Bounded request examples
Documentation source:
collector/examples/README.mdat revisiond127243ba072cb8572df536d40a31be3935e0106.
Every file is non-secret, validates offline, creates a new download when collected, and leaves raw
and processed outputs under .download/ until a maintainer approves promotion. Live record
contents can change; the expected result below describes boundary and evidence, not fixed titles
or counts beyond the declared maximum.
| Request | Prerequisites and scope | Expected result and caveat |
|---|---|---|
openalex-author.json | OPENALEX_API_KEY; one OpenAlex author ID | At most one author record under the new run directory; algorithmic identity and changing affiliations are limitations. |
openalex-work.json | OPENALEX_API_KEY; one OpenAlex work ID | At most one source-native work record; available abstract/citation fields remain OpenAlex assertions. |
dblp-person.json | No credential; one stable DBLP PID | One person/profile assertion with linked publication keys; DBLP's CS-centered coverage is explicit. |
dblp-publication.json | No credential; bounded query, maximum five | Zero through five publication results with a completion summary; DBLP supplies no abstracts. |
orcid-record.json | ORCID_CLIENT_ID and ORCID_CLIENT_SECRET; one ORCID iD | Public record fields only; absence may reflect visibility and the release defaults to metadata-only. |
arxiv-preprint.json | No credential; one arXiv ID | At most one preprint with ordered author assertions; no researcher profile, PDF, or source archive. |
arxiv-category-cs-ai.json | No credential; complete cs.AI OAI-PMH set | All metadata records exposed in the category when the set is exhausted; the null record limit is allowed only for arXiv OAI-PMH, and collection remains sequential and resumable. |
Validate all examples without credentials or network access:
for request_file in examples/requests/*.json; do
uv run --locked scope-collect validate-request "$request_file"
done
For fixture-backed expected behavior, run uv run --locked pytest -m "not live" -q. A live
download writes raw pages, checkpoints, and an input-bundle seal to
.download/downloads/<download-id>/. A separately selected recipe writes a reviewable candidate to
.download/processing/<processing-id>/candidate/. Status 0 is complete, 3 is safely resumable, and
other exit meanings are documented in the package README. No example processes, promotes, stages,
or commits data automatically.
DBLP examples cover stable person PID lookup (dblp-person.json), bounded person discovery
(dblp-person-search.json), and bounded publication discovery (dblp-publication.json).
ORCID examples cover public records (orcid-record.json) and public work assertions
(orcid-work.json).