The entity, relationship, and scoring vocabulary behind /v1/extract, /v1/career-graph, and /v1/job-match — published openly so other tools can build against it as a shared standard, not just a private implementation detail.
curl https://api.careerstudiomax.com/api/transformer/v1/ontology
The 8 entity types /v1/extract recognizes. Career-graph nodes use a 6-type subset — DATE and SALARY are modeled as properties, not standalone nodes.
Directed edges between entities, with their expected domain (source type) and range (target type).
| Relation | Domain | Range |
|---|
Every entity and relationship from /v1/extract carries its own confidence — relationships also carry evidence, a verbatim quote from the source text. Every node and edge in a /v1/career-graph result carries a provenance object.
The five dimensions every /v1/career-graph result scores, each 0–100.
The verdict enum and the four dimensions a /v1/job-match score breaks down into.
Published from day one, not added after the first breaking change.
Scheme: semver (MAJOR.MINOR.PATCH).
Adding a new entity type, relationship type, or score dimension is additive — MINOR version bump, nothing breaks.
Renaming or removing one is MAJOR, with a minimum 90-day deprecation window: both the old and new identifiers are accepted, and affected API responses carry a deprecation signal — the same pattern already used for this API's own model-name migrations.
PATCH covers description/documentation wording only — no schema effect, safe to ignore.