Apple's Glyph uses LLM agents to tag enterprise data columns

Enterprise data lakes accumulate tables faster than human stewards can document or classify them, a documentation debt that the authors say undermines data discovery, access control, and regulatory compliance. Glyph addresses this by framing two coupled problems, column description generation and column type annotation for data classification, as cooperating LLM agents orchestrated as stateful graphs.

To write descriptions, the Descriptor grounds generation in the pipeline source code that produces each column, retrieved on demand from an enterprise GitHub via a reasoning-acting tool loop (active Retrieval-Augmented Generation). For labeling, the Tagger assigns labels from a governed 275-leaf Data Classification Ontology by running three complementary strategies in parallel: a description tagger, a line-of-business regex tagger, and a metadata tagger backed by a fine-tuned contrastive encoder over a vector database. Their ranked outputs are fused with Reciprocal Rank Fusion (RRF).

The system's engineering decisions are meant to distinguish it from prior column-type-annotation work and from commercial value/regex sensitivity scanners: a value-free and code-grounded design, per-tag provenance, and graceful degradation. The main test will be whether the reported recall-weighted F2 results and ablation studies translate into dependable cataloging for data stewards and governance teams.