Active incident

support-agent/refund is the likely broken agent

Refund ranker release pulled full CRM context into the prompt, then retries amplified cost and latency.

Inspect suspect span

Open failures

Across 3 agents

Retry waste

$0.181

83% of suspect run cost

P95 latency

15.0s

3x over guardrail

Next span

span_5

ContextLengthExceeded

Local sources

Demo mode uses seeded Paperclip-shaped data. Optional connectors do not block incident review.

TRACEIQ_DATA_MODE=demo

Paperclip

demo

Seeded Paperclip-shaped demo data is active.

Hermes

offline

Connector pending. Set HERMES_API_URL for future live runtime events.

Ironclaw

offline

Connector pending. Set IRONCLAW_API_URL for future policy signals.

TraceIQ ingest

demo

Local API is the normalization boundary for future connectors.

24h run health

Volume, errors, median cost, p95 latency

17:00

18:00

19:00

20:00

21:00

22:00

23:00

00:00

01:00

02:00

03:00

04:00

05:00

06:00

07:00

08:00

09:00

10:00

11:00

12:00

13:00

14:00

15:00

16:00

Incident strip

refund-ranker release

traceiq-refund-ranker-2026-05-23.1

support-agent/refund

94%

Post-release refund failures clustered around CRM context expansion

Limit CRM rows before refund ranking and retry only the failed section.

98 risk scoretr_i9j0k1l2

Start here: support-agent/refund

Refund ranker pulled the full CRM export after the release, causing context errors and retries.

Cost

$0.531

Latency

21.6s

Tokens

17,696

Evidence 1

Release traceiq-refund-ranker-2026-05-23.1 preceded the failure cluster

Evidence 2

CRM tool returned 847 rows

Evidence 3

Healthy comparison trace stayed under the token guardrail

Actionable findings

View all

Retry cost34m ago

One run spent $0.181 on retries after two failed LLM spans.

Next: span_5

Error spike1h ago

Context length errors repeated across report-generator.

Next: span_5

Latency2h ago

Final summarization exceeded the 5s latency guard.

Next: span_4

Runs ranked by suspicion

Open trace queue

Risk	Agent	Likely cause	Release	Delta	Cost	Latency	Next
99	support-agent/refund	ContextLengthExceeded: prompt exceeds 16,384 token limit at Generate Win/Loss Analysis	traceiq-refund-ranker-2026-05-23.1	+218% / +184%	$0.531	21.6s	span_7
94	report-generator	ContextLengthExceeded: notes payload exceeded model limit at Generate save-plan section	n/a	n/a / n/a	$0.247	11.4s	span_5
57	contract-summarizer	Latency exceeded guardrail around Generate final summary	n/a	n/a / n/a	$0.112	14.5s	span_4
49	onboarding-flow	ToolTimeoutError: profile API exceeded 5s timeout at Fetch profile	n/a	n/a / n/a	$0.007	5.9s	span_1
46	search-agent	Latency exceeded guardrail around Rank competitors	n/a	n/a / n/a	$0.089	10.0s	span_4