[Gap]: Document console usage and event retention for both gateways #1243

What needs documentation?

Document retention for the usage metrics and event history backing the Connector Gateway and AI Gateway console views, including defaults, supported configuration, and cleanup behavior.

Cover:

• Both gateways default to four weeks of event retention, configurable from 1–12 weeks.
• Retention uses weekly UTC partitions: four complete previous weeks plus the current week, with hourly cleanup by default, rather than an exact rolling 28-day deletion cutoff.
• AI Gateway usage/spend retention is configured through enterprise-manager.appConfig.budgets.retentionWeeks.
• Connector Gateway tool-call usage/activity retention currently uses the Enterprise Manager environment variable ENTERPRISE_METERING_PARTMAN_RETENTION_WEEKS, supplied through enterprise-manager.extraEnv when using Helm. Document a dedicated Helm setting when it becomes available in a released version.
• Budget totals and dimension counters survive event partition cleanup; explain which views remain available and which depend on retained events.
• Captured AI conversation text, when enabled, has independent retention: seven days after the last turn by default, configurable through appConfig.budgets.captureInactivityWindow, capped at thirty days.
• External audit-log retention is controlled by the operator's log pipeline/SIEM and is separate from console-data retention.

Explain that selecting a three-month chart window does not extend stored history. The Connector Gateway Tool Usage page currently describes a three-month trend without explaining the four-week storage default.

Context and references

The relevant documentation includes the Connector Gateway Tool Usage and connector Activity guidance, the AI Gateway budget guides, and the Enterprise Manager Helm reference. Both gateways already have audit-log forwarding guides that describe externally managed retention, but those guides do not cover console-data storage.

Use case

As a platform operator, I need to know how much usage and activity history the console keeps, how to configure that window, and whether older aggregate totals remain available. This helps me size database storage, interpret incomplete chart history, and decide whether I need an external audit archive for longer retention.

Additional context

Documentation coverage checked on October 15, 2026. The docs contain references to retention clamping in generated API descriptions and generic appConfig/extraEnv Helm guidance, but no operator-facing explanation of these defaults and settings.

Document the supported configuration for the released platform version. Document a dedicated Connector Gateway retention Helm setting when available; existing retention behavior can be documented independently.