Skip to main content

Plan: Ingest kommune-level helt ledige (fully unemployed) statistics

Ingests NAV's HL060 table — monthly count and share (of the labour force) of people registered as fully unemployed, per kommune — as Atlas's fourth NAV-adjacent source. Unlike nav-sykefravaer (PLAN-013, which turned out to require SSB instead of NAV), this candidate's own mechanism matches the investigation's description: NAV's own site genuinely publishes a dedicated kommune-level table (HL060 "Fylke og kommune"), the same NAV-Excel shape as nav-uforetrygd and nav-aap, reached from a real HTML anchor the same way nav-aap's download link was.

IMPLEMENTATION RULES: Before implementing this plan, read and follow:

Status: Completed​

Goal: Add nav-helt-ledige as a served Atlas source — the short-tail labour-market signal slotting between nav-uforetrygd (long-tail disability outcome) and nav-aap (transitional work-assessment benefit), per the investigation's own framing. Plugs Report #5 (Income & Welfare) and adds an axis to Report #4 (Mental-Health Triangulation — acute unemployment correlates with mental health).

Last Updated: 2026-10-02

Investigation: INVESTIGATE-new-norwegian-public-sources.md §C.4 ([Q45], [Q37])

Prerequisites: None. nav is already a valid publishers.yaml provider, topics.yaml already has social validated (nav-uforetrygd/nav-aap), and the existing monthly_sources_refresh Dagster job already exists and fits this cadence directly — no new job.

Phase 1: Confirm the real shape — DONE (verified 2026-10-02)​

Checked live rather than trusting the investigation's description, the same discipline applied to every prior source — but this time the description held up structurally, which is itself worth confirming explicitly rather than assuming after two sources in a row (husbanken-bostotte, nav-sykefravaer) where it didn't.

Tasks​

  • 1.1 Find the live download page and the real download link. Confirmed — https://www.nav.no/no/nav-og-samfunn/statistikk/arbeidssokere-og-stillinger-statistikk/helt-ledige lists ~8 numbered tables (HL050-HL090), each a different breakdown (age×sex, education, occupation, duration/long-term, immigrant background, fylke-only yearly). Only HL060 ("Helt ledige. Fylke og kommune. Tidsserie måned") is kommune-resolved — same "exactly one kommune-grain table per NAV statistics page" pattern as PST302/AAP155. The download link is a real <a href="/_/attachment/download/<uuid>:<hash>/202608_HL060%20...xlsx"> anchor directly in the server-rendered HTML — same simpler-than-PST302 discovery shape nav-aap found, not __NEXT_DATA__ JSON.

  • 1.2 Licence — same blanket NAV statement already confirmed twice this session. The workbook's own first cell states "Kilde: NAV" directly, and the file is served from www.nav.no's own attachment system — no SSB involvement this time (unlike nav-sykefravaer). CC BY 4.0 applies, per the same statistics-practices page fetched during nav-aap's Phase 1 (nav.no/.../praksis-rutiner-og-retningslinjer-rundt-offisiell-og-offentlig-statistikk-fra-nav) — re-confirmed as applying to this specific table by checking its own "Kilde: NAV" provenance, not assumed by family resemblance alone.

  • 1.3 Download the live HL060 workbook (period: January-August 2026) and inspect its real structure. 5 sheets: "0. Om tabellene" (methodology), "1. Fylke Antall", "2. Fylke Prosent av arbeidsstyr", "3. Kommune Antall", "4. Kommune Prosent av arbeidsst". Only the two Kommune sheets are in scope.

    A fourth distinct pivot shape within the NAV family this session — not identical to PST302's, AAP155's, or (not applicable) sykefravær's SSB shape:

    B6:  "Oslo - Oslove"                        ← bare fylke header, NO "I alt", NO code, NO data
    B7:  (blank) | Januar | ... | August         ← month header row, repeats per fylke block
    B8:  "I alt Oslo - Oslove" | 12125 | ...     ← fylke total, "I alt <name>" — NO digit code at all
    B9:  "0301 Oslo - Oslove" | 12125 | ...      ← the one kommune in this fylke, real values
    B10: "Rogaland"                              ← next fylke header
    ...
    B13: "1101 Eigersund" | 197 | ...            ← ordinary kommune row
    ...
    B422: "Svalbard og øvrige områder"           ← a 17th "fylke" block
    B424: "I alt Svalbard og øvrige områder" | 13 | ...
    B425: "2100 Svalbard" | 13 | ...             ← Svalbard's own real pseudo-kommune code
    B427: "Ukjent"                               ← an 18th "fylke" block, no numeric code anywhere
    B429: "I alt Ukjent" | 13 | ...
    B430: "Ukjent" | 13 | ...                    ← the "kommune"-level child, same literal "Ukjent"
                                                     label, same values as its own "I alt" row

    Simpler to classify than PST302 or AAP155: neither the bare fylke header row nor the "I alt " total row carries any digit at all here (unlike AAP155's "I alt 03 Oslo", which embeds the 2-digit fylke code). A single rule — "label starts with exactly 4 digits" — correctly admits only real kommune rows and the Svalbard pseudo-kommune, with no "I alt" prefix check needed at all. Confirmed by scanning the entire sheet: zero digit-leading labels outside that shape.

    Both sentinel shapes already have precedent from this session, no new one needed: 2100 Svalbard matches classify_region_code's existing ^21\d{2}$ branch exactly (same as udir-gsi/husbanken-bostotte's finding); the bare literal "Ukjent" label with no numeric code at all is structurally identical to nav-aap's own Ukjent finding (PLAN-012 [Q1]) and falls through to the macro's existing unknown branch the same way.

    358 distinct 4-digit-leading codes: 357 real kommuner + 2100 Svalbard — confirmed by counting, not assumed.

  • 1.4 Confirm numeric cell typing and the suppression marker. Confirmed — native numeric cells (same as every NAV-Excel source this session). Suppression marker is the literal *, same convention as nav-uforetrygd/nav-aap — but a different legal citation: the workbook's own methodology sheet states "I henhold til Statistikklovens § 7-1 ... mindre enn 4 ... «»"* — § 7-1, not AAP155's § 2-6. Same <4-person threshold, different paragraph of the same law; worth recording exactly rather than assuming every NAV table cites the same section. 15 kommuner show real suppression in the live sample (Utsira, Kvitsøy, Modalen, Tydal, Røst, and others — genuinely small kommuner), confirmed cell-by-cell (a single kommune row can mix real numbers and * across different months in the same row).

  • 1.5 Check the archive / historical-backfill shape. An archive page exists (.../helt-ledige/arkiv-helt-ledige_kap, HTTP 200) but a static-HTML scan found no HL060 reference in it — same inconclusive result as nav-aap's archive check, not the same as nav-uforetrygd's clean one-file-per-year pattern. [Q1] Historical backfill is deferred, same decision as every prior NAV-Excel source — v1 ingests the live current-year file only.

  • 1.6 Checked data.norge.no for a cleaner distribution — same conclusion as every prior NAV table: queried the real search API (search.api.fellesdatakatalog.digdir.no/search) for HL060: the ten results returned are all unrelated Brønnøysund/Nasjonalbiblioteket entries, not NAV statistics. www.nav.no is the real, current, sole publication surface.

  • 1.7 Two real trend-break notes in the workbook's own methodology sheet — represent, don't correct. (1) A January 2024 kommune/fylke reform footnote — fylker split/changed, figures "presented under the currently valid kommune/fylke structure" per NAV's own note. (2) A labour- force-methodology break from November 2018 (new registration method) and April 2025 (a larger modernisation of NAV's labour-market statistics) — two distinct trend breaks in the same series, not one. Both are upstream's own documented account of its data, same "carry verbatim into methodology_notes" discipline as ssb-12451's [Q7].

Validation​

✅ Confirmed 2026-10-02. Real file downloaded and inspected directly (both sheets, full row count, not a sample) — same discipline as every prior source.

Open Questions​

  • [Q1] Historical backfill — deferred, same decision as every prior NAV-Excel source. HL060's archive mechanism is inconclusive from a static scan (see Phase 1.5) — unlike nav-uforetrygd's clean archive, more like nav-aap's inconclusive one. v1 ingests the live current-year file only; revisit if a real consumer need for historical helt-ledige data emerges.
  • [Q2] Reuse code from nav-aap, or write fresh? nav-aap's own [Q3] decided to write fresh rather than adapt nav-uforetrygd's parser, because the real shapes differed enough that porting would import unneeded complexity. Here, HL060's row-classification rule ("exactly 4 leading digits, nothing else") is simpler than both PST302's and AAP155's — recommendation: write fresh again, following the same pattern (discovery tier, pure-function parser, golden-file tests against a real uncut workbook) rather than importing either prior parser's code.
  • [Q3] topics.yaml category — social, matching the sibling NAV sources. Not expected to be a real decision, but confirm against the file before writing the manifest, not after, per nav-aap's own Phase 2 note about catching this before a CI failure rather than from one.
  • [Q4] Presentation sensitivity. Per-kommune unemployment, like uføretrygd and sykefravær, can be politically charged in small kommuner — same presentation_policy: 'sensitive' consideration the investigation's [Q7] (on the original NAV family entry) already flagged. Not resolved here; a Phase 2/3 modelling decision.
  • [Q5] The Ukjent bucket's values are IDENTICAL between its own row and its fylke-level "I alt Ukjent" row (confirmed: both show 13,14,15,15,13,13,15,20) — i.e., "Ukjent" is simultaneously its own "fylke" and its own sole "kommune" in this table's pivot structure, unlike nav-aap where Ukjent was a genuine kommune-level leaf under no separate rollup. Recommendation: treat it the same as every other kommune-shaped row structurally (one row = one region_code), which naturally produces exactly this 1:1 duplication without any special-casing — not a defect to fix, just worth naming so a future reader isn't surprised to see the same numbers twice under slightly different row positions.

Phase 2: Ingest module + raw table — DONE (verified 2026-10-02)​

Tasks​

  • 2.1 Created atlas-data/ingest/src/sources/nav-helt-ledige/:
    • discoverWorkbookPath — found the real <a href> on the helt-ledige page, matched HL060. Same shape as nav-aap's discovery function, written fresh per [Q2], confirmed live — the path segment is download, not inline (AAP155's segment), confirmed not assumed.
    • parse.ts — pure functions reading "3. Kommune Antall"/"4. Kommune Prosent av arbeidsst", classifying rows by "label starts with exactly 4 digits" alone (simpler than both nav-uforetrygd's and nav-aap's classifiers — no "I alt" prefix check needed at all, confirmed per Phase 1.3's finding).
    • fetch_retry.ts — copied from nav-aap, same convention as every prior source.
    • manifest.yml — source_id: nav-helt-ledige, provider: nav, periodicity: P1M, eu_theme: SOCI, tags.topic: social ([Q3], confirmed against topics.yaml), license: CC BY 4.0 (confirmed directly against this table's own "Kilde: NAV" provenance, Phase 1.2).
    • README.md and __tests__/ — golden-file tests against the real captured workbook (full file, not trimmed, same precedent as nav-uforetrygd/nav-aap's fixtures) covering: an ordinary kommune row (Eigersund, 1101), the Svalbard pseudo-kommune (2100), the bare Ukjent block, and a suppressed row (1151 Utsira, mixed real/suppressed cells in the same row). 26 tests, all passing against the real downloaded fixture.
  • 2.2 Migration 063_raw_nav_helt_ledige.sql — raw.nav_helt_ledige(region_code, category_format, year, month, value, values_json, loaded_at), same shape as raw.nav_aap, PK (region_code, category_format, year, month).
  • 2.3 Dagster registration — added to the existing monthly_sources_refresh job (_MONTHLY_SOURCE_IDS in schedules.py, OTHER_SOURCES/asset group in raw_other.py). No new job — the third source on it. dagster definitions validate passes.

Validation​

cd atlas-data/ingest && npm run ingest:nav-helt-ledige   # the REAL npm-run invocation

🔴 Run it this exact way, not a direct tsx call. Ran it this exact way, against the live workbook and then against local Postgres.

Confirmed, not assumed: a real run against the live workbook returns 359 distinct regions (357 kommuner + Svalbard's 2100 + Ukjent), 8 months each, zero rows silently dropped — antall_rows: 2872, prosent_rows: 2872, rows_written: 5744 into local Postgres. One finding caught during real-data verification, not assumed from the plan: Ukjent IS present in BOTH sheets here (unlike nav-aap, where it's Antall-only) — its Prosent-sheet cells are all NAV's own suppression marker *, not omitted. Documented in parse.ts, manifest.yml and this file.

Phase 3: dbt staging and marts — DONE (verified 2026-10-02)​

Tasks​

  • 3.1 Added raw.nav_helt_ledige to models/indicators/sources.yml — ingest_cadence: monthly, freshness bounds matching cadence.MONTHLY_FRESHNESS exactly.
  • 3.2 indicators__nav_helt_ledige.sql — kommune_nr/region_kind via region_code_to_kommune_nr/classify_region_code from the first commit. Passed clean on the first dbt build, as expected — both sentinel shapes resolved exactly as predicted (see Validation below).
  • 3.3 Documented columns in schema.yml; validated against local Postgres loaded with the real ingest — 17/17 data tests pass on the first build.
  • 3.4 mart_indicators__nav_helt_ledige.sql api passthrough + marts/api/schema.yml entry.
  • 3.5 dbt build --select indicators__nav_helt_ledige mart_indicators__nav_helt_ledige against real loaded data — PASS=19 WARN=0 ERROR=0 SKIP=0 TOTAL=19.

Validation​

Real local Postgres, not an empty schema. Explicitly confirmed by direct query:

 region_code | kommune_nr | region_kind | count
-------------+------------+-------------+-------
 2100        |            | svalbard    |    16
 Ukjent      |            | unknown     |    16

2100 resolves to region_kind='svalbard'/kommune_nr=NULL and Ukjent resolves to region_kind='unknown'/kommune_nr=NULL, exactly as predicted — no macro change needed, matching the precedent from udir-gsi/husbanken-bostotte (svalbard) and nav-aap (unknown).

Full dbt check-suite (all 17 scripts) and a full dbt build both ran clean — the only non-pass results (extracted_columns_are_still_populated, raw_sources_were_refreshed_recently, the mart_atlas_inventory SKIP cascade) are pre-existing environmental staleness in the long-lived local scratch Postgres, confirmed unrelated to this source: nav_helt_ledige does not appear in either failing test's result set, and mart_brreg_enhet (the SKIP's root) builds 14/14 clean in isolation. Website build (npm run build) also passed clean: 51 sources, 88 relations, no broken links.

Phase 4: Deploy and verify arrival — DONE (verified 2026-10-02)​

PR #523 merged to main at e9e33a5. Image build (run 37070426629) succeeded; both digests read from the release's own uis-artifact.json, not reconstructed, and cross-checked against the full build log before sending:

artifact_digest  sha256:f91d7dee689f2cbbeb326d130703524a5e4e21bc8d8332832cba0490dd9a7ff4  ghcr.io/terchris/atlas-data/uis:v20261002-e9e33a5
image_digest     sha256:4f32b9a13b4e3b4e2ff5dc9d54ef05b26ab6148d550f05598e34a166eb811c7c  ghcr.io/terchris/atlas-data:v20261002-e9e33a5

lands-with.sh 57bb5f9..e9e33a5 output was clean (no false-positive this time — this range starts right after ssb-12451's own close-out commit, so only nav-helt-ledige's own ingest directory changed): monthly_sources_refresh (re-fetch) then transform_and_publish, plus a template-info pin (this range changes atlas-data/template-info.yaml's counts).

Two tasks filed, per the pin/run split ([[a-pin-and-a-run-are-two-actions]]):

  • Deploy request to imac: urb-agents#1813 — names raw.nav_helt_ledige (5,744 rows as of today's validation run, caveat stated: NAV's live file grows a column every month), both relations to check row counts on (indicators__nav_helt_ledige, atlas_inventory).
  • Pin nomination to ops-dev: urb-agents#1814 — both digests copied verbatim from uis-artifact.json, not reconstructed, per the one prior incident where an unlabelled digest was refused at the catalogue.

Also found and fixed before merge, not after: CI's render-template-info.sh caught a stale "59 raw BASE TABLEs" claim in template-info.yaml (migrations now create 60) — fixed in a follow-up commit on the same PR before merge. While fixing it, found and corrected a second, adjacent false claim in the same paragraph ("EVERY NUMBER IN IT IS NOW WRITTEN BY uis/generate-holdings.py" — checked directly: that script has zero substitutions for "raw" anywhere, so the raw-table figure is hand-maintained and had already drifted stale before this PR, independent of it).

imac's deploy report (urb-agents#1813): monthly_sources_refresh SUCCESS (76.7s) — raw.nav_helt_ledige row count 5,744, exact match to prediction. transform_and_publish SUCCESS (474.3s) — watched PostgreSQL's log live throughout for #1810-style corruption; zero events. Regression check against 6 prior sources (nav_uforetrygd, nav_aap, bufdir_barnevern, imdi_bosetting, udir_gsi, ssb_12451, husbanken_bostotte) — all unchanged. One real discrepancy flagged rather than silently reconciled: this plan's own Phase 4 request had predicted atlas_inventory would also carry a mart_indicators__nav_helt_ledige entry; imac found that endpoint doesn't exist (404) and isn't in inventory.

ops-dev's pin report (urb-agents#1814): tag v20261002-e9e33a5 pushed to dev-templates main, catalogue text regenerated (this bump also absorbed 11 days of accumulated drift: nav-aap, nav-uforetrygd and PR #482 had landed since the last catalogue bump). Separately flagged, not resolved: template-info.yaml's operational.automation text ("ships stopped") may be stale against PR #482's default_status=RUNNING, measured live in urb-agents#1794 — filed as a cross-cutting note in 1PRIORITY.md rather than left to die in a closed bus task, since it's unrelated to this source.

Both independently re-verified against the live public API before closing either task — neither report was taken on its own word:

  • GET /indicators__nav_helt_ledige?limit=1 → real row (Oslo, January, 12125 — matches the captured live workbook exactly); Content-Range / atlas_inventory both confirm 5744/5744.
  • GET /meta_sources?source_id=eq.nav-helt-ledige → served_as: ["indicators__nav_helt_ledige"], non-empty.
  • Spot-checked Svalbard (2100 → region_kind=svalbard, kommune_nr=NULL, [13,9,11,14,20,19,16,18]), Ukjent (→ unknown, NULL, [13,14,15,15,13,13,15,20]), and Utsira's mixed real/suppressed row (1151 → [null,null,null,null,4,null,4,null]) — all three match the live HL060 workbook and the golden-fixture tests exactly.
  • Resolved the mart_indicators__ discrepancy as a mistake in this plan's own Phase 4 write-up, not a defect: GET /mart_indicators__nav_aap — an existing source shipped weeks ago — returns the identical 404 from the live API, and its own atlas_inventory also lists only indicators__nav_aap. The mart_indicators__* dbt model selects from indicators__* and is never itself exposed as a separate api_v1 endpoint under that name; this plan's row-count prediction was simply wrong, carried from mart_atlas_inventory.sql's {relation, mart} dict where mart only names the dbt model. Worth remembering for the next source's Phase 4 write-up.

Both bus tasks closed with the verification evidence attached as comments.

Acceptance Criteria​

  • The mechanism is verified live, and matches the investigation's description this time — NAV's own HL060 table, kommune-resolved, confirmed by direct download and inspection.
  • Licence independently confirmed for this specific candidate — CC BY 4.0, verified against this table's own "Kilde: NAV" provenance, not assumed by family resemblance to nav-uforetrygd/nav-aap alone.
  • nav-helt-ledige ingests cleanly from the live HL060 workbook with zero rows silently dropped, including the Svalbard pseudo-kommune and the bare Ukjent bucket both represented — 5,744 rows written (359 regions × 8 months × 2 sheets).
  • raw.nav_helt_ledige stores region_code as NAV publishes it, including the literal Ukjent label with no numeric code — represented, not normalised away.
  • indicators__nav_helt_ledige and mart_indicators__nav_helt_ledige build and test clean against real loaded data, with 2100 and Ukjent resolving through classify_region_code exactly as predicted (PASS=19 WARN=0 ERROR=0).
  • nav-helt-ledige appears in meta_sources.served_as after a real deploy, independently verified via live curl — served_as: ["indicators__nav_helt_ledige"].
  • Golden-file tests cover: an ordinary kommune row, the Svalbard pseudo-kommune, the Ukjent block, and a suppressed row with mixed real/suppressed cells in the same row — 26 tests, all passing.
  • The investigation and 1PRIORITY.md are updated to mark this candidate shipped.

Implementation Notes​

  • Write a fresh parser — this table's classification rule is the simplest of the NAV-Excel family so far. See [Q2]. "Exactly 4 leading digits" alone correctly separates kommune rows from every header/total row in this table, with no "I alt" prefix check needed — porting either nav-uforetrygd's or nav-aap's parser would import a check this table doesn't need.
  • No new Dagster job. monthly_sources_refresh already exists; this is a registration, not new infrastructure — the third source on it after nav-uforetrygd/nav-aap.
  • Two sentinel shapes, both already have precedent — confirm, don't assume. 2100 Svalbard matches udir-gsi/husbanken-bostotte's finding; bare Ukjent matches nav-aap's. Verify both resolve as predicted with a real dbt test during Phase 3, same discipline as every prior source.
  • A different Statistikklov paragraph than AAP155's. § 7-1 here, § 2-6 there — don't assume every NAV table cites the same section; read each table's own methodology sheet.
  • Historical backfill genuinely unresolved, not just deferred by convention — see [Q1].

Outcome​

Shipped end to end, 2026-10-02: ingest (5,744 rows, zero dropped) → dbt staging and api_v1 publication → live cluster deploy → independently re-verified arrival. Atlas's fourth NAV-adjacent source and third on monthly_sources_refresh. Unlike nav-sykefravaer (PLAN-013), the investigation's mechanism held exactly as described this time — NAV's own HL060 table is genuinely kommune-resolved — worth confirming explicitly rather than assuming "the description held" after two sources in a row where it hadn't.

A fourth distinct pivot shape within the NAV-Excel family, but the simplest to classify yet: neither the bare fylke header row nor the "I alt <name>" rollup carries any digit at all, so "label starts with exactly 4 digits" alone separates kommune rows — a fresh parser, not adapted from nav-aap's or nav-uforetrygd's (per [Q2]). Both sentinel shapes already had precedent and resolved exactly as predicted on the first dbt build, with zero relationship-test failures: the Svalbard pseudo-kommune 2100 through classify_region_code's existing svalbard branch, and the literal Ukjent through its existing unknown branch — confirmed directly against both the local build and the live API, not assumed to work.

One genuinely new finding, caught against real data rather than assumed from the investigation: unlike nav-aap, where Ukjent exists only in the Antall sheet, here it is present in BOTH sheets — its Prosent-sheet cells are all NAV's own suppression marker, not omitted. A second, structural finding ([Q5]): the Ukjent block's fylke-level header row and its kommune-level leaf row carry the identical bare label "Ukjent" — the header row is correctly dropped by the ingest's existing hasData guard, with no special-casing added.

Found and fixed before merge, not after: CI's render-template-info.sh caught a stale "59 raw BASE TABLEs" claim in template-info.yaml left over from before this PR (migrations now create 60). While fixing it, found and corrected an adjacent false claim in the same paragraph — it said every number there was "now written by uis/generate-holdings.py", but that script has zero substitutions for the raw-table figure; it was hand-maintained and had already drifted stale independent of this change.

One deploy-verification mistake, caught and corrected rather than silently reconciled: this plan's own Phase 4 deploy request predicted a second mart_indicators__nav_helt_ledige inventory entry that doesn't exist. imac flagged the discrepancy instead of quietly resolving it; checking mart_indicators__nav_aap (an existing, long-shipped sibling) confirmed the same 404 there too — the mart_indicators__* view is never itself an exposed api_v1 endpoint, so the prediction was wrong in the request template, not a defect in the deploy. Also surfaced, filed separately rather than left in a closed bus task: ops-dev flagged that template-info.yaml's own operational.automation text may be stale against PR #482's default_status=RUNNING behaviour — unrelated to this source, tracked as a cross-cutting note in 1PRIORITY.md.

Plugs the short-tail labour-market signal slotting between nav-uforetrygd (long-tail disability outcome) and nav-aap (transitional work-assessment benefit) — Report #5 (Income & Welfare) and an axis for Report #4 (Mental-Health Triangulation).

Files to Modify​

  • atlas-data/ingest/src/sources/nav-helt-ledige/manifest.yml (new)
  • atlas-data/ingest/src/sources/nav-helt-ledige/index.ts (new)
  • atlas-data/ingest/src/sources/nav-helt-ledige/parse.ts (new)
  • atlas-data/ingest/src/sources/nav-helt-ledige/fetch_retry.ts (new, copied)
  • atlas-data/ingest/src/sources/nav-helt-ledige/README.md (new)
  • atlas-data/ingest/src/sources/nav-helt-ledige/__tests__/ (new)
  • atlas-data/ingest/package.json (ingest:nav-helt-ledige script — add this BEFORE shipping, verify with npm run ingest:nav-helt-ledige, not a direct tsx call)
  • atlas-data/migrations/<next>_raw_nav_helt_ledige.sql (new)
  • atlas-data/dagster/atlas_data/assets/raw_other.py, schedules.py (asset registration — monthly cadence, existing job)
  • atlas-data/dbt/models/indicators/sources.yml, indicators__nav_helt_ledige.sql (new), schema.yml
  • atlas-data/dbt/models/marts/api/mart_indicators__nav_helt_ledige.sql (new), schema.yml
  • atlas-data/dbt/models/marts/api/mart_atlas_inventory.sql (depends_on + counted list)
  • atlas-data/dbt/scripts/generate_api_v1.py (SCHEMA_COMMENT's nav listing — bump the count, add alongside nav-uforetrygd/nav-aap)
  • atlas-data/template-info.yaml, website/docs/developers/index.md (regenerated counts)
  • website/docs/ai-developer/plans/backlog/INVESTIGATE-new-norwegian-public-sources.md (mark shipped)
  • website/docs/ai-developer/plans/backlog/1PRIORITY.md (mark shipped)