feat: Introduce WorkflowExecution Domain Entity and Repository, Replace WorkflowRun Direct Usage, and Unify Stream Response Logic (#20067)
Signed-off-by: -LAN- <laipz8200@outlook.com>pull/19751/head
-LAN-
11 months ago
committed by
GitHub
parent
7d230acf40
commit
d31235ca13
@ -0,0 +1,591 @@
|
||||
import logging
|
||||
import time
|
||||
from collections.abc import Generator
|
||||
from typing import Optional, Union
|
||||
|
||||
from sqlalchemy import select
|
||||
from sqlalchemy.orm import Session
|
||||
|
||||
from constants.tts_auto_play_timeout import TTS_AUTO_PLAY_TIMEOUT, TTS_AUTO_PLAY_YIELD_CPU_TIME
|
||||
from core.app.apps.base_app_queue_manager import AppQueueManager
|
||||
from core.app.entities.app_invoke_entities import (
|
||||
InvokeFrom,
|
||||
WorkflowAppGenerateEntity,
|
||||
)
|
||||
from core.app.entities.queue_entities import (
|
||||
QueueAgentLogEvent,
|
||||
QueueErrorEvent,
|
||||
QueueIterationCompletedEvent,
|
||||
QueueIterationNextEvent,
|
||||
QueueIterationStartEvent,
|
||||
QueueLoopCompletedEvent,
|
||||
QueueLoopNextEvent,
|
||||
QueueLoopStartEvent,
|
||||
QueueNodeExceptionEvent,
|
||||
QueueNodeFailedEvent,
|
||||
QueueNodeInIterationFailedEvent,
|
||||
QueueNodeInLoopFailedEvent,
|
||||
QueueNodeRetryEvent,
|
||||
QueueNodeStartedEvent,
|
||||
QueueNodeSucceededEvent,
|
||||
QueueParallelBranchRunFailedEvent,
|
||||
QueueParallelBranchRunStartedEvent,
|
||||
QueueParallelBranchRunSucceededEvent,
|
||||
QueuePingEvent,
|
||||
QueueStopEvent,
|
||||
QueueTextChunkEvent,
|
||||
QueueWorkflowFailedEvent,
|
||||
QueueWorkflowPartialSuccessEvent,
|
||||
QueueWorkflowStartedEvent,
|
||||
QueueWorkflowSucceededEvent,
|
||||
)
|
||||
from core.app.entities.task_entities import (
|
||||
ErrorStreamResponse,
|
||||
MessageAudioEndStreamResponse,
|
||||
MessageAudioStreamResponse,
|
||||
StreamResponse,
|
||||
TextChunkStreamResponse,
|
||||
WorkflowAppBlockingResponse,
|
||||
WorkflowAppStreamResponse,
|
||||
WorkflowFinishStreamResponse,
|
||||
WorkflowStartStreamResponse,
|
||||
WorkflowTaskState,
|
||||
)
|
||||
from core.app.task_pipeline.based_generate_task_pipeline import BasedGenerateTaskPipeline
|
||||
from core.base.tts import AppGeneratorTTSPublisher, AudioTrunk
|
||||
from core.ops.ops_trace_manager import TraceQueueManager
|
||||
from core.workflow.entities.workflow_execution_entities import WorkflowExecution
|
||||
from core.workflow.enums import SystemVariableKey
|
||||
from core.workflow.repository.workflow_execution_repository import WorkflowExecutionRepository
|
||||
from core.workflow.repository.workflow_node_execution_repository import WorkflowNodeExecutionRepository
|
||||
from core.workflow.workflow_cycle_manager import WorkflowCycleManager
|
||||
from extensions.ext_database import db
|
||||
from models.account import Account
|
||||
from models.enums import CreatorUserRole
|
||||
from models.model import EndUser
|
||||
from models.workflow import (
|
||||
Workflow,
|
||||
WorkflowAppLog,
|
||||
WorkflowAppLogCreatedFrom,
|
||||
WorkflowRun,
|
||||
WorkflowRunStatus,
|
||||
)
|
||||
|
||||
logger = logging.getLogger(__name__)
|
||||
|
||||
|
||||
class WorkflowAppGenerateTaskPipeline:
|
||||
"""
|
||||
WorkflowAppGenerateTaskPipeline is a class that generate stream output and state management for Application.
|
||||
"""
|
||||
|
||||
def __init__(
|
||||
self,
|
||||
application_generate_entity: WorkflowAppGenerateEntity,
|
||||
workflow: Workflow,
|
||||
queue_manager: AppQueueManager,
|
||||
user: Union[Account, EndUser],
|
||||
stream: bool,
|
||||
workflow_execution_repository: WorkflowExecutionRepository,
|
||||
workflow_node_execution_repository: WorkflowNodeExecutionRepository,
|
||||
) -> None:
|
||||
self._base_task_pipeline = BasedGenerateTaskPipeline(
|
||||
application_generate_entity=application_generate_entity,
|
||||
queue_manager=queue_manager,
|
||||
stream=stream,
|
||||
)
|
||||
|
||||
if isinstance(user, EndUser):
|
||||
self._user_id = user.id
|
||||
user_session_id = user.session_id
|
||||
self._created_by_role = CreatorUserRole.END_USER
|
||||
elif isinstance(user, Account):
|
||||
self._user_id = user.id
|
||||
user_session_id = user.id
|
||||
self._created_by_role = CreatorUserRole.ACCOUNT
|
||||
else:
|
||||
raise ValueError(f"Invalid user type: {type(user)}")
|
||||
|
||||
self._workflow_cycle_manager = WorkflowCycleManager(
|
||||
application_generate_entity=application_generate_entity,
|
||||
workflow_system_variables={
|
||||
SystemVariableKey.FILES: application_generate_entity.files,
|
||||
SystemVariableKey.USER_ID: user_session_id,
|
||||
SystemVariableKey.APP_ID: application_generate_entity.app_config.app_id,
|
||||
SystemVariableKey.WORKFLOW_ID: workflow.id,
|
||||
SystemVariableKey.WORKFLOW_RUN_ID: application_generate_entity.workflow_run_id,
|
||||
},
|
||||
workflow_execution_repository=workflow_execution_repository,
|
||||
workflow_node_execution_repository=workflow_node_execution_repository,
|
||||
)
|
||||
|
||||
self._application_generate_entity = application_generate_entity
|
||||
self._workflow_id = workflow.id
|
||||
self._workflow_features_dict = workflow.features_dict
|
||||
self._task_state = WorkflowTaskState()
|
||||
self._workflow_run_id = ""
|
||||
|
||||
def process(self) -> Union[WorkflowAppBlockingResponse, Generator[WorkflowAppStreamResponse, None, None]]:
|
||||
"""
|
||||
Process generate task pipeline.
|
||||
:return:
|
||||
"""
|
||||
generator = self._wrapper_process_stream_response(trace_manager=self._application_generate_entity.trace_manager)
|
||||
if self._base_task_pipeline._stream:
|
||||
return self._to_stream_response(generator)
|
||||
else:
|
||||
return self._to_blocking_response(generator)
|
||||
|
||||
def _to_blocking_response(self, generator: Generator[StreamResponse, None, None]) -> WorkflowAppBlockingResponse:
|
||||
"""
|
||||
To blocking response.
|
||||
:return:
|
||||
"""
|
||||
for stream_response in generator:
|
||||
if isinstance(stream_response, ErrorStreamResponse):
|
||||
raise stream_response.err
|
||||
elif isinstance(stream_response, WorkflowFinishStreamResponse):
|
||||
response = WorkflowAppBlockingResponse(
|
||||
task_id=self._application_generate_entity.task_id,
|
||||
workflow_run_id=stream_response.data.id,
|
||||
data=WorkflowAppBlockingResponse.Data(
|
||||
id=stream_response.data.id,
|
||||
workflow_id=stream_response.data.workflow_id,
|
||||
status=stream_response.data.status,
|
||||
outputs=stream_response.data.outputs,
|
||||
error=stream_response.data.error,
|
||||
elapsed_time=stream_response.data.elapsed_time,
|
||||
total_tokens=stream_response.data.total_tokens,
|
||||
total_steps=stream_response.data.total_steps,
|
||||
created_at=int(stream_response.data.created_at),
|
||||
finished_at=int(stream_response.data.finished_at),
|
||||
),
|
||||
)
|
||||
|
||||
return response
|
||||
else:
|
||||
continue
|
||||
|
||||
raise ValueError("queue listening stopped unexpectedly.")
|
||||
|
||||
def _to_stream_response(
|
||||
self, generator: Generator[StreamResponse, None, None]
|
||||
) -> Generator[WorkflowAppStreamResponse, None, None]:
|
||||
"""
|
||||
To stream response.
|
||||
:return:
|
||||
"""
|
||||
workflow_run_id = None
|
||||
for stream_response in generator:
|
||||
if isinstance(stream_response, WorkflowStartStreamResponse):
|
||||
workflow_run_id = stream_response.workflow_run_id
|
||||
|
||||
yield WorkflowAppStreamResponse(workflow_run_id=workflow_run_id, stream_response=stream_response)
|
||||
|
||||
def _listen_audio_msg(self, publisher: AppGeneratorTTSPublisher | None, task_id: str):
|
||||
if not publisher:
|
||||
return None
|
||||
audio_msg = publisher.check_and_get_audio()
|
||||
if audio_msg and isinstance(audio_msg, AudioTrunk) and audio_msg.status != "finish":
|
||||
return MessageAudioStreamResponse(audio=audio_msg.audio, task_id=task_id)
|
||||
return None
|
||||
|
||||
def _wrapper_process_stream_response(
|
||||
self, trace_manager: Optional[TraceQueueManager] = None
|
||||
) -> Generator[StreamResponse, None, None]:
|
||||
tts_publisher = None
|
||||
task_id = self._application_generate_entity.task_id
|
||||
tenant_id = self._application_generate_entity.app_config.tenant_id
|
||||
features_dict = self._workflow_features_dict
|
||||
|
||||
if (
|
||||
features_dict.get("text_to_speech")
|
||||
and features_dict["text_to_speech"].get("enabled")
|
||||
and features_dict["text_to_speech"].get("autoPlay") == "enabled"
|
||||
):
|
||||
tts_publisher = AppGeneratorTTSPublisher(
|
||||
tenant_id, features_dict["text_to_speech"].get("voice"), features_dict["text_to_speech"].get("language")
|
||||
)
|
||||
|
||||
for response in self._process_stream_response(tts_publisher=tts_publisher, trace_manager=trace_manager):
|
||||
while True:
|
||||
audio_response = self._listen_audio_msg(publisher=tts_publisher, task_id=task_id)
|
||||
if audio_response:
|
||||
yield audio_response
|
||||
else:
|
||||
break
|
||||
yield response
|
||||
|
||||
start_listener_time = time.time()
|
||||
while (time.time() - start_listener_time) < TTS_AUTO_PLAY_TIMEOUT:
|
||||
try:
|
||||
if not tts_publisher:
|
||||
break
|
||||
audio_trunk = tts_publisher.check_and_get_audio()
|
||||
if audio_trunk is None:
|
||||
# release cpu
|
||||
# sleep 20 ms ( 40ms => 1280 byte audio file,20ms => 640 byte audio file)
|
||||
time.sleep(TTS_AUTO_PLAY_YIELD_CPU_TIME)
|
||||
continue
|
||||
if audio_trunk.status == "finish":
|
||||
break
|
||||
else:
|
||||
yield MessageAudioStreamResponse(audio=audio_trunk.audio, task_id=task_id)
|
||||
except Exception:
|
||||
logger.exception(f"Fails to get audio trunk, task_id: {task_id}")
|
||||
break
|
||||
if tts_publisher:
|
||||
yield MessageAudioEndStreamResponse(audio="", task_id=task_id)
|
||||
|
||||
def _process_stream_response(
|
||||
self,
|
||||
tts_publisher: Optional[AppGeneratorTTSPublisher] = None,
|
||||
trace_manager: Optional[TraceQueueManager] = None,
|
||||
) -> Generator[StreamResponse, None, None]:
|
||||
"""
|
||||
Process stream response.
|
||||
:return:
|
||||
"""
|
||||
graph_runtime_state = None
|
||||
|
||||
for queue_message in self._base_task_pipeline._queue_manager.listen():
|
||||
event = queue_message.event
|
||||
|
||||
if isinstance(event, QueuePingEvent):
|
||||
yield self._base_task_pipeline._ping_stream_response()
|
||||
elif isinstance(event, QueueErrorEvent):
|
||||
err = self._base_task_pipeline._handle_error(event=event)
|
||||
yield self._base_task_pipeline._error_to_stream_response(err)
|
||||
break
|
||||
elif isinstance(event, QueueWorkflowStartedEvent):
|
||||
# override graph runtime state
|
||||
graph_runtime_state = event.graph_runtime_state
|
||||
|
||||
with Session(db.engine, expire_on_commit=False) as session:
|
||||
# init workflow run
|
||||
workflow_execution = self._workflow_cycle_manager.handle_workflow_run_start(
|
||||
session=session,
|
||||
workflow_id=self._workflow_id,
|
||||
)
|
||||
self._workflow_run_id = workflow_execution.id
|
||||
start_resp = self._workflow_cycle_manager.workflow_start_to_stream_response(
|
||||
task_id=self._application_generate_entity.task_id,
|
||||
workflow_execution=workflow_execution,
|
||||
)
|
||||
|
||||
yield start_resp
|
||||
elif isinstance(
|
||||
event,
|
||||
QueueNodeRetryEvent,
|
||||
):
|
||||
if not self._workflow_run_id:
|
||||
raise ValueError("workflow run not initialized.")
|
||||
with Session(db.engine, expire_on_commit=False) as session:
|
||||
workflow_node_execution = self._workflow_cycle_manager.handle_workflow_node_execution_retried(
|
||||
workflow_execution_id=self._workflow_run_id,
|
||||
event=event,
|
||||
)
|
||||
response = self._workflow_cycle_manager.workflow_node_retry_to_stream_response(
|
||||
event=event,
|
||||
task_id=self._application_generate_entity.task_id,
|
||||
workflow_node_execution=workflow_node_execution,
|
||||
)
|
||||
session.commit()
|
||||
|
||||
if response:
|
||||
yield response
|
||||
elif isinstance(event, QueueNodeStartedEvent):
|
||||
if not self._workflow_run_id:
|
||||
raise ValueError("workflow run not initialized.")
|
||||
|
||||
workflow_node_execution = self._workflow_cycle_manager.handle_node_execution_start(
|
||||
workflow_execution_id=self._workflow_run_id, event=event
|
||||
)
|
||||
node_start_response = self._workflow_cycle_manager.workflow_node_start_to_stream_response(
|
||||
event=event,
|
||||
task_id=self._application_generate_entity.task_id,
|
||||
workflow_node_execution=workflow_node_execution,
|
||||
)
|
||||
|
||||
if node_start_response:
|
||||
yield node_start_response
|
||||
elif isinstance(event, QueueNodeSucceededEvent):
|
||||
workflow_node_execution = self._workflow_cycle_manager.handle_workflow_node_execution_success(
|
||||
event=event
|
||||
)
|
||||
node_success_response = self._workflow_cycle_manager.workflow_node_finish_to_stream_response(
|
||||
event=event,
|
||||
task_id=self._application_generate_entity.task_id,
|
||||
workflow_node_execution=workflow_node_execution,
|
||||
)
|
||||
|
||||
if node_success_response:
|
||||
yield node_success_response
|
||||
elif isinstance(
|
||||
event,
|
||||
QueueNodeFailedEvent
|
||||
| QueueNodeInIterationFailedEvent
|
||||
| QueueNodeInLoopFailedEvent
|
||||
| QueueNodeExceptionEvent,
|
||||
):
|
||||
workflow_node_execution = self._workflow_cycle_manager.handle_workflow_node_execution_failed(
|
||||
event=event,
|
||||
)
|
||||
node_failed_response = self._workflow_cycle_manager.workflow_node_finish_to_stream_response(
|
||||
event=event,
|
||||
task_id=self._application_generate_entity.task_id,
|
||||
workflow_node_execution=workflow_node_execution,
|
||||
)
|
||||
|
||||
if node_failed_response:
|
||||
yield node_failed_response
|
||||
|
||||
elif isinstance(event, QueueParallelBranchRunStartedEvent):
|
||||
if not self._workflow_run_id:
|
||||
raise ValueError("workflow run not initialized.")
|
||||
|
||||
parallel_start_resp = self._workflow_cycle_manager.workflow_parallel_branch_start_to_stream_response(
|
||||
task_id=self._application_generate_entity.task_id,
|
||||
workflow_execution_id=self._workflow_run_id,
|
||||
event=event,
|
||||
)
|
||||
|
||||
yield parallel_start_resp
|
||||
|
||||
elif isinstance(event, QueueParallelBranchRunSucceededEvent | QueueParallelBranchRunFailedEvent):
|
||||
if not self._workflow_run_id:
|
||||
raise ValueError("workflow run not initialized.")
|
||||
|
||||
parallel_finish_resp = (
|
||||
self._workflow_cycle_manager.workflow_parallel_branch_finished_to_stream_response(
|
||||
task_id=self._application_generate_entity.task_id,
|
||||
workflow_execution_id=self._workflow_run_id,
|
||||
event=event,
|
||||
)
|
||||
)
|
||||
|
||||
yield parallel_finish_resp
|
||||
|
||||
elif isinstance(event, QueueIterationStartEvent):
|
||||
if not self._workflow_run_id:
|
||||
raise ValueError("workflow run not initialized.")
|
||||
|
||||
iter_start_resp = self._workflow_cycle_manager.workflow_iteration_start_to_stream_response(
|
||||
task_id=self._application_generate_entity.task_id,
|
||||
workflow_execution_id=self._workflow_run_id,
|
||||
event=event,
|
||||
)
|
||||
|
||||
yield iter_start_resp
|
||||
|
||||
elif isinstance(event, QueueIterationNextEvent):
|
||||
if not self._workflow_run_id:
|
||||
raise ValueError("workflow run not initialized.")
|
||||
|
||||
iter_next_resp = self._workflow_cycle_manager.workflow_iteration_next_to_stream_response(
|
||||
task_id=self._application_generate_entity.task_id,
|
||||
workflow_execution_id=self._workflow_run_id,
|
||||
event=event,
|
||||
)
|
||||
|
||||
yield iter_next_resp
|
||||
|
||||
elif isinstance(event, QueueIterationCompletedEvent):
|
||||
if not self._workflow_run_id:
|
||||
raise ValueError("workflow run not initialized.")
|
||||
|
||||
iter_finish_resp = self._workflow_cycle_manager.workflow_iteration_completed_to_stream_response(
|
||||
task_id=self._application_generate_entity.task_id,
|
||||
workflow_execution_id=self._workflow_run_id,
|
||||
event=event,
|
||||
)
|
||||
|
||||
yield iter_finish_resp
|
||||
|
||||
elif isinstance(event, QueueLoopStartEvent):
|
||||
if not self._workflow_run_id:
|
||||
raise ValueError("workflow run not initialized.")
|
||||
|
||||
loop_start_resp = self._workflow_cycle_manager.workflow_loop_start_to_stream_response(
|
||||
task_id=self._application_generate_entity.task_id,
|
||||
workflow_execution_id=self._workflow_run_id,
|
||||
event=event,
|
||||
)
|
||||
|
||||
yield loop_start_resp
|
||||
|
||||
elif isinstance(event, QueueLoopNextEvent):
|
||||
if not self._workflow_run_id:
|
||||
raise ValueError("workflow run not initialized.")
|
||||
|
||||
loop_next_resp = self._workflow_cycle_manager.workflow_loop_next_to_stream_response(
|
||||
task_id=self._application_generate_entity.task_id,
|
||||
workflow_execution_id=self._workflow_run_id,
|
||||
event=event,
|
||||
)
|
||||
|
||||
yield loop_next_resp
|
||||
|
||||
elif isinstance(event, QueueLoopCompletedEvent):
|
||||
if not self._workflow_run_id:
|
||||
raise ValueError("workflow run not initialized.")
|
||||
|
||||
loop_finish_resp = self._workflow_cycle_manager.workflow_loop_completed_to_stream_response(
|
||||
task_id=self._application_generate_entity.task_id,
|
||||
workflow_execution_id=self._workflow_run_id,
|
||||
event=event,
|
||||
)
|
||||
|
||||
yield loop_finish_resp
|
||||
|
||||
elif isinstance(event, QueueWorkflowSucceededEvent):
|
||||
if not self._workflow_run_id:
|
||||
raise ValueError("workflow run not initialized.")
|
||||
if not graph_runtime_state:
|
||||
raise ValueError("graph runtime state not initialized.")
|
||||
|
||||
with Session(db.engine, expire_on_commit=False) as session:
|
||||
workflow_execution = self._workflow_cycle_manager.handle_workflow_run_success(
|
||||
workflow_run_id=self._workflow_run_id,
|
||||
total_tokens=graph_runtime_state.total_tokens,
|
||||
total_steps=graph_runtime_state.node_run_steps,
|
||||
outputs=event.outputs,
|
||||
conversation_id=None,
|
||||
trace_manager=trace_manager,
|
||||
)
|
||||
|
||||
# save workflow app log
|
||||
self._save_workflow_app_log(session=session, workflow_execution=workflow_execution)
|
||||
|
||||
workflow_finish_resp = self._workflow_cycle_manager.workflow_finish_to_stream_response(
|
||||
session=session,
|
||||
task_id=self._application_generate_entity.task_id,
|
||||
workflow_execution=workflow_execution,
|
||||
)
|
||||
session.commit()
|
||||
|
||||
yield workflow_finish_resp
|
||||
elif isinstance(event, QueueWorkflowPartialSuccessEvent):
|
||||
if not self._workflow_run_id:
|
||||
raise ValueError("workflow run not initialized.")
|
||||
if not graph_runtime_state:
|
||||
raise ValueError("graph runtime state not initialized.")
|
||||
|
||||
with Session(db.engine, expire_on_commit=False) as session:
|
||||
workflow_execution = self._workflow_cycle_manager.handle_workflow_run_partial_success(
|
||||
workflow_run_id=self._workflow_run_id,
|
||||
total_tokens=graph_runtime_state.total_tokens,
|
||||
total_steps=graph_runtime_state.node_run_steps,
|
||||
outputs=event.outputs,
|
||||
exceptions_count=event.exceptions_count,
|
||||
conversation_id=None,
|
||||
trace_manager=trace_manager,
|
||||
)
|
||||
|
||||
# save workflow app log
|
||||
self._save_workflow_app_log(session=session, workflow_execution=workflow_execution)
|
||||
|
||||
workflow_finish_resp = self._workflow_cycle_manager.workflow_finish_to_stream_response(
|
||||
session=session,
|
||||
task_id=self._application_generate_entity.task_id,
|
||||
workflow_execution=workflow_execution,
|
||||
)
|
||||
session.commit()
|
||||
|
||||
yield workflow_finish_resp
|
||||
elif isinstance(event, QueueWorkflowFailedEvent | QueueStopEvent):
|
||||
if not self._workflow_run_id:
|
||||
raise ValueError("workflow run not initialized.")
|
||||
if not graph_runtime_state:
|
||||
raise ValueError("graph runtime state not initialized.")
|
||||
|
||||
with Session(db.engine, expire_on_commit=False) as session:
|
||||
workflow_execution = self._workflow_cycle_manager.handle_workflow_run_failed(
|
||||
workflow_run_id=self._workflow_run_id,
|
||||
total_tokens=graph_runtime_state.total_tokens,
|
||||
total_steps=graph_runtime_state.node_run_steps,
|
||||
status=WorkflowRunStatus.FAILED
|
||||
if isinstance(event, QueueWorkflowFailedEvent)
|
||||
else WorkflowRunStatus.STOPPED,
|
||||
error_message=event.error
|
||||
if isinstance(event, QueueWorkflowFailedEvent)
|
||||
else event.get_stop_reason(),
|
||||
conversation_id=None,
|
||||
trace_manager=trace_manager,
|
||||
exceptions_count=event.exceptions_count if isinstance(event, QueueWorkflowFailedEvent) else 0,
|
||||
)
|
||||
|
||||
# save workflow app log
|
||||
self._save_workflow_app_log(session=session, workflow_execution=workflow_execution)
|
||||
|
||||
workflow_finish_resp = self._workflow_cycle_manager.workflow_finish_to_stream_response(
|
||||
session=session,
|
||||
task_id=self._application_generate_entity.task_id,
|
||||
workflow_execution=workflow_execution,
|
||||
)
|
||||
session.commit()
|
||||
|
||||
yield workflow_finish_resp
|
||||
elif isinstance(event, QueueTextChunkEvent):
|
||||
delta_text = event.text
|
||||
if delta_text is None:
|
||||
continue
|
||||
|
||||
# only publish tts message at text chunk streaming
|
||||
if tts_publisher:
|
||||
tts_publisher.publish(queue_message)
|
||||
|
||||
self._task_state.answer += delta_text
|
||||
yield self._text_chunk_to_stream_response(
|
||||
delta_text, from_variable_selector=event.from_variable_selector
|
||||
)
|
||||
elif isinstance(event, QueueAgentLogEvent):
|
||||
yield self._workflow_cycle_manager.handle_agent_log(
|
||||
task_id=self._application_generate_entity.task_id, event=event
|
||||
)
|
||||
else:
|
||||
continue
|
||||
|
||||
if tts_publisher:
|
||||
tts_publisher.publish(None)
|
||||
|
||||
def _save_workflow_app_log(self, *, session: Session, workflow_execution: WorkflowExecution) -> None:
|
||||
workflow_run = session.scalar(select(WorkflowRun).where(WorkflowRun.id == workflow_execution.id))
|
||||
assert workflow_run is not None
|
||||
invoke_from = self._application_generate_entity.invoke_from
|
||||
if invoke_from == InvokeFrom.SERVICE_API:
|
||||
created_from = WorkflowAppLogCreatedFrom.SERVICE_API
|
||||
elif invoke_from == InvokeFrom.EXPLORE:
|
||||
created_from = WorkflowAppLogCreatedFrom.INSTALLED_APP
|
||||
elif invoke_from == InvokeFrom.WEB_APP:
|
||||
created_from = WorkflowAppLogCreatedFrom.WEB_APP
|
||||
else:
|
||||
# not save log for debugging
|
||||
return
|
||||
|
||||
workflow_app_log = WorkflowAppLog()
|
||||
workflow_app_log.tenant_id = workflow_run.tenant_id
|
||||
workflow_app_log.app_id = workflow_run.app_id
|
||||
workflow_app_log.workflow_id = workflow_run.workflow_id
|
||||
workflow_app_log.workflow_run_id = workflow_run.id
|
||||
workflow_app_log.created_from = created_from.value
|
||||
workflow_app_log.created_by_role = self._created_by_role
|
||||
workflow_app_log.created_by = self._user_id
|
||||
|
||||
session.add(workflow_app_log)
|
||||
session.commit()
|
||||
|
||||
def _text_chunk_to_stream_response(
|
||||
self, text: str, from_variable_selector: Optional[list[str]] = None
|
||||
) -> TextChunkStreamResponse:
|
||||
"""
|
||||
Handle completed event.
|
||||
:param text: text
|
||||
:return:
|
||||
"""
|
||||
response = TextChunkStreamResponse(
|
||||
task_id=self._application_generate_entity.task_id,
|
||||
data=TextChunkStreamResponse.Data(text=text, from_variable_selector=from_variable_selector),
|
||||
)
|
||||
|
||||
return response
|
||||
@ -0,0 +1,242 @@
|
||||
"""
|
||||
SQLAlchemy implementation of the WorkflowExecutionRepository.
|
||||
"""
|
||||
|
||||
import json
|
||||
import logging
|
||||
from typing import Optional, Union
|
||||
|
||||
from sqlalchemy import select
|
||||
from sqlalchemy.engine import Engine
|
||||
from sqlalchemy.orm import sessionmaker
|
||||
|
||||
from core.workflow.entities.workflow_execution_entities import (
|
||||
WorkflowExecution,
|
||||
WorkflowExecutionStatus,
|
||||
WorkflowType,
|
||||
)
|
||||
from core.workflow.repository.workflow_execution_repository import WorkflowExecutionRepository
|
||||
from models import (
|
||||
Account,
|
||||
CreatorUserRole,
|
||||
EndUser,
|
||||
WorkflowRun,
|
||||
)
|
||||
from models.enums import WorkflowRunTriggeredFrom
|
||||
|
||||
logger = logging.getLogger(__name__)
|
||||
|
||||
|
||||
class SQLAlchemyWorkflowExecutionRepository(WorkflowExecutionRepository):
|
||||
"""
|
||||
SQLAlchemy implementation of the WorkflowExecutionRepository interface.
|
||||
|
||||
This implementation supports multi-tenancy by filtering operations based on tenant_id.
|
||||
Each method creates its own session, handles the transaction, and commits changes
|
||||
to the database. This prevents long-running connections in the workflow core.
|
||||
|
||||
This implementation also includes an in-memory cache for workflow executions to improve
|
||||
performance by reducing database queries.
|
||||
"""
|
||||
|
||||
def __init__(
|
||||
self,
|
||||
session_factory: sessionmaker | Engine,
|
||||
user: Union[Account, EndUser],
|
||||
app_id: Optional[str],
|
||||
triggered_from: Optional[WorkflowRunTriggeredFrom],
|
||||
):
|
||||
"""
|
||||
Initialize the repository with a SQLAlchemy sessionmaker or engine and context information.
|
||||
|
||||
Args:
|
||||
session_factory: SQLAlchemy sessionmaker or engine for creating sessions
|
||||
user: Account or EndUser object containing tenant_id, user ID, and role information
|
||||
app_id: App ID for filtering by application (can be None)
|
||||
triggered_from: Source of the execution trigger (DEBUGGING or APP_RUN)
|
||||
"""
|
||||
# If an engine is provided, create a sessionmaker from it
|
||||
if isinstance(session_factory, Engine):
|
||||
self._session_factory = sessionmaker(bind=session_factory, expire_on_commit=False)
|
||||
elif isinstance(session_factory, sessionmaker):
|
||||
self._session_factory = session_factory
|
||||
else:
|
||||
raise ValueError(
|
||||
f"Invalid session_factory type {type(session_factory).__name__}; expected sessionmaker or Engine"
|
||||
)
|
||||
|
||||
# Extract tenant_id from user
|
||||
tenant_id: str | None = user.tenant_id if isinstance(user, EndUser) else user.current_tenant_id
|
||||
if not tenant_id:
|
||||
raise ValueError("User must have a tenant_id or current_tenant_id")
|
||||
self._tenant_id = tenant_id
|
||||
|
||||
# Store app context
|
||||
self._app_id = app_id
|
||||
|
||||
# Extract user context
|
||||
self._triggered_from = triggered_from
|
||||
self._creator_user_id = user.id
|
||||
|
||||
# Determine user role based on user type
|
||||
self._creator_user_role = CreatorUserRole.ACCOUNT if isinstance(user, Account) else CreatorUserRole.END_USER
|
||||
|
||||
# Initialize in-memory cache for workflow executions
|
||||
# Key: execution_id, Value: WorkflowRun (DB model)
|
||||
self._execution_cache: dict[str, WorkflowRun] = {}
|
||||
|
||||
def _to_domain_model(self, db_model: WorkflowRun) -> WorkflowExecution:
|
||||
"""
|
||||
Convert a database model to a domain model.
|
||||
|
||||
Args:
|
||||
db_model: The database model to convert
|
||||
|
||||
Returns:
|
||||
The domain model
|
||||
"""
|
||||
# Parse JSON fields
|
||||
inputs = db_model.inputs_dict
|
||||
outputs = db_model.outputs_dict
|
||||
graph = db_model.graph_dict
|
||||
|
||||
# Convert status to domain enum
|
||||
status = WorkflowExecutionStatus(db_model.status)
|
||||
|
||||
return WorkflowExecution(
|
||||
id=db_model.id,
|
||||
workflow_id=db_model.workflow_id,
|
||||
sequence_number=db_model.sequence_number,
|
||||
type=WorkflowType(db_model.type),
|
||||
workflow_version=db_model.version,
|
||||
graph=graph,
|
||||
inputs=inputs,
|
||||
outputs=outputs,
|
||||
status=status,
|
||||
error_message=db_model.error or "",
|
||||
total_tokens=db_model.total_tokens,
|
||||
total_steps=db_model.total_steps,
|
||||
exceptions_count=db_model.exceptions_count,
|
||||
started_at=db_model.created_at,
|
||||
finished_at=db_model.finished_at,
|
||||
)
|
||||
|
||||
def _to_db_model(self, domain_model: WorkflowExecution) -> WorkflowRun:
|
||||
"""
|
||||
Convert a domain model to a database model.
|
||||
|
||||
Args:
|
||||
domain_model: The domain model to convert
|
||||
|
||||
Returns:
|
||||
The database model
|
||||
"""
|
||||
# Use values from constructor if provided
|
||||
if not self._triggered_from:
|
||||
raise ValueError("triggered_from is required in repository constructor")
|
||||
if not self._creator_user_id:
|
||||
raise ValueError("created_by is required in repository constructor")
|
||||
if not self._creator_user_role:
|
||||
raise ValueError("created_by_role is required in repository constructor")
|
||||
|
||||
db_model = WorkflowRun()
|
||||
db_model.id = domain_model.id
|
||||
db_model.tenant_id = self._tenant_id
|
||||
if self._app_id is not None:
|
||||
db_model.app_id = self._app_id
|
||||
db_model.workflow_id = domain_model.workflow_id
|
||||
db_model.triggered_from = self._triggered_from
|
||||
db_model.sequence_number = domain_model.sequence_number
|
||||
db_model.type = domain_model.type
|
||||
db_model.version = domain_model.workflow_version
|
||||
db_model.graph = json.dumps(domain_model.graph) if domain_model.graph else None
|
||||
db_model.inputs = json.dumps(domain_model.inputs) if domain_model.inputs else None
|
||||
db_model.outputs = json.dumps(domain_model.outputs) if domain_model.outputs else None
|
||||
db_model.status = domain_model.status
|
||||
db_model.error = domain_model.error_message if domain_model.error_message else None
|
||||
db_model.total_tokens = domain_model.total_tokens
|
||||
db_model.total_steps = domain_model.total_steps
|
||||
db_model.exceptions_count = domain_model.exceptions_count
|
||||
db_model.created_by_role = self._creator_user_role
|
||||
db_model.created_by = self._creator_user_id
|
||||
db_model.created_at = domain_model.started_at
|
||||
db_model.finished_at = domain_model.finished_at
|
||||
|
||||
# Calculate elapsed time if finished_at is available
|
||||
if domain_model.finished_at:
|
||||
db_model.elapsed_time = (domain_model.finished_at - domain_model.started_at).total_seconds()
|
||||
else:
|
||||
db_model.elapsed_time = 0
|
||||
|
||||
return db_model
|
||||
|
||||
def save(self, execution: WorkflowExecution) -> None:
|
||||
"""
|
||||
Save or update a WorkflowExecution domain entity to the database.
|
||||
|
||||
This method serves as a domain-to-database adapter that:
|
||||
1. Converts the domain entity to its database representation
|
||||
2. Persists the database model using SQLAlchemy's merge operation
|
||||
3. Maintains proper multi-tenancy by including tenant context during conversion
|
||||
4. Updates the in-memory cache for faster subsequent lookups
|
||||
|
||||
The method handles both creating new records and updating existing ones through
|
||||
SQLAlchemy's merge operation.
|
||||
|
||||
Args:
|
||||
execution: The WorkflowExecution domain entity to persist
|
||||
"""
|
||||
# Convert domain model to database model using tenant context and other attributes
|
||||
db_model = self._to_db_model(execution)
|
||||
|
||||
# Create a new database session
|
||||
with self._session_factory() as session:
|
||||
# SQLAlchemy merge intelligently handles both insert and update operations
|
||||
# based on the presence of the primary key
|
||||
session.merge(db_model)
|
||||
session.commit()
|
||||
|
||||
# Update the in-memory cache for faster subsequent lookups
|
||||
logger.debug(f"Updating cache for execution_id: {db_model.id}")
|
||||
self._execution_cache[db_model.id] = db_model
|
||||
|
||||
def get(self, execution_id: str) -> Optional[WorkflowExecution]:
|
||||
"""
|
||||
Retrieve a WorkflowExecution by its ID.
|
||||
|
||||
First checks the in-memory cache, and if not found, queries the database.
|
||||
If found in the database, adds it to the cache for future lookups.
|
||||
|
||||
Args:
|
||||
execution_id: The workflow execution ID
|
||||
|
||||
Returns:
|
||||
The WorkflowExecution instance if found, None otherwise
|
||||
"""
|
||||
# First check the cache
|
||||
if execution_id in self._execution_cache:
|
||||
logger.debug(f"Cache hit for execution_id: {execution_id}")
|
||||
# Convert cached DB model to domain model
|
||||
cached_db_model = self._execution_cache[execution_id]
|
||||
return self._to_domain_model(cached_db_model)
|
||||
|
||||
# If not in cache, query the database
|
||||
logger.debug(f"Cache miss for execution_id: {execution_id}, querying database")
|
||||
with self._session_factory() as session:
|
||||
stmt = select(WorkflowRun).where(
|
||||
WorkflowRun.id == execution_id,
|
||||
WorkflowRun.tenant_id == self._tenant_id,
|
||||
)
|
||||
|
||||
if self._app_id:
|
||||
stmt = stmt.where(WorkflowRun.app_id == self._app_id)
|
||||
|
||||
db_model = session.scalar(stmt)
|
||||
if db_model:
|
||||
# Add DB model to cache
|
||||
self._execution_cache[execution_id] = db_model
|
||||
|
||||
# Convert to domain model and return
|
||||
return self._to_domain_model(db_model)
|
||||
|
||||
return None
|
||||
@ -0,0 +1,91 @@
|
||||
"""
|
||||
Domain entities for workflow execution.
|
||||
|
||||
Models are independent of the storage mechanism and don't contain
|
||||
implementation details like tenant_id, app_id, etc.
|
||||
"""
|
||||
|
||||
from collections.abc import Mapping
|
||||
from datetime import UTC, datetime
|
||||
from enum import StrEnum
|
||||
from typing import Any, Optional
|
||||
|
||||
from pydantic import BaseModel, Field
|
||||
|
||||
|
||||
class WorkflowType(StrEnum):
|
||||
"""
|
||||
Workflow Type Enum for domain layer
|
||||
"""
|
||||
|
||||
WORKFLOW = "workflow"
|
||||
CHAT = "chat"
|
||||
|
||||
|
||||
class WorkflowExecutionStatus(StrEnum):
|
||||
RUNNING = "running"
|
||||
SUCCEEDED = "succeeded"
|
||||
FAILED = "failed"
|
||||
STOPPED = "stopped"
|
||||
PARTIAL_SUCCEEDED = "partial-succeeded"
|
||||
|
||||
|
||||
class WorkflowExecution(BaseModel):
|
||||
"""
|
||||
Domain model for workflow execution based on WorkflowRun but without
|
||||
user, tenant, and app attributes.
|
||||
"""
|
||||
|
||||
id: str = Field(...)
|
||||
workflow_id: str = Field(...)
|
||||
workflow_version: str = Field(...)
|
||||
sequence_number: int = Field(...)
|
||||
|
||||
type: WorkflowType = Field(...)
|
||||
graph: Mapping[str, Any] = Field(...)
|
||||
|
||||
inputs: Mapping[str, Any] = Field(...)
|
||||
outputs: Optional[Mapping[str, Any]] = None
|
||||
|
||||
status: WorkflowExecutionStatus = WorkflowExecutionStatus.RUNNING
|
||||
error_message: str = Field(default="")
|
||||
total_tokens: int = Field(default=0)
|
||||
total_steps: int = Field(default=0)
|
||||
exceptions_count: int = Field(default=0)
|
||||
|
||||
started_at: datetime = Field(...)
|
||||
finished_at: Optional[datetime] = None
|
||||
|
||||
@property
|
||||
def elapsed_time(self) -> float:
|
||||
"""
|
||||
Calculate elapsed time in seconds.
|
||||
If workflow is not finished, use current time.
|
||||
"""
|
||||
end_time = self.finished_at or datetime.now(UTC).replace(tzinfo=None)
|
||||
return (end_time - self.started_at).total_seconds()
|
||||
|
||||
@classmethod
|
||||
def new(
|
||||
cls,
|
||||
*,
|
||||
id: str,
|
||||
workflow_id: str,
|
||||
sequence_number: int,
|
||||
type: WorkflowType,
|
||||
workflow_version: str,
|
||||
graph: Mapping[str, Any],
|
||||
inputs: Mapping[str, Any],
|
||||
started_at: datetime,
|
||||
) -> "WorkflowExecution":
|
||||
return WorkflowExecution(
|
||||
id=id,
|
||||
workflow_id=workflow_id,
|
||||
sequence_number=sequence_number,
|
||||
type=type,
|
||||
workflow_version=workflow_version,
|
||||
graph=graph,
|
||||
inputs=inputs,
|
||||
status=WorkflowExecutionStatus.RUNNING,
|
||||
started_at=started_at,
|
||||
)
|
||||
@ -0,0 +1,42 @@
|
||||
from typing import Optional, Protocol
|
||||
|
||||
from core.workflow.entities.workflow_execution_entities import WorkflowExecution
|
||||
|
||||
|
||||
class WorkflowExecutionRepository(Protocol):
|
||||
"""
|
||||
Repository interface for WorkflowExecution.
|
||||
|
||||
This interface defines the contract for accessing and manipulating
|
||||
WorkflowExecution data, regardless of the underlying storage mechanism.
|
||||
|
||||
Note: Domain-specific concepts like multi-tenancy (tenant_id), application context (app_id),
|
||||
and other implementation details should be handled at the implementation level, not in
|
||||
the core interface. This keeps the core domain model clean and independent of specific
|
||||
application domains or deployment scenarios.
|
||||
"""
|
||||
|
||||
def save(self, execution: WorkflowExecution) -> None:
|
||||
"""
|
||||
Save or update a WorkflowExecution instance.
|
||||
|
||||
This method handles both creating new records and updating existing ones.
|
||||
The implementation should determine whether to create or update based on
|
||||
the execution's ID or other identifying fields.
|
||||
|
||||
Args:
|
||||
execution: The WorkflowExecution instance to save or update
|
||||
"""
|
||||
...
|
||||
|
||||
def get(self, execution_id: str) -> Optional[WorkflowExecution]:
|
||||
"""
|
||||
Retrieve a WorkflowExecution by its ID.
|
||||
|
||||
Args:
|
||||
execution_id: The workflow execution ID
|
||||
|
||||
Returns:
|
||||
The WorkflowExecution instance if found, None otherwise
|
||||
"""
|
||||
...
|
||||
Loading…
Reference in New Issue