Files
next-tjwater-server/app/api/v1/endpoints/operational_tasks.py
T

682 lines
22 KiB
Python

"""Project-scoped APIs for operational tasks, evidence and approved actions."""
from __future__ import annotations
from datetime import datetime
from typing import Literal
from uuid import UUID
from fastapi import APIRouter, Depends, Header, HTTPException, Query, status
from psycopg import AsyncConnection
from app.auth.permissions import (
OPERATIONAL_TASK_EXECUTE,
OPERATIONAL_TASK_EXTERNAL_SYNC,
OPERATIONAL_TASK_VIEW,
OPERATIONAL_TASK_WRITE,
SCADA_VIEW,
SIMULATION_VIEW,
permissions_for_context,
require_permission,
)
from app.auth.project_dependencies import (
ProjectContext,
get_project_context,
get_project_pg_connection,
)
from app.domain.domain_errors import DomainError
from app.domain.schemas.operational_data import TaskDataQuery, TaskDataSnapshot
from app.domain.schemas.operational_task import (
ActionDecisionCommand,
ActionExecuteCommand,
ActionProposalCommand,
ActionReconcileCommand,
WorkOrderDraftUpdateCommand,
WorkOrderConfirmDispatchCommand,
TaskAccessRequest,
TaskAccessRequestCreate,
TaskAccessRequestDecision,
AgentArtifactsCommand,
AgentArtifactsResult,
AgentRunCancelCommand,
AgentRunCommand,
AgentRunDispatchFailureCommand,
AgentRunFailureCommand,
AgentTaskContext,
DispatchDueSchedulesCommand,
DispatchDueSchedulesResult,
EvidenceRecord,
ExternalFeedbackCommand,
ExternalFeedbackResult,
ExternalSourceWithdrawalCommand,
HumanApprovalAction,
HumanEvidenceCommand,
NotificationAcknowledgeCommand,
OperationalTaskCreate,
OperationalTaskDetail,
OperationalTaskNotification,
OperationalTaskNotificationPage,
OperationalTaskPage,
OperationalTaskRetryCommand,
OperationalTaskRun,
OperationalTaskRunPage,
OperationalTaskSchedule,
OperationalTaskScheduleCreate,
OperationalTaskSchedulePage,
OperationalTaskScheduleUpdate,
RecoverExpiredLeasesCommand,
RecoverExpiredLeasesResult,
TaskStatus,
TaskDataQueryResult,
TaskType,
)
from app.infra.db.postgresql.operational_tasks import OperationalTaskRepository
from app.infra.integrations.operational_actions import (
OperationalActionDispatcher,
get_operational_action_dispatcher,
)
from app.infra.integrations.operational_data import TaskDataError, read_task_data
from app.services.operational_tasks import OperationalTaskService
router = APIRouter()
def _service(ctx: ProjectContext, conn: AsyncConnection) -> OperationalTaskService:
return OperationalTaskService(
OperationalTaskRepository(conn, ctx.project_id),
ctx,
_task_permissions(ctx),
)
def _task_permissions(ctx: ProjectContext) -> set[str]:
permissions = set(permissions_for_context(ctx))
if ctx.system_role == "service":
permissions.difference_update({SCADA_VIEW, SIMULATION_VIEW})
return permissions
def _raise_domain(exc: DomainError):
raise HTTPException(status_code=exc.status, detail=exc.payload()) from exc
@router.post(
"/operational-tasks",
response_model=OperationalTaskDetail,
status_code=status.HTTP_201_CREATED,
)
async def create_operational_task(
command: OperationalTaskCreate,
ctx: ProjectContext = Depends(require_permission(OPERATIONAL_TASK_WRITE)),
conn: AsyncConnection = Depends(get_project_pg_connection),
):
try:
return await _service(ctx, conn).create(command)
except DomainError as exc:
_raise_domain(exc)
@router.get("/operational-tasks", response_model=OperationalTaskPage)
async def list_operational_tasks(
task_type: TaskType | None = None,
task_status: TaskStatus | None = Query(default=None, alias="status"),
limit: int = Query(default=50, ge=1, le=100),
offset: int = Query(default=0, ge=0),
peripheral_work_order_id: UUID | None = Query(
default=None, description="Only tasks sourced from or linked to this peripheral work order",
),
ctx: ProjectContext = Depends(require_permission(OPERATIONAL_TASK_VIEW)),
conn: AsyncConnection = Depends(get_project_pg_connection),
):
try:
items, next_offset = await _service(ctx, conn).list(
task_type=task_type, status=task_status, limit=limit, offset=offset,
**({"peripheral_work_order_id": peripheral_work_order_id}
if peripheral_work_order_id is not None else {}),
)
return OperationalTaskPage(
items=items, limit=limit, offset=offset, next_offset=next_offset
)
except DomainError as exc:
_raise_domain(exc)
@router.get("/operational-task-runs", response_model=OperationalTaskRunPage)
async def list_operational_task_runs(
run_status: Literal[
"queued", "claimed", "running", "succeeded", "failed", "timed_out", "cancelled"
] | None = Query(default=None, alias="status"),
limit: int = Query(default=50, ge=1, le=100),
cursor: str | None = None,
ctx: ProjectContext = Depends(require_permission(OPERATIONAL_TASK_VIEW)),
conn: AsyncConnection = Depends(get_project_pg_connection),
):
cursor_created_at = None
cursor_run_id = None
if cursor:
try:
timestamp, run_id = cursor.rsplit("|", 1)
cursor_created_at = datetime.fromisoformat(timestamp)
cursor_run_id = UUID(run_id)
except (ValueError, TypeError) as exc:
raise HTTPException(status_code=400, detail="Invalid run cursor") from exc
try:
items, next_cursor = await _service(ctx, conn).list_runs(
status=run_status,
limit=limit,
cursor_created_at=cursor_created_at,
cursor_run_id=cursor_run_id,
)
return OperationalTaskRunPage(items=items, next_cursor=next_cursor)
except DomainError as exc:
_raise_domain(exc)
@router.post(
"/operational-tasks/schedules",
response_model=OperationalTaskSchedule,
status_code=status.HTTP_201_CREATED,
)
async def create_operational_schedule(
command: OperationalTaskScheduleCreate,
ctx: ProjectContext = Depends(require_permission(OPERATIONAL_TASK_WRITE)),
conn: AsyncConnection = Depends(get_project_pg_connection),
):
try:
return await _service(ctx, conn).create_schedule(command)
except DomainError as exc:
_raise_domain(exc)
@router.get(
"/operational-tasks/schedules", response_model=OperationalTaskSchedulePage
)
async def list_operational_schedules(
ctx: ProjectContext = Depends(require_permission(OPERATIONAL_TASK_VIEW)),
conn: AsyncConnection = Depends(get_project_pg_connection),
):
try:
return OperationalTaskSchedulePage(items=await _service(ctx, conn).list_schedules())
except DomainError as exc:
_raise_domain(exc)
@router.patch(
"/operational-tasks/schedules/{schedule_id}",
response_model=OperationalTaskSchedule,
)
async def update_operational_schedule(
schedule_id: UUID,
command: OperationalTaskScheduleUpdate,
ctx: ProjectContext = Depends(require_permission(OPERATIONAL_TASK_WRITE)),
conn: AsyncConnection = Depends(get_project_pg_connection),
):
try:
return await _service(ctx, conn).update_schedule(schedule_id, command)
except DomainError as exc:
_raise_domain(exc)
@router.post(
"/operational-tasks/schedules/dispatch-due",
response_model=DispatchDueSchedulesResult,
)
async def dispatch_due_operational_schedules(
command: DispatchDueSchedulesCommand,
ctx: ProjectContext = Depends(require_permission(OPERATIONAL_TASK_EXECUTE)),
conn: AsyncConnection = Depends(get_project_pg_connection),
):
try:
return await _service(ctx, conn).dispatch_due_schedules(command.limit)
except DomainError as exc:
_raise_domain(exc)
@router.post(
"/operational-tasks/leases/recover",
response_model=RecoverExpiredLeasesResult,
)
async def recover_operational_task_leases(
command: RecoverExpiredLeasesCommand,
ctx: ProjectContext = Depends(require_permission(OPERATIONAL_TASK_EXECUTE)),
conn: AsyncConnection = Depends(get_project_pg_connection),
):
try:
return await _service(ctx, conn).recover_expired_leases(command.limit)
except DomainError as exc:
_raise_domain(exc)
@router.get(
"/operational-task-notifications",
response_model=OperationalTaskNotificationPage,
)
async def list_operational_task_notifications(
unread_only: bool = False,
limit: int = Query(default=50, ge=1, le=100),
offset: int = Query(default=0, ge=0),
ctx: ProjectContext = Depends(require_permission(OPERATIONAL_TASK_VIEW)),
conn: AsyncConnection = Depends(get_project_pg_connection),
):
try:
items, next_offset = await _service(ctx, conn).list_notifications(
unread_only=unread_only, limit=limit, offset=offset
)
return OperationalTaskNotificationPage(
items=items, limit=limit, offset=offset, next_offset=next_offset
)
except DomainError as exc:
_raise_domain(exc)
@router.post(
"/operational-task-notifications/{notification_id}/acknowledge",
response_model=OperationalTaskNotification,
)
async def acknowledge_operational_task_notification(
notification_id: UUID,
command: NotificationAcknowledgeCommand,
ctx: ProjectContext = Depends(require_permission(OPERATIONAL_TASK_VIEW)),
conn: AsyncConnection = Depends(get_project_pg_connection),
):
try:
return await _service(ctx, conn).acknowledge_notification(notification_id, command)
except DomainError as exc:
_raise_domain(exc)
@router.get("/operational-tasks/{task_id}", response_model=OperationalTaskDetail)
async def get_operational_task(
task_id: UUID,
ctx: ProjectContext = Depends(require_permission(OPERATIONAL_TASK_VIEW)),
conn: AsyncConnection = Depends(get_project_pg_connection),
):
try:
return await _service(ctx, conn).detail(task_id)
except DomainError as exc:
_raise_domain(exc)
@router.post("/operational-tasks/{task_id}/retry", response_model=OperationalTaskDetail)
async def retry_operational_task(
task_id: UUID,
command: OperationalTaskRetryCommand,
ctx: ProjectContext = Depends(require_permission(OPERATIONAL_TASK_WRITE)),
conn: AsyncConnection = Depends(get_project_pg_connection),
):
try:
return await _service(ctx, conn).retry_task(task_id, command)
except DomainError as exc:
_raise_domain(exc)
@router.post(
"/operational-tasks/{task_id}/runs/{run_id}/claim",
response_model=OperationalTaskRun,
)
async def claim_operational_task_run(
task_id: UUID,
run_id: UUID,
command: AgentRunCommand,
ctx: ProjectContext = Depends(require_permission(OPERATIONAL_TASK_EXECUTE)),
conn: AsyncConnection = Depends(get_project_pg_connection),
):
try:
return await _service(ctx, conn).claim_run(task_id, run_id, command)
except DomainError as exc:
_raise_domain(exc)
@router.post(
"/operational-tasks/{task_id}/runs/{run_id}/heartbeat",
response_model=OperationalTaskRun,
)
async def heartbeat_operational_task_run(
task_id: UUID,
run_id: UUID,
command: AgentRunCommand,
ctx: ProjectContext = Depends(require_permission(OPERATIONAL_TASK_EXECUTE)),
conn: AsyncConnection = Depends(get_project_pg_connection),
):
try:
return await _service(ctx, conn).heartbeat_run(task_id, run_id, command)
except DomainError as exc:
_raise_domain(exc)
@router.post(
"/operational-tasks/{task_id}/runs/{run_id}/complete",
response_model=OperationalTaskDetail,
)
async def complete_operational_task_run(
task_id: UUID,
run_id: UUID,
command: AgentRunCommand,
ctx: ProjectContext = Depends(require_permission(OPERATIONAL_TASK_EXECUTE)),
conn: AsyncConnection = Depends(get_project_pg_connection),
):
try:
return await _service(ctx, conn).complete_run(task_id, run_id, command)
except DomainError as exc:
_raise_domain(exc)
@router.post(
"/operational-tasks/{task_id}/runs/{run_id}/fail",
response_model=OperationalTaskDetail,
)
async def fail_operational_task_run(
task_id: UUID,
run_id: UUID,
command: AgentRunFailureCommand,
ctx: ProjectContext = Depends(require_permission(OPERATIONAL_TASK_EXECUTE)),
conn: AsyncConnection = Depends(get_project_pg_connection),
):
try:
return await _service(ctx, conn).fail_run(task_id, run_id, command)
except DomainError as exc:
_raise_domain(exc)
@router.post(
"/operational-tasks/{task_id}/runs/{run_id}/cancel",
response_model=OperationalTaskDetail,
)
async def cancel_operational_task_run(
task_id: UUID,
run_id: UUID,
command: AgentRunCancelCommand,
ctx: ProjectContext = Depends(require_permission(OPERATIONAL_TASK_WRITE)),
conn: AsyncConnection = Depends(get_project_pg_connection),
):
try:
return await _service(ctx, conn).cancel_run(task_id, run_id, command)
except DomainError as exc:
_raise_domain(exc)
@router.post(
"/operational-tasks/{task_id}/runs/{run_id}/dispatch-failure",
response_model=OperationalTaskRun,
)
async def record_operational_task_run_dispatch_failure(
task_id: UUID,
run_id: UUID,
command: AgentRunDispatchFailureCommand,
ctx: ProjectContext = Depends(require_permission(OPERATIONAL_TASK_EXECUTE)),
conn: AsyncConnection = Depends(get_project_pg_connection),
):
try:
return await _service(ctx, conn).record_dispatch_failure(task_id, run_id, command)
except DomainError as exc:
_raise_domain(exc)
@router.get(
"/operational-tasks/{task_id}/agent-context", response_model=AgentTaskContext
)
async def get_operational_task_agent_context(
task_id: UUID,
ctx: ProjectContext = Depends(require_permission(OPERATIONAL_TASK_VIEW)),
conn: AsyncConnection = Depends(get_project_pg_connection),
):
try:
return await _service(ctx, conn).agent_context(task_id)
except DomainError as exc:
_raise_domain(exc)
@router.post(
"/operational-tasks/{task_id}/agent-artifacts",
response_model=AgentArtifactsResult,
)
async def publish_operational_task_artifacts(
task_id: UUID,
command: AgentArtifactsCommand,
ctx: ProjectContext = Depends(require_permission(OPERATIONAL_TASK_EXECUTE)),
conn: AsyncConnection = Depends(get_project_pg_connection),
):
try:
return await _service(ctx, conn).publish_artifacts(task_id, command)
except DomainError as exc:
_raise_domain(exc)
@router.post(
"/operational-tasks/{task_id}/data-query", response_model=TaskDataQueryResult
)
async def query_operational_task_data(
task_id: UUID,
query: TaskDataQuery,
authorization: str | None = Header(default=None, alias="Authorization"),
ctx: ProjectContext = Depends(require_permission(OPERATIONAL_TASK_EXECUTE)),
conn: AsyncConnection = Depends(get_project_pg_connection),
):
service = _service(ctx, conn)
try:
if not await service.can_read_source(task_id, query.source):
raise DomainError("PERMISSION_DENIED", "Task has no approved access to this data source", status=403)
source_permission = SCADA_VIEW if query.source.startswith("scada_") else SIMULATION_VIEW
snapshot = await read_task_data(
context=ctx, query=query, credential=authorization,
delegated=source_permission not in _task_permissions(ctx),
)
evidence, stored_snapshot = await service.record_data_snapshot(task_id, snapshot)
return TaskDataQueryResult(evidence=evidence, snapshot=stored_snapshot)
except TaskDataError as exc:
raise HTTPException(status_code=exc.status, detail=exc.detail) from exc
except DomainError as exc:
_raise_domain(exc)
@router.post(
"/operational-tasks/{task_id}/access-requests",
response_model=TaskAccessRequest,
status_code=status.HTTP_201_CREATED,
)
async def request_operational_task_access(
task_id: UUID,
command: TaskAccessRequestCreate,
ctx: ProjectContext = Depends(require_permission(OPERATIONAL_TASK_EXECUTE)),
conn: AsyncConnection = Depends(get_project_pg_connection),
):
try:
return await _service(ctx, conn).request_access(task_id, command)
except DomainError as exc:
_raise_domain(exc)
@router.post(
"/operational-tasks/{task_id}/access-requests/{access_request_id}/decision",
response_model=TaskAccessRequest,
)
async def decide_operational_task_access(
task_id: UUID,
access_request_id: UUID,
command: TaskAccessRequestDecision,
ctx: ProjectContext = Depends(require_permission(OPERATIONAL_TASK_VIEW)),
conn: AsyncConnection = Depends(get_project_pg_connection),
):
try:
return await _service(ctx, conn).decide_access(task_id, access_request_id, command)
except DomainError as exc:
_raise_domain(exc)
@router.post(
"/operational-tasks/{task_id}/human-evidence",
response_model=OperationalTaskDetail,
)
async def add_operational_task_human_evidence(
task_id: UUID,
command: HumanEvidenceCommand,
ctx: ProjectContext = Depends(require_permission(OPERATIONAL_TASK_WRITE)),
conn: AsyncConnection = Depends(get_project_pg_connection),
):
try:
return await _service(ctx, conn).add_human_evidence(task_id, command)
except DomainError as exc:
_raise_domain(exc)
@router.get(
"/operational-tasks/{task_id}/evidence/{evidence_id}/snapshot",
response_model=TaskDataSnapshot,
)
async def get_operational_task_data_snapshot(
task_id: UUID,
evidence_id: str,
ctx: ProjectContext = Depends(require_permission(OPERATIONAL_TASK_VIEW)),
conn: AsyncConnection = Depends(get_project_pg_connection),
):
try:
return await _service(ctx, conn).data_snapshot(task_id, evidence_id)
except DomainError as exc:
_raise_domain(exc)
@router.post(
"/operational-tasks/{task_id}/actions",
response_model=HumanApprovalAction,
status_code=status.HTTP_201_CREATED,
)
async def propose_operational_task_action(
task_id: UUID,
command: ActionProposalCommand,
ctx: ProjectContext = Depends(require_permission(OPERATIONAL_TASK_WRITE)),
conn: AsyncConnection = Depends(get_project_pg_connection),
):
try:
return await _service(ctx, conn).propose_action(task_id, command)
except DomainError as exc:
_raise_domain(exc)
@router.post(
"/operational-tasks/{task_id}/actions/{action_id}/decision",
response_model=HumanApprovalAction,
)
async def decide_operational_task_action(
task_id: UUID,
action_id: UUID,
command: ActionDecisionCommand,
ctx: ProjectContext = Depends(require_permission(OPERATIONAL_TASK_VIEW)),
conn: AsyncConnection = Depends(get_project_pg_connection),
):
try:
return await _service(ctx, conn).decide_action(task_id, action_id, command)
except DomainError as exc:
_raise_domain(exc)
@router.patch(
"/operational-tasks/{task_id}/actions/{action_id}/draft",
response_model=HumanApprovalAction,
)
async def update_operational_work_order_draft(
task_id: UUID,
action_id: UUID,
command: WorkOrderDraftUpdateCommand,
ctx: ProjectContext = Depends(require_permission(OPERATIONAL_TASK_VIEW)),
conn: AsyncConnection = Depends(get_project_pg_connection),
):
try:
return await _service(ctx, conn).update_work_order_draft(task_id, action_id, command)
except DomainError as exc:
_raise_domain(exc)
@router.post(
"/operational-tasks/{task_id}/actions/{action_id}/confirm-dispatch",
response_model=HumanApprovalAction,
)
async def confirm_operational_work_order_dispatch(
task_id: UUID,
action_id: UUID,
command: WorkOrderConfirmDispatchCommand,
authorization: str | None = Header(default=None, alias="Authorization"),
ctx: ProjectContext = Depends(require_permission(OPERATIONAL_TASK_VIEW)),
conn: AsyncConnection = Depends(get_project_pg_connection),
dispatcher: OperationalActionDispatcher = Depends(get_operational_action_dispatcher),
):
try:
return await _service(ctx, conn).confirm_work_order_dispatch(
task_id, action_id, command, dispatcher, authorization
)
except DomainError as exc:
_raise_domain(exc)
@router.post(
"/operational-tasks/{task_id}/actions/{action_id}/execute",
response_model=HumanApprovalAction,
)
async def execute_operational_task_action(
task_id: UUID,
action_id: UUID,
command: ActionExecuteCommand,
authorization: str | None = Header(default=None, alias="Authorization"),
ctx: ProjectContext = Depends(require_permission(OPERATIONAL_TASK_VIEW)),
conn: AsyncConnection = Depends(get_project_pg_connection),
dispatcher: OperationalActionDispatcher = Depends(get_operational_action_dispatcher),
):
try:
return await _service(ctx, conn).submit_action(
task_id, action_id, command, dispatcher, authorization
)
except DomainError as exc:
_raise_domain(exc)
@router.post(
"/operational-tasks/{task_id}/actions/{action_id}/reconcile",
response_model=HumanApprovalAction,
)
async def reconcile_operational_task_action(
task_id: UUID,
action_id: UUID,
command: ActionReconcileCommand,
authorization: str | None = Header(default=None, alias="Authorization"),
ctx: ProjectContext = Depends(require_permission(OPERATIONAL_TASK_VIEW)),
conn: AsyncConnection = Depends(get_project_pg_connection),
dispatcher: OperationalActionDispatcher = Depends(get_operational_action_dispatcher),
):
try:
return await _service(ctx, conn).reconcile_action(
task_id, action_id, command, dispatcher, authorization
)
except DomainError as exc:
_raise_domain(exc)
@router.post(
"/operational-tasks/{task_id}/external-feedback",
response_model=ExternalFeedbackResult,
)
async def synchronize_operational_task_feedback(
task_id: UUID,
command: ExternalFeedbackCommand,
ctx: ProjectContext = Depends(require_permission(OPERATIONAL_TASK_EXTERNAL_SYNC)),
conn: AsyncConnection = Depends(get_project_pg_connection),
):
try:
return await _service(ctx, conn).external_feedback(task_id, command)
except DomainError as exc:
_raise_domain(exc)
@router.post(
"/operational-tasks/{task_id}/withdraw-external-source",
response_model=OperationalTaskDetail,
)
async def withdraw_operational_task_external_source(
task_id: UUID,
command: ExternalSourceWithdrawalCommand,
ctx: ProjectContext = Depends(require_permission(OPERATIONAL_TASK_EXTERNAL_SYNC)),
conn: AsyncConnection = Depends(get_project_pg_connection),
):
try:
return await _service(ctx, conn).withdraw_external_source(task_id, command)
except DomainError as exc:
_raise_domain(exc)