personal memory agent
Add startup grace period to supervisor health checks
Fixes false "offline" alerts on startup by adding an ever_received flag
to track whether we've received at least one observe.status event. Health
checks now return healthy (empty list) until the first status is received,
giving the observer time to start up before alerting.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
+20
-2
+11
-2
tests/test_supervisor.py
+11
-2
tests/test_supervisor.py
···
14
14
15
15
# Reset state for clean test
16
16
mod._observe_status_state["last_ts"] = 0.0
17
+
mod._observe_status_state["ever_received"] = False
17
18
mod._observe_status_state["activity_active"] = False
18
19
mod._observe_status_state["screencast_recording"] = False
19
20
mod._observe_status_state["files_growing"] = False
20
21
21
-
# No status received yet - both should be stale
22
+
# Startup grace period: no status ever received - returns healthy (no alerts)
23
+
stale = mod.check_health(threshold=60)
24
+
assert stale == [] # Grace period - don't alert until first status received
25
+
26
+
# After first status received, stale timestamp triggers alerts
27
+
mod._observe_status_state["ever_received"] = True
28
+
mod._observe_status_state["last_ts"] = 0.0 # Very old timestamp
22
29
stale = mod.check_health(threshold=60)
23
30
assert sorted(stale) == ["hear", "see"]
24
31
25
-
# Simulate receiving a status event (user inactive)
32
+
# Simulate receiving a fresh status event (user inactive)
26
33
mod._observe_status_state["last_ts"] = time.time()
27
34
mod._observe_status_state["activity_active"] = False
28
35
stale = mod.check_health(threshold=60)
···
57
64
58
65
# Reset state
59
66
mod._observe_status_state["last_ts"] = 0.0
67
+
mod._observe_status_state["ever_received"] = False
60
68
mod._observe_status_state["activity_active"] = False
61
69
mod._observe_status_state["screencast_recording"] = False
62
70
mod._observe_status_state["files_growing"] = False
···
71
79
mod._handle_observe_status(message)
72
80
73
81
assert mod._observe_status_state["last_ts"] > 0
82
+
assert mod._observe_status_state["ever_received"] is True # Grace period ended
74
83
assert mod._observe_status_state["activity_active"] is True
75
84
assert mod._observe_status_state["screencast_recording"] is True
76
85
assert mod._observe_status_state["files_growing"] is True
+9
think/supervisor.py
+9
think/supervisor.py
···
103
103
# Observe status state for health monitoring (updated from observe.status events)
104
104
_observe_status_state: dict = {
105
105
"last_ts": 0.0, # Timestamp of last observe.status event
106
+
"ever_received": False, # Whether we've received at least one status event
106
107
"activity_active": False, # Whether user is active (not idle/locked)
107
108
"screencast_recording": False, # Whether screencast is running
108
109
"files_growing": False, # Whether screencast files are growing
···
219
220
Health is derived from the last observe.status Callosum event:
220
221
- hear: Stale if no status received within threshold
221
222
- see: Stale if active AND (not recording OR files not growing)
223
+
224
+
During startup grace period (before first status event received),
225
+
returns empty list to avoid false alerts while observer is starting.
222
226
"""
227
+
# Grace period: don't alert until we've received at least one status event
228
+
if not _observe_status_state["ever_received"]:
229
+
return []
230
+
223
231
now = time.time()
224
232
stale: list[str] = []
225
233
···
1038
1046
1039
1047
# Update observe status state for health checking
1040
1048
_observe_status_state["last_ts"] = time.time()
1049
+
_observe_status_state["ever_received"] = True
1041
1050
1042
1051
activity = message.get("activity", {})
1043
1052
_observe_status_state["activity_active"] = activity.get("active", False)