Skip to content

us-house-bills: 119th Congress rolling-window bills pack - #12

Open
JSv4 wants to merge 3 commits into
mainfrom
feat/us-house-bills-pack
Open

us-house-bills: 119th Congress rolling-window bills pack#12
JSv4 wants to merge 3 commits into
mainfrom
feat/us-house-bills-pack

Conversation

@JSv4

@JSv4 JSv4 commented Aug 26, 2026

Copy link
Copy Markdown
Contributor

New base pack for proposed Federal legislation (House side), designed against the OpenContracts house-bills core seams (Open-Source-Legal/OpenContracts#2279).

What's in the pack

  • hr namespace (authority_type: bill — a proposed bill is not a statute), keys hr:119-<number>; unqualified grammar keys (hr:1234) fold onto the current congress via equivalence rows (seeded here, continuously upserted by the ingest drain, re-pointed each congress).
  • Corpus house-bills-119 at PROPOSED weight: persona (proposed-not-law, rolling-window silence semantics, version naming, AMENDS-walk methodology), charter, metadata schema (13 Fieldset columns), 6-bill live seed built from govinfo bulkdata on 2026-08-26 with verified AMENDS/CITES relationships parsed from GPO's parsable-cite markup.
  • providers/house_bill_provider.py — on-demand gap-fill for cited-but-absent bills (frontier/crawl path). Status-driven version choice (BILLSTATUS textVersions, not listing mtime — GPO re-touches old files). Vendored stdlib-only converter at pack root, relative imports per the copy-to-port rule.
  • validate_pack.py: bill added to the authority-type vocabulary (mirror of core's AUTHORITY_TYPE_BILL).

Continuous content

The private us-house-bill-feed repo harvests the rolling window (BILLS + BILLSTATUS bulkdata listings, lastModified-driven) and pushes to POST /api/worker-uploads/authority-sections/ under a corpus token.

Validation

  • scripts/validate_pack.py us-house-bills → OK
  • Canonical load_authority_pack --check → pack valid, 1 corpus converges
  • Real install on the seams branch: corpus public, persona applied (1,998 chars), hr namespace corpus-bound, 6 equivalences, 22 relationships all verified=True, Fieldset created, 6 documents
  • Provider discovered via AUTHORITY_PACK_PATHS, can_handle accepts hr:119-41 / refuses hr:41, locate plan correct

Depends on OpenContracts PR #2279 (grammar + vocabulary + ingest endpoint).

…ht, verified AMENDS edges

- New base pack us-house-bills: hr namespace (authority_type bill), corpus
  house-bills-119 with persona/charter/metadata schema, 6-bill live seed,
  and an on-demand HouseBillSourceProvider (govinfo bulkdata, status-driven
  version choice).
- validate_pack.py: add 'bill' to the authority-type vocabulary (mirrors
  OpenContracts AUTHORITY_TYPE_BILL from the house-bills core seams).
- Continuous content path: Open-Source-Legal/us-house-bill-feed pushes
  rolling-window payloads to /api/worker-uploads/authority-sections/.

Validated: scripts/validate_pack.py OK; canonical load_authority_pack --check
and a real install verified against OpenContracts feature/us-house-bills-pack
(corpus+persona+namespace+equivalences+22 verified relationships+Fieldset+6
docs; provider discovered via AUTHORITY_PACK_PATHS, locate plan correct).
JSv4 added 2 commits August 26, 2026 17:15
Live GUI walkthrough found the corpus agent declaring H.R. 41 outside the
rolling window while the bill sat in the corpus: 'H.R. 41' is an identifier,
not a meaning, so semantic search missed and the silence rule (correctly)
refused to invent an answer — from a wrong premise. Rule 2a requires an
exact/title lookup before any absence conclusion.
gpt-4.1 read the silence rule and answered 'outside the window' for a
bill sitting in the corpus, without one tool call (usage: requests=1,
tool_calls=0). The exact-lookup guidance must be a hard precondition,
not advice.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant