[ system / observability ]· synthetic demo data
Execution trace.
Metadata, not reasoning.
What the backend records for a run: mode, health, latency, calls, retrieval usage, governance events and termination status. Prompts, chain-of-thought, raw documents, embeddings, tool arguments and secrets are never shown.
[ run ]
Execution modeSingle agent
Execution healthDemo telemetry not recorded
Execution latency4.86 s
Human-wait latencywaiting…
Termination statuspending_review
Model calls2
Tool calls9
Provider-reported tokens4,990
Vector / graph evidence4 / 1
RetriesDemo telemetry not recorded
TimeoutsDemo telemetry not recorded
CostUnavailable · no token pricing configured
[ reliability configuration · repository-verified ]
Model timeout30 s
Embedding timeout30 s
Graph extraction timeout30 s
Tool timeout10 s
Provider maximum retries2
Provider retry base delay250 ms
Maximum human revisions2
Configuration limits, not per-run telemetry. Run values above are synthetic demo telemetry for the shared demo run.
[ execution nodes · latency waterfall ]
Context420 ms
Retrieval880 ms
Strategy140 ms
Analysis1620 ms
Risk760 ms
Governance220 ms
Evaluation640 ms
Recommendation180 ms
[ governance events ]
- risk_gateRisk gate triggeredCTRL-017 four-eyes · Customer-outcome impact · Escalation influence · policy CTRL-017
- interruptExecution paused → pending_reviewLangGraph interrupt · evaluation and release wait for a human decision
[ evaluation ]
Evaluation runs after human approval.