LLM-Driven Extraction of Unstructured Data — Built for API Deployments & ETL Pipeline Workflows
Python ai-agentsdata-engineeringdocument-aigenerative-aiidpjson-extractionllmmcp-serverocrpdf-extractionprompt-engineeringstructured-output
from rest_framework import serializers
from workflow_manager.file_execution.models import (
WorkflowFileExecution as FileExecution,
)
from workflow_manager.workflow_v2.enums import ExecutionStatus
from .models import WorkflowFileExecution
class WorkflowFileExecutionSerializer(serializers.ModelSerializer):
class Meta:
model = WorkflowFileExecution
fields = "__all__"
class FileCentricExecutionSerializer(serializers.ModelSerializer):
INIT_STATUS_MSG = "Waiting for a worker to pick up file's execution..."
DEFAULT_STATUS_MSG = (
"No status message available yet, please check again after a few minutes."
)
status_msg = serializers.SerializerMethodField()
file_size = serializers.ReadOnlyField(source="pretty_file_size")
execution_time = serializers.ReadOnlyField(source="pretty_execution_time")
class Meta:
model = FileExecution
exclude = ["file_hash"]
def get_status_msg(self, obj: FileExecution) -> dict[str, any] | None:
if obj.status in [ExecutionStatus.PENDING.value]:
return self.INIT_STATUS_MSG
elif obj.status == ExecutionStatus.ERROR.value:
return obj.execution_error
# Use pre-annotated field from viewset queryset to avoid N+1 queries
if hasattr(obj, "latest_log_data") and obj.latest_log_data:
return obj.latest_log_data.get("log", self.DEFAULT_STATUS_MSG)
return self.DEFAULT_STATUS_MSG