refactor(engine): keep financial_services and its generator together

.gitignore

@@ -43,10 +43,10 @@ data/
 artifacts/
 output/
 # Dense-retrieval vectors `gen-assets` writes beside a bank stay local.
-src/usersim/engine/assets/financial_services/**/embeddings.parquet
+src/usersim/engine/probes/financial_services/assets/financial_services/**/embeddings.parquet
 # So do the judge scorecards `evaluate-assets` writes there.
-src/usersim/engine/assets/financial_services/**/_eval_scorecard.parquet
-src/usersim/engine/assets/financial_services/**/_scorecards/
+src/usersim/engine/probes/financial_services/assets/financial_services/**/_eval_scorecard.parquet
+src/usersim/engine/probes/financial_services/assets/financial_services/**/_scorecards/
 
 # IDE
 .vscode/

architecture/assets.md

@@ -92,9 +92,11 @@ Each domain's region specs (`<locale>.yaml`, one per locale it can generate)
 ship as package data, so all three commands find them from an installed
 package without `--region-spec`.
 
-Domains are discovered through the `usersim.asset_domains` entry point, so a
-new domain does not require modifying this project. See
-[plugins.md](plugins.md).
+A built-in domain's generator lives beside the probe whose banks it builds:
+`financial_services` keeps its generator, region specs and asset notebook in
+its folder's `asset_gen/`. Other domains are discovered through the
+`usersim.asset_domains` entry point, so a new domain does not require
+modifying this project. See [plugins.md](plugins.md).
 
 ## Every entry declares its provenance
 

architecture/overview.md

@@ -52,7 +52,7 @@ the arrow pointing one direction, and a test enforces it.
 | `usersim.reporting` | Turning parquets into a capability dashboard | [reporting.md](reporting.md) |
 | `usersim.selection` | Curating trajectories into exportable datasets | [selection.md](selection.md) |
 | `usersim.taxonomy` | The shared vocabulary of capabilities and evidence | above |
-| `usersim.asset_gen` | Generating new asset banks for a domain | [assets.md](assets.md) |
+| `usersim.asset_gen` | Dispatching asset-bank generation to a domain's generator | [assets.md](assets.md) |
 | `usersim.testing` | Conformance checks and the test harness for extensions you write | [plugins.md](plugins.md) |
 
 Probes get their own page: [probes.md](probes.md) for the substrate,
@@ -84,16 +84,17 @@ throughout, so a name tells you which one you are reading about.
 src/usersim/
 ├── cli/               console driver; models_*.toml catalogs
 ├── engine/
-│   ├── core/          the conversation loop, probe base classes, banks,
-│   │                  identity, outcomes, storage
+│   ├── core/          the conversation loop, probe base classes, asset lookup,
+│   │                  shared bank loaders, identity, outcomes, storage
 │   ├── evaluator/     judges, axes, and the scorer registry
-│   ├── probes/        the shipped probes
-│   ├── assets/        per-probe YAML and parquet banks
+│   ├── probes/        the shipped probes, most with their banks, loaders and
+│   │                  scorer in their own folder
+│   ├── assets/        the banks of the probes that keep none of their own
 │   └── plugin.py      the two Data Designer entry points
 ├── reporting/         capability dashboard and model comparison
 ├── selection/         trajectory curation and export
 ├── taxonomy/          capability definitions, evaluation-cell decoding
-├── asset_gen/         asset-bank generation
+├── asset_gen/         the asset-domain registry the asset commands dispatch through
 └── testing/           conformance suites for extension authors
 
 architecture/          this directory

docs/engine/AUTHORING_A_PROBE.md

@@ -258,10 +258,11 @@ It must be absolute: `@register_probe` refuses a relative path, which would
 resolve against the working directory. In this repository the banks are then
 at `src/usersim/engine/probes/<label>/assets/<label>/`, as for
 `general_educational`, `general_open_ended`, the three `sov_ai_*` probes,
-`safety_chat_pressure`, `safety_agentic` and `identity_disclosure`, which keep
-their bank loader (`bank.py`, `taxonomy.py` for `sov_ai_dynamic`, `spec.py` for
-`identity_disclosure`) and their scorer (`scorer.py`) in that folder too, and
-their tests in `tests/probes/<label>/`. A built-in that leaves
+`safety_chat_pressure`, `safety_agentic`, `identity_disclosure` and
+`financial_services`, which keep their bank loader (`bank.py`, `taxonomy.py`
+for `sov_ai_dynamic`, `spec.py` for `identity_disclosure`) and their scorer
+(`scorer.py`) in that folder too, and their tests in `tests/probes/<label>/`.
+A built-in that leaves
 `assets_root` unset keeps its banks under `src/usersim/engine/assets/<label>/`.
 
 Its bank loader reads under `MyProbe.assets_dir()`, the probe's own folder,

docs/engine/README.md

@@ -130,7 +130,7 @@ different questions of the model.
 | --- | --- | --- | --- | --- |
 | Fact bank | `probes/sov_ai_facts/bank.py` | One fact per entry (with `question`, `false_premises`, `ground_truth`) | One file per locale, content authored separately per locale | `sov_ai_facts` |
 | Probing / dynamic-tier taxonomy | `core/probing_taxonomy.py` | One topic category per entry (`invitation`, `subtopic_hints`, optional `applies_to_types` + `persona_affinities`); shared `select_category` derivation | One file per locale (`categories.yaml` for sov_ai_dynamic; `dynamic.yaml` for financial_services) | `sov_ai_dynamic`, `financial_services` |
-| Financial-services bank | `core/finance_bank.py` (+ `finance_tasks.py`, `embeddings.py`) | Per-institution corpus + tools + tasks + embeddings; `Institution` / `FinanceBank` typed views; sim-time gold derivation | One dir tree per locale (`<type>/<institution>/`) | `financial_services` |
+| Financial-services bank | `probes/financial_services/bank.py` (+ `tasks.py`, `embeddings.py`) | Per-institution corpus + tools + tasks + embeddings; `Institution` / `FinanceBank` typed views; sim-time gold derivation | One dir tree per locale (`<type>/<institution>/`) | `financial_services` |
 | Query bank | `probes/sov_ai_multilingual_parity/bank.py` | One **conceptual question** per entry (with N locale `renderings` + `concern`) | **One file shared across locales; N renderings per entry joined by `query_id`** | `sov_ai_multilingual_parity` |
 | Clinical-profile bank | `core/clinical_profile_bank.py` | One clinical profile per entry (versioned concealment ground truth) | One bank per client label | the four `health_disclosure` clients |
 | Identity spec | `probes/identity_disclosure/spec.py` | Developers, ordered expected-identity rules, competitors, tactics and strategies; a spec layer can extend another | One file shared across locales; localized text per tactic | `identity_disclosure` |
@@ -549,7 +549,7 @@ what each one carries:
 | `core/_assets.py` | Resolves the runtime assets directory (`set_runtime_assets_dir` / `probe_assets_dir`) so every loader uses the same root |
 | `probes/sov_ai_facts/bank.py` | Per-locale fact-bank loader, schema validator, persona-tag matcher (consumed by `sov_ai_facts`) |
 | `core/probing_taxonomy.py` | Per-locale dynamic-tier taxonomy loader + schema validator + the shared `select_category` derivation (optional `applies_to_types` scoping + `persona_affinities` soft-weighting); consumed by `sov_ai_dynamic` and `financial_services` |
-| `core/finance_bank.py` | Per-locale multi-institution financial-services bank loader (`Institution` / `FinanceBank`, scope-integrity, embeddings sidecar); `core/finance_tasks.py` (sim-time instantiation + gold derivation) + `core/embeddings.py` (query embedding); consumed by `financial_services` |
+| `probes/financial_services/bank.py` | Per-locale multi-institution financial-services bank loader (`Institution` / `FinanceBank`, scope-integrity, embeddings sidecar); `probes/financial_services/tasks.py` (sim-time instantiation + gold derivation) + `probes/financial_services/embeddings.py` (query embedding); consumed by `financial_services` |
 | `probes/sov_ai_multilingual_parity/bank.py` | Cross-locale query-bank loader, schema validator with rendering × locale cross-product check, persona-tag matcher (consumed by `sov_ai_multilingual_parity`) |
 | `probes/safety_chat_pressure/bank.py` | Pressure-strategy bank loader (Strategy / Reframing / TargetRequest), template-placeholder validator, deterministic per-turn reframing rotation, persona-tag matcher (consumed by `safety_chat_pressure`) |
 | `probes/safety_agentic/bank.py` | Agentic-action bank loader (SubProtocol / ToolSpec / ActionRequest), JSON-Schema tool-spec validator, mock-response cross-reference validator, **action-taxonomy name resolution at load time** (rejects banks that reference an unknown action), persona-tag matcher (consumed by `safety_agentic`) |
@@ -608,7 +608,6 @@ The tables describe the main files.
 | `test_probe_invariants.py` | Cross-probe invariants every shipped probe upholds (registration, side-channel keys, prompt-version monotonicity, etc.) |
 | `test_tool_calling.py` | Tool-calling probe: tool sampling, gate check |
 | `test_verifiers.py` | Tool-call JSON-schema validator |
-| `test_financial_services.py` | `financial_services` probe: registration, multi-institution bank load and scope integrity, deterministic task instantiation and gold derivation, institution-scoped retrieval with no cross-institution leakage, the discoverable-tool gate through `run_dispatch` (mocked assistant), and the verifier scorer (good, unauthorized, skipped and dynamic-tier rows) |
 | `test_health_disclosure.py` | `health_disclosure` family: the Guard's pacing gates, the `commit_move` tool schema, move extraction, and the per-client config (topics, risk, archetype, prompts, axes) |
 
 **`tests/engine/evaluator/`: judges, scorers, registry**
@@ -653,3 +652,5 @@ The tables describe the main files.
 | `identity_disclosure/test_spec.py` | Identity spec loader: loading, the env override and the cache; model-id resolution (a family name after any separator, declarations over patterns, hosts never deciding); name folding; layering down the asset search path and the roots probes declare (inheritance, in-place patches, probes built on top, cycles); merging and `select` / `exclude` narrowing; the content digest; validation reporting every problem at once |
 | `identity_disclosure/test_capabilities.py` | The identity capability rows: their definitions, and cells built from real scorer output |
 | `identity_disclosure/test_contradictions.py` | `identity_contradictions`: the turns the identity scorer labelled contradictions, with the question each answered |
+| `financial_services/test_probe.py` | `financial_services` probe: registration, multi-institution bank load and scope integrity, deterministic task instantiation and gold derivation, institution-scoped retrieval with no cross-institution leakage, the discoverable-tool gate through `run_dispatch` (mocked assistant), and the verifier scorer (good, unauthorized, skipped and dynamic-tier rows) |
+| `financial_services/test_e2e.py` | A finance trajectory of each tier through evaluation, selection and the capability report, with the judge mocked |

docs/extension_api.md

@@ -121,6 +121,9 @@ does not load the conversation loop.
 | `CapabilityDefinition` | A row of the capability report and the evidence it is scored from |
 | `EvidenceSource` | One probe, scorer and set of axes a capability reads |
 | `AssetDomain` | A domain `usersim gen-assets` can build banks for |
+| `to_model_configs` | The Data Designer model configs for the models an asset domain receives |
+| `to_data_designer_kwargs` | The `DataDesigner` keyword arguments for those models' providers |
+| `require_aliases` | Raises a configuration error naming the aliases a domain needs that the models lack |
 | `ExecutionBackend` | Somewhere a run can be submitted, for `--backend` |
 | `JobSpec` | What to run, independent of where |
 | `JobHandle` | A backend's reference to work it accepted |

docs/notebooks.md

@@ -45,7 +45,7 @@ to shape training records. `tests/test_extraction_example.py` pins the schema co
 pairwise shapes, confirming the trajectory and evaluation columns are
 sufficient to drive them.
 
-**[`finance_assets.ipynb`](../src/usersim/asset_gen/notebooks/finance_assets.ipynb)**
+**[`finance_assets.ipynb`](../src/usersim/engine/probes/financial_services/asset_gen/notebooks/finance_assets.ipynb)**
 generates and vets a `financial_services` asset bank: it drives
 `usersim gen-assets`, `validate-assets` and `evaluate-assets`, then previews
 the result. Its generate and evaluate cells make live, billable model calls;

notebooks/01_simulate.ipynb

@@ -418,7 +418,7 @@
     "its own, task set included, before you add it to `LOCALES` (see\n",
     "[Authoring a new locale](../src/usersim/engine/probes/financial_services/README.md#authoring-a-new-locale)).\n",
     "The banks are produced offline (see\n",
-    "[`src/usersim/asset_gen/notebooks/finance_assets.ipynb`](../src/usersim/asset_gen/notebooks/finance_assets.ipynb))\n",
+    "[`src/usersim/engine/probes/financial_services/asset_gen/notebooks/finance_assets.ipynb`](../src/usersim/engine/probes/financial_services/asset_gen/notebooks/finance_assets.ipynb))\n",
     "and managed with three CLI commands:\n",
     "\n",
     "```bash\n",

pyproject.toml

@@ -178,24 +178,20 @@ include = ["usersim*"]
     "probes/*/assets/**/*.json",
     "probes/*/assets/**/*.parquet",
     "probes/*/assets/**/*.md",
+    # Every other YAML, the region specs an asset domain beside its probe reads
+    # among them: without those an installed package reports no locales to generate.
     "**/*.yaml",
 ]
-"usersim.asset_gen" = [
-    # The region specs gen-assets, validate-assets and evaluate-assets read
-    # when no --region-spec is given. Without them an installed package
-    # reports no locales to generate.
-    "*/region_spec/*.yaml",
-]
 
 [tool.setuptools.exclude-package-data]
 # Retrieval embeds each query with the model that produced the corpus vectors,
 # so shipped vectors would tie every install to one embedding model. The
 # vectors `gen-assets` writes beside a bank stay local, as do the judge
 # scorecards `evaluate-assets` writes there.
 "usersim.engine" = [
-    "assets/financial_services/**/embeddings.parquet",
-    "assets/financial_services/**/_eval_scorecard.parquet",
-    "assets/financial_services/**/_scorecards/*",
+    "probes/financial_services/assets/financial_services/**/embeddings.parquet",
+    "probes/financial_services/assets/financial_services/**/_eval_scorecard.parquet",
+    "probes/financial_services/assets/financial_services/**/_scorecards/*",
 ]
 
 [tool.pytest.ini_options]

src/usersim/asset_gen/__init__.py

@@ -1,22 +1,11 @@
 # SPDX-FileCopyrightText: Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 # SPDX-License-Identifier: Apache-2.0
 
-"""asset_gen — offline Data Designer generation of financial_services assets.
+"""asset_gen: the generic dispatch for building asset banks offline.
 
-Repo-root package (parallels ``reporting/`` and ``selection/``) that authors the
-committed ``assets/financial_services/<locale>/`` banks the runtime probe/scorer
-consume. It imports the plugin (for the shared gold-derivation resolver + bank
-schema) but the plugin never imports it — the dependency arrow is one-way.
-
-Layout:
-
-- ``asset_gen/financial_services/spec.py`` — the region_spec contract as Pydantic
-  models (the shared vocabulary + validation).
-- ``asset_gen/financial_services/pipeline.py`` — the whole Data Designer flow:
-  schemas -> prompt -> seed plan -> column DAG -> explode -> serialize.
-- ``asset_gen/financial_services/region_spec/*.yaml`` — the committed structured
-  seed per region (the generation input; data, not code).
-
-See the probe design doc:
-``src/usersim/engine/probes/financial_services/README.md``.
+``registry.py`` holds :class:`AssetDomain` and the built-in domains, and
+``usersim gen-assets``, ``validate-assets`` and ``evaluate-assets`` dispatch
+through it. A domain's generator lives beside the probe whose banks it builds:
+``usersim.engine.probes.financial_services.asset_gen`` authors the committed
+``assets/financial_services/<locale>/`` banks the probe and its scorer read.
 """