personal memory agent
convey/chat: surface talent_errored on watchdog timeout
Record a terminal talent_errored chat-stream event before removing timed-out spawned talents so result/<use_id> no longer stays stuck active after the watchdog fires. Also relax the exec-dispatch runtime test to assert routing contract instead of full prompt formatting, and add a regression test for timed-out talent result state.
+92
-16
+6
convey/chat.py
+6
convey/chat.py
+86
-16
tests/test_chat_runtime.py
+86
-16
tests/test_chat_runtime.py
···
386
386
assert sol_messages[-1]["requested_task"] == "research it"
387
387
assert spawned_events[-1]["name"] == "exec"
388
388
assert spawned_events[-1]["task"] == "research it"
389
-
assert spawn_calls == [
390
-
{
391
-
"prompt": "Task: research it\n\nContext hints:\n{'k': 'v'}\n\n"
392
-
"Location: app=sol path=/app/sol facet=work\n\n"
393
-
"Recent chat:\n**Sol**: I am looking into that.",
394
-
"name": "exec",
395
-
"provider": None,
396
-
"config": {
397
-
"app": "sol",
398
-
"path": "/app/sol",
399
-
"facet": "work",
400
-
"chat_parent_use_id": "1713625500000",
401
-
},
402
-
"use_id": spawned_events[-1]["use_id"],
403
-
}
404
-
]
389
+
assert len(spawn_calls) == 1
390
+
spawn_call = spawn_calls[0]
391
+
assert spawn_call["name"] == "exec"
392
+
assert spawn_call["use_id"] == spawned_events[-1]["use_id"]
393
+
assert spawn_call["config"] == {
394
+
"app": "sol",
395
+
"path": "/app/sol",
396
+
"facet": "work",
397
+
"chat_parent_use_id": "1713625500000",
398
+
}
399
+
assert "research it" in str(spawn_call["prompt"])
405
400
assert len(timers) == 2
406
401
assert timers[0].cancelled is True
407
402
with chat._state_lock:
···
514
509
assert chat._current_chat_use_id is None
515
510
assert chat._current_chat_state is None
516
511
assert "1713629000001" not in chat._watchdog_timers
512
+
513
+
514
+
def test_chat_watchdog_marks_timed_out_talent_result_as_errored(tmp_path, monkeypatch):
515
+
import convey.chat as chat
516
+
517
+
_setup_journal(tmp_path, monkeypatch)
518
+
_reset_chat_state(chat)
519
+
timers = _install_fake_timers(monkeypatch)
520
+
521
+
monkeypatch.setattr("convey.chat._emit_cortex_event", lambda *args, **kwargs: None)
522
+
monkeypatch.setattr("convey.chat._emit_error", lambda *args, **kwargs: None)
523
+
monkeypatch.setattr(
524
+
"convey.utils.spawn_agent", lambda *args, **kwargs: kwargs["use_id"]
525
+
)
526
+
527
+
with chat._state_lock:
528
+
logical_use_id = chat._reserve_use_id_locked()
529
+
talent_use_id = chat._reserve_use_id_locked()
530
+
531
+
append_chat_event(
532
+
"talent_spawned",
533
+
use_id=talent_use_id,
534
+
name="exec",
535
+
task="summarize",
536
+
started_at=int(talent_use_id),
537
+
)
538
+
539
+
with chat._state_lock:
540
+
chat._current_chat_use_id = logical_use_id
541
+
chat._current_chat_state = {
542
+
"raw_use_id": None,
543
+
"trigger": {"type": "owner_message", "message": "help"},
544
+
"location": {"app": "sol", "path": "/app/sol", "facet": "work"},
545
+
"retry_count": 0,
546
+
}
547
+
chat._active_talents[talent_use_id] = {
548
+
"chat_use_id": logical_use_id,
549
+
"target": "exec",
550
+
"task": "summarize",
551
+
"location": {"app": "sol", "path": "/app/sol", "facet": "work"},
552
+
}
553
+
554
+
chat._run_next_action(
555
+
{
556
+
"kind": "talent",
557
+
"logical_use_id": logical_use_id,
558
+
"target": "exec",
559
+
"use_id": talent_use_id,
560
+
"task": "summarize",
561
+
"context": {},
562
+
"location": {"app": "sol", "path": "/app/sol", "facet": "work"},
563
+
}
564
+
)
565
+
566
+
timers[-1].fire()
567
+
568
+
assert chat._read_result_state(talent_use_id) == {
569
+
"state": "errored",
570
+
"reason": "chat took too long — try again",
571
+
"task": "summarize",
572
+
}
573
+
parent_errors = [
574
+
event
575
+
for event in read_chat_events(chat._today_day())
576
+
if event["kind"] == "chat_error"
577
+
]
578
+
talent_errors = [
579
+
event
580
+
for event in read_chat_events(chat._today_day())
581
+
if event["kind"] == "talent_errored"
582
+
]
583
+
assert parent_errors[-1]["use_id"] == logical_use_id
584
+
assert parent_errors[-1]["reason"] == "chat took too long — try again"
585
+
assert talent_errors[-1]["use_id"] == talent_use_id
586
+
assert talent_errors[-1]["reason"] == "chat took too long — try again"
517
587
518
588
519
589
def test_cortex_finish_logs_warning_for_unrouteable_use_id(