Mastra Traces Gain Outcomes, Stop Reasons

This moves AI observability beyond binary success-or-error reporting toward actionable execution diagnostics.

• –Separates failures, cancellations, limits, guardrail blocks, and interruptions
• –Enables trace and span queries grouped by outcome or stop reason
• –Custom values let teams encode domain-specific agent behavior
• –Stop-reason context adds compact details such as max-step limits
• –Capability checks preserve compatibility with older storage adapters