Self-improving AI Employees for everyone
Python ai-agentsai-employeeai-employeesassistantautonomous-agentsco-employeescompanygrok-botinstinctloop-agentsmuseown-your-datapersonalself-hosted
"""Metadata-only Canvas lifecycle event.
A UI lifecycle notification, not a workflow trigger. No node type registers a
canary consumer for ``com.opencompany.canvas.*``, so routing it through
``services.events.dispatch.emit`` would run a Temporal Visibility
``ListWorkflowExecutions`` query that is guaranteed to match nothing — once
per ``canvas`` tool call. It is broadcast straight to connected WebSocket
clients instead, the canonical plugin pattern (see ``nodes/context/_events.py``).
The payload is identity + revision only. It carries no item content, because
the broadcast fans out to every connected socket; panels refetch through the
authorized ``canvas_list`` handler, which is where ownership is enforced.
"""
from __future__ import annotations
from typing import Optional
from services.events.envelope import WorkflowEvent
_WIRE_KEY = "canvas_updated"
def canvas_updated(
*,
workflow_id: Optional[str],
node_id: str,
revision: int,
) -> WorkflowEvent:
return WorkflowEvent(
source="opencompany://nodes/canvas",
type="com.opencompany.canvas.updated",
subject=node_id,
workflow_id=workflow_id,
data={
"workflow_id": workflow_id,
"node_id": node_id,
"revision": revision,
},
)
async def dispatch_canvas_updated(
*,
workflow_id: Optional[str],
node_id: str,
revision: int,
) -> None:
from services.status_broadcaster import get_status_broadcaster
event = canvas_updated(
workflow_id=workflow_id, node_id=node_id, revision=revision
)
await get_status_broadcaster().broadcast(
{
"type": _WIRE_KEY,
"data": event.model_dump(mode="json", exclude_none=True),
}
)
__all__ = ["canvas_updated", "dispatch_canvas_updated"]