feat: add escalation tool #303
Conversation
radu-mocanu
commented
Dec 3, 2025
radu-mocanu
commented
- add escalation tool
| output_schema = response_schema or END_EXECUTION_TOOL.args_schema | ||
| validated = output_schema.model_validate(args) | ||
| return validated.model_dump() | ||
| return _handle_end_execution(tool_call["args"], response_schema) |
There was a problem hiding this comment.
Should we make these into tools instead? That way, the LLM gets the tool definitions without additional change.
There was a problem hiding this comment.
I think it might be better to leave the terminate node as a 'special' tool for separation of concerns.
| from pydantic import BaseModel, Field | ||
|
|
||
|
|
||
| class AgentTerminationSource(StrEnum): |
There was a problem hiding this comment.
Need more information here. AFAIK escalations states are submit and reject. IMO the model should be the one deciding to terminate. Otherwise, any potential cleanup may get skipped.
There was a problem hiding this comment.
this is the way low code currently works. without this we won t be able to reach parity
| else None | ||
| ) | ||
| if outcome and outcome == EscalationAction.END: | ||
| return Command( |
There was a problem hiding this comment.
This feels like an anti-pattern. Tools should ideally not be aware of state.
There was a problem hiding this comment.
same as above. the escalations can be configured to terminate the execution on reject
There was a problem hiding this comment.
In that case, can we port over this logic to ToolNode?
There was a problem hiding this comment.
this is already being converted to a ToolNode
| ) | ||
| output_model: type[BaseModel] = jsonschema_to_pydantic(channel.output_schema) | ||
|
|
||
| # only works for UserEmail type as value is GUID for UserId or GroupId for other types |
There was a problem hiding this comment.
should be better to extend the api to accept recipient type and to update this: https://github.com/UiPath/uipath-python/blob/379b7b310dca36cd6bcf9889bdbc2c81a4d0966d/src/uipath/platform/action_center/_tasks_service.py#L125-L125 ?
3 participants
