Description
Add useful workload and operational context around the communications assistants while retaining direct navigation to Agent Observability for conversation internals.
Acceptance Criteria
- #1 Workload tab shows real scenario route language and activity distributions.
- #2 Execution health tab shows supported operational metrics without fabricated results or empty panels.
- #3 Dashboard queries and all tabs are verified live and source is saved.
Definition of Done
- #1 just check
Implementation Plan
Add Overview/Request mix/Execution health tabs using verified existing metric histograms and workflow logs. Deduplicate run/conversation ids to handle terminal-log replay. Use normal operational dashboard language per user correction, preserving provenance in raw records and runbook. Query all panels, render each tab and save private source.
Implementation Notes
User correction incorporated: normal operational wording and no synthetic/real-Bedrock presentation commentary in dashboard; provenance retained in source/raw records. Built Request mix6panels and Execution health6panels around original Overview. Verified4products/3languages/9channel-formats/7routes and P95histograms. Terminal-log replay discovered; batch counter now counts distinct run_id (3), workflow mix uses distinct conversation_id. All18data queries return populated results, including actual0execution failures and1output-cap hit.
One temporary workflow aggregation expression failed parsing during verification; corrected before final deployment and all18data queries now pass with results. Request mix and Execution health screenshots show populated panels; output-limit count1 is amber and true execution failure0 stays green. Removed provenance/implementation commentary from visible titles and intro per user correction; raw records remain unchanged. Duplicate terminal-log ingestion recorded separately; run/workflow counts use distinct ids.
Final render confirms Overview normal wording and populated corrected workflow query; Request mix and Execution health show all intended data.19panels referenced exactly once with unique ids. Added replay-safe distinct counts. CodeRabbit skipped for declarative dashboard JSON per policy; actual query and render validation are proof. Private just check running as repository gate.
Final Summary
Added3tabs with19panels: original Overview plus Request mix(activity,routes,stages,products,languages,channels) and Execution health(completed workflows,actual failures,output caps,unique processing runs,P95duration,stop reasons). All18data queries return real results;3tabs rendered/inspected and independent review found no material issues. Fixed temporary query syntax issue before final proof; unique-id aggregation prevents replayed terminal logs inflating workflow/run counts. User presentation correction applied: normal operational wording, no synthetic/real-Bedrock commentary in visible dashboard; raw records/provenance unchanged. Private just check passed;CodeRabbit skipped declarative dashboard config. Saved private source commita385e97. Collector replay cause tracked in AIO-0015 (deduplicate replayed terminal records from completed Jobs).