hermes-agent/tests/gateway/test_agents_command_delegations.py
teknium1 b792bd0529 feat(delegation): structured stall metadata + live per-child status in /agents
Completes #51690 on top of the salvaged #60378 timeout metadata:

- async_delegation: terminal 'stalled' events now carry structured
  stall context (stalled_after_quiet_seconds, stall_threshold_seconds,
  stall_phase idle|in_tool, stall_grace_seconds) on both single and
  batch paths, persisted in the durable row so restart-restored events
  keep it. Mirrors the sync path's timeout_seconds/timed_out_after_
  seconds/timeout_phase from #60378.
- list_async_delegations(): exposes seconds_since_progress and live
  children_activity (per-child api_calls, current_tool,
  seconds_since_activity) sampled from the dispatch's progress_fn
  outside the records lock; private monitor bookkeeping and callables
  never leak.
- /agents (CLI + gateway): background delegations render per-child
  activity rows, quiet-time hints, and the stalling state; gateway
  section is new (previously async delegations were invisible there).
  New locale key gateway.agents.background_delegations in all 17
  catalogs.

Tests: stall-metadata event shape, live-listing projection, gateway
/agents rendering (real registry dispatch, sabotage-verified), sync
timeout metadata fields, non-timeout None contract.
2026-07-26 17:13:52 -07:00

119 lines
3.7 KiB
Python

"""Gateway /agents surfaces background delegations with live activity (#51690).
Drives the REAL GatewayRunner._handle_agents_command against a REAL
async-delegation registry dispatch (no mocked list function), so the test
covers the whole projection: registry record → list_async_delegations()
live sampling → /agents rendering.
"""
import threading
import time
import pytest
from tools import async_delegation as ad
from tools.process_registry import process_registry
@pytest.fixture(autouse=True)
def _clean_state():
ad._reset_for_tests()
while not process_registry.completion_queue.empty():
process_registry.completion_queue.get_nowait()
yield
deadline = time.monotonic() + 2.0
while ad.active_count() and time.monotonic() < deadline:
time.sleep(0.02)
ad._reset_for_tests()
while not process_registry.completion_queue.empty():
process_registry.completion_queue.get_nowait()
def _make_runner():
from gateway.run import GatewayRunner
runner = object.__new__(GatewayRunner)
runner._running_agents = {}
runner._running_agents_ts = {}
runner._background_tasks = set()
runner._session_key_for_source = lambda source: "agent:main:test:dm:1"
return runner
class _Event:
source = None
@pytest.mark.asyncio
async def test_agents_command_lists_background_delegation_with_activity():
gate = threading.Event()
base_ts = time.time() - 8.0
res = ad.dispatch_async_delegation(
goal="research the delegation stall monitor",
context=None, toolsets=None, role="leaf", model="m",
session_key="agent:main:test:dm:1", max_async_children=1,
runner=lambda: {} if gate.wait(timeout=10) else {},
progress_fn=lambda: (((2, "web_search", base_ts),), True),
)
assert res["status"] == "dispatched"
try:
runner = _make_runner()
out = await runner._handle_agents_command(_Event())
finally:
gate.set()
assert res["delegation_id"] in out
assert "running" in out
assert "research the delegation stall monitor" in out
# Live per-child activity sampled from progress_fn.
assert "2 api calls" in out
assert "web_search" in out
@pytest.mark.asyncio
async def test_agents_command_marks_stalling_delegation(monkeypatch):
monkeypatch.setattr(ad, "_STALE_CHECK_INTERVAL", 0.03)
monkeypatch.setattr(ad, "_STALE_IDLE_SECONDS", 0.1)
# Long grace so the record stays in 'stalling' while we render.
monkeypatch.setattr(ad, "_STALL_GRACE_SECONDS", 30.0)
gate = threading.Event()
res = ad.dispatch_async_delegation(
goal="wedged child", context=None, toolsets=None, role="leaf",
model="m", session_key="agent:main:test:dm:1", max_async_children=1,
runner=lambda: {} if gate.wait(timeout=10) else {},
progress_fn=lambda: ((0, None), False),
)
assert res["status"] == "dispatched"
try:
deadline = time.monotonic() + 5.0
while time.monotonic() < deadline:
items = ad.list_async_delegations()
if any(
d["delegation_id"] == res["delegation_id"]
and d.get("status") == "stalling"
for d in items
):
break
time.sleep(0.02)
else:
pytest.fail("delegation never reached stalling state")
runner = _make_runner()
out = await runner._handle_agents_command(_Event())
finally:
gate.set()
assert res["delegation_id"] in out
assert "stalling" in out
assert "no progress" in out
@pytest.mark.asyncio
async def test_agents_command_no_delegations_keeps_none_message():
runner = _make_runner()
out = await runner._handle_agents_command(_Event())
assert "No active agents" in out