Observability
Observability in Subbasis focuses on execution visibility, traceability, and review confidence.
What this means
Observability is the operational visibility layer for governed execution. It helps teams understand not only whether workflows run, but why outcomes happen.
What to observe
- Workflow progression
- Approval bottlenecks
- Retry and recovery patterns
- Connector and model usage
- Evidence completeness
What traces contain
Traces can include:
- workflow execution transitions
- agent task boundaries
- approval events
- connector actions and outcomes
- retries/failures/resumes
- model resource usage markers
- evidence linkage references
Trace ingestion and max trace size are plan/deployment dimensions.
How evaluation works
Evaluation workflows let teams inspect quality and policy alignment over time:
- compare expected vs actual outcomes
- inspect frequent failure patterns
- refine prompts/skills/policies
- validate approval and evidence behavior
How exports support review and BI
Bulk export and structured output paths support:
- audit-readiness preparation
- operational reviews
- external BI analysis
- incident retrospectives
Plan limits
Observability depth can vary by:
- trace ingestion limits
- max trace size
- retention/export scope
- evaluation workflow capacity
Related docs
Continue with Data export.