extraction.json Fields

Complete reference for every field in extraction.json — the primary output of the extract command and the file all query commands read.

Top-Level Structure

{
  "schema_version": "1.0",
  "bill": { ... },
  "provisions": [ ... ],
  "summary": { ... },
  "chunk_map": [ ... ]
}

BillInfo (bill)

Example (H.R. 9468):

{
  "identifier": "H.R. 9468",
  "classification": "supplemental",
  "short_title": "Veterans Benefits Continuity and Accountability Supplemental Appropriations Act, 2024",
  "fiscal_years": [2024],
  "divisions": [],
  "public_law": null
}

Provisions (provisions)

An array of provision objects. Each provision has a provision_type field that determines which type-specific fields are present, plus the common fields shared by all types.

See Provision Types for the complete type-by-type reference including type-specific fields and examples.

Common Fields (All Provision Types)

CrossReference

Amount

Dollar amounts appear throughout the schema — on appropriation, rescission, limitation, directed_spending, mandatory_spending_extension, and other provision types. CR substitutions have new_amount and old_amount instead of a single amount.

Each amount has three sub-fields:

AmountValue (value)

Tagged by the kind field:

Amount Semantics (semantics)

Text As Written (text_as_written)

The verbatim dollar string from the bill text (e.g., "$2,285,513,000"). Used by the verification pipeline — this exact string is searched for in the source XML.

Complete Amount Example

{
  "value": {
    "kind": "specific",
    "dollars": 2285513000
  },
  "semantics": "new_budget_authority",
  "text_as_written": "$2,285,513,000"
}

Detail Level (Appropriation Type Only)

The detail_level field on appropriation provisions indicates structural position in the funding hierarchy:

The compute_totals() function uses detail_level to prevent double-counting. Sub-allocations and proviso amounts are breakdowns of a parent appropriation, not additional money.

Proviso

Conditions attached to appropriations via “Provided, That” clauses:

Earmark

Community project funding or directed spending items:

CrAnomaly

Anomaly entries within a continuing_resolution_baseline provision:

ExtractionSummary (summary)

LLM-produced self-check totals. These are diagnostic only — budget authority displayed by the summary command is always computed from individual provisions, never from these fields.

Chunk Map (chunk_map)

Links provisions to the extraction chunks they came from. For single-chunk bills (like H.R. 9468), this is an empty array. For multi-chunk bills, each entry maps a chunk ID (ULID) to a range of provision indices:

[
  {
    "chunk_id": "01JRWN9T5RR0JTQ6C9FYYE96A8",
    "label": "A-I",
    "provision_start": 0,
    "provision_end": 42
  },
  {
    "chunk_id": "01JRWNA2B3C4D5E6F7G8H9J0K1",
    "label": "A-II",
    "provision_start": 42,
    "provision_end": 95
  }
]

This enables full audit trails — you can trace any provision back to the specific chunk and LLM call that produced it.

Complete Minimal Example (H.R. 9468)

{
  "schema_version": "1.0",
  "bill": {
    "identifier": "H.R. 9468",
    "classification": "supplemental",
    "short_title": "Veterans Benefits Continuity and Accountability Supplemental Appropriations Act, 2024",
    "fiscal_years": [2024],
    "divisions": [],
    "public_law": null
  },
  "provisions": [
    {
      "provision_type": "appropriation",
      "account_name": "Compensation and Pensions",
      "agency": "Department of Veterans Affairs",
      "program": null,
      "amount": {
        "value": { "kind": "specific", "dollars": 2285513000 },
        "semantics": "new_budget_authority",
        "text_as_written": "$2,285,513,000"
      },
      "fiscal_year": 2024,
      "availability": "to remain available until expended",
      "provisos": [],
      "earmarks": [],
      "detail_level": "top_level",
      "parent_account": null,
      "section": "",
      "division": null,
      "title": null,
      "confidence": 0.99,
      "raw_text": "For an additional amount for ''Compensation and Pensions'', $2,285,513,000, to remain available until expended.",
      "notes": [
        "Supplemental appropriation under Veterans Benefits Administration heading",
        "No-year funding"
      ],
      "cross_references": []
    },
    {
      "provision_type": "appropriation",
      "account_name": "Readjustment Benefits",
      "agency": "Department of Veterans Affairs",
      "program": null,
      "amount": {
        "value": { "kind": "specific", "dollars": 596969000 },
        "semantics": "new_budget_authority",
        "text_as_written": "$596,969,000"
      },
      "fiscal_year": 2024,
      "availability": "to remain available until expended",
      "provisos": [],
      "earmarks": [],
      "detail_level": "top_level",
      "parent_account": null,
      "section": "",
      "division": null,
      "title": null,
      "confidence": 0.99,
      "raw_text": "For an additional amount for ''Readjustment Benefits'', $596,969,000, to remain available until expended.",
      "notes": [
        "Supplemental appropriation under Veterans Benefits Administration heading",
        "No-year funding"
      ],
      "cross_references": []
    },
    {
      "provision_type": "rider",
      "description": "Establishes that each amount appropriated or made available by this Act is in addition to amounts otherwise appropriated for the fiscal year involved.",
      "policy_area": null,
      "section": "SEC. 101",
      "division": null,
      "title": null,
      "confidence": 0.98,
      "raw_text": "SEC. 101. Each amount appropriated or made available by this Act is in addition to amounts otherwise appropriated for the fiscal year involved.",
      "notes": [],
      "cross_references": []
    },
    {
      "provision_type": "directive",
      "description": "Requires the Secretary of Veterans Affairs to submit a report detailing corrections the Department will make to improve forecasting, data quality, and budget assumptions.",
      "deadlines": ["30 days after enactment"],
      "section": "SEC. 103",
      "division": null,
      "title": null,
      "confidence": 0.97,
      "raw_text": "SEC. 103. (a) Not later than 30 days after the date of enactment of this Act, the Secretary of Veterans Affairs shall submit to the Committees on App",
      "notes": [],
      "cross_references": []
    }
  ],
  "summary": {
    "total_provisions": 7,
    "by_division": {},
    "by_type": {
      "appropriation": 2,
      "rider": 2,
      "directive": 3
    },
    "total_budget_authority": 2882482000,
    "total_rescissions": 0,
    "sections_with_no_provisions": [],
    "flagged_issues": []
  },
  "chunk_map": []
}

Note: The example above is abbreviated — the actual H.R. 9468 extraction has 7 provisions (2 appropriations, 2 riders, 3 directives). Only 4 are shown here for brevity.

Accessing extraction.json

From the CLI

All query commands (search, summary, compare, audit) read extraction.json automatically. You don’t need to interact with the file directly for normal use.

From Python

import json

with open("data/118-hr9468/extraction.json") as f:
    data = json.load(f)

# Bill info
print(data["bill"]["identifier"])  # "H.R. 9468"

# Provisions
for p in data["provisions"]:
    ptype = p["provision_type"]
    if ptype == "appropriation":
        dollars = p["amount"]["value"]["dollars"]
        account = p["account_name"]
        print(f"{account}: ${dollars:,}")

From Rust (Library API)

#![allow(unused)]
fn main() {
use congress_appropriations::load_bills;
use std::path::Path;

let bills = load_bills(Path::new("examples"))?;
for bill in &bills {
    println!("{}: {} provisions",
        bill.extraction.bill.identifier,
        bill.extraction.provisions.len());
}
}

See Use the Library API from Rust for the full guide.

Schema Versioning

The schema_version field tracks the extraction data format. When the schema evolves (new fields, renamed fields), the upgrade command migrates existing data to the latest version without re-extraction.

The upgrade command adds schema_version to pre-versioned files and applies any necessary field migrations. See Upgrade Extraction Data.

Related References

• Provision Types — type-by-type field reference with examples
• verification.json Fields — the verification report that accompanies each extraction
• embeddings.json Fields — embedding metadata
• Data Directory Layout — where extraction.json fits in the file hierarchy