Platform
Observability
Monitor workflow health with run history, logs, and metrics designed for debugging and performance tuning.
Run insights
Run timelines
Step-by-step execution history with inputs and outputs.
Error tracing
Stack traces, retries, and failure snapshots per block.
Cost and usage
Token, provider cost, and time spent per workflow.
Quality signals
Human ratings, eval scores, and regression tracking.
Metrics to monitor
Latency p50/p95 per block and per workflow.
Failure rate by dependency and provider.
Token usage, cache hit ratio, and cost budgets.
Queue depth, concurrency saturation, and timeouts.
SLO definitions
Set success, latency, and cost budgets per workflow.
Alert routing
Send alerts to Slack, PagerDuty, or email escalation paths.
Release monitoring
Track early-run quality for new deployments.
Debugging toolkit
Replay executions with the same inputs.
Compare outputs across versions or providers.
Inspect intermediate state and block payloads.
Export logs and evidence for incident review.
Optimization loop
- Identify slow or costly blocks using run history.
- Enable caching or parallelization where safe.
- Tune prompts, providers, and tool timeouts.
- Deploy changes and watch metrics for improvement.