Merge pull request 'fix(sources): resolve arXiv IDs via arXiv DOI (OpenAlex /works/arxiv: 404s)' (#5) from fix/openalex-arxiv-doi into main

fix(sources): resolve arXiv IDs via arXiv DOI (OpenAlex /works/arxiv: 404s)

codex/sources/openalex.py

@@ -25,17 +25,21 @@ _BASE = "https://api.openalex.org"
 
 
 def _resolve_id(id: str) -> str:
-    """Prefix a bare DOI or arXiv ID for the OpenAlex /works/ endpoint.
+    """Resolve a bare DOI or arXiv ID for the OpenAlex /works/ endpoint.
 
-    OpenAlex accepts: doi:<doi>, arxiv:<id>, W-IDs, or full https:// URLs.
-    Bare DOIs (starting with "10.") and bare arXiv IDs require a prefix.
+    OpenAlex resolves DOIs, W-IDs and full https:// URLs in the /works/ path,
+    but NOT the ``arxiv:`` namespace (that path 404s). arXiv papers are looked
+    up via their arXiv DOI ``10.48550/arXiv.<id>`` instead — accepted by
+    OpenAlex for both modern (``2301.07041``) and legacy (``math/0603097``) IDs.
     """
     s = id.strip()
-    if s.startswith(("W", "https://", "doi:", "arxiv:")):
+    if s.lower().startswith("arxiv:"):
+        s = s[len("arxiv:") :]
+    if s.startswith(("W", "https://", "doi:")):
         return s
     if s.startswith("10."):
         return f"doi:{s}"
-    return f"arxiv:{s}"
+    return f"doi:10.48550/arXiv.{s}"
 
 
 def _is_retryable(exc: BaseException) -> bool: