3Sign up
Seeded demo for a fictional SaaS — nothing here is a real account. Sign up to see your own
helio-prodDashboardTraces

Traces (Agent Runs)

Browse every agent run — pick one to see its per-span composition, reroutes, and loop fingerprint.

View session

inbox-router

Run 2 of 2 LOOPEnforce
BlockedEnforceRule: Loop: same prompt ×4Model: gpt-4o

A later model call on this already-running trace was denied before it reached the provider. Earlier calls ran and were billed; only the blocked call never hit the provider.

4 repeats (limit 4)

Plan

free

Model calls

4

Tokens

153.6k→480

Total cost

$0.38880

Duration

18.40s

Usage by modality

Chat4 callsgpt-4o153,600→480TOKENS$0.38880

Models: gpt-4o

Cost vs workflow baseline

expensive outlier

This run $0.39 — expensive outlier

Among the most expensive 5% of the last 1,284 runs

$0.0041Typical run: $0.0094 – $0.03$0.07

Timeline

9 spans · 18.40s total

inbox-router
18.40s
gpt-4o
1.82s
fetch_conversations
240ms
gpt-4o
1.76s
fetch_conversations
210ms
gpt-4o
1.79s
fetch_conversations
198ms
gpt-4o
1.81s
gpt-4o
0ms

9 spans

Loop fingerprint sequence

HASH_CYCLEEnforcecycle length 1 · repeats 4

1

#a10fee

38.4k

—

2

#a10fee

38.4k

0

3

#a10fee

38.4k

0

4

#a10fee

38.4k

0

5

—

—

Highlighted: the repeating fingerprint the detector matched (4 calls × 1). Each chip is one model call the detector evaluated, in start-time order; matching hashes mean identical prompts. Structural agent, tool and chain rows are not shown — the detector never sees them. The dashed chip is the call the firewall denied before it reached the provider.