Files
stack/tests/rex/comments
kert 452f1587b8
Some checks are pending
CI / lint (push) Successful in 1m12s
CI / notebooks-smoke (push) Successful in 1m35s
Deploy / notebooks (push) Has been skipped
Deploy / zotero (push) Has been skipped
Deploy / docs (push) Has been skipped
Deploy / api (push) Has been skipped
Deploy / mc (push) Has been skipped
Infra CI / notebooks (push) Successful in 1m13s
Infra CI / zotero (push) Successful in 20s
Infra CI / docs (push) Successful in 1m43s
Infra CI / api (push) Successful in 59s
Infra CI / mc (push) Successful in 13s
Deploy / report (push) Successful in 13s
CI / test (push) Has started running
feat(comments): LLM position/theme classifier for rulemaking comments (refs #254)
rex.comments.classify: relevance filter (skin-sub terms, CTP, Q41xx/
C527x code families), forced-tool-call classification via the
provider-agnostic prisma.llm layer — position (5-point scale on the
reclassification proposal), themes (9 tags), stakeholder type (#255's
per-comment half), rule provisions (#256), and commenter/organization
from letterhead/signature (seeds #416). Enum-validated so a model
glitch can't contaminate the table.

dev/scripts/classify_comments.py: per-docket driver — reads comment
text from bib (combined.md note preferred, inline abstract fallback),
filters to the relevant subset (CMS-2025-0304: 384 of 14,092),
classifies resumably into a JSONL cache, loads
skin_subs.rulemaking_comments via duckdb_batch (replacing the DEMO
pilot rows).

18 unit tests (fake provider — no network). Live run smoke-tested to
the API boundary; the at-scale run is blocked on Anthropic API
credits.
2026-07-10 22:05:59 -04:00
..