@@ -26,6 +26,10 @@ node_modules/
 *.db
 *.sqlite3
 
+# Module-managed runtime state (e.g. uploaded dataset files,
+# default storage_dir for SM_DATASETS_STORAGE_DIR).
+var/
+
 # file_storage filesystem backend default root (override via SM_FILE_STORAGE_FS_ROOT_PATH).
 uploads/
 
 
@@ -75,15 +75,12 @@ def emit_frontend_types(registry: I18nRegistry, project_root: Path) -> None:
         logger.exception("Failed to write generated-resources.ts — frontend types will be stale")
 
 
-_RESOURCES_HEADER = """\
-// AUTO-GENERATED by simple_module_hosting.i18n_manifest — do not edit by hand.
-// Regenerate by booting the host in development mode.
-"""
+_RESOURCES_HEADER = (
+    "// AUTO-GENERATED by simple_module_hosting.i18n_manifest — do not edit by hand.\n"
+    "// Regenerate by booting the host in development mode."
+)
 
-_KEYS_HEADER = """\
-// AUTO-GENERATED by simple_module_hosting.i18n_manifest — do not edit by hand.
-// Regenerate by booting the host in development mode.
-"""
+_KEYS_HEADER = _RESOURCES_HEADER
 
 _PLURAL_SUFFIXES: tuple[str, ...] = tuple(f"_{c}" for c in PLURAL_CATEGORIES)
 
@@ -125,7 +122,7 @@ def _render_resources(keys: list[str]) -> str:
 def _render_keys(keys: list[str]) -> str:
     """Render the nested ``keys`` constant tree."""
     tree = _build_key_tree(keys)
-    return f"{_KEYS_HEADER}\nexport const keys = {_serialize(tree, indent=0)} as const;\n"
+    return f"{_KEYS_HEADER}\n\nexport const keys = {_serialize(tree, indent=0)} as const;\n"
 
 
 def _build_key_tree(flat_keys: list[str]) -> dict[str, Any]:
 
@@ -76,7 +76,12 @@ async def check_broken():
         assert "connection refused" in data["checks"]["broken_service"]["detail"]
 
     async def test_ready_no_checks_is_healthy(self, client: httpx.AsyncClient):
+        # Modules may register their own health checks (e.g. gis_datasets
+        # probes its storage dir). Assert the aggregate is healthy and
+        # every reported check is healthy, rather than requiring zero
+        # checks — which would break whenever a new module joins boot.
         resp = await client.get("/health/ready")
         data = resp.json()
         assert data["status"] == "healthy"
-        assert data["checks"] == {}
+        for name, payload in data["checks"].items():
+            assert payload["status"] == "healthy", f"{name} not healthy: {payload}"
@@ -0,0 +1,69 @@
+"""add datasets table
+
+Revision ID: dad9a134290f
+Revises: b7e1af4c9d02
+Create Date: 2026-04-19 11:51:19.145277
+"""
+
+from collections.abc import Sequence
+
+import sqlalchemy as sa
+from alembic import op
+
+# revision identifiers, used by Alembic.
+revision: str = "dad9a134290f"
+down_revision: str | None = "b7e1af4c9d02"
+branch_labels: str | Sequence[str] | None = ("datasets",)
+depends_on: str | Sequence[str] | None = None
+
+
+def upgrade() -> None:
+    # ### commands auto generated by Alembic - please adjust! ###
+    op.create_table(
+        "datasets_dataset",
+        sa.Column("is_deleted", sa.Boolean(), nullable=False),
+        sa.Column("deleted_at", sa.DateTime(timezone=True), nullable=True),
+        sa.Column("deleted_by", sa.String(length=255), nullable=True),
+        sa.Column(
+            "created_at",
+            sa.DateTime(timezone=True),
+            server_default=sa.text("(CURRENT_TIMESTAMP)"),
+            nullable=False,
+        ),
+        sa.Column("updated_at", sa.DateTime(timezone=True), nullable=True),
+        sa.Column("created_by", sa.String(length=255), nullable=True),
+        sa.Column("updated_by", sa.String(length=255), nullable=True),
+        sa.Column("id", sa.Integer(), nullable=False),
+        sa.Column("name", sa.String(length=200), nullable=False),
+        sa.Column("slug", sa.String(length=200), nullable=False),
+        sa.Column("kind", sa.String(length=32), nullable=False),
+        sa.Column("description", sa.String(length=2000), nullable=True),
+        sa.Column("original_filename", sa.String(length=255), nullable=False),
+        sa.Column("mime_type", sa.String(length=127), nullable=True),
+        sa.Column("size_bytes", sa.Integer(), nullable=False),
+        sa.Column("storage_key", sa.String(length=512), nullable=False),
+        sa.Column("crs", sa.String(length=64), nullable=True),
+        sa.Column("bbox_min_x", sa.Float(), nullable=True),
+        sa.Column("bbox_min_y", sa.Float(), nullable=True),
+        sa.Column("bbox_max_x", sa.Float(), nullable=True),
+        sa.Column("bbox_max_y", sa.Float(), nullable=True),
+        sa.Column("feature_count", sa.Integer(), nullable=True),
+        sa.Column("band_count", sa.Integer(), nullable=True),
+        sa.Column("extraction_status", sa.String(length=16), nullable=False),
+        sa.PrimaryKeyConstraint("id", name=op.f("pk_datasets_dataset")),
+    )
+    op.create_index(
+        "ix_datasets_dataset_is_deleted", "datasets_dataset", ["is_deleted"], unique=False
+    )
+    op.create_index(op.f("ix_datasets_dataset_kind"), "datasets_dataset", ["kind"], unique=False)
+    op.create_index(op.f("ix_datasets_dataset_slug"), "datasets_dataset", ["slug"], unique=True)
+    # ### end Alembic commands ###
+
+
+def downgrade() -> None:
+    # ### commands auto generated by Alembic - please adjust! ###
+    op.drop_index(op.f("ix_datasets_dataset_slug"), table_name="datasets_dataset")
+    op.drop_index(op.f("ix_datasets_dataset_kind"), table_name="datasets_dataset")
+    op.drop_index("ix_datasets_dataset_is_deleted", table_name="datasets_dataset")
+    op.drop_table("datasets_dataset")
+    # ### end Alembic commands ###
@@ -8,6 +8,7 @@ dependencies = [
     "auth",
     "dashboard",
     "products",
+    "datasets",
     "permissions",
     "background-tasks",
     "file-storage",
@@ -21,6 +22,7 @@ simple-module-hosting = { workspace = true }
 auth = { workspace = true }
 products = { workspace = true }
 dashboard = { workspace = true }
+datasets = { workspace = true }
 permissions = { workspace = true }
 background-tasks = { workspace = true }
 file-storage = { workspace = true }
 
@@ -0,0 +1 @@
+"""Datasets module."""
@@ -0,0 +1,146 @@
+"""Stable identifiers for the datasets module.
+
+Every string that identifies a permission, role, page, module dependency,
+event, route, or config key lives here. Everything else in the module
+imports from this file, so ``rg`` can prove the module has no scattered
+magic strings.
+"""
+
+from __future__ import annotations
+
+from typing import Final
+
+# ── Module identity ──────────────────────────────────────────────────
+MODULE_NAME: Final = "datasets"
+MODULE_PASCAL: Final = "Datasets"
+MODULE_DISPLAY_NAME: Final = "Datasets"
+
+# ── Configuration ────────────────────────────────────────────────────
+ENV_PREFIX: Final = "SM_DATASETS_"
+
+# ── Routing ──────────────────────────────────────────────────────────
+ROUTE_PREFIX_API: Final = "/api/datasets"
+ROUTE_PREFIX_VIEW: Final = "/datasets"
+
+# REST sub-paths (joined to ROUTE_PREFIX_API by the router)
+PATH_DOWNLOAD: Final = "/{dataset_id}/download"
+PATH_DATASET: Final = "/{dataset_id}"
+
+# View routes (relative to ROUTE_PREFIX_VIEW, used in redirects)
+REDIRECT_BROWSE: Final = "/datasets/"
+
+# ── Module dependencies (for ModuleMeta.depends_on) ──────────────────
+MODULE_FILE_STORAGE: Final = "FileStorage"
+MODULE_BACKGROUND_TASKS: Final = "BackgroundTasks"
+
+# ── i18n ─────────────────────────────────────────────────────────────
+LOCALE_NAMESPACE: Final = MODULE_NAME
+
+# ── Inertia page identifiers ─────────────────────────────────────────
+PAGE_BROWSE: Final = "Datasets/Browse"
+PAGE_CREATE: Final = "Datasets/Create"
+PAGE_EDIT: Final = "Datasets/Edit"
+PAGE_SHOW: Final = "Datasets/Show"
+
+# ── Permissions ──────────────────────────────────────────────────────
+PERM_DATASETS_VIEW: Final = "datasets.view"
+PERM_DATASETS_UPLOAD: Final = "datasets.upload"
+PERM_DATASETS_EDIT: Final = "datasets.edit"
+PERM_DATASETS_DELETE: Final = "datasets.delete"
+
+PERMISSION_GROUP: Final = "Datasets"
+ALL_PERMISSIONS: Final = (
+    PERM_DATASETS_VIEW,
+    PERM_DATASETS_UPLOAD,
+    PERM_DATASETS_EDIT,
+    PERM_DATASETS_DELETE,
+)
+
+# ── Menu ─────────────────────────────────────────────────────────────
+MENU_LABEL: Final = "Datasets"
+MENU_ICON: Final = "layers"
+MENU_ORDER: Final = 40
+
+# ── Celery tasks ─────────────────────────────────────────────────────
+TASK_EXTRACT_METADATA: Final = "datasets.extract_metadata"
+
+# ── Feature flags ────────────────────────────────────────────────────
+FLAG_AUTO_EXTRACT: Final = "datasets.auto_extract"
+FLAG_ALLOW_RASTER_UPLOADS: Final = "datasets.allow_raster_uploads"
+
+# ── Settings keys (registered with the settings module) ──────────────
+SETTING_MAX_UPLOAD_MB: Final = "datasets.max_upload_mb"
+SETTING_DEFAULT_KIND: Final = "datasets.default_kind"
+
+# ── Roles ────────────────────────────────────────────────────────────
+ROLE_USER: Final = "user"
+ROLE_ADMIN: Final = "admin"
+
+# Permissions granted to plain ``user`` role via ``map_role``. ``admin``
+# inherits everything through the framework's wildcard, so we don't list
+# it explicitly here.
+USER_ROLE_PERMISSIONS: Final = (
+    PERM_DATASETS_VIEW,
+    PERM_DATASETS_UPLOAD,
+)
+
+# ── Health check ─────────────────────────────────────────────────────
+HEALTH_CHECK_STORAGE: Final = "datasets.storage"
+
+# ── Storage ──────────────────────────────────────────────────────────
+STORAGE_KEY_PREFIX: Final = "datasets/"
+
+# ── Database ─────────────────────────────────────────────────────────
+SCHEMA_NAME: Final = "datasets"
+TABLE_DATASET: Final = "datasets_dataset"
+
+# ── Defaults ─────────────────────────────────────────────────────────
+DEFAULT_MAX_UPLOAD_MB: Final = 256
+DEFAULT_UPLOAD_CHUNK_SIZE: Final = 1024 * 1024  # 1 MB
+DEFAULT_PRESIGN_TTL_SECONDS: Final = 300  # 5 minutes
+DEFAULT_FALLBACK_FILENAME: Final = "upload.bin"
+DEFAULT_MIME_TYPE: Final = "application/octet-stream"
+DEFAULT_GEOJSON_CRS: Final = "EPSG:4326"
+
+
+class DatasetKind:
+    """Built-in dataset kinds.
+
+    The ``kind`` column is a free-form string so new providers can register
+    their own labels, but these are the ones the shipped extractors and
+    the UI know about.
+    """
+
+    VECTOR_GEOJSON: Final = "vector_geojson"
+    VECTOR_SHAPEFILE: Final = "vector_shapefile"
+    VECTOR_KML: Final = "vector_kml"
+    RASTER_GEOTIFF: Final = "raster_geotiff"
+    TABULAR_CSV: Final = "tabular_csv"
+    OTHER: Final = "other"
+
+
+ALL_KINDS: Final = (
+    DatasetKind.VECTOR_GEOJSON,
+    DatasetKind.VECTOR_SHAPEFILE,
+    DatasetKind.VECTOR_KML,
+    DatasetKind.RASTER_GEOTIFF,
+    DatasetKind.TABULAR_CSV,
+    DatasetKind.OTHER,
+)
+
+
+class ExtractionStatus:
+    """Values stored in ``Dataset.extraction_status``.
+
+    ``pending`` is set by the upload endpoint; ``ok``/``partial``/``failed``
+    are set by the Celery worker once extraction completes. ``manual`` is
+    the fallback for kinds the extractor doesn't understand; ``not_found``
+    is only used as a task-result marker (never stored).
+    """
+
+    PENDING: Final = "pending"
+    OK: Final = "ok"
+    PARTIAL: Final = "partial"
+    FAILED: Final = "failed"
+    MANUAL: Final = "manual"
+    NOT_FOUND: Final = "not_found"
@@ -0,0 +1,51 @@
+"""Datasets contracts — public interface for other modules.
+
+Downstream modules should import from here, not from
+``datasets.models`` or ``datasets.service`` internals. The SM009
+diagnostic enforces framework→plugin purity; this package is the
+supported surface for plugin→plugin coupling::
+
+    from datasets.contracts import (
+        DatasetOut,            # DTO returned by service lookups
+        DatasetFile,           # handle for stored file access
+        DatasetUploaded,       # event subscribers listen for
+        download_url,          # URL helper for UIs
+    )
+
+For the FastAPI dependency, prefer
+``from datasets.deps import DatasetServiceDep``. Type-hint against the
+concrete ``DatasetService`` — the module is single-impl, so there's no
+Protocol abstraction (matching the framework's "ship a Protocol only
+for real extension points" rule).
+"""
+
+from datasets.contracts.events import DatasetDeleted, DatasetUploaded
+from datasets.contracts.files import DatasetFile
+from datasets.contracts.schemas import (
+    KIND_VALUES,
+    DatasetKind,
+    DatasetOut,
+    DatasetUpdate,
+)
+from datasets.contracts.urls import (
+    API_PREFIX,
+    VIEW_PREFIX,
+    detail_url,
+    download_url,
+    show_url,
+)
+
+__all__ = [
+    "API_PREFIX",
+    "KIND_VALUES",
+    "VIEW_PREFIX",
+    "DatasetDeleted",
+    "DatasetFile",
+    "DatasetKind",
+    "DatasetOut",
+    "DatasetUpdate",
+    "DatasetUploaded",
+    "detail_url",
+    "download_url",
+    "show_url",
+]
@@ -0,0 +1,28 @@
+"""Domain events emitted by the Datasets module.
+
+Events carry the handful of fields subscribers most often route on
+(``slug``, ``kind``) so downstream handlers don't have to make a round
+trip to ``IDatasetService`` just to decide whether the event is for
+them. Anything beyond this — CRS, bbox, feature counts — is still a
+``get_by_id`` away.
+"""
+
+from __future__ import annotations
+
+from dataclasses import dataclass
+
+from simple_module_core.events import Event
+
+
+@dataclass
+class DatasetUploaded(Event):
+    dataset_id: int
+    name: str
+    slug: str
+    kind: str
+
+
+@dataclass
+class DatasetDeleted(Event):
+    dataset_id: int
+    slug: str