Omi 24h Health Report — Full Trace Edition

Generated: 2026-07-26T08:39:49.527356297Z | Window: 24h | Each log pattern includes up to 10 full entries for root cause tracing

CTO Triage Table

P Signal Tier Impact Code Entry Point
P1 Desktop-backend 5xx P99 90s timeout cascade user_facing Desktop users frozen for 90s Desktop-backend upstream call chain
P1 iOS crashes 3,676 (153/hr) critical 6.5x Android rate Crashlytics → http_pool_manager.dart
P1 Backend-integration truncated responses (408x) user_facing Request timeouts / app exit backend-integration Cloud Run timeout
P2 WebSocket auth: 627x Token expired + 24x cert errors user_facing Disconnected from transcription utils/other/endpoints.py:WebSocket auth
P2 Conversation finalization: 196x stale reconciliation user_facing Orphaned conversations services/conversation_finalization.py:135
P2 GPU capacity: 6/7 allocated, 1 remaining user_facing Transcription/diarization at risk if demand spikes GKE autoscaler, node pools

Service Health Matrix

Service Status Issues
desktop-backend CRITICAL No issues
backend-integration CRITICAL No issues
backend-sync DEGRADED No issues
pusher DEGRADED No issues
diarizer HEALTHY No issues
parakeet HEALTHY No issues
vad HEALTHY No issues
backend-listen HEALTHY No issues
agent-proxy HEALTHY No issues

Mobile Crashes (24h)

Platform Total Per Hour
iOS 3,676 153/hr
Android 563 23/hr

Per-Status Latency (Gap 1)

Service Status Count P50 P99 Analysis
desktop-backend2xx2,446379ms20.3sNormal
5xx6090.0s90.2sGateway timeout — 100% are 90s waits (4.4x vs 2xx)
backend2xx40,76745ms2.8sNormal
backend-sync2xx282776ms29.0sNormal
backend-integration2xx1,36211ms1.3sNormal

GPU Capacity (Gap 3)

MetricValueStatus
GPU Allocation6 allocated / 7 total (1 remaining)WARNING
VRAM (max)23.5%OK
Node (pool)VRAM %
vad-pool-v2-47a302d1-71u523.5%
diarizer-pool-037ee134-qt5h22.6%
parakeet-pool-c1f29adf-9lzh16.3%
parakeet-pool-c1f29adf-kwxz16.1%
vad-pool-v2-47a302d1-brjg10.1%
nllb-translation-poo-bf276826-g5f65.1%
vad-pool-v2-47a302d1-5jr20.0%

Top Error Endpoints

ServicePathCountTier
backend-integration/v1/mcp/sse35,114internal
backend-integration/v1/mcp/conversations?limit=330,346internal
backend-sync/v2/sync-local-files27,019user_facing
pusherhttps://audio.soundbrain.cc/webhook23,100known_noise
pusherhttps://omi-assist-api-prod.dakom.workers.dev/omi20,316known_noise
pusherhttps://omi.supermemory.ai/webhook10,548known_noise
backend-sync/v2/audio-merge-jobs/run6,732user_facing
backend-sync/v2/sync-jobs/run3,532user_facing
backend-sync/v2/sync-local-files2,426user_facing
backend-sync/v1/conversation-finalization-jobs/run2,113user_facing

Slowest Endpoints

ServicePathAvg LatencyCount
pusher/v1/trigger/listen17.5s48,313
parakeet/v2/transcribe3.5s91,823
parakeet/v3/stream12.0s3,873
parakeet/v1/transcribe296ms1,540

Abnormal Log Patterns — Full Trace (20 patterns, up to 10 logs each)

Each pattern shows up to 10 raw log entries with timestamps, pod names, and full text for code tracing. Click "Show logs" to expand.

#1 [CRITICAL] backend-listen — 627x in 24h
utils.other.endpoints
ERROR:utils.other.endpoints:WebSocket auth failed: code=4001 error=Token expired, <HEX> < <HEX>
Show 10 logs ▼
#2 [CRITICAL] backend-integration — 408x in 24h
Truncated response body. Usually implies that the request timed out or the application exited before the response was fi
Show 10 logs ▼
#3 [CRITICAL] backend-integration — 127x in 24h
services.conversation_finalization
ERROR:services.conversation_finalization:stale processing conversation reconciliation query failed
Show 10 logs ▼
#4 [CRITICAL] backend-integration — 125x in 24h
/app/services/conversation_finalization.py:135
Traceback (most recent call last):
Show 10 logs ▼
#5 [WARNING] backend-sync — 72x in 24h
/app/services/conversation_finalization.py:135
Traceback (most recent call last):
Show 10 logs ▼
#6 [WARNING] backend-sync — 69x in 24h
services.conversation_finalization
ERROR:services.conversation_finalization:stale processing conversation reconciliation query failed
Show 10 logs ▼
#7 [WARNING] pusher — 61x in 24h
utils.app_integrations
ERROR:utils.app_integrations:App 01K8FRZA3XV2T4N1VD9CMTW8HV is not proactive_notification or data invalid CInQQMaPJvZfaP
Show 10 logs ▼
#8 [WARNING] backend-listen — 24x in 24h
utils.other.endpoints
ERROR:utils.other.endpoints:WebSocket auth failed: code=4001 error=Certificate for key id <HEX> not found.
Show 10 logs ▼
#9 [INFO] backend-listen — 14x in 24h
utils.stt.streaming
ERROR:utils.stt.streaming:Parakeet WS recv error: no close frame received or sent
Show 10 logs ▼
#10 [INFO] crashlytics — 13x in 24h
uid=<HEX> platform=ios device=iPhone17,2 version=1.0.544 package
uid=<HEX> platform=ios device=iPhone17,2 version=1.0.544 package:omi/backend/http/http_pool_manager.dart - HttpPoolManag
Show 3 logs ▼
#11 [INFO] pusher — 12x in 24h
utils.webhooks
ERROR:utils.webhooks:Error sending audio bytes to developer webhook:
Show 10 logs ▼
#12 [INFO] pusher — 9x in 24h
utils.app_integrations
ERROR:utils.app_integrations:App supermemory-for-omi-01JE0442S4S20TRAHVYAXGQ2RQ is not proactive_notification or data in
Show 10 logs ▼
#13 [INFO] crashlytics — 8x in 24h
uid=<HEX> platform=ios device=iPhone18,2 version=1.0.544 package
uid=<HEX> platform=ios device=iPhone18,2 version=1.0.544 package:omi/backend/http/http_pool_manager.dart - HttpPoolManag
Show 3 logs ▼
#14 [INFO] pusher — 7x in 24h
utils.app_integrations
ERROR:utils.app_integrations:Plugin integration error:
Show 10 logs ▼
#15 [INFO] backend-sync — 6x in 24h
utils.llm.conversation_processing
ERROR:utils.llm.conversation_processing:Error determining memory discard: Invalid json output: discard = False
Show 10 logs ▼
#16 [INFO] backend-sync — 6x in 24h
utils.sync.pipeline
ERROR:utils.sync.pipeline:event=sync_transcription_job outcome=invalid_input status=failed provider=unknown model=unknow
Show 10 logs ▼
#17 [INFO] backend-listen — 5x in 24h
routers.listen.runtime
ERROR:routers.listen.runtime:Listen event delivery failed type=RuntimeError
Show 10 logs ▼
#18 [INFO] pusher — 5x in 24h
utils.app_integrations
ERROR:utils.app_integrations:App supermemory-for-omi-01JE0442S4S20TRAHVYAXGQ2RQ is not proactive_notification or data in
Show 10 logs ▼
#19 [INFO] backend-sync — 5x in 24h
utils.conversations.finalizer
ERROR:utils.conversations.finalizer:persisted conversation finalization failed uid=MVNZQHN42xeKJsXjfm2F5ZpK1sx1 conversa
Show 10 logs ▼
#20 [INFO] pusher — 5x in 24h
utils.app_integrations
ERROR:utils.app_integrations:App supermemory-for-omi-01JE0442S4S20TRAHVYAXGQ2RQ is not proactive_notification or data in
Show 10 logs ▼

Infrastructure Health

CheckStatusDetails
Grafana AlertsCRITICAL76 alert instances, 7 firing, 1 pending
Cloud Run 5xxWARNING8427 total 5xx across 10 services
Pod LifecycleCRITICAL3061 events (Unhealthy: 1803, Killing: 324, Started: 323)
AutoscalerWARNING1134 events (1134 FailedScheduling, 0 scale failures)
DeploysWARNING18 deploy runs, 2 failed
Node MaintOKno events
HPAOKno rescales

Signal Tier Classification (Gap 2)

TierWeightActive Signals
critical_user_facing3.0xGPU exhaustion, transcription pipeline
user_facing1.0xDesktop timeouts, sync 5xx, WebSocket auth, iOS crashes
internal0.3xMCP SSE 401s, Cloud Armor
known_noise0.1xPusher webhooks, firmware 404s

Detection Gap Coverage

GapStatusCoverage
1. Per-status latencyLIVEbeast --per-status + Grafana 5 panels + 2 alerts
2. Route tiersLIVEbeast --tier-filter + 19 route rules
3. GPU capacityLIVEbeast --gpu + Grafana 10 panels + 3 alerts

Sources: beast omi ops health-report --since 24h --drill-down | beast omi ops signal detect --per-status --gpu | Cloud Logging (GKE k8s_container) + BQ (Cloud Run stderr) for full trace logs
Patterns with full logs: 20/20 | Grafana dashboards: Per-Status-Code Latency + GPU Capacity & Health | 5 alert rules

- mon, by AI for @beastoin