luckiest-co 1.0.47 → 1.0.49

package/skills/trends/scripts/lib/arxiv.py

@@ -61,13 +61,21 @@ def _today() -> datetime:
     return datetime.now(timezone.utc)
 
 
-def _build_search_query(topic: str) -> str:
-    """Quote the topic so arXiv treats it as a phrase across all fields.
+def _build_search_query(topic: str, *, quoted: bool = True) -> str:
+    """Build the arXiv search-query string for ``topic``.
 
-    Inner double-quotes are stripped (arXiv has no phrase-escaping); the outer
-    quotes plus ``all:`` give a phrase-scoped relevance search.
+    Quoted (default): phrase-scoped exact match across all fields. Precise
+    for topics that genuinely appear as a phrase in a title/abstract, but a
+    natural-language multi-word topic ("AI video generation advances") almost
+    never appears verbatim, so it returns zero results (#908). Unquoted uses
+    an AND-conjoined clause for every individual term as a fallback retry.
+
+    Inner double-quotes are stripped (arXiv has no phrase-escaping) either way.
     """
-    return f'all:"{_clean_phrase(topic)}"'
+    phrase = _clean_phrase(topic)
+    if quoted:
+        return f'all:"{phrase}"'
+    return " AND ".join(f'all:"{term}"' for term in phrase.split())
 
 
 def _clean_phrase(topic: str) -> str:
@@ -75,12 +83,12 @@ def _clean_phrase(topic: str) -> str:
     return " ".join(topic.replace('"', " ").split())
 
 
-def _build_search_args(topic: str, limit: int) -> List[str]:
+def _build_search_args(topic: str, limit: int, *, quoted: bool = True) -> List[str]:
     return [
         CLI_BIN,
         "query",
         "--search-query",
-        _build_search_query(topic),
+        _build_search_query(topic, quoted=quoted),
         "--sort-by",
         "relevance",
         "--max-results",
@@ -118,13 +126,18 @@ def _run_cli(cmd: List[str], timeout: int) -> Dict[str, Any]:
 
     stdout = result.stdout or ""
     if not stdout.strip():
-        return {"results": []}
+        _log("CLI returned empty stdout")
+        return {"results": [], "error": "empty stdout"}
     try:
         data = json.loads(stdout)
     except json.JSONDecodeError as exc:
         _log(f"JSON decode failed: {exc}")
         return {"results": [], "error": f"json decode: {exc}"}
 
+    if not _is_entry_envelope(data):
+        _log("CLI returned an unrecognized JSON response")
+        return {"results": [], "error": "unrecognized JSON response"}
+
     return {"results": _extract_entries(data)}
 
 
@@ -150,6 +163,20 @@ def _extract_entries(data: Any) -> List[Dict[str, Any]]:
     return []
 
 
+def _is_entry_envelope(data: Any) -> bool:
+    """Return whether ``data`` has one of the supported entry-list shapes."""
+    if isinstance(data, list):
+        return True
+    if not isinstance(data, dict):
+        return False
+    results = data.get("results")
+    return (
+        isinstance(results, list)
+        or (isinstance(results, dict) and isinstance(results.get("entries"), list))
+        or isinstance(data.get("entries"), list)
+    )
+
+
 def search_arxiv(
     topic: str,
     from_date: str,
@@ -172,6 +199,13 @@ def search_arxiv(
     _log(f"query '{topic}' (relevance, max={limit})")
     response = _run_cli(cmd, timeout=SEARCH_TIMEOUT)
     _log(f"found {len(response.get('results') or [])} entries")
+    # Retry a clean zero-result phrase match with individually quoted AND terms.
+    # CLI failures, malformed responses, and missing binaries skip the retry.
+    if not response.get("error") and not response.get("results"):
+        retry_cmd = _build_search_args(topic, limit, quoted=False)
+        _log(f"quoted phrase matched nothing; retrying unquoted for '{topic}'")
+        response = _run_cli(retry_cmd, timeout=SEARCH_TIMEOUT)
+        _log(f"unquoted retry found {len(response.get('results') or [])} entries")
     return response