Congress from GovInfo

By default, Congress (US) syncs from LegiScan, like every other state. This page covers the opt-in alternative: the govinfo provider, which reads the official, public-domain Bill Status data that GovInfo publishes for the Government Publishing Office (GPO) and the Library of Congress. It spends no LegiScan calls on Congress, and it adds what LegiScan's Congress data lacks: Congressional Research Service summaries, the official relationship between identical House and Senate bills, member votes named by Bioguide id, and committee codes that stay the same from one Congress to the next.

Nothing here is required. A deployment that does nothing keeps syncing Congress from LegiScan, and the rest of this page doesn't apply to it. The provider guide explains how any provider is claimed, and docs/internal/sync-pipeline.md in the repository has the maintainer's detail.

What it needs

• Bills and votes need no key. Bills come from GovInfo's Bill Status bulk files. Each House and Senate roll call a bill links comes from the House Clerk's and the Senate's own files, which name every member's vote.

• Committee meetings need an optional key. Hearings, markups, and meetings come from the Congress.gov API, which needs a free api.data.gov key. Without one, Congress's bills and votes still sync, and every calendar for Congress stays empty. Sign up at api.congress.gov/sign-up, then set it as a secret on central, from inside central/:

  wrangler secret put CONGRESS_API_KEY --env legiscan

  Use your central's environment name if it isn't legiscan. Central sends the key only in an X-Api-Key header to api.congress.gov, never in an address, a log, or anything it stores. It asks for no more than 10 requests a minute.

Turn it on when no one tracks Congress yet

If no instance has Congress bills linked, a claim is enough. Set the key first if you want the calendar (see Committee meetings), then claim US with the admin secret:

curl -X POST "$CENTRAL/api/admin/state-providers/US" \
  -H "x-admin-secret: $ADMIN_SECRET" -H "content-type: application/json" \
  -d '{"provider":"govinfo"}'

Central answers 200, and the state starts syncing from GovInfo once at least one active instance covers US. The hourly job does its first full read at its next full pass (5, 13, and 23 Eastern), or you can start it now with POST /api/admin/providers/govinfo/sync, which reads in full. A full read of the 119th Congress downloads eight zip files, 54 MB in all, and takes a few seconds of CPU, so no cpu_ms limit needs to change.

If an instance does track Congress bills from LegiScan, central answers 409 and names the instances and how many bills they track. The row doesn't change. Only a cutover moves those bills.

Move a deployment that already tracks Congress

A cutover moves Congress's bills and legislators to GovInfo without writing to any instance's database. Each matched bill keeps its row and its handle, so every instance's copy and every link to it stay as they are. There is no dashboard control for it: the claim and the cutover are admin routes, called with x-admin-secret, and the States page on the dashboard shows the result ("Claimed" or "Cut over", with the date).

1. Set CONGRESS_API_KEY first, if you want committee meetings. Central starts reading meetings only once GovInfo owns US. Setting the key before you switch starts the meeting read at the switch and shortens the window described under Committee meetings.

2. Run the dry run. It reads GovInfo's files for the current Congress and matches them to the bills and legislators central holds, and it writes nothing but the API call log:

   curl -X POST "$CENTRAL/api/admin/state-providers/US/cutover" \
     -H "x-admin-secret: $ADMIN_SECRET" -H "content-type: application/json" \
     -d '{"provider":"govinfo"}' > us-cutover-dry-run.json

3. Read the report. It lists the bills that matched and the ones that didn't on each side, the legislators matched by Bioguide id and by name (name matches are weak when chamber and district don't both confirm), any ambiguous or conflicting matches, the tracked bills whose keywords the new titles won't match, and, when committee meetings have been read, the calendar entries that keep their identity, are cancelled, or are added. Legislators are matched only through the sponsors of the bills the plan reads, so a legislator with no tracked bill is not in the report and gets a second people row later.

4. Confirm. The same request with ?confirm=true carries out the cutover and starts GovInfo's first sync. The response says how many bills and legislators moved, and whether that sync worked. If it reports an error, the cutover stands, and the hourly sync tries again. That sync covers bills only. Committee meetings start with the next hourly body-event job, so run POST /api/admin/providers/govinfo/sync once right after you confirm, which also starts the meeting read and shortens the window described under Committee meetings.

5. Undo it if needed. POST /api/admin/state-providers/US/undo-cutover is a dry run too, and ?confirm=true puts Congress back on LegiScan.

The first sync queues a first ingest for every moved bill. GovInfo has no ingest queue of its own, so those ingests share central's queue with LegiScan's bills and take a while to drain. They make no LegiScan calls. After a cutover, ask the admin of each instance that covers Congress to re-pick the committees in their event rules, because GovInfo mints its own committee ids. The sync pipeline notes cover what a cutover does to calendars, ids, and queues in full.

Committee meetings and the first read

Central reads committee meetings once for each state GovInfo owns that at least one instance covers, and then keeps them in its provider_cache table, so it never reads a meeting twice. The 119th Congress had 2,767 meetings in October 2026. A pass reads at most 60 meeting details an hour, paced at 10 requests a minute, so the first full read takes about one to two days after the key is set and US is switched.

Until that read finishes, GovInfo can't say a bill has no meeting, because the meeting may be one not read yet. Central marks the read as finished by writing a warm/<congress> row for govinfo into provider_cache (warm/119 for the 119th Congress), and until it exists:

• Bills sync with no calendar at all, which central treats as no word on the entries a bill already has.
• A bill that moved in a cutover keeps the hearings it carried from LegiScan, but only for so long. After three ingests with no calendar, central stops waiting and cancels them, so a busy bill that is ingested that often before the read finishes loses its LegiScan hearings. The meetings the read finds are added to its calendar afterward.

Check for the row with:

npx wrangler d1 execute central-bills-ls --remote --env legiscan \
  --command "SELECT key, updated_at FROM provider_cache WHERE provider = 'govinfo' AND key LIKE 'warm/%'"

Use your central's database and environment names. When the row exists, every record changes once, so tracked bills are ingested again, now with their calendars. A meeting that can't be read three passes in a row is set aside and the operators get one email, and it counts as read for this purpose, so one broken meeting can't hold every calendar back.

The deadline

Do the cutover before January 3, 2027. Two things happen around then:

• The 120th Congress begins at noon on January 3, 2027. From that day GovInfo lists the 120th Congress and no longer the 119th, so a cutover plans against 120th bills only. LegiScan's 119th bills would stay with LegiScan, frozen. Finish before the new Congress begins.
• LegiScan creates its 120th Congress session around the same day. Each session gets a URL slug that is unique within its state, and a slug never changes once assigned. Whichever session is created first takes the plain slug, 120th-congress, and the other gets 120th-congress-2. Once GovInfo owns US, LegiScan no longer syncs it, so a cutover that lands before LegiScan creates the 120th leaves the 120th Congress the plain slug. If LegiScan has already created its session, that frozen session holds the plain slug, and every 120th Congress link ends in -2.

The session slug

A cutover doesn't hand LegiScan's session slug to GovInfo. LegiScan's 119th Congress session keeps 119th-congress, and GovInfo's 119th Congress session becomes 119th-congress-2, so a moved bill's URL changes to /US/119th-congress-2/HB1 the next time it reaches an instance. Old links keep working: each instance falls back to the slug the session's name asks for, as long as no other bill answers to the same address. The same happens on a claim if LegiScan has synced Congress here, since its session already holds the plain slug. A deployment that never did gives GovInfo's session 119th-congress.

The credit

An instance whose coverage includes US shows this credit in its footer's "Data:" line once US is claimed, beside the credits for its other states' providers:

GovInfo Bill Status (GPO and Library of Congress), public domain; roll calls from the House Clerk and the Senate

The name links to GovInfo's Bill Status bulk data. It names the sources and doesn't imply that FloorVote publishes official data. Until US is claimed, the instance credits LegiScan for Congress instead.