Merge remote-tracking branch 'origin/master' into dev/4.11.x

# Conflicts:
#	src/langbot/pkg/pipeline/process/handlers/chat.py
#	tests/integration/persistence/test_workspace_migration.py
This commit is contained in:
huanghuoguoguo
2026-08-01 09:31:23 +08:00
57 changed files with 2755 additions and 134 deletions
@@ -418,6 +418,7 @@ class TestUserServiceGenerateJwtToken:
assert token is not None
class TestUserServiceVerifyJwtToken:
"""Tests for verify_jwt_token method."""
@@ -149,6 +149,36 @@ async def test_session_scope_matches_exact_tenant_placement_and_principal():
assert sessions == {}
async def test_support_admin_sessions_are_scoped_to_the_persisted_grant():
def support_context(grant_jti_hash: str) -> RequestContext:
return RequestContext(
instance_uuid='instance-test',
placement_generation=1,
request_id='request-test',
auth_type='support-admin',
principal=PrincipalContext(
principal_type=PrincipalType.SUPPORT_ADMIN,
actor_account_uuid='support-actor',
support_session_id=grant_jti_hash,
),
workspace=WorkspaceContext(
workspace_uuid='workspace-a',
membership_uuid=None,
role='owner',
permissions=frozenset({'resource.manage'}),
),
)
first_context = support_context('a' * 64)
second_context = support_context('b' * 64)
sessions: dict[str, dict] = {'session-test': {'status': 'waiting'}}
_bind_session_scope(sessions['session-test'], first_context)
assert _get_owned_session(sessions, 'session-test', second_context) is None
assert _pop_owned_session(sessions, 'session-test', second_context) is None
assert _get_owned_session(sessions, 'session-test', first_context) is sessions['session-test']
async def test_session_capacity_evicts_oldest_session_in_same_workspace():
owner_context = _request_context()
sessions: dict[str, dict] = {}
@@ -46,6 +46,20 @@ def plugin_router_cls():
yield PluginsRouterGroup
@pytest.mark.asyncio
async def test_authenticated_plugin_resource_fences_injected_workspace_context(plugin_router_cls):
connector = SimpleNamespace(
require_workspace_context=AsyncMock(return_value=CONTEXT),
)
router = object.__new__(plugin_router_cls)
router.ap = SimpleNamespace(plugin_connector=connector)
result = await router._require_authenticated_plugin_runtime_context(CONTEXT)
assert result == CONTEXT
connector.require_workspace_context.assert_awaited_once_with(CONTEXT)
@pytest.mark.asyncio
async def test_public_plugin_asset_route_is_disabled_for_multi_workspace_policy(plugin_router_cls):
connector = SimpleNamespace(
@@ -11,6 +11,7 @@ from cryptography.hazmat.primitives import serialization
from cryptography.hazmat.primitives.asymmetric.ed25519 import Ed25519PrivateKey
from langbot.pkg.cloud.launch import SpaceLaunchError, SpaceLaunchService
from langbot.pkg.cloud.support_admin import SupportAdminReplayError
pytestmark = pytest.mark.asyncio
@@ -57,9 +58,21 @@ def _service(private_key: Ed25519PrivateKey, *, now: int) -> SpaceLaunchService:
encoding=serialization.Encoding.Raw,
format=serialization.PublicFormat.Raw,
)
consumed: set[str] = set()
class DurableSupportAdminService:
async def consume_launch_grant(self, **kwargs):
grant_hash = kwargs['grant_jti_hash']
if grant_hash in consumed:
raise SupportAdminReplayError('already consumed')
consumed.add(grant_hash)
return SimpleNamespace(token='support-admin-token')
app = SimpleNamespace(
deployment=SimpleNamespace(multi_workspace_enabled=True, verification_key_id=KEY_ID),
workspace_service=SimpleNamespace(instance_uuid=INSTANCE_UUID),
logger=SimpleNamespace(info=lambda *args, **kwargs: None),
support_admin_session_service=DurableSupportAdminService(),
instance_config=SimpleNamespace(
data={
'space': {
@@ -86,6 +99,81 @@ async def test_consumes_valid_workspace_launch_assertion_once():
await service.consume_assertion(token, expected_workspace_uuid=WORKSPACE_UUID)
async def test_consumes_admin_owner_launch_once_and_validates_claims():
private_key = Ed25519PrivateKey.generate()
now = int(time.time())
service = _service(private_key, now=now)
claims = _claims(now=now)
claims['kind'] = 'workspace.support_admin_launch'
claims['payload'].update(
{
'launch_mode': 'support_admin',
'principal_type': 'support_admin',
'actor_account_uuid': '33333333-3333-4333-8333-333333333333',
'effective_role': 'owner',
}
)
claims['payload'].pop('account_uuid')
token = _sign(private_key, claims)
launch = await service.consume_assertion(token, expected_workspace_uuid=WORKSPACE_UUID)
assert launch == {
'workspace_uuid': WORKSPACE_UUID,
'launch_mode': 'support_admin',
'actor_account_uuid': '33333333-3333-4333-8333-333333333333',
'effective_role': 'owner',
'grant_jti_hash': launch['grant_jti_hash'],
'support_admin_token': 'support-admin-token',
}
with pytest.raises(SpaceLaunchError, match='already been consumed'):
await service.consume_assertion(token, expected_workspace_uuid=WORKSPACE_UUID)
invalid = _claims(now=now)
invalid['kind'] = 'workspace.support_admin_launch'
invalid['payload'].update(
{
'launch_mode': 'support_admin',
'principal_type': 'support_admin',
'actor_account_uuid': '33333333-3333-4333-8333-333333333333',
'effective_role': 'member',
}
)
invalid['payload'].pop('account_uuid')
with pytest.raises(SpaceLaunchError, match='effective role'):
await service.consume_assertion(_sign(private_key, invalid), expected_workspace_uuid=WORKSPACE_UUID)
too_long = _claims(now=now)
too_long['kind'] = 'workspace.support_admin_launch'
too_long['exp'] = now + 91
too_long['payload'].update(
{
'launch_mode': 'support_admin',
'principal_type': 'support_admin',
'actor_account_uuid': '33333333-3333-4333-8333-333333333333',
'effective_role': 'owner',
}
)
too_long['payload'].pop('account_uuid')
with pytest.raises(SpaceLaunchError, match='lifetime exceeds 90 seconds'):
await service.consume_assertion(_sign(private_key, too_long), expected_workspace_uuid=WORKSPACE_UUID)
impersonating = _claims(now=now)
impersonating['kind'] = 'workspace.support_admin_launch'
impersonating['payload'].update(
{
'launch_mode': 'support_admin',
'principal_type': 'support_admin',
'actor_account_uuid': '33333333-3333-4333-8333-333333333333',
'effective_role': 'owner',
}
)
with pytest.raises(SpaceLaunchError, match='customer Account'):
await service.consume_assertion(_sign(private_key, impersonating), expected_workspace_uuid=WORKSPACE_UUID)
async def test_replay_cache_does_not_scan_all_live_assertions(monkeypatch):
private_key = Ed25519PrivateKey.generate()
now = int(time.time())
@@ -1,7 +1,8 @@
"""Tests for Telegram Dify form callback helpers."""
import json
from unittest.mock import AsyncMock, MagicMock
from contextlib import asynccontextmanager
from unittest.mock import AsyncMock, MagicMock, patch
import pytest
from telegram import ForceReply
@@ -9,8 +10,10 @@ from telegram import ForceReply
import langbot_plugin.api.entities.builtin.platform.entities as platform_entities
import langbot_plugin.api.entities.builtin.platform.events as platform_events
import langbot_plugin.api.entities.builtin.platform.message as platform_message
from langbot.pkg.platform.sources.telegram import (
TelegramAdapter,
TelegramMessageConverter,
_decode_telegram_base64_limited,
_telegram_form_action_from_callback,
_telegram_select_field_options,
@@ -26,6 +29,70 @@ def test_telegram_base64_decode_is_bounded(monkeypatch):
_decode_telegram_base64_limited('A' * 12)
TELEGRAM_BOT_TOKEN = '123456789:AAExampleBotTokenThatMustNotLeak'
TELEGRAM_FILE_URL = f'https://api.telegram.org/file/bot{TELEGRAM_BOT_TOKEN}/photos/file_0.jpg'
@pytest.mark.asyncio
async def test_telegram_photo_does_not_expose_bot_token_in_image_url():
"""Regression test for the Telegram bot-token leak.
telegram.Bot builds file.file_path as
https://api.telegram.org/file/bot<TOKEN>/<path>, embedding the bot token.
The converter must not copy that URL into Image.url, or the token leaks to
the monitoring DB, dashboard and every installed plugin via the message
chain. Only base64 (which carries no token) may be stored.
"""
tg_file = MagicMock()
tg_file.file_path = TELEGRAM_FILE_URL
photo_size = MagicMock()
photo_size.get_file = AsyncMock(return_value=tg_file)
message = MagicMock()
message.text = None
message.caption = None
message.photo = [photo_size]
message.voice = None
message.document = None
response = MagicMock()
response.headers = {}
async def iter_chunked(_chunk_size):
yield b'\xff\xd8\xff\xe0jpeg-bytes'
response.content.iter_chunked = iter_chunked
@asynccontextmanager
async def fake_get(url):
yield response
fake_session = MagicMock()
fake_session.get = fake_get
with patch(
'langbot.pkg.platform.sources.telegram.httpclient.get_session',
return_value=fake_session,
):
chain = await TelegramMessageConverter.target2yiri(message, MagicMock(), 'bot-account')
images = [c for c in chain if isinstance(c, platform_message.Image)]
assert len(images) == 1
image = images[0]
# The token-bearing URL must not be retained anywhere on the component.
assert not image.url
assert image.base64 is not None
assert image.base64.startswith('data:image/jpeg;base64,')
# Belt-and-suspenders: the token must not appear in the serialized chain
# (this is what gets persisted to the monitoring DB and sent to plugins).
serialized = json.dumps(chain.model_dump(), ensure_ascii=False)
assert TELEGRAM_BOT_TOKEN not in serialized
assert 'api.telegram.org/file/bot' not in serialized
def _select_form_data() -> dict:
return {
'_current_input_field': 'choice',
+57 -13
View File
@@ -60,10 +60,12 @@ class TestBuildHeartbeatPayload:
async def test_payload_shape(self):
heartbeat = get_heartbeat_module()
ap = make_app()
payload = await heartbeat.build_heartbeat_payload(ap)
payload = await heartbeat.build_heartbeat_payload(ap, workspace_uuid='workspace-a')
assert payload['event_type'] == 'instance_heartbeat'
assert payload['query_id'] == ''
assert payload['workspace_uuid'] == 'workspace-a'
assert 'instance_id' not in payload
assert 'instance_create_ts' in payload
assert 'timestamp' in payload
f = payload['features']
@@ -86,7 +88,7 @@ class TestBuildHeartbeatPayload:
@pytest.mark.asyncio
async def test_payload_is_json_serializable(self):
heartbeat = get_heartbeat_module()
payload = await heartbeat.build_heartbeat_payload(make_app())
payload = await heartbeat.build_heartbeat_payload(make_app(), workspace_uuid='workspace-a')
json.dumps(payload)
@pytest.mark.asyncio
@@ -94,7 +96,7 @@ class TestBuildHeartbeatPayload:
heartbeat = get_heartbeat_module()
ap = make_app()
ap.persistence_mgr.execute_async = AsyncMock(side_effect=RuntimeError('db down'))
payload = await heartbeat.build_heartbeat_payload(ap)
payload = await heartbeat.build_heartbeat_payload(ap, workspace_uuid='workspace-a')
assert payload['features']['pipeline_count'] == -1
@pytest.mark.asyncio
@@ -106,31 +108,73 @@ class TestBuildHeartbeatPayload:
side_effect=AssertionError('Cloud heartbeat must not issue per-tenant COUNTs')
)
ap.pipeline_mgr = SimpleNamespace(
_pipelines_by_key={'pipeline-a': object(), 'pipeline-b': object()},
_pipelines_by_key={
('instance-a', 'workspace-a', 'pipeline-a'): object(),
('instance-a', 'workspace-a', 'pipeline-b'): object(),
},
)
adapter_a = type('WorkspaceAAdapter', (), {})()
adapter_b = type('WorkspaceBAdapter', (), {})()
ap.platform_mgr._bots_by_key = {
('instance-a', 'workspace-a', 'bot-a'): SimpleNamespace(enable=True, adapter=adapter_a),
}
ap.tool_mgr = SimpleNamespace(
mcp_tool_loader=SimpleNamespace(
_sessions={'mcp-a': object(), 'mcp-b': object(), 'mcp-c': object()},
_sessions={
('instance-a', 'workspace-a', 1, 'mcp-a'): object(),
('instance-a', 'workspace-a', 1, 'mcp-b'): object(),
('instance-a', 'workspace-a', 1, 'mcp-c'): object(),
},
),
)
ap.rag_mgr = SimpleNamespace(
knowledge_bases={'kb-a': object()},
knowledge_bases={('workspace-a', 'kb-a'): object()},
)
ap.plugin_connector._workspace_installations = {
'workspace-a': {'plugin-a', 'plugin-b'},
}
ap.skill_mgr._skills_by_scope = {
('instance-a', 'workspace-a', 1): {'skill-a': {}, 'skill-b': {}},
('instance-a', 'workspace-b', 1): {'skill-c': {}},
}
ap.workspace_service.list_active_execution_bindings = AsyncMock(
return_value=[
SimpleNamespace(workspace_uuid='workspace-a', placement_generation=7),
SimpleNamespace(workspace_uuid='workspace-b', placement_generation=9),
],
)
ap.platform_mgr._bots_by_key[('instance-a', 'workspace-b', 'bot-b')] = SimpleNamespace(
enable=True, adapter=adapter_b
)
payload = await heartbeat.build_heartbeat_payload(ap)
payloads = await heartbeat.build_heartbeat_payloads(ap)
features = payload['features']
assert features['pipeline_count'] == 2
assert features['mcp_server_count'] == 3
assert features['knowledge_base_count'] == 1
assert features['bot_count'] == 1
assert [payload['workspace_uuid'] for payload in payloads] == ['workspace-a', 'workspace-b']
assert all('instance_id' not in payload for payload in payloads)
by_workspace = {payload['workspace_uuid']: payload['features'] for payload in payloads}
assert by_workspace['workspace-a']['pipeline_count'] == 2
assert by_workspace['workspace-a']['mcp_server_count'] == 3
assert by_workspace['workspace-a']['knowledge_base_count'] == 1
assert by_workspace['workspace-a']['bot_count'] == 1
assert by_workspace['workspace-a']['plugin_count'] == 2
assert by_workspace['workspace-a']['extension_count'] == 5
assert by_workspace['workspace-a']['skill_count'] == 2
assert by_workspace['workspace-a']['execution_generation'] == 7
assert by_workspace['workspace-a']['adapters'] == ['WorkspaceAAdapter']
assert by_workspace['workspace-b']['bot_count'] == 1
assert by_workspace['workspace-b']['pipeline_count'] == 0
assert by_workspace['workspace-b']['skill_count'] == 1
assert by_workspace['workspace-b']['execution_generation'] == 9
assert by_workspace['workspace-b']['adapters'] == ['WorkspaceBAdapter']
assert 'workspace_resources' not in by_workspace['workspace-a']
ap.persistence_mgr.execute_async.assert_not_awaited()
ap.workspace_service.list_active_execution_bindings.assert_awaited_once()
@pytest.mark.asyncio
async def test_no_user_content_fields(self):
"""The heartbeat must never carry message content / credentials keys."""
heartbeat = get_heartbeat_module()
payload = await heartbeat.build_heartbeat_payload(make_app())
payload = await heartbeat.build_heartbeat_payload(make_app(), workspace_uuid='workspace-a')
flat = json.dumps(payload).lower()
for forbidden in ('api_key', 'password', 'token', 'message_content'):
assert forbidden not in flat
@@ -569,6 +569,33 @@ class TestHTTPScenarios:
await manager.send({'query_id': 'test'})
class TestTelemetryManagedRuntimeAuthentication:
@pytest.mark.asyncio
async def test_send_includes_managed_runtime_token_header(self):
telemetry = get_telemetry_module()
mock_app = Mock()
mock_app.logger = Mock()
manager = telemetry.TelemetryManager(mock_app)
manager.telemetry_config = {'url': 'https://example.com'}
captured = {}
async def mock_post(url, json, headers):
captured['headers'] = headers
return Mock(status_code=200, text='', json=Mock(return_value={'code': 0}))
mock_client = Mock()
mock_client.post = mock_post
mock_client.__aenter__ = AsyncMock(return_value=mock_client)
mock_client.__aexit__ = AsyncMock(return_value=None)
with (
patch.dict('os.environ', {'LANGBOT_TELEMETRY_INGEST_TOKEN': 'managed-runtime-secret'}),
patch.object(httpx, 'AsyncClient', return_value=mock_client),
):
await manager.send({'event_type': 'instance_heartbeat'})
assert captured['headers'] == {'X-LangBot-Telemetry-Token': 'managed-runtime-secret'}
class TestStartSendTask:
"""Tests for start_send_task() method."""
@@ -0,0 +1,31 @@
from __future__ import annotations
import uuid
from types import SimpleNamespace
def test_standard_oss_instance_id_aligns_to_embedded_uuid():
from langbot.pkg.workspace.identity import workspace_uuid_from_instance_id
instance_uuid = "a711d9e4-0953-443f-a0e9-7dd50193a79f"
assert workspace_uuid_from_instance_id(instance_uuid) == instance_uuid
assert workspace_uuid_from_instance_id(f"instance_{instance_uuid}") == instance_uuid
def test_custom_legacy_instance_id_maps_to_stable_valid_uuid():
from langbot.pkg.workspace.identity import workspace_uuid_from_instance_id
first = workspace_uuid_from_instance_id("instance_migration_test")
second = workspace_uuid_from_instance_id("instance_migration_test")
assert first == second
assert str(uuid.UUID(first)) == first
def test_query_telemetry_identity_uses_execution_workspace_only():
from langbot.pkg.telemetry.identity import workspace_identity
identity = workspace_identity(SimpleNamespace(workspace_uuid="workspace-a", instance_uuid="instance-a"))
assert identity == {"workspace_uuid": "workspace-a"}
@@ -26,6 +26,7 @@ from langbot.pkg.workspace import (
WorkspaceOwnerAlreadyExistsError,
WorkspaceService,
)
from langbot.pkg.workspace.identity import workspace_uuid_from_instance_id
from langbot.pkg.workspace.policy import CloudWorkspacePolicy
@@ -113,6 +114,7 @@ async def test_ensure_singleton_workspace_is_idempotent(workspace_test_context):
first = await service.ensure_singleton_workspace()
second = await service.ensure_singleton_workspace()
assert first.uuid == workspace_uuid_from_instance_id('instance_service_test')
assert second.uuid == first.uuid
async with session_factory() as session:
assert await session.scalar(sqlalchemy.select(sqlalchemy.func.count()).select_from(Workspace)) == 1