach used ?filter=title.search:X|abstract.search:X — a strict
-    field-level filter requiring exact token matches in title/abstract only, which
-    returned irrelevant or zero results. The ?search= parameter uses OpenAlex's
-    full relevance engine across all fields, identical to the browser link.
+    Search OpenAlex using the ?search= relevance engine.
+    This matches the clickable OpenAlex link shown to users.
+    The previous ?filter=title.search:X|abstract.search:X approach was a strict
+    field-level filter that returned irrelevant results. The ?search= parameter
+    uses OpenAlex full relevance ranking across all fields.
     """
     try:
-        from urllib.parse import quote
+        from urllib.parse import quote as _oa_quote
 
-        # Strip boolean operators and retrieval filler — keep only keywords.
-        # Matches what the user-facing OpenAlex link does.
+        # Strip boolean operators and common retrieval filler words.
         clean = re.sub(r'\b(AND|OR|NOT)\b|[()"\']', ' ', topic, flags=re.IGNORECASE)
         clean = re.sub(
-            r'\b(find|search|show|get|give|recent|research on|articles? on'
-            r'|papers? on|literature on|studies on|overview of|summary of|review of)\b',
+            r'\b(find|search|show|get|give|research on|articles on'
+            r'|papers on|literature on|studies on|overview of|summary of|review of)\b',
             ' ', clean, flags=re.IGNORECASE,
         )
         clean = re.sub(r'\s+', ' ', clean).strip()[:150]
-        encoded = quote(clean, safe='')
+        encoded = _oa_quote(clean, safe='')
 
         select = "id,title,authorships,publication_year,primary_location,abstract_inverted_index,doi,cited_by_count,open_access"
 
-        # FIX: use ?search= — same as the clickable OpenAlex link
         url = (
             f"https://api.openalex.org/works"
             f"?search={encoded}"
@@ -1581,8 +1588,8 @@ async def _search_primo_catalog(context: SearchContextPayload, limit: int = 5) -
         boolean = clean_topic or context.topic
     result = await search_primo(
         SearchRequest(
-            query=context.topic,       # plain topic as q= fallback
-            boolean_query=boolean,     # FIX: boolean goes here so _boolean_to_primo_params() fires
+            query=context.topic,
+            boolean_query=boolean,
             source="primo",
             limit=limit,
             peer_reviewed=context.peer_reviewed,
@@ -2382,7 +2389,7 @@ async def _research_snapshot(context: SearchContextPayload, model: str) -> Tuple
 
     # Parallel fetch: OpenAlex (fast, free) + PRIMO
     papers, (primo_result, primo_url) = await _asyncio.gather(
-        _openalex_search(context.ai_tool_query or context.topic, limit=8),  # FIX: richer natural query
+        _openalex_search(context.ai_tool_query or context.topic, limit=8),
         _search_primo_catalog(context, limit=6),
     )
 
@@ -2814,7 +2821,7 @@ async def _run_search_mode(
         if use_openalex:
             (result, source_url), oa_papers = await _aio.gather(
                 _search_pubmed_articles(context, limit=6),
-                _openalex_search(context.ai_tool_query or context.topic, limit=5),  # FIX: richer natural query
+                _openalex_search(context.ai_tool_query or context.topic, limit=5),
             )
         else:
             result, source_url = await _search_pubmed_articles(context, limit=6)
@@ -2824,7 +2831,7 @@ async def _run_search_mode(
         if use_openalex:
             (result, source_url), oa_papers = await _aio.gather(
                 _search_primo_catalog(context, limit=6),
-                _openalex_search(context.ai_tool_query or context.topic, limit=5),  # FIX: richer natural query
+                _openalex_search(context.ai_tool_query or context.topic, limit=5),
             )
         else:
             result, source_url = await _search_primo_catalog(context, limit=6)