Machine-readable companion: /docs-for-robots

Arc/docs

Observability

Logs, Analytics, And Observability

Arc stores request-level metadata, trace-level rollups, and diagnostic heuristics so you can reason about cost, latency, and route behavior with more context than a raw provider dashboard.

Request Logs

timestamp

datetime

When Arc recorded the request.

route / model / provider

metadata

Which policy object and upstream model handled the call.

tokens / cost / latency

metrics

Direct cost and performance indicators for the request.

status code

int

Arc or provider result code.

complexity breakdown

json

Optional heuristic explanation for smart-tier style routing decisions.

latency breakdown

json

Optional stage timings inside the proxy path for newer requests.

Workflow Traces

When workflow mode is active, Arc groups requests into traces and spans. That lets you inspect one agent run as a unit instead of manually correlating many request rows.

Operational Signals

Arc surfaces complexity heuristics, route rollups, cost charts, latency trends, and traffic anomalies. The strongest operational use cases are model cost control, fallback gaps, and latency regression detection.