feat: define governed tenant erasure contracts
Module Package Release / publish-packages (push) Successful in 12s

This commit is contained in:
2026-08-24 16:00:06 +02:00
parent 9cb2080938
commit 9a3008002d
12 changed files with 725 additions and 20 deletions
+467
View File
@@ -0,0 +1,467 @@
from __future__ import annotations
from collections.abc import Mapping
from dataclasses import dataclass, field
from datetime import UTC, datetime
from typing import Literal, Protocol, runtime_checkable
TENANT_ERASURE_PROVIDER_CAPABILITY_PREFIX = "tenancy.erasure_provider."
TenantErasureDisposition = Literal[
"erase",
"retain",
"legal_hold",
"external_cleanup",
"key_destroy",
"backup_expiry",
"unavailable",
]
TenantErasureStepKind = Literal[
"export",
"erase",
"retain",
"external_cleanup",
"key_destroy",
"backup_expiry",
"verify",
]
TenantErasureResultState = Literal[
"completed",
"pending",
"blocked",
"outcome_unknown",
]
_DISPOSITIONS = frozenset(
{
"erase",
"retain",
"legal_hold",
"external_cleanup",
"key_destroy",
"backup_expiry",
"unavailable",
}
)
_STEP_KINDS = frozenset(
{
"export",
"erase",
"retain",
"external_cleanup",
"key_destroy",
"backup_expiry",
"verify",
}
)
_RESULT_STATES = frozenset(
{"completed", "pending", "blocked", "outcome_unknown"}
)
def _text(value: str, label: str, *, maximum: int) -> str:
normalized = value.strip()
if (
not normalized
or len(normalized) > maximum
or any(ord(character) < 32 for character in normalized)
):
raise ValueError(f"Tenant erasure {label} is invalid.")
return normalized
def _texts(
values: tuple[str, ...],
label: str,
*,
maximum_items: int = 100,
maximum_length: int = 500,
) -> tuple[str, ...]:
if len(values) > maximum_items:
raise ValueError(f"Tenant erasure {label} has too many entries.")
normalized = tuple(
_text(value, label, maximum=maximum_length) for value in values
)
if len(normalized) != len(set(normalized)):
raise ValueError(f"Tenant erasure {label} contains duplicates.")
return normalized
def _metrics(values: Mapping[str, int]) -> dict[str, int]:
if len(values) > 30:
raise ValueError("Tenant erasure metrics has too many entries.")
normalized: dict[str, int] = {}
for key, value in values.items():
normalized_key = _text(key, "metric key", maximum=80)
if type(value) is not int or value < 0:
raise ValueError("Tenant erasure metric values must be non-negative integers.")
normalized[normalized_key] = value
return normalized
@dataclass(frozen=True, slots=True)
class TenantErasureResource:
resource_type: str
count: int
disposition: TenantErasureDisposition
summary: str
governance_ref: str | None = None
external: bool = False
def __post_init__(self) -> None:
_text(self.resource_type, "resource type", maximum=120)
_text(self.summary, "resource summary", maximum=1000)
if type(self.count) is not int or self.count < 0:
raise ValueError("Tenant erasure resource count is invalid.")
if self.disposition not in _DISPOSITIONS:
raise ValueError("Tenant erasure resource disposition is invalid.")
if self.governance_ref is not None:
_text(self.governance_ref, "governance reference", maximum=300)
def to_dict(self) -> dict[str, object]:
return {
"resource_type": self.resource_type,
"count": self.count,
"disposition": self.disposition,
"summary": self.summary,
"governance_ref": self.governance_ref,
"external": self.external,
}
@dataclass(frozen=True, slots=True)
class TenantErasureStep:
step_id: str
kind: TenantErasureStepKind
summary: str
destructive: bool
irreversible: bool
requires_reconciliation: bool = False
depends_on: tuple[str, ...] = ()
def __post_init__(self) -> None:
_text(self.step_id, "step id", maximum=160)
_text(self.summary, "step summary", maximum=1000)
if self.kind not in _STEP_KINDS:
raise ValueError("Tenant erasure step kind is invalid.")
_texts(self.depends_on, "step dependencies", maximum_length=160)
if self.step_id in self.depends_on:
raise ValueError("Tenant erasure step cannot depend on itself.")
if self.irreversible and not self.destructive:
raise ValueError("An irreversible tenant erasure step must be destructive.")
def to_dict(self) -> dict[str, object]:
return {
"step_id": self.step_id,
"kind": self.kind,
"summary": self.summary,
"destructive": self.destructive,
"irreversible": self.irreversible,
"requires_reconciliation": self.requires_reconciliation,
"depends_on": list(self.depends_on),
}
@dataclass(frozen=True, slots=True)
class TenantErasurePreview:
module_id: str
complete: bool
resources: tuple[TenantErasureResource, ...] = ()
steps: tuple[TenantErasureStep, ...] = ()
blockers: tuple[str, ...] = ()
warnings: tuple[str, ...] = ()
provider_revision: str = "1"
def __post_init__(self) -> None:
_text(self.module_id, "module id", maximum=120)
_text(self.provider_revision, "provider revision", maximum=120)
_texts(self.blockers, "blockers", maximum_length=1000)
_texts(self.warnings, "warnings", maximum_length=1000)
if len(self.resources) > 500 or len(self.steps) > 500:
raise ValueError("Tenant erasure preview is too large.")
resource_types = [item.resource_type for item in self.resources]
if len(resource_types) != len(set(resource_types)):
raise ValueError("Tenant erasure preview repeats a resource type.")
resources_requiring_action = tuple(
item for item in self.resources if item.count > 0
)
if resources_requiring_action and not self.steps and not self.blockers:
raise ValueError(
"Tenant erasure resources require steps or an explicit blocker."
)
if any(
item.count > 0 and item.disposition == "unavailable"
for item in self.resources
) and not self.blockers:
raise ValueError(
"Unavailable tenant erasure resources require an explicit blocker."
)
if not self.complete and not self.blockers:
raise ValueError(
"An incomplete tenant erasure preview requires an explicit blocker."
)
step_ids = [item.step_id for item in self.steps]
if len(step_ids) != len(set(step_ids)):
raise ValueError("Tenant erasure preview repeats a step id.")
known_step_ids = set(step_ids)
if any(
dependency not in known_step_ids
for step in self.steps
for dependency in step.depends_on
):
raise ValueError("Tenant erasure step references an unknown dependency.")
remaining = {
step.step_id: set(step.depends_on)
for step in self.steps
}
resolved: set[str] = set()
while remaining:
ready = sorted(
step_id
for step_id, dependencies in remaining.items()
if dependencies.issubset(resolved)
)
if not ready:
raise ValueError("Tenant erasure step dependencies contain a cycle.")
resolved.update(ready)
for step_id in ready:
remaining.pop(step_id)
@property
def allowed(self) -> bool:
return self.complete and not self.blockers
def to_dict(self) -> dict[str, object]:
return {
"module_id": self.module_id,
"complete": self.complete,
"allowed": self.allowed,
"provider_revision": self.provider_revision,
"resources": [item.to_dict() for item in self.resources],
"steps": [item.to_dict() for item in self.steps],
"blockers": list(self.blockers),
"warnings": list(self.warnings),
}
@dataclass(frozen=True, slots=True)
class TenantErasureStepResult:
state: TenantErasureResultState
summary: str
receipt_ref: str | None = None
metrics: Mapping[str, int] = field(default_factory=dict)
def __post_init__(self) -> None:
if self.state not in _RESULT_STATES:
raise ValueError("Tenant erasure result state is invalid.")
_text(self.summary, "result summary", maximum=1000)
if self.receipt_ref is not None:
_text(self.receipt_ref, "receipt reference", maximum=500)
_metrics(self.metrics)
def to_dict(self) -> dict[str, object]:
return {
"state": self.state,
"summary": self.summary,
"receipt_ref": self.receipt_ref,
"metrics": dict(sorted(_metrics(self.metrics).items())),
}
@runtime_checkable
class TenantErasureProvider(Protocol):
module_id: str
def preview_tenant_erasure(
self,
session: object,
tenant_id: str,
) -> TenantErasurePreview:
...
def execute_tenant_erasure_step(
self,
session: object,
tenant_id: str,
step_id: str,
idempotency_key: str,
) -> TenantErasureStepResult:
...
def reconcile_tenant_erasure_step(
self,
session: object,
tenant_id: str,
step_id: str,
idempotency_key: str,
) -> TenantErasureStepResult:
...
@dataclass(frozen=True, slots=True)
class TenantErasureInventory:
tenant_id: str
generated_at: datetime
complete: bool
modules: tuple[TenantErasurePreview, ...]
@property
def allowed(self) -> bool:
return self.complete and all(item.allowed for item in self.modules)
def to_dict(self) -> dict[str, object]:
generated_at = self.generated_at
if generated_at.tzinfo is None:
generated_at = generated_at.replace(tzinfo=UTC)
return {
"schema_version": 1,
"tenant_id": self.tenant_id,
"generated_at": generated_at.astimezone(UTC).isoformat(),
"complete": self.complete,
"allowed": self.allowed,
"modules": [item.to_dict() for item in self.modules],
}
def tenant_erasure_providers(registry: object) -> dict[str, TenantErasureProvider]:
capability_names = getattr(registry, "capability_names", None)
capability = getattr(registry, "capability", None)
if not callable(capability_names) or not callable(capability):
raise ValueError("Tenant erasure requires a module registry.")
providers: dict[str, TenantErasureProvider] = {}
for capability_name in sorted(capability_names()):
if not capability_name.startswith(TENANT_ERASURE_PROVIDER_CAPABILITY_PREFIX):
continue
expected_module_id = capability_name.removeprefix(
TENANT_ERASURE_PROVIDER_CAPABILITY_PREFIX
)
provider = capability(capability_name)
if not isinstance(provider, TenantErasureProvider):
raise TypeError(
f"Tenant erasure provider {expected_module_id or 'unknown'} is invalid."
)
module_id = _text(provider.module_id, "provider module id", maximum=120)
if module_id != expected_module_id or module_id in providers:
raise ValueError("Tenant erasure provider identity is invalid.")
providers[module_id] = provider
return providers
def collect_tenant_erasure_inventory(
registry: object,
session: object,
tenant_id: str,
*,
observed_at: datetime | None = None,
) -> TenantErasureInventory:
normalized_tenant_id = _text(tenant_id, "tenant id", maximum=120)
manifests = getattr(registry, "manifests", None)
summary_providers = getattr(registry, "tenant_summary_providers", None)
if not callable(manifests) or not callable(summary_providers):
raise ValueError("Tenant erasure inventory requires a module registry.")
provider_by_module = tenant_erasure_providers(registry)
summary_by_module = dict(summary_providers())
manifest_ids = {
str(manifest.id)
for manifest in manifests()
if getattr(manifest, "id", None)
}
module_ids = manifest_ids | set(summary_by_module) | set(provider_by_module)
previews: list[TenantErasurePreview] = []
complete = True
for module_id in sorted(module_ids):
provider = provider_by_module.get(module_id)
if provider is not None:
try:
preview = provider.preview_tenant_erasure(session, normalized_tenant_id)
if not isinstance(preview, TenantErasurePreview):
raise TypeError("provider returned an invalid preview")
if preview.module_id != module_id:
raise ValueError("provider returned another module's preview")
except Exception as exc:
complete = False
preview = TenantErasurePreview(
module_id=module_id,
complete=False,
blockers=(
f"{type(exc).__name__}: provider preview could not be completed",
),
)
previews.append(preview)
complete = complete and preview.complete
continue
summary_provider = summary_by_module.get(module_id)
if summary_provider is None:
previews.append(
TenantErasurePreview(
module_id=module_id,
complete=True,
warnings=(
"Module declares no tenant-owned summary or erasure provider; no tenant persistence is in scope.",
),
provider_revision="manifest-no-tenant-data",
)
)
continue
try:
raw_counts = summary_provider(session, normalized_tenant_id)
counts = _metrics({str(key): int(value) for key, value in raw_counts.items()})
resources = tuple(
TenantErasureResource(
resource_type=resource_type,
count=count,
disposition="unavailable" if count else "erase",
summary=(
"Tenant-owned data requires a module erasure provider."
if count
else "The module reported no tenant-owned records."
),
)
for resource_type, count in sorted(counts.items())
)
blockers = (
("Tenant-owned data exists but the module has no erasure provider.",)
if any(counts.values())
else ()
)
preview = TenantErasurePreview(
module_id=module_id,
complete=True,
resources=resources,
blockers=blockers,
provider_revision="tenant-summary-fallback",
)
except Exception as exc:
complete = False
preview = TenantErasurePreview(
module_id=module_id,
complete=False,
blockers=(
f"{type(exc).__name__}: tenant summary could not be completed",
),
provider_revision="tenant-summary-fallback",
)
previews.append(preview)
timestamp = observed_at or datetime.now(UTC)
if timestamp.tzinfo is None:
timestamp = timestamp.replace(tzinfo=UTC)
return TenantErasureInventory(
tenant_id=normalized_tenant_id,
generated_at=timestamp,
complete=complete,
modules=tuple(previews),
)
__all__ = [
"TENANT_ERASURE_PROVIDER_CAPABILITY_PREFIX",
"TenantErasureInventory",
"TenantErasurePreview",
"TenantErasureProvider",
"TenantErasureResource",
"TenantErasureStep",
"TenantErasureStepResult",
"collect_tenant_erasure_inventory",
"tenant_erasure_providers",
]
@@ -27,6 +27,7 @@ LEGACY_TO_MODULE_SCOPES: dict[str, str] = {
"system:tenants:create": "access:tenant:create",
"system:tenants:update": "access:tenant:update",
"system:tenants:suspend": "access:tenant:suspend",
"system:tenants:erase": "access:tenant:erase",
"system:accounts:read": "access:account:read",
"system:accounts:create": "access:account:create",
"system:accounts:update": "access:account:update",
@@ -78,6 +78,7 @@ SYSTEM_PERMISSIONS: tuple[PermissionDefinition, ...] = (
PermissionDefinition("system:tenants:create", "Create tenants", "Create new tenant spaces.", "System administration", "system"),
PermissionDefinition("system:tenants:update", "Update tenants", "Edit tenant metadata and governance overrides.", "System administration", "system"),
PermissionDefinition("system:tenants:suspend", "Suspend tenants", "Activate or suspend tenant spaces while preserving evidence.", "System administration", "system"),
PermissionDefinition("system:tenants:erase", "Erase tenants", "Preview, approve, execute, and reconcile governed destructive tenant erasure.", "System administration", "system"),
PermissionDefinition("system:accounts:read", "View accounts", "List global login accounts and memberships.", "System administration", "system"),
PermissionDefinition("system:accounts:create", "Create accounts", "Create global login accounts.", "System administration", "system"),
PermissionDefinition("system:accounts:update", "Update accounts", "Edit global account metadata.", "System administration", "system"),