Task: Generate data-scope populate archetypes from dataset org-models

• Only slovaris has a live model. It uses two legacy shapes: the model.json batch manifest (files[] -> {name,model,template}) and the bare-filename map get_template_mappings (country_currency.json etc.). Driving 8 templates: catalog, methodology, dataset, dataset_dependency, tag, flag, currency, country populate (+ a master include).
• Shape #3 (*_data.json / dataset.code) is DEAD at runtime — zero *_data.json files exist; get_populate_template_mappings / get_non_iso_currency_template_mappings and the is_data_model dispatch branch never fire. The committed fpml/, refdata/, dq/ populate .sql are orphaned artefacts whose models were removed.
• sql_populate_function_refdata is schema, not data — migrates into ores.sql.schema (prime origin's consolidation area; out of scope here).
• seeder.sh is broken — it execs a nonexistent projects/ores.codegen/src/generator.py; must be repointed at the graph path.
• Engine seam (archetype-granularity activation + #+default: disabled) is already built in B8. Facet stub ores.sql.populate.org exists (empty).
• Heavy data ENRICHMENT lives in core.py (flag SVGs + currency defaults ~1483, dataset dependency expansion ~1526) tied to the JSON files — must be preserved.

1. Retire the dead shape #3 rather than migrate it: remove get_populate_template_mappings, get_non_iso_currency_template_mappings, the is_data_model dispatch branch + dataset.code output block, and the dead orphan docs (sql_non_iso_currency_populate, sql_dataset_refdata, sql_populate_refdata). No models -> output-neutral.
2. Migrate the live slovaris archetypes (8 + master include) into ores.sql.populate as addressed, #+default: disabled nodes; #+output: projects/ores.sql/populate/{dataset}/{prefix}_<x>_populate.sql.
3. Dataset-as-org-model (per B8 Decision B): promote datasets/slovaris/dataset_overview.org to the model — #+type: ores.codegen.dataset, a drawer enabling its ores.sql.populate.* archetypes, and #+ keys for prefix/dir; payloads referenced (the existing JSON files), not inlined, so the core.py enrichment is reused by a new dataset loader.
4. Engine: resolve_output_path gains a dataset branch ({dataset},{prefix}); a dataset loader feeds the referenced JSON payloads (+ enrichment) into the template context.
5. Repoint seeder.sh at the graph (codegen.sh generate --model dataset_overview.org --address ores.sql.populate).
6. Verify byte-identical against the committed solvaris_* outputs (note the slovaris-dir / solvaris-prefix divergence — standardise as part of this).

prime origin is concurrently consolidating the schema create templates (B1+B7+ temporal) on feature/migrate-country-unified-model, heavily editing the SAME generate_from_model dispatch ladder + mapping functions in core.py. To avoid core.py merge conflicts, sequence: do the conflict-free doc-graph + dataset-doc work here now (library/templates + projects/ores.seeder only), and land the core.py engine changes (steps 1,4,5 retirement/loader/branch) AFTER prime origin's schema consolidation merges, then rebase.

Removed from core.py (all provably dead — zero *_data.json models exist): get_populate_template_mappings, get_non_iso_currency_template_mappings, is_entity_data_model, the is_data_model variable, the is_data_model dispatch branch, the data return in get_model_type, the dataset.code output-naming branch, and the ~87-line is_data_model enrichment block. Verified: 26 tests pass, module imports, and the live slovaris legacy path still generates all 9 solvaris_* files unchanged (output-neutral).

Deferred to B9/B10 (doc-side, entangled): the 3 now-dead orphan docs + .mustache (sql_non_iso_currency_populate, sql_dataset_refdata, sql_populate_refdata) and their doc-links in currency.org, sql_populate.org, architecture.org. Left as dead artefacts for now to avoid currency.org conflicts with prime origin and because legacy-literate-doc retirement is B10's scope.

DONE (committed):

• Step 4 primitives: ores.codegen.dataset -> model_type dataset; resolve_output_path dataset branch ({dataset}/{prefix}); load_graph captures #+data_source on archetypes. 27 tests.

REMAINING WIRING — approach (B), reuses the existing per-file enrichment (lower risk than a merged-context refactor):

1. Put #+data_source on each of the 9 populate archetype docs — the dataset- relative payload file the template renders from: catalog->catalogs.json, methodology->manifest.json, dataset->datasets.json, dataset_dependency->datasets.json, tag->tags.json, flag/currency/country->country_currency.json, batch_execute->model.json (the manifest, for {{#model.files}}).
2. load_org_dataset_model (org_loader): read the promoted dataset doc -> {dataset: {name, prefix}} (for resolve_output_path) + the dataset dir; wire into load_model routing.
3. resolve_targets: thread each unit's data_source through; read the dataset name/prefix from the model for output resolution.
4. _generate_single: for dataset units, call the EXISTING generate_from_model(dataset_dir/data_source, target_template, target_output=, prefix=)= — the legacy per-file path — so all enrichment (flag SVGs, currency defaults, dataset-dependency expansion, manifest methodology lift) is reused unchanged.
5. Promote datasets/slovaris/dataset_overview.org: #+type: ores.codegen.dataset, #+name:/#+prefix: (solvaris), drawer enabling its 9 ores.sql.populate.* archetypes (default-off facet -> opt-in).
6. Repoint seeder.sh at codegen.sh generate --model dataset_overview.org --address ores.sql.populate.
7. Verify byte-identical vs committed solvaris_* (standardise the slovaris-dir / solvaris-prefix divergence as part of this).

This is the end-to-end wiring + verification — best done as one focused pass.

Executed approach (B) end to end; the data-scope path now generates through the graph:

• load_org_dataset_model (org_loader) -> {dataset: {name, prefix}}; routed in load_model on #+type: ores.codegen.dataset.
• resolve_targets threads each archetype's #+data_source: into its unit; _generate_single renders dataset units from dataset_dir/data_source via the existing per-file generate_from_model (is_processing_batch=True suppresses the model.json batch dispatch), reusing all enrichment.
• 9 archetype docs carry #+data_source: (catalog->catalogs.json, methodology->manifest.json, dataset/dataset_dependency->datasets.json, tag->tags.json, flag/currency/country->country_currency.json, batch_execute->model.json).
• dataset_overview.org promoted: #+type: ores.codegen.dataset, #+name:/#+prefix: solvaris, drawer :ores.sql.populate.enabled: true (opts the default-off facet in -> all 9 archetypes enable).
• seeder.sh repointed at codegen.sh generate --model dataset_overview.org --address ores.sql.populate; generate_solvaris_refdata.sh already redirects here.

STANDARDISATION (full, per decision): the populate tree had two divergent committed dirs — live populate/slovaris/ (current-code output) and stale populate/solvaris/ (May, different code), with the DB actually \ir'ing solvaris/solvaris_populate.sql. Resolved to a single populate/solvaris/: dir==prefix==solvaris; master renamed \{prefix}.sql -> \{prefix}_populate.sql (in the archetype #+output: and the generate_from_model master special case) to match every sibling <dir>/<dir>_populate.sql master and the existing catalogues_populate.sql \ir (no consumer change needed). Deleted the stale slovaris/ dir.

VERIFICATION: the 8 populate files are byte-identical to current-code output; the master is byte-identical content, renamed to solvaris_populate.sql. seeder.sh generate slovaris is idempotent. 30 codegen tests pass (added tests/test_data_scope.py for the loader, output-path branch, and unit threading).

SCHEMA-DRIFT FIX (surfaced by the DB-load verification): standardising to the live seeder output broke compass db recreate — the currency template emitted column asset_class, but the schema had renamed it to monetary_nature (the previously-loaded solvaris/ dir was an older, schema-matching artefact; the live template output had never been validated against the schema). A full static cross-check (every insert column-list vs create table, every upsert call's arg-count vs its signature) found this was the only mismatch across all 8 archetypes. Fixed in ores.sql.populate.currency.org (insert target column + summary query references; the internal VALUES-CTE alias and the {{asset_class}} data lookup are schema-independent and unchanged), re-tangled, regenerated.

DB VERIFICATION: compass db recreate -y -k now completes clean — solvaris populates 100 currencies + 100 countries; monetary_nature carries 4 distinct values. Seed loads end to end.

Remaining: raise the PR.