Update src/api/agent.py

src/api/agent.py

@@ -1399,27 +1399,35 @@ 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 title.search + abstract.search filters for precision.
-    Targets only title and abstract fields — more relevant than general search.
-    Free, no API key, typically < 2s.
+    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.
     """
     try:
         from urllib.parse import quote
 
-        # Strip boolean operators — OpenAlex filter search uses its own relevance engine
+        # Strip boolean operators and retrieval filler — keep only keywords.
+        # Matches what the user-facing OpenAlex link does.
         clean = re.sub(r'\b(AND|OR|NOT)\b|[()"\']', ' ', topic, flags=re.IGNORECASE)
-        clean = re.sub(r'\s+', ' ', clean).strip()[:200]
+        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',
+            ' ', clean, flags=re.IGNORECASE,
+        )
+        clean = re.sub(r'\s+', ' ', clean).strip()[:150]
         encoded = quote(clean, safe='')
 
-        # Build URL manually — urlencode encodes | and : which breaks OpenAlex filter syntax
-        filter_str = (
-            f"title.search:{encoded}|abstract.search:{encoded},"
-            f"has_abstract:true,type:article,is_paratext:false"
-        )
         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"?filter={filter_str}"
+            f"?search={encoded}"
+            f"&filter=has_abstract:true,type:article,is_paratext:false"
             f"&sort=relevance_score:desc"
             f"&per_page={limit}"
             f"&select={select}"
@@ -1429,38 +1437,11 @@ async def _openalex_search(topic: str, limit: int = 8) -> List[dict]:
         async with httpx.AsyncClient(timeout=10) as client:
             r = await client.get(url, headers=headers)
             if r.status_code != 200:
-                logger.warning(f"OpenAlex title+abstract search {r.status_code} — retrying with general search")
-                # Fallback: general search
-                url_fb = (
-                    f"https://api.openalex.org/works"
-                    f"?search={encoded}"
-                    f"&filter=has_abstract:true,type:article,is_paratext:false"
-                    f"&sort=relevance_score:desc"
-                    f"&per_page={limit}"
-                    f"&select={select}"
-                )
-                r = await client.get(url_fb, headers=headers)
-                if r.status_code != 200:
-                    logger.warning(f"OpenAlex fallback also failed: {r.status_code}")
-                    return []
+                logger.warning(f"OpenAlex search failed: {r.status_code}")
+                return []
 
             works = r.json().get("results", [])
 
-            # If title+abstract returns nothing, retry with general search
-            if not works:
-                logger.info("OpenAlex title+abstract returned 0 — retrying with general search")
-                url_fb = (
-                    f"https://api.openalex.org/works"
-                    f"?search={encoded}"
-                    f"&filter=has_abstract:true,type:article,is_paratext:false"
-                    f"&sort=relevance_score:desc"
-                    f"&per_page={limit}"
-                    f"&select={select}"
-                )
-                r2 = await client.get(url_fb, headers=headers)
-                if r2.status_code == 200:
-                    works = r2.json().get("results", [])
-
             results = []
             for work in works:
                 title = work.get("title") or "Untitled"
@@ -1600,8 +1581,8 @@ 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,       # plain topic as q= fallback
+            boolean_query=boolean,     # FIX: boolean goes here so _boolean_to_primo_params() fires
             source="primo",
             limit=limit,
             peer_reviewed=context.peer_reviewed,
@@ -2401,7 +2382,7 @@ 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: use richer natural query
+        _openalex_search(context.ai_tool_query or context.topic, limit=8),  # FIX: richer natural query
         _search_primo_catalog(context, limit=6),
     )
 
@@ -2833,7 +2814,7 @@ 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: use richer natural query
+                _openalex_search(context.ai_tool_query or context.topic, limit=5),  # FIX: richer natural query
             )
         else:
             result, source_url = await _search_pubmed_articles(context, limit=6)
@@ -2843,7 +2824,7 @@ 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: use richer natural query
+                _openalex_search(context.ai_tool_query or context.topic, limit=5),  # FIX: richer natural query
             )
         else:
             result, source_url = await _search_primo_catalog(context, limit=6)