omniseek_coauthors
Map co-authorship networks from author names or IDs to reveal advisors, top collaborators, pairwise connections, bridges, and sub-communities from joint-paper counts.
Instructions
Use WHEN you want WHO a researcher collaborates with — advisor + closest collaborators by joint-paper count, or how a paper's author group is connected (WebSearch cannot build this). One LAYER, not the whole graph — co-authorship is one edge type; YOU overlay the others (advising, institution cohort, citation, code, social) and judge what each connection MEANS.
Pass author NAMES and/or ids (from omniseek_resolve_identity). A brand-new arXiv paper is not in the graph yet, so this reconstructs from each author's PRIOR work: • N=1 -> that author's frequency-ranked coauthor neighborhood. The advisor + closest collaborators surface by joint-paper count (e.g. Yi R. Fung -> Heng Ji ~51x = her PhD advisor, no advisor field needed — YOU read that signal). • N>1 (e.g. a paper's whole author list) -> additionally the PAIRWISE prior joint-work edges among them (with the actual joint paper titles as evidence) + BRIDGE collaborators (people who co-authored with >=2 of the inputs but are not in the set). This is the "how is this author group actually connected" reconstruction.
Each input may be a NAME, an id, or '+'-joined ids ("id1+id2") for ONE person SPLIT
across ids — their works are MERGED (OpenAlex/S2 routinely split a junior's recent papers;
merging recovers the complete network). Each becomes a node with resolved,
ambiguous + alternatives (juniors often need source="s2", a paper anchor, or an
explicit id — the node note says so when unresolved). The output also carries cooc:
which of the network's top external coauthors co-appear on the same papers, i.e. the
SUB-COMMUNITY structure (an ego's distinct 'research worlds'). Mechanical throughout:
"these two share these N papers" is a fact; advisor-vs-peer, what a cluster MEANS, is YOUR
judgment. For the citation/influence layer use omniseek_field_skeleton; for the others, assemble
from the dossier recipe (github, bluesky, exa, cdp_fulltext, omniseek_read).
hints / papers are parallel lists for per-author disambiguation (an institution
hint, or a known paper that pins a common-name junior).
Returns: {source, n_authors, nodes:[{query, resolved, ambiguous, alternatives, works_seen,
top_coauthors:[{id,name,joint}], degraded?}], edges:[{a,b,joint_count,papers:[{title,year,id}]}],
bridges:[{id,name,shared_by,total_joint}], cooc:[{a,b,n}], degraded?}. (top_coauthors/bridges
carry a representative id you can harvest and pass back to omniseek_coauthors to drill that
person.) A top-level/node degraded means that author's OpenAlex lookup FAILED (rate-limited
/ upstream down): an empty graph is then missing-data to RETRY, not "no collaborators".
Input Schema
| Name | Required | Description | Default |
|---|---|---|---|
| hints | No | ||
| papers | No | ||
| source | No | openalex | |
| authors | Yes |