estel/dify
1
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity

feat(workflow): domain model for workflow node execution (#19430)

Browse Source 
Signed-off-by: -LAN- <laipz8200@outlook.com>
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
This commit is contained in:
-LAN-
2025-05-17 00:56:16 +08:00
committed by GitHub
parent aeceb200ec
commit 4977bb21ec
31 changed files with 1108 additions and 483 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
api/core/ops/langfuse_trace/entities/langfuse_trace_entity.py
View File

@@ -1,3 +1,4 @@
from collections.abc import Mapping
from datetime import datetime
from enum import StrEnum
from typing import Any, Optional, Union
@@ -155,10 +156,10 @@ class LangfuseSpan(BaseModel):
description="The status message of the span. Additional field for context of the event. E.g. the error "
"message of an error event.",
)
input: Optional[Union[str, dict[str, Any], list, None]] = Field(
input: Optional[Union[str, Mapping[str, Any], list, None]] = Field(
default=None, description="The input of the span. Can be any JSON object."
)
output: Optional[Union[str, dict[str, Any], list, None]] = Field(
output: Optional[Union[str, Mapping[str, Any], list, None]] = Field(
default=None, description="The output of the span. Can be any JSON object."
)
version: Optional[str] = Field(

48
api/core/ops/langfuse_trace/langfuse_trace.py
View File

@@ -1,11 +1,10 @@
import json
import logging
import os
from datetime import datetime, timedelta
from typing import Optional
from langfuse import Langfuse # type: ignore
from sqlalchemy.orm import sessionmaker
from sqlalchemy.orm import Session, sessionmaker
from core.ops.base_trace_instance import BaseTraceInstance
from core.ops.entities.config_entity import LangfuseConfig
@@ -30,8 +29,9 @@ from core.ops.langfuse_trace.entities.langfuse_trace_entity import (
)
from core.ops.utils import filter_none_values
from core.repositories import SQLAlchemyWorkflowNodeExecutionRepository
from core.workflow.nodes.enums import NodeType
from extensions.ext_database import db
from models.model import EndUser
from models import Account, App, EndUser, WorkflowNodeExecutionTriggeredFrom
logger = logging.getLogger(__name__)
@@ -113,8 +113,29 @@ class LangFuseDataTrace(BaseTraceInstance):
# through workflow_run_id get all_nodes_execution using repository
session_factory = sessionmaker(bind=db.engine)
# Find the app's creator account
with Session(db.engine, expire_on_commit=False) as session:
# Get the app to find its creator
app_id = trace_info.metadata.get("app_id")
if not app_id:
raise ValueError("No app_id found in trace_info metadata")
app = session.query(App).filter(App.id == app_id).first()
if not app:
raise ValueError(f"App with id {app_id} not found")
if not app.created_by:
raise ValueError(f"App with id {app_id} has no creator (created_by is None)")
service_account = session.query(Account).filter(Account.id == app.created_by).first()
if not service_account:
raise ValueError(f"Creator account with id {app.created_by} not found for app {app_id}")
workflow_node_execution_repository = SQLAlchemyWorkflowNodeExecutionRepository(
session_factory=session_factory, tenant_id=trace_info.tenant_id
session_factory=session_factory,
user=service_account,
app_id=trace_info.metadata.get("app_id"),
triggered_from=WorkflowNodeExecutionTriggeredFrom.WORKFLOW_RUN,
)
# Get all executions for this workflow run
@@ -124,23 +145,22 @@ class LangFuseDataTrace(BaseTraceInstance):
for node_execution in workflow_node_executions:
node_execution_id = node_execution.id
tenant_id = node_execution.tenant_id
app_id = node_execution.app_id
tenant_id = trace_info.tenant_id # Use from trace_info instead
app_id = trace_info.metadata.get("app_id") # Use from trace_info instead
node_name = node_execution.title
node_type = node_execution.node_type
status = node_execution.status
if node_type == "llm":
inputs = (
json.loads(node_execution.process_data).get("prompts", {}) if node_execution.process_data else {}
)
if node_type == NodeType.LLM:
inputs = node_execution.process_data.get("prompts", {}) if node_execution.process_data else {}
else:
inputs = json.loads(node_execution.inputs) if node_execution.inputs else {}
outputs = json.loads(node_execution.outputs) if node_execution.outputs else {}
inputs = node_execution.inputs if node_execution.inputs else {}
outputs = node_execution.outputs if node_execution.outputs else {}
created_at = node_execution.created_at or datetime.now()
elapsed_time = node_execution.elapsed_time
finished_at = created_at + timedelta(seconds=elapsed_time)
metadata = json.loads(node_execution.execution_metadata) if node_execution.execution_metadata else {}
execution_metadata = node_execution.metadata if node_execution.metadata else {}
metadata = {str(k): v for k, v in execution_metadata.items()}
metadata.update(
{
"workflow_run_id": trace_info.workflow_run_id,
@@ -152,7 +172,7 @@ class LangFuseDataTrace(BaseTraceInstance):
"status": status,
}
)
process_data = json.loads(node_execution.process_data) if node_execution.process_data else {}
process_data = node_execution.process_data if node_execution.process_data else {}
model_provider = process_data.get("model_provider", None)
model_name = process_data.get("model_name", None)
if model_provider is not None and model_name is not None:

5
api/core/ops/langsmith_trace/entities/langsmith_trace_entity.py
View File

@@ -1,3 +1,4 @@
from collections.abc import Mapping
from datetime import datetime
from enum import StrEnum
from typing import Any, Optional, Union
@@ -30,8 +31,8 @@ class LangSmithMultiModel(BaseModel):
class LangSmithRunModel(LangSmithTokenUsage, LangSmithMultiModel):
name: Optional[str] = Field(..., description="Name of the run")
inputs: Optional[Union[str, dict[str, Any], list, None]] = Field(None, description="Inputs of the run")
outputs: Optional[Union[str, dict[str, Any], list, None]] = Field(None, description="Outputs of the run")
inputs: Optional[Union[str, Mapping[str, Any], list, None]] = Field(None, description="Inputs of the run")
outputs: Optional[Union[str, Mapping[str, Any], list, None]] = Field(None, description="Outputs of the run")
run_type: LangSmithRunType = Field(..., description="Type of the run")
start_time: Optional[datetime | str] = Field(None, description="Start time of the run")
end_time: Optional[datetime | str] = Field(None, description="End time of the run")

56
api/core/ops/langsmith_trace/langsmith_trace.py
View File

@@ -1,4 +1,3 @@
import json
import logging
import os
import uuid
@@ -7,7 +6,7 @@ from typing import Optional, cast
from langsmith import Client
from langsmith.schemas import RunBase
from sqlalchemy.orm import sessionmaker
from sqlalchemy.orm import Session, sessionmaker
from core.ops.base_trace_instance import BaseTraceInstance
from core.ops.entities.config_entity import LangSmithConfig
@@ -29,8 +28,10 @@ from core.ops.langsmith_trace.entities.langsmith_trace_entity import (
)
from core.ops.utils import filter_none_values, generate_dotted_order
from core.repositories import SQLAlchemyWorkflowNodeExecutionRepository
from core.workflow.entities.node_entities import NodeRunMetadataKey
from core.workflow.nodes.enums import NodeType
from extensions.ext_database import db
from models.model import EndUser, MessageFile
from models import Account, App, EndUser, MessageFile, WorkflowNodeExecutionTriggeredFrom
logger = logging.getLogger(__name__)
@@ -137,8 +138,29 @@ class LangSmithDataTrace(BaseTraceInstance):
# through workflow_run_id get all_nodes_execution using repository
session_factory = sessionmaker(bind=db.engine)
# Find the app's creator account
with Session(db.engine, expire_on_commit=False) as session:
# Get the app to find its creator
app_id = trace_info.metadata.get("app_id")
if not app_id:
raise ValueError("No app_id found in trace_info metadata")
app = session.query(App).filter(App.id == app_id).first()
if not app:
raise ValueError(f"App with id {app_id} not found")
if not app.created_by:
raise ValueError(f"App with id {app_id} has no creator (created_by is None)")
service_account = session.query(Account).filter(Account.id == app.created_by).first()
if not service_account:
raise ValueError(f"Creator account with id {app.created_by} not found for app {app_id}")
workflow_node_execution_repository = SQLAlchemyWorkflowNodeExecutionRepository(
session_factory=session_factory, tenant_id=trace_info.tenant_id, app_id=trace_info.metadata.get("app_id")
session_factory=session_factory,
user=service_account,
app_id=trace_info.metadata.get("app_id"),
triggered_from=WorkflowNodeExecutionTriggeredFrom.WORKFLOW_RUN,
)
# Get all executions for this workflow run
@@ -148,27 +170,23 @@ class LangSmithDataTrace(BaseTraceInstance):
for node_execution in workflow_node_executions:
node_execution_id = node_execution.id
tenant_id = node_execution.tenant_id
app_id = node_execution.app_id
tenant_id = trace_info.tenant_id # Use from trace_info instead
app_id = trace_info.metadata.get("app_id") # Use from trace_info instead
node_name = node_execution.title
node_type = node_execution.node_type
status = node_execution.status
if node_type == "llm":
inputs = (
json.loads(node_execution.process_data).get("prompts", {}) if node_execution.process_data else {}
)
if node_type == NodeType.LLM:
inputs = node_execution.process_data.get("prompts", {}) if node_execution.process_data else {}
else:
inputs = json.loads(node_execution.inputs) if node_execution.inputs else {}
outputs = json.loads(node_execution.outputs) if node_execution.outputs else {}
inputs = node_execution.inputs if node_execution.inputs else {}
outputs = node_execution.outputs if node_execution.outputs else {}
created_at = node_execution.created_at or datetime.now()
elapsed_time = node_execution.elapsed_time
finished_at = created_at + timedelta(seconds=elapsed_time)
execution_metadata = (
json.loads(node_execution.execution_metadata) if node_execution.execution_metadata else {}
)
node_total_tokens = execution_metadata.get("total_tokens", 0)
metadata = execution_metadata.copy()
execution_metadata = node_execution.metadata if node_execution.metadata else {}
node_total_tokens = execution_metadata.get(NodeRunMetadataKey.TOTAL_TOKENS) or 0
metadata = {str(key): value for key, value in execution_metadata.items()}
metadata.update(
{
"workflow_run_id": trace_info.workflow_run_id,
@@ -181,7 +199,7 @@ class LangSmithDataTrace(BaseTraceInstance):
}
)
process_data = json.loads(node_execution.process_data) if node_execution.process_data else {}
process_data = node_execution.process_data if node_execution.process_data else {}
if process_data and process_data.get("model_mode") == "chat":
run_type = LangSmithRunType.llm
@@ -191,7 +209,7 @@ class LangSmithDataTrace(BaseTraceInstance):
"ls_model_name": process_data.get("model_name", ""),
}
)
elif node_type == "knowledge-retrieval":
elif node_type == NodeType.KNOWLEDGE_RETRIEVAL:
run_type = LangSmithRunType.retriever
else:
run_type = LangSmithRunType.tool

54
api/core/ops/opik_trace/opik_trace.py
View File

@@ -1,4 +1,3 @@
import json
import logging
import os
import uuid
@@ -7,7 +6,7 @@ from typing import Optional, cast
from opik import Opik, Trace
from opik.id_helpers import uuid4_to_uuid7
from sqlalchemy.orm import sessionmaker
from sqlalchemy.orm import Session, sessionmaker
from core.ops.base_trace_instance import BaseTraceInstance
from core.ops.entities.config_entity import OpikConfig
@@ -23,8 +22,10 @@ from core.ops.entities.trace_entity import (
WorkflowTraceInfo,
)
from core.repositories import SQLAlchemyWorkflowNodeExecutionRepository
from core.workflow.entities.node_entities import NodeRunMetadataKey
from core.workflow.nodes.enums import NodeType
from extensions.ext_database import db
from models.model import EndUser, MessageFile
from models import Account, App, EndUser, MessageFile, WorkflowNodeExecutionTriggeredFrom
logger = logging.getLogger(__name__)
@@ -150,8 +151,29 @@ class OpikDataTrace(BaseTraceInstance):
# through workflow_run_id get all_nodes_execution using repository
session_factory = sessionmaker(bind=db.engine)
# Find the app's creator account
with Session(db.engine, expire_on_commit=False) as session:
# Get the app to find its creator
app_id = trace_info.metadata.get("app_id")
if not app_id:
raise ValueError("No app_id found in trace_info metadata")
app = session.query(App).filter(App.id == app_id).first()
if not app:
raise ValueError(f"App with id {app_id} not found")
if not app.created_by:
raise ValueError(f"App with id {app_id} has no creator (created_by is None)")
service_account = session.query(Account).filter(Account.id == app.created_by).first()
if not service_account:
raise ValueError(f"Creator account with id {app.created_by} not found for app {app_id}")
workflow_node_execution_repository = SQLAlchemyWorkflowNodeExecutionRepository(
session_factory=session_factory, tenant_id=trace_info.tenant_id, app_id=trace_info.metadata.get("app_id")
session_factory=session_factory,
user=service_account,
app_id=trace_info.metadata.get("app_id"),
triggered_from=WorkflowNodeExecutionTriggeredFrom.WORKFLOW_RUN,
)
# Get all executions for this workflow run
@@ -161,26 +183,22 @@ class OpikDataTrace(BaseTraceInstance):
for node_execution in workflow_node_executions:
node_execution_id = node_execution.id
tenant_id = node_execution.tenant_id
app_id = node_execution.app_id
tenant_id = trace_info.tenant_id # Use from trace_info instead
app_id = trace_info.metadata.get("app_id") # Use from trace_info instead
node_name = node_execution.title
node_type = node_execution.node_type
status = node_execution.status
if node_type == "llm":
inputs = (
json.loads(node_execution.process_data).get("prompts", {}) if node_execution.process_data else {}
)
if node_type == NodeType.LLM:
inputs = node_execution.process_data.get("prompts", {}) if node_execution.process_data else {}
else:
inputs = json.loads(node_execution.inputs) if node_execution.inputs else {}
outputs = json.loads(node_execution.outputs) if node_execution.outputs else {}
inputs = node_execution.inputs if node_execution.inputs else {}
outputs = node_execution.outputs if node_execution.outputs else {}
created_at = node_execution.created_at or datetime.now()
elapsed_time = node_execution.elapsed_time
finished_at = created_at + timedelta(seconds=elapsed_time)
execution_metadata = (
json.loads(node_execution.execution_metadata) if node_execution.execution_metadata else {}
)
metadata = execution_metadata.copy()
execution_metadata = node_execution.metadata if node_execution.metadata else {}
metadata = {str(k): v for k, v in execution_metadata.items()}
metadata.update(
{
"workflow_run_id": trace_info.workflow_run_id,
@@ -193,7 +211,7 @@ class OpikDataTrace(BaseTraceInstance):
}
)
process_data = json.loads(node_execution.process_data) if node_execution.process_data else {}
process_data = node_execution.process_data if node_execution.process_data else {}
provider = None
model = None
@@ -226,7 +244,7 @@ class OpikDataTrace(BaseTraceInstance):
parent_span_id = trace_info.workflow_app_log_id or trace_info.workflow_run_id
if not total_tokens:
total_tokens = execution_metadata.get("total_tokens", 0)
total_tokens = execution_metadata.get(NodeRunMetadataKey.TOTAL_TOKENS) or 0
span_data = {
"trace_id": opik_trace_id,

5
api/core/ops/weave_trace/entities/weave_trace_entity.py
View File

@@ -1,3 +1,4 @@
from collections.abc import Mapping
from typing import Any, Optional, Union
from pydantic import BaseModel, Field, field_validator
@@ -19,8 +20,8 @@ class WeaveMultiModel(BaseModel):
class WeaveTraceModel(WeaveTokenUsage, WeaveMultiModel):
id: str = Field(..., description="ID of the trace")
op: str = Field(..., description="Name of the operation")
inputs: Optional[Union[str, dict[str, Any], list, None]] = Field(None, description="Inputs of the trace")
outputs: Optional[Union[str, dict[str, Any], list, None]] = Field(None, description="Outputs of the trace")
inputs: Optional[Union[str, Mapping[str, Any], list, None]] = Field(None, description="Inputs of the trace")
outputs: Optional[Union[str, Mapping[str, Any], list, None]] = Field(None, description="Outputs of the trace")
attributes: Optional[Union[str, dict[str, Any], list, None]] = Field(
None, description="Metadata and attributes associated with trace"
)

89
api/core/ops/weave_trace/weave_trace.py
View File

@@ -1,4 +1,3 @@
import json
import logging
import os
import uuid
@@ -7,6 +6,7 @@ from typing import Any, Optional, cast
import wandb
import weave
from sqlalchemy.orm import Session, sessionmaker
from core.ops.base_trace_instance import BaseTraceInstance
from core.ops.entities.config_entity import WeaveConfig
@@ -22,9 +22,11 @@ from core.ops.entities.trace_entity import (
WorkflowTraceInfo,
)
from core.ops.weave_trace.entities.weave_trace_entity import WeaveTraceModel
from core.repositories import SQLAlchemyWorkflowNodeExecutionRepository
from core.workflow.entities.node_entities import NodeRunMetadataKey
from core.workflow.nodes.enums import NodeType
from extensions.ext_database import db
from models.model import EndUser, MessageFile
from models.workflow import WorkflowNodeExecution
from models import Account, App, EndUser, MessageFile, WorkflowNodeExecutionTriggeredFrom
logger = logging.getLogger(__name__)
@@ -128,58 +130,57 @@ class WeaveDataTrace(BaseTraceInstance):
self.start_call(workflow_run, parent_run_id=trace_info.message_id)
# through workflow_run_id get all_nodes_execution
workflow_nodes_execution_id_records = (
db.session.query(WorkflowNodeExecution.id)
.filter(WorkflowNodeExecution.workflow_run_id == trace_info.workflow_run_id)
.all()
# through workflow_run_id get all_nodes_execution using repository
session_factory = sessionmaker(bind=db.engine)
# Find the app's creator account
with Session(db.engine, expire_on_commit=False) as session:
# Get the app to find its creator
app_id = trace_info.metadata.get("app_id")
if not app_id:
raise ValueError("No app_id found in trace_info metadata")
app = session.query(App).filter(App.id == app_id).first()
if not app:
raise ValueError(f"App with id {app_id} not found")
if not app.created_by:
raise ValueError(f"App with id {app_id} has no creator (created_by is None)")
service_account = session.query(Account).filter(Account.id == app.created_by).first()
if not service_account:
raise ValueError(f"Creator account with id {app.created_by} not found for app {app_id}")
workflow_node_execution_repository = SQLAlchemyWorkflowNodeExecutionRepository(
session_factory=session_factory,
user=service_account,
app_id=trace_info.metadata.get("app_id"),
triggered_from=WorkflowNodeExecutionTriggeredFrom.WORKFLOW_RUN,
)
for node_execution_id_record in workflow_nodes_execution_id_records:
node_execution = (
db.session.query(
WorkflowNodeExecution.id,
WorkflowNodeExecution.tenant_id,
WorkflowNodeExecution.app_id,
WorkflowNodeExecution.title,
WorkflowNodeExecution.node_type,
WorkflowNodeExecution.status,
WorkflowNodeExecution.inputs,
WorkflowNodeExecution.outputs,
WorkflowNodeExecution.created_at,
WorkflowNodeExecution.elapsed_time,
WorkflowNodeExecution.process_data,
WorkflowNodeExecution.execution_metadata,
)
.filter(WorkflowNodeExecution.id == node_execution_id_record.id)
.first()
)
if not node_execution:
continue
# Get all executions for this workflow run
workflow_node_executions = workflow_node_execution_repository.get_by_workflow_run(
workflow_run_id=trace_info.workflow_run_id
)
for node_execution in workflow_node_executions:
node_execution_id = node_execution.id
tenant_id = node_execution.tenant_id
app_id = node_execution.app_id
tenant_id = trace_info.tenant_id # Use from trace_info instead
app_id = trace_info.metadata.get("app_id") # Use from trace_info instead
node_name = node_execution.title
node_type = node_execution.node_type
status = node_execution.status
if node_type == "llm":
inputs = (
json.loads(node_execution.process_data).get("prompts", {}) if node_execution.process_data else {}
)
if node_type == NodeType.LLM:
inputs = node_execution.process_data.get("prompts", {}) if node_execution.process_data else {}
else:
inputs = json.loads(node_execution.inputs) if node_execution.inputs else {}
outputs = json.loads(node_execution.outputs) if node_execution.outputs else {}
inputs = node_execution.inputs if node_execution.inputs else {}
outputs = node_execution.outputs if node_execution.outputs else {}
created_at = node_execution.created_at or datetime.now()
elapsed_time = node_execution.elapsed_time
finished_at = created_at + timedelta(seconds=elapsed_time)
execution_metadata = (
json.loads(node_execution.execution_metadata) if node_execution.execution_metadata else {}
)
node_total_tokens = execution_metadata.get("total_tokens", 0)
attributes = execution_metadata.copy()
execution_metadata = node_execution.metadata if node_execution.metadata else {}
node_total_tokens = execution_metadata.get(NodeRunMetadataKey.TOTAL_TOKENS) or 0
attributes = {str(k): v for k, v in execution_metadata.items()}
attributes.update(
{
"workflow_run_id": trace_info.workflow_run_id,
@@ -192,7 +193,7 @@ class WeaveDataTrace(BaseTraceInstance):
}
)
process_data = json.loads(node_execution.process_data) if node_execution.process_data else {}
process_data = node_execution.process_data if node_execution.process_data else {}
if process_data and process_data.get("model_mode") == "chat":
attributes.update(
{