@mailwoman/corpus 9.4.0 → 10.1.0

package/lib/adapters/state-hi-schools/workbook.ts

@@ -1,35 +0,0 @@
-/**
- * @copyright Sister Software
- * @license AGPL-3.0
- * @author Teffen Ellis, et al.
- */
-
-import type { XLSXCellValue } from "spliterator"
-
-/**
- * The two worksheets that together form the public-school directory.
- */
-export const STATE_HI_SCHOOL_SHEETS = ["HIDOE", "PCS"] as const
-
-/**
- * Columns required to construct a corpus row. Fetch validation uses the same contract as the adapter.
- */
-export const STATE_HI_SCHOOL_REQUIRED_COLUMNS = ["code", "name", "address", "city", "zip"] as const
-
-/**
- * Workbook row fields consumed by the adapter. XLSX cells are typed; the legacy CSV path supplies strings.
- */
-export interface HiSchoolRow {
-	code: XLSXCellValue | undefined
-	name: XLSXCellValue | undefined
-	address: XLSXCellValue | undefined
-	city: XLSXCellValue | undefined
-	zip: XLSXCellValue | undefined
-}
-
-/**
- * Convert a typed workbook cell or legacy CSV field to the trimmed text used by the corpus row.
- */
-export function schoolCellText(value: XLSXCellValue | undefined): string {
-	return value === null || value === undefined ? "" : String(value).trim()
-}

package/lib/adapters/state-ia-contractors/adapter.ts

@@ -1,144 +0,0 @@
-/**
- * @copyright Sister Software
- * @license AGPL-3.0
- * @author Teffen Ellis, et al.
- *
- *   `state-ia-contractors`: Iowa Active Construction Contractor Registrations CSV consumer.
- *
- *   Iowa Workforce Development publishes a public registry of active construction contractors. Each
- *   row carries a business name, street address, city/state/zip, and contact info.
- *
- *   The adapter consumes the CSV the operator pre-downloads via `fetch-state-sources.ts`.
- *
- *   Output: one row per contractor with `venue` (business name) and address quad `(house_number,
- *   street, locality, region, postcode)`.
- *
- *   License: stamped `"Public Domain"` per Iowa state government open-data terms.
- */
-
-import { isPresent } from "@mailwoman/core/objects"
-import { reconcileComponents } from "@mailwoman/formatter"
-import { CSVSpliterator } from "spliterator"
-
-import { splitStreetLine, stableSourceID } from "#adapters/utils"
-import { lookupStateAbbreviation } from "#codex/us-fips-state"
-import type { AdapterOptions, CanonicalRow, CorpusAdapter } from "#types"
-
-/**
- * Registry id for this adapter. Stamped into every row it emits, so a corpus record can be traced back to the dataset
- * it came from.
- */
-export const STATE_IA_CONTRACTORS_ADAPTER_ID = "state-ia-contractors"
-/**
- * License carried by this source (Public Domain), attached to each row so downstream consumers inherit the terms rather
- * than having to look them up.
- */
-export const STATE_IA_CONTRACTORS_DEFAULT_LICENSE = "Public Domain"
-
-interface IaContractorRow {
-	"Registration #": string
-	"Business Name": string
-	"Address 1": string
-	"Address 2": string
-	City: string
-	State: string
-	"Zip Code": string
-	"First Name": string
-	"Last Name": string
-}
-
-export function createStateIaContractorsAdapter(): CorpusAdapter {
-	return {
-		id: STATE_IA_CONTRACTORS_ADAPTER_ID,
-		defaultLicense: STATE_IA_CONTRACTORS_DEFAULT_LICENSE,
-		description:
-			"Iowa Active Construction Contractor Registrations — business name + full street address (public-domain).",
-
-		async *rows(opts: AdapterOptions): AsyncIterable<CanonicalRow> {
-			if (opts.country && opts.country !== "US") {
-				throw new Error(`state-ia-contractors adapter: only US supported, got country=${opts.country}`)
-			}
-
-			const rows = CSVSpliterator.fromAsync(opts.inputPath, {
-				mode: "object",
-				normalizeKeys: false,
-				enableQuoteHandling: true,
-			})
-
-			let emitted = 0
-
-			for await (const record of rows as AsyncIterable<IaContractorRow>) {
-				if (opts.signal?.aborted) break
-
-				if (opts.limit !== undefined && emitted >= opts.limit) break
-
-				const businessName = (record["Business Name"] ?? "").trim()
-				const address1 = (record["Address 1"] ?? "").trim()
-				const address2 = (record["Address 2"] ?? "").trim()
-				const city = (record.City ?? "").trim()
-				const stateAbbr = (record.State ?? "").trim()
-				const zip = (record["Zip Code"] ?? "").trim()
-
-				if (!city || !zip) continue
-
-				const state = lookupStateAbbreviation(stateAbbr)
-
-				if (!state) continue
-
-				const fullAddress = [address1, address2].filter(isPresent).join(" ")
-				const split = splitStreetLine(fullAddress)
-
-				if (!split) continue
-
-				const venue = businessName || undefined
-
-				const components: CanonicalRow["components"] = {
-					...(venue ? { venue } : {}),
-					...(split.house_number ? { house_number: split.house_number } : {}),
-					street: split.street,
-					locality: city,
-					region: state.abbreviation,
-					postcode: zip,
-				}
-
-				const streetPart = [split.house_number, split.street].filter(isPresent).join(" ").trim()
-
-				const raw = [
-					venue,
-					streetPart,
-					[city, [stateAbbr, zip].filter(isPresent).join(" ")].filter(isPresent).join(", "),
-				]
-					.filter(isPresent)
-					.join(", ")
-
-				const aligned = reconcileComponents(components, raw)
-
-				if (Object.keys(aligned).length <= 2) continue
-
-				const regNum = (record["Registration #"] ?? "").trim()
-
-				const sourceID = regNum
-					? `${STATE_IA_CONTRACTORS_ADAPTER_ID}-${regNum}`
-					: stableSourceID(STATE_IA_CONTRACTORS_ADAPTER_ID, aligned)
-
-				yield {
-					raw,
-					components: aligned,
-					country: "US",
-					locale: "en-US",
-					source: STATE_IA_CONTRACTORS_ADAPTER_ID,
-					source_id: sourceID,
-					corpus_version: "",
-					license: STATE_IA_CONTRACTORS_DEFAULT_LICENSE,
-				}
-
-				emitted++
-			}
-		},
-	}
-}
-
-/**
- * The configured adapter instance registered with the corpus builder.
- */
-export const stateIaContractorsAdapter = createStateIaContractorsAdapter()

package/lib/adapters/state-ny-notaries/adapter.ts

@@ -1,141 +0,0 @@
-/**
- * @copyright Sister Software
- * @license AGPL-3.0
- * @author Teffen Ellis, et al.
- *
- *   `state-ny-notaries`: New York Commissioned Notaries CSV consumer.
- *
- *   The New York Department of State publishes a registry of commissioned notaries public. Each row
- *   optionally carries a business name and business address (~1-5% fill rate).
- *
- *   The adapter consumes the CSV the operator pre-downloads via `fetch-state-sources.ts`. Column
- *   names match the data.ny.gov export header (note: some columns have leading spaces).
- *
- *   License: stamped `"Public Domain"` per New York state government open-data terms.
- */
-
-import { isPresent } from "@mailwoman/core/objects"
-import { reconcileComponents } from "@mailwoman/formatter"
-import { CSVSpliterator } from "spliterator"
-
-import { splitStreetLine, stableSourceID } from "#adapters/utils"
-import { lookupStateAbbreviation } from "#codex/us-fips-state"
-import type { AdapterOptions, CanonicalRow, CorpusAdapter } from "#types"
-
-/**
- * Registry id for this adapter. Stamped into every row it emits, so a corpus record can be traced back to the dataset
- * it came from.
- */
-export const STATE_NY_NOTARIES_ADAPTER_ID = "state-ny-notaries"
-/**
- * License carried by this source (Public Domain), attached to each row so downstream consumers inherit the terms rather
- * than having to look them up.
- */
-export const STATE_NY_NOTARIES_DEFAULT_LICENSE = "Public Domain"
-
-export function createStateNyNotariesAdapter(): CorpusAdapter {
-	return {
-		id: STATE_NY_NOTARIES_ADAPTER_ID,
-		defaultLicense: STATE_NY_NOTARIES_DEFAULT_LICENSE,
-		description: "New York Commissioned Notaries — name + optional business address (public-domain).",
-
-		async *rows(opts: AdapterOptions): AsyncIterable<CanonicalRow> {
-			if (opts.country && opts.country !== "US") {
-				throw new Error(`state-ny-notaries adapter: only US supported, got country=${opts.country}`)
-			}
-
-			const rows = CSVSpliterator.fromAsync(opts.inputPath, {
-				mode: "object",
-				normalizeKeys: false,
-				enableQuoteHandling: true,
-			})
-
-			let emitted = 0
-
-			for await (const rawRecord of rows as AsyncIterable<Record<string, string>>) {
-				if (opts.signal?.aborted) break
-
-				if (opts.limit !== undefined && emitted >= opts.limit) break
-
-				// NY CSV has columns with leading spaces, so we normalize by trimming keys.
-				const record: Record<string, string> = {}
-
-				for (const key of Object.keys(rawRecord)) {
-					record[key.trim()] = rawRecord[key] ?? ""
-				}
-
-				const holderName = (record["Commission Holder Name"] ?? "").trim()
-				const businessName = (record["Business Name (if available)"] ?? "").trim()
-				const address1 = (record["Business Address 1 (if available)"] ?? "").trim()
-				const address2 = (record["Business Address 2 (if available)"] ?? "").trim()
-				const city = (record["Business City (if available)"] ?? "").trim()
-				const stateAbbr = (record["Business State (if available)"] ?? "").trim()
-				const zip = (record["Business Zip (if available)"] ?? "").trim()
-				const county = (record["Commissioned County"] ?? "").trim()
-
-				if (!city || !stateAbbr || !zip) continue
-
-				if (!address1 && !address2) continue
-
-				const state = lookupStateAbbreviation(stateAbbr)
-
-				if (!state) continue
-
-				const fullAddress = [address1, address2].filter(isPresent).join(" ")
-				const split = splitStreetLine(fullAddress)
-
-				if (!split) continue
-
-				const venue = businessName || holderName || undefined
-
-				const components: CanonicalRow["components"] = {
-					...(venue ? { venue } : {}),
-					...(split.house_number ? { house_number: split.house_number } : {}),
-					street: split.street,
-					locality: city,
-					region: state.abbreviation,
-					postcode: zip,
-					...(county ? { subregion: county } : {}),
-				}
-
-				const streetPart = [split.house_number, split.street].filter(isPresent).join(" ").trim()
-
-				const raw = [
-					venue,
-					streetPart,
-					[city, [stateAbbr, zip].filter(isPresent).join(" ")].filter(isPresent).join(", "),
-				]
-					.filter(isPresent)
-					.join(", ")
-
-				const aligned = reconcileComponents(components, raw)
-
-				if (Object.keys(aligned).length <= 2) continue
-
-				const commNum = (record["Commission Number (UID)"] ?? "").trim()
-
-				const sourceID = commNum
-					? `${STATE_NY_NOTARIES_ADAPTER_ID}-${commNum}`
-					: stableSourceID(STATE_NY_NOTARIES_ADAPTER_ID, aligned)
-
-				yield {
-					raw,
-					components: aligned,
-					country: "US",
-					locale: "en-US",
-					source: STATE_NY_NOTARIES_ADAPTER_ID,
-					source_id: sourceID,
-					corpus_version: "",
-					license: STATE_NY_NOTARIES_DEFAULT_LICENSE,
-				}
-
-				emitted++
-			}
-		},
-	}
-}
-
-/**
- * The configured adapter instance registered with the corpus builder.
- */
-export const stateNyNotariesAdapter = createStateNyNotariesAdapter()

package/lib/adapters/state-tx-notaries/adapter.ts

@@ -1,157 +0,0 @@
-/**
- * @copyright Sister Software
- * @license AGPL-3.0
- * @author Teffen Ellis, et al.
- *
- *   `state-tx-notaries`: Texas Notary Public Commissions CSV consumer.
- *
- *   The Texas Secretary of State publishes a registry of commissioned notaries public. Each row
- *   optionally carries a mailing address in free-form text (often multi-line with embedded
- *   city/state/zip). Address fill rate is ~5-10%.
- *
- *   The adapter parses the embedded `Address` field for city/state/zip using a trailing `"CITY, ST
- *   ZIP"` pattern.
- *
- *   License: stamped `"Public Domain"` per Texas state government open-data terms.
- */
-
-import { isPresent } from "@mailwoman/core/objects"
-import { reconcileComponents } from "@mailwoman/formatter"
-import { CSVSpliterator } from "spliterator"
-
-import { splitStreetLine, stableSourceID } from "#adapters/utils"
-import { lookupStateAbbreviation } from "#codex/us-fips-state"
-import type { AdapterOptions, CanonicalRow, CorpusAdapter } from "#types"
-
-/**
- * Registry id for this adapter. Stamped into every row it emits, so a corpus record can be traced back to the dataset
- * it came from.
- */
-export const STATE_TX_NOTARIES_ADAPTER_ID = "state-tx-notaries"
-/**
- * License carried by this source (Public Domain), attached to each row so downstream consumers inherit the terms rather
- * than having to look them up.
- */
-export const STATE_TX_NOTARIES_DEFAULT_LICENSE = "Public Domain"
-
-/**
- * Match trailing "CITY, ST ZIP" or "CITY, ST" at the end of an address line.
- */
-const CITY_STATE_ZIP_SUFFIX = /[,]?\s*([^,]+),\s*([A-Z]{2})\s*(\d{5}(?:-\d{4})?)?\s*$/i
-
-interface TxNotaryRow {
-	"Notary ID": string
-	"First Name": string
-	"Last Name": string
-	Address: string
-}
-
-export function createStateTxNotariesAdapter(): CorpusAdapter {
-	return {
-		id: STATE_TX_NOTARIES_ADAPTER_ID,
-		defaultLicense: STATE_TX_NOTARIES_DEFAULT_LICENSE,
-		description:
-			"Texas Notary Public Commissions — name + mailing address with embedded city/state/zip (public-domain).",
-
-		async *rows(opts: AdapterOptions): AsyncIterable<CanonicalRow> {
-			if (opts.country && opts.country !== "US") {
-				throw new Error(`state-tx-notaries adapter: only US supported, got country=${opts.country}`)
-			}
-
-			const rows = CSVSpliterator.fromAsync(opts.inputPath, {
-				mode: "object",
-				normalizeKeys: false,
-				enableQuoteHandling: true,
-			})
-
-			let emitted = 0
-
-			for await (const record of rows as AsyncIterable<TxNotaryRow>) {
-				if (opts.signal?.aborted) break
-
-				if (opts.limit !== undefined && emitted >= opts.limit) break
-
-				const rawAddress = (record.Address ?? "").trim()
-
-				if (!rawAddress) continue
-
-				const firstName = (record["First Name"] ?? "").trim()
-				const lastName = (record["Last Name"] ?? "").trim()
-				const notaryID = (record["Notary ID"] ?? "").trim()
-
-				// Parse embedded city/state/zip from the trailing portion of the address.
-				// Addresses look like: "1215 MCMILLAN DR\nCEDAR HILL, TX 75104"
-				const addrSingleLine = rawAddress.replaceAll("\n", ", ")
-				const cszMatch = CITY_STATE_ZIP_SUFFIX.exec(addrSingleLine)
-
-				if (!cszMatch) continue
-
-				const city = (cszMatch[1] ?? "").trim()
-				const stateAbbr = (cszMatch[2] ?? "").trim()
-				const zip = (cszMatch[3] ?? "").trim()
-
-				if (!city || !stateAbbr) continue
-
-				const state = lookupStateAbbreviation(stateAbbr)
-
-				if (!state) continue
-
-				// Extract the street portion (everything before the city/state/zip)
-				const streetPortion = addrSingleLine.slice(0, cszMatch.index).replace(/,\s*$/, "").trim()
-
-				if (!streetPortion) continue
-
-				const split = splitStreetLine(streetPortion)
-
-				if (!split) continue
-
-				const venue = [firstName, lastName].filter(isPresent).join(" ") || undefined
-
-				const components: CanonicalRow["components"] = {
-					...(venue ? { venue } : {}),
-					...(split.house_number ? { house_number: split.house_number } : {}),
-					street: split.street,
-					locality: city,
-					region: state.abbreviation,
-					...(zip ? { postcode: zip } : {}),
-				}
-
-				const streetPart = [split.house_number, split.street].filter(isPresent).join(" ").trim()
-
-				const raw = [
-					venue,
-					streetPart,
-					[city, [stateAbbr, zip].filter(isPresent).join(" ")].filter(isPresent).join(", "),
-				]
-					.filter(isPresent)
-					.join(", ")
-
-				const aligned = reconcileComponents(components, raw)
-
-				if (Object.keys(aligned).length <= 2) continue
-
-				const sourceID = notaryID
-					? `${STATE_TX_NOTARIES_ADAPTER_ID}-${notaryID}`
-					: stableSourceID(STATE_TX_NOTARIES_ADAPTER_ID, aligned)
-
-				yield {
-					raw,
-					components: aligned,
-					country: "US",
-					locale: "en-US",
-					source: STATE_TX_NOTARIES_ADAPTER_ID,
-					source_id: sourceID,
-					corpus_version: "",
-					license: STATE_TX_NOTARIES_DEFAULT_LICENSE,
-				}
-
-				emitted++
-			}
-		},
-	}
-}
-
-/**
- * The configured adapter instance registered with the corpus builder.
- */
-export const stateTxNotariesAdapter = createStateTxNotariesAdapter()

package/lib/adapters/synth-po-box/adapter.ts

@@ -1,204 +0,0 @@
-/**
- * @copyright Sister Software
- * @license AGPL-3.0
- * @author Teffen Ellis, et al.
- *
- *   `synth-po-box`: PO box / PMB / Apartado / BP synthesizer adapter.
- *
- *   Consumes a JSONL stream of (locality, region, postcode, country) tuples — typically extracted
- *   from existing corpus output (TIGER/NAD/BAN/WOF) — and emits synthetic PO box training rows. See
- *   `@mailwoman/corpus/synthesizers/po-box` for the per-locale templates and number-noise logic.
- *
- *   Why an adapter and not an augmenter:
- *
- *   - Per USPS Pub 28 / DMM 508, a PO box delivery line is mutually exclusive with a street line.
- *       Synthesizing PO boxes by mutating a street row would teach the model an invalid pattern.
- *       The clean shape is: read just (locality, region, postcode, country) and produce a fresh
- *       PO-box-shaped row.
- *   - Per-DeepSeek (3-turn consult, 2026-05-28): PMB rows that COMBINE a street line with a PMB number
- *       ARE valid (CMRA addresses). Those are produced when `pmbRatio > 0` AND the input tuple
- *       carries a `street` field.
- */
-
-import { tryParsingJSON } from "@mailwoman/core/json"
-import { makeLcg } from "@mailwoman/core/random"
-import { TextSpliterator } from "spliterator"
-
-import { stableSourceID } from "#adapters/utils"
-import {
-	countryToLocale,
-	REGION_OPTIONAL_LOCALES,
-	synthesizeMilitaryPoBoxRow,
-	synthesizePoBoxRow,
-	type PoBoxBaseTuple,
-} from "#synthesizers/po-box"
-import type { AdapterOptions, CanonicalRow, CorpusAdapter } from "#types"
-
-/**
- * Registry id for this adapter. Stamped into every row it emits, so a corpus record can be traced back to the dataset
- * it came from.
- */
-export const SYNTH_PO_BOX_ADAPTER_ID = "synth-po-box"
-/**
- * License for the synthetic PO-box rows. The output is generated, but it inherits the terms of the real tuples it is
- * derived from, so the attribution travels with it.
- */
-export const SYNTH_PO_BOX_LICENSE = "Synthetic — derived from CC-BY / public-domain input tuples"
-
-export interface PoBoxInputRow extends PoBoxBaseTuple {
-	street?: string
-	houseNumber?: string
-}
-
-export interface SynthPoBoxAdapterOptions {
-	/**
-	 * How many PO box variants to emit per input tuple. Each variant picks a different leader (and possibly a different
-	 * number / noise level). Default 1.
-	 */
-	variantsPerInput?: number
-	/**
-	 * Probability (0..1) of emitting a PMB-with-street variant when both the input has a street and the locale supports
-	 * PMB. Default 0.15.
-	 */
-	pmbRatio?: number
-	/**
-	 * Deterministic seed for reproducible synthesis. Default Date.now().
-	 */
-	seed?: number
-	/**
-	 * Probability (0..1), evaluated per input tuple, of ALSO emitting one US military/diplomatic PO-box row
-	 * (`PSC/CMR/Unit <id> Box <box>, APO/FPO/DPO AA/AE/AP <zip>`, #517). These rows are self-contained — they draw no
-	 * field from the input tuple, so military volume scales with the input stream size. Default 0 (off) — the adapter's
-	 * contract is "one row per input"; the corpus build recipe opts in to seed the rare-but-real military class without
-	 * changing the default.
-	 */
-	militaryRatio?: number
-}
-
-export function createSynthPoBoxAdapter(opts: SynthPoBoxAdapterOptions = {}): CorpusAdapter {
-	const variantsPerInput = opts.variantsPerInput ?? 1
-	const pmbRatio = opts.pmbRatio ?? 0.15
-	const militaryRatio = opts.militaryRatio ?? 0
-
-	return {
-		id: SYNTH_PO_BOX_ADAPTER_ID,
-		defaultLicense: SYNTH_PO_BOX_LICENSE,
-		description:
-			"Synthetic PO box / PMB / Apartado / Boîte Postale rows. Consumes JSONL of (locality, region, postcode, country) tuples and emits locale-appropriate PO box variants.",
-
-		async *rows(options: AdapterOptions): AsyncIterable<CanonicalRow> {
-			const random = makeLcg(opts.seed ?? Date.now())
-
-			// TextSpliterator streams string lines; the per-line tryParsingJSON below keeps this reader
-			// tolerant of malformed rows (skipped++), so TextSpliterator + a non-throwing parse — not
-			// JSONSpliterator, which would throw on the first bad line.
-			const lines = TextSpliterator.fromAsync(options.inputPath)
-
-			let emitted = 0
-			let skipped = 0
-			let militarySeq = 0
-
-			for await (const line of lines) {
-				if (options.signal?.aborted) break
-
-				if (options.limit !== undefined && emitted >= options.limit) break
-
-				const trimmed = line.trim()
-
-				if (!trimmed) continue
-
-				const input = tryParsingJSON<PoBoxInputRow>(trimmed)
-
-				if (input === null) {
-					skipped++
-
-					continue
-				}
-
-				// Region is required EXCEPT for region-less locales (NZ: `Private Bag 12, Auckland 1010`
-				// has no region token, #517). synthesizePoBoxRow handles region absence; the guard just
-				// must not discard those tuples as "missing region".
-				const regionOptional = input.country ? REGION_OPTIONAL_LOCALES.has(countryToLocale(input.country)) : false
-
-				if (!input.locality || !input.postcode || !input.country || (!input.region && !regionOptional)) {
-					skipped++
-
-					continue
-				}
-
-				if (options.country && options.country !== input.country) continue
-
-				for (let v = 0; v < variantsPerInput; v++) {
-					const synth = synthesizePoBoxRow(input, { random, pmbRatio })
-
-					if (!synth) continue
-
-					// Include `v` in dependent_locality slot to vary the digest across variants;
-					// stableSourceID only accepts ComponentTag keys.
-					const sourceID = stableSourceID(SYNTH_PO_BOX_ADAPTER_ID, {
-						locality: `${input.locality}#${v}`,
-						region: input.region,
-						postcode: input.postcode,
-						country: input.country,
-					})
-
-					yield {
-						raw: synth.raw,
-						components: synth.components,
-						country: input.country,
-						locale: synth.locale,
-						source: SYNTH_PO_BOX_ADAPTER_ID,
-						source_id: sourceID,
-						corpus_version: "",
-						license: SYNTH_PO_BOX_LICENSE,
-					}
-
-					emitted++
-
-					if (options.limit !== undefined && emitted >= options.limit) break
-				}
-
-				// US military/diplomatic PO-box rows (#517): self-contained — draw nothing from the input
-				// tuple — emitted per input line with probability `militaryRatio` (off by default, so the
-				// default random stream and output are byte-identical). Military volume scales with the
-				// stream rather than the US-tuple count. US-only: suppressed under a non-US country filter
-				// and counted against `limit` like any other row.
-				const militaryAllowed = !options.country || options.country === "US"
-
-				if (
-					militaryRatio > 0 &&
-					militaryAllowed &&
-					(options.limit === undefined || emitted < options.limit) &&
-					random() < militaryRatio
-				) {
-					const mil = synthesizeMilitaryPoBoxRow({ random })
-
-					const sourceID = stableSourceID(SYNTH_PO_BOX_ADAPTER_ID, {
-						po_box: `${mil.components.po_box}#mil${militarySeq++}`,
-						locality: mil.components.locality!,
-						region: mil.components.region!,
-						postcode: mil.components.postcode!,
-					})
-
-					yield {
-						raw: mil.raw,
-						components: mil.components,
-						country: "US",
-						locale: mil.locale,
-						source: SYNTH_PO_BOX_ADAPTER_ID,
-						source_id: sourceID,
-						corpus_version: "",
-						license: SYNTH_PO_BOX_LICENSE,
-					}
-
-					emitted++
-				}
-			}
-		},
-	}
-}
-
-/**
- * The configured adapter instance registered with the corpus builder.
- */
-export const synthPoBoxAdapter = createSynthPoBoxAdapter()