Generate a proposed hierarchical tag taxonomy for a content source WITHOUT
persisting anything.
Primary path: embed a sample of the source's content items, cluster them via
ClusteringEngine (which names clusters with an LLM), and turn each
named cluster into a top-level taxonomy node.
Fallback path: when no embeddings can be produced (no embedding model, empty
texts, etc.), run the "Generate Seed Taxonomy" AI prompt over a sample of the
source's content item texts and use its proposed tree.
Parameters
sourceID: string
The ContentSource ID whose items seed the taxonomy.
Generate a proposed hierarchical tag taxonomy for a content source WITHOUT persisting anything.
Primary path: embed a sample of the source's content items, cluster them via ClusteringEngine (which names clusters with an LLM), and turn each named cluster into a top-level taxonomy node.
Fallback path: when no embeddings can be produced (no embedding model, empty texts, etc.), run the "Generate Seed Taxonomy" AI prompt over a sample of the source's content item texts and use its proposed tree.