Add ciphertext-safe Postbox search source

This commit is contained in:
2026-08-04 03:03:27 +02:00
parent a97eb3bcef
commit 889b5e61f0
4 changed files with 490 additions and 0 deletions
+32
View File
@@ -44,9 +44,11 @@ from govoplan_core.core.postbox import (
CAPABILITY_POSTBOX_MESSAGES,
CAPABILITY_POSTBOX_ROUTING,
)
from govoplan_core.core.search import SearchSourceProviderRegistration
from govoplan_core.core.views import ViewSurface
from govoplan_core.db.base import Base
from govoplan_postbox.backend.db import models as postbox_models
from govoplan_postbox.backend.search_source import create_postbox_search_source
MODULE_ID = "postbox"
@@ -222,6 +224,7 @@ manifest = ModuleManifest(
"portal",
"views",
"workflow_engine",
"search",
),
required_capabilities=(
CAPABILITY_AUTH_PRINCIPAL_RESOLVER,
@@ -266,9 +269,21 @@ manifest = ModuleManifest(
version_max_exclusive="2.0.0",
optional=True,
),
ModuleInterfaceRequirement(
name="search.source",
version_min="1.0.0",
version_max_exclusive="2.0.0",
optional=True,
),
),
permissions=PERMISSIONS,
role_templates=ROLE_TEMPLATES,
search_sources=(
SearchSourceProviderRegistration(
id="postbox.messages",
factory=create_postbox_search_source,
),
),
nav_items=(
NavItem(
path="/postbox",
@@ -363,6 +378,23 @@ manifest = ModuleManifest(
CAPABILITY_POSTBOX_ROUTING: _configure,
},
documentation=(
DocumentationTopic(
id="postbox.search.messages",
title="Search authorized Postbox messages",
summary="Expose Postbox subjects and permitted plaintext content to permission-aware platform Search.",
body=(
"When Search is installed, Postbox contributes message subjects, sender labels, and plaintext "
"content only. Ciphertext and key material are never indexed. Every result is tenant-bounded and "
"rechecks current function assignment, Postbox binding, classification, acting context, and generic "
"read authority without recording a message read. Committed deliveries and message changes update "
"the derived index through the durable platform event path."
),
layer="configured",
documentation_types=("admin", "user"),
audience=("administrator", "user", "campaign_manager"),
related_modules=("search", "idm", "encryption"),
order=34,
),
DocumentationTopic(
id="postbox.function-bound-containers",
title="Function-bound Postboxes",
@@ -0,0 +1,296 @@
from __future__ import annotations
from collections.abc import Mapping, Sequence
from urllib.parse import quote
from sqlalchemy import func, select
from sqlalchemy.orm import Session
from govoplan_core.auth import ApiPrincipal
from govoplan_core.core.events import PlatformEvent
from govoplan_core.core.modules import ModuleContext
from govoplan_core.core.postbox import PostboxActorRef
from govoplan_core.core.search import (
SearchAuthorizationRequest,
SearchBackfillPage,
SearchBackfillRequest,
SearchDocument,
SearchIndexChange,
SearchResourceReference,
SearchResourceType,
)
from govoplan_postbox.backend.db.models import (
Postbox,
PostboxMessage,
PostboxRoute,
)
from govoplan_postbox.backend.service import PostboxService
PROVIDER_ID = "postbox.messages"
RESOURCE_TYPE = "postbox_message"
READ_SCOPE = "postbox:postbox:read"
CONFIDENTIAL_SCOPE = "postbox:classification:confidential"
RESTRICTED_SCOPE = "postbox:classification:restricted"
class PostboxSearchSource:
def __init__(self, service: PostboxService) -> None:
self.service = service
def resource_types(self) -> Sequence[SearchResourceType]:
return (
SearchResourceType(
provider_id=PROVIDER_ID,
module_id="postbox",
resource_type=RESOURCE_TYPE,
label="Postbox messages",
requires_authorization_recheck=True,
),
)
def backfill(
self,
session: object,
*,
request: SearchBackfillRequest,
) -> SearchBackfillPage:
_assert_source(request.provider_id, request.resource_type)
db = _session(session)
statement = (
select(PostboxMessage, Postbox)
.join(Postbox, Postbox.id == PostboxMessage.postbox_id)
.where(
PostboxMessage.tenant_id == request.tenant_id,
Postbox.tenant_id == request.tenant_id,
)
)
if request.cursor:
statement = statement.where(PostboxMessage.id > request.cursor)
rows = list(
db.execute(
statement.order_by(PostboxMessage.id).limit(request.limit + 1)
).all()
)
has_more = len(rows) > request.limit
selected = rows[: request.limit]
high_watermark = db.scalar(
select(func.max(PostboxMessage.updated_at)).where(
PostboxMessage.tenant_id == request.tenant_id
)
)
return SearchBackfillPage(
documents=tuple(
_document(message, postbox=postbox)
for message, postbox in selected
),
next_cursor=(
selected[-1][0].id if has_more and selected else None
),
complete=not has_more,
high_watermark=(
high_watermark.isoformat()
if high_watermark is not None
else None
),
)
def authorize(
self,
session: object,
principal: object,
*,
requests: Sequence[SearchAuthorizationRequest],
) -> Mapping[str, bool]:
decisions = {item.reference.key: False for item in requests}
if not isinstance(principal, ApiPrincipal) or not principal.has(READ_SCOPE):
return decisions
actor = _actor(principal)
db = _session(session)
for item in requests:
reference = item.reference
if (
reference.tenant_id != principal.tenant_id
or reference.module_id != "postbox"
or reference.resource_type != RESOURCE_TYPE
):
continue
try:
decisions[reference.key] = self.service.can_read_message(
db,
tenant_id=principal.tenant_id,
message_id=reference.resource_id,
actor=actor,
)
except (RuntimeError, ValueError):
decisions[reference.key] = False
return decisions
def index_changes_for_event(
self,
session: object,
*,
event: PlatformEvent,
delivery_key: str,
) -> Sequence[SearchIndexChange]:
if (
event.module_id != "postbox"
or event.tenant is None
or event.resource is None
or event.resource.id is None
or event.resource.type not in {RESOURCE_TYPE, "postbox_route"}
):
return ()
db = _session(session)
message_id = event.resource.id
if event.resource.type == "postbox_route":
route = db.get(PostboxRoute, event.resource.id)
if (
route is None
or route.tenant_id != event.tenant.id
or route.target_message_id is None
):
return ()
message_id = route.target_message_id
row = db.get(PostboxMessage, message_id)
postbox = (
db.get(Postbox, row.postbox_id)
if row is not None and row.tenant_id == event.tenant.id
else None
)
deleted = (
row is None
or row.tenant_id != event.tenant.id
or postbox is None
)
cursor = event.event_id
document = (
None
if deleted
else _document(row, postbox=postbox, change_cursor=cursor)
)
reference = SearchResourceReference(
tenant_id=event.tenant.id,
module_id="postbox",
resource_type=RESOURCE_TYPE,
resource_id=message_id,
)
return (
SearchIndexChange(
change_id=f"{delivery_key}:{PROVIDER_ID}:{message_id}",
provider_id=PROVIDER_ID,
kind="delete" if deleted else "upsert",
reference=reference,
source_revision=(
document.source_revision if document is not None else cursor
),
cursor=cursor,
document=document,
occurred_at=event.occurred_at,
),
)
def create_postbox_search_source(
context: ModuleContext,
) -> PostboxSearchSource:
return PostboxSearchSource(PostboxService.from_registry(context.registry))
def _actor(principal: ApiPrincipal) -> PostboxActorRef:
selected = principal.acting_assignment_id
if selected is None and len(principal.function_assignment_ids) == 1:
selected = next(iter(principal.function_assignment_ids))
classifications = {"public", "internal"}
if principal.has(CONFIDENTIAL_SCOPE):
classifications.add("confidential")
if principal.has(RESTRICTED_SCOPE):
classifications.update(("confidential", "restricted"))
return PostboxActorRef(
account_id=principal.account_id,
identity_id=principal.identity_id,
selected_assignment_id=selected,
acting_for_account_id=principal.acting_for_account_id,
authorized_actions=frozenset({"discover", "read"}),
authorized_classifications=frozenset(classifications),
)
def _document(
message: PostboxMessage,
*,
postbox: Postbox,
change_cursor: str | None = None,
) -> SearchDocument:
updated_at = message.updated_at or message.created_at
body = (
message.body_text
if message.encryption_profile == "plaintext_v1"
and message.body_ciphertext is None
else None
)
tokens = [f"scope:{READ_SCOPE}"]
if message.classification == "confidential":
tokens.append(f"scope:{CONFIDENTIAL_SCOPE}")
elif message.classification == "restricted":
tokens.append(f"scope:{RESTRICTED_SCOPE}")
return SearchDocument(
tenant_id=message.tenant_id,
module_id="postbox",
provider_id=PROVIDER_ID,
resource_type=RESOURCE_TYPE,
resource_id=message.id,
title=message.subject[:500],
url=f"/postbox?messageId={quote(message.id, safe='')}",
summary=(message.sender_label or postbox.name)[:4000],
body=body[:200_000] if body else None,
keywords=tuple(
item[:200]
for item in (
postbox.name,
message.status,
message.classification,
message.producer_module or "",
)
if item
),
visibility="restricted",
acl_tokens=tuple(dict.fromkeys(tokens)),
metadata={
"postbox_id": message.postbox_id,
"postbox_name": postbox.name,
"status": message.status,
"classification": message.classification,
"sender_label": message.sender_label,
"delivered_at": message.delivered_at.isoformat(),
"withdrawn_at": (
message.withdrawn_at.isoformat()
if message.withdrawn_at
else None
),
"encrypted": body is None and message.body_ciphertext is not None,
},
source_revision=f"{message.status}:{updated_at.isoformat()}",
change_cursor=change_cursor,
source_updated_at=updated_at,
requires_authorization_recheck=True,
)
def _assert_source(provider_id: str, resource_type: str) -> None:
if provider_id != PROVIDER_ID or resource_type != RESOURCE_TYPE:
raise ValueError("Unsupported Postbox search source.")
def _session(value: object) -> Session:
if not isinstance(value, Session):
raise TypeError("Postbox search requires a SQLAlchemy session.")
return value
__all__ = [
"PROVIDER_ID",
"PostboxSearchSource",
"RESOURCE_TYPE",
"create_postbox_search_source",
]
+26
View File
@@ -845,6 +845,32 @@ class PostboxService:
)
return self._message_ref(message, account_id=actor.account_id)
def can_read_message(
self,
session: object,
*,
tenant_id: str,
message_id: str,
actor: PostboxActorRef,
) -> bool:
"""Recheck message access without creating a read or denial event."""
db = _session(session)
message = self._get_message(
db,
tenant_id=tenant_id,
message_id=message_id,
required=False,
)
if message is None:
return False
return self._message_access_decision(
db,
message=message,
actor=actor,
action="read",
).allowed
def mark_message(
self,
session: object,