Self-improving AI Employees for everyone
Python ai-agentsai-employeeai-employeesassistantautonomous-agentsco-employeescompanygrok-botinstinctloop-agentsmuseown-your-datapersonalself-hosted
"""Admitted recipe metadata, intrinsic submission and terminal lifecycle."""
from contextlib import asynccontextmanager
from copy import deepcopy
from types import SimpleNamespace
from unittest.mock import AsyncMock
import asyncio
import pytest
from services.employees.team_runtime import employee_runtime_plan, admit_employee_runtime_job
from services.temporal.agent_workflow import AgentWorkflow, _trusted_tool_scope
from nodes.tool.task_manager import _execute_task_manager
def admitted():
plan = {"version": 2, "lead_node_id": "lead", "talk_node_id": "talk", "members": [{"node_id": "child"}],
"delivery_node_ids": ["gate", "send"], "talk_delivery_node_ids": ["chat"]}
nodes = [{"id": name, "type": kind} for name, kind in (("lead", "ai_employee"), ("talk", "chatAgent"), ("child", "aiAgent"), ("gate", "approvalGate"), ("send", "telegramSend"), ("chat", "chatReply"))]
nodes[0]["data"] = {"employee_recipe_version": 2, "employee_team_plan": deepcopy(plan)}
return {"workflow_id": "wf", "node_id": "lead", "execution_id": "origin", "nodes": nodes, "edges": [],
"generation": 4, "data_scope_id": "original-scope", "user_id": "owner", "parameter_snapshot": {"send": {"recipient": "original"}},
"inputs": {"trigger": {"recipient": "original"}}}, plan
def generation(context):
"""A database holding the generation's admitted graph and nothing saved."""
control = SimpleNamespace(graph_snapshot={"nodes": context["nodes"], "edges": context["edges"]})
return SimpleNamespace(get_workflow_control_generation=AsyncMock(return_value=control),
get_session=AsyncMock(side_effect=AssertionError("must not read saved metadata")))
async def test_plan_is_from_admitted_graph_even_after_saved_rollback():
context, original = admitted()
database = generation(context)
# A run holds only what its trigger reaches; the plan names nodes beyond it.
run = {**context, "nodes": [node for node in context["nodes"] if node["id"] == "lead"]}
assert await employee_runtime_plan(database, run) == original
database.get_workflow_control_generation.assert_awaited_with("wf", 4)
context["nodes"][0]["data"]["employee_team_plan"]["delivery_node_ids"].append("unactivated")
with pytest.raises(ValueError, match="outside its admitted graph"):
await employee_runtime_plan(database, run)
async def test_admission_preserves_original_outputs_and_review_reuses_job(monkeypatch):
context, plan = admitted()
create = AsyncMock(return_value=SimpleNamespace(id="job"))
monkeypatch.setattr("services.employees.jobs.create_job", create)
await admit_employee_runtime_job(object(), context, plan, "Owner mission")
assert context["employee_job_id"] == "job"
assert create.await_args.args[1].outputs == {"trigger": {"recipient": "original"}}
assert create.await_args.kwargs["dispatch"] is False
@asynccontextmanager
async def session():
yield SimpleNamespace(get=AsyncMock(return_value=SimpleNamespace(workflow_id="wf", team_id="team")),
execute=AsyncMock(return_value=SimpleNamespace(scalar_one_or_none=lambda: SimpleNamespace(id="job", workflow_id="wf", lead_node_id="lead", team_id=None, source={}))))
review = {**context, "execution_id": "later-review"}
review.pop("employee_job_id")
await admit_employee_runtime_job(SimpleNamespace(get_session=session), review, plan, "Review submitted work", {"team_id": "team", "task_id": "task"})
assert review["employee_job_id"] == "job"
create.assert_awaited_once()
async def test_intrinsic_talk_submission_ignores_model_destinations(monkeypatch):
context, plan = admitted()
scope = {**context, "parent_node_id": "talk", "node_id": "builtin_task_manager_talk"}
create = AsyncMock(return_value=SimpleNamespace(id="job", state="queued"))
dispatch = AsyncMock()
monkeypatch.setattr("services.plugin.deps.get_database", lambda: generation(context))
monkeypatch.setattr("services.agent_team.get_agent_team_service", lambda: object())
monkeypatch.setattr("services.employees.jobs.create_job", create)
monkeypatch.setattr("services.employees.jobs.dispatch_job", dispatch)
result = await _execute_task_manager({"operation": "submit_job", "mission": "Do research", "lead_node_id": "forged", "delivery_node_ids": ["forged"]}, scope)
assert result["job_id"] == "job"
assert create.await_args.kwargs["lead_node_id"] == "lead"
assert create.await_args.kwargs["delivery_node_ids"] == ["chat"]
assert create.await_args.args[1].raw["generation"] == 4
assert create.await_args.args[1].raw["data_scope_id"] == "original-scope"
def test_trusted_tool_scope_carries_managed_job_and_original_generation():
context, _ = admitted()
context["employee_job_id"] = "server-job"
scope = _trusted_tool_scope(context)
for key in ("generation", "data_scope_id", "user_id", "parameter_snapshot", "employee_job_id"):
assert scope[key] == context[key]
assert scope["outputs"] == context["inputs"]
@pytest.mark.parametrize("mode", ["failed", "cancelled", "acknowledged"])
async def test_runtime_terminal_cleanup_preserves_live_delegated_acknowledgement(monkeypatch, mode):
context, _ = admitted()
context["employee_job_id"] = "job"
agent = AgentWorkflow()
if mode == "cancelled":
agent._run_impl = AsyncMock(side_effect=asyncio.CancelledError())
else:
agent._run_impl = AsyncMock(return_value={"success": mode == "acknowledged"})
execute = AsyncMock(return_value={"saved": True})
monkeypatch.setattr("services.temporal.agent_workflow.workflow.patched", lambda _: True)
monkeypatch.setattr("services.temporal.agent_workflow.workflow.execute_activity", execute)
if mode == "cancelled":
with pytest.raises(asyncio.CancelledError):
await agent.run(context)
else:
await agent.run(context)
if mode == "acknowledged":
execute.assert_not_awaited()
else:
assert execute.await_args.args[0] == "employee.job.failed"
assert execute.await_args.args[1]["cancelled"] == (mode == "cancelled")