Labelling & Taxonomy
Labelling is what keeps a multi-domain knowledge base from bleeding across topics. Every chunk is assigned one or more labels drawn from the service-category taxonomy, and at query time the retriever uses those labels to return only on-topic chunks. Without labels, a question about one service could pull in unrelated chunks from another.
Where this lives now
The full labelling reference (developer and operator angles, including the
configuration table, output contract, file-scope labels, failure modes,
verification steps, and the document-repository /api/labels surface) has
moved to
RAG pipeline → Data labelling strategy.
Manage the taxonomy
Knowledge managers curate the service-category hierarchy through the Admin
Dashboard → Knowledge Hierarchy tab, which edits the same backend CRUD
endpoints (/api/service-categories/*) documented in the canonical reference.
For guidance on diagnosing label quality (per-chunk Final labels log entries,
the >80 % labelled threshold, probing the live vLLM, and the
document-repository label-tree CRUD), see the
Diagnosing label quality
section of the canonical reference.
For per-chunk log-schema details and the canonical message strings to grep for, see the Ingestion log reference.