Add distinct AI agent sessions Service Ping metrics

What does this MR do and why?

Work item 607935 asks for Service Ping metrics covering AI audit event activity, including:

• Number of distinct AI agent sessions audited

This MR adds that count:

Metric	Time frames	How
counts.count_distinct_ai_agent_sessions_with_audit_events	28d, all	batched COUNT(DISTINCT workflow_id) on ai_audit_events in PostgreSQL
counts.count_distinct_ai_agent_sessions_with_audit_events_clickhouse	28d, all	uniqExact(workflow_id) in ClickHouse

An AI agent session is a Ai::DuoWorkflows::Workflow; every AI audit event carries its workflow_id, so distinct workflow ids give the number of sessions that actually produced audit activity — which event totals alone cannot tell apart from a few very chatty sessions. The PostgreSQL and ClickHouse pair follows the merged count_ai_audit_events / count_ai_audit_events_clickhouse precedent: instances with ClickHouse analytics store AI audit events only there, everyone else uses the PostgreSQL fallback table.

References

• https://gitlab.com/gitlab-org/gitlab/-/work_items/607935

Database review

The PostgreSQL metric batches with distinct_count (batch size 10,000) over unfiltered MIN/MAX(workflow_id) bounds.The production ai_audit_events table is empty (GitLab.com writes AI audit events to ClickHouse; the PostgreSQL table serves instances without ClickHouse analytics), so the Database Lab clone was seeded with 200,000 events for the plans (session).

-- Batch bounds
SELECT MIN("ai_audit_events"."workflow_id") FROM "ai_audit_events";
SELECT MAX("ai_audit_events"."workflow_id") FROM "ai_audit_events";

-- Per-batch, all time frame
SELECT COUNT(DISTINCT "ai_audit_events"."workflow_id") FROM "ai_audit_events"
  WHERE "ai_audit_events"."workflow_id" >= 1 AND "ai_audit_events"."workflow_id" < 10001;

-- Per-batch, 28d time frame, batch overlapping the window
SELECT COUNT(DISTINCT "ai_audit_events"."workflow_id") FROM "ai_audit_events"
  WHERE "ai_audit_events"."created_at" BETWEEN '2026-07-20 00:00:00' AND '2026-08-17 00:00:00'
    AND "ai_audit_events"."workflow_id" >= 10001 AND "ai_audit_events"."workflow_id" < 20002;

-- Per-batch, 28d time frame, historical batch fully outside the window
SELECT COUNT(DISTINCT "ai_audit_events"."workflow_id") FROM "ai_audit_events"
  WHERE "ai_audit_events"."created_at" BETWEEN '2026-07-20 00:00:00' AND '2026-08-17 00:00:00'
    AND "ai_audit_events"."workflow_id" >= 1 AND "ai_audit_events"."workflow_id" < 10001;

Query	Plan	Execution	Buffers	Rows aggregated
MIN(workflow_id)	158592	0.46 ms	19 hits	1
MAX(workflow_id)	158593	0.45 ms	26 hits	1
Batch, all time	158594	28.0 ms	525 hits (4.1 MiB)	99,999
Batch, 28d, overlapping	158595	17.2 ms	443 hits (3.5 MiB)	51,852
Batch, 28d, outside window	158597	0.17 ms	10 hits	0

Observations:

• Every query resolves through index-only scans on the per-partition (workflow_id, created_at DESC, id DESC) indexes with zero heap fetches and no seq scans over populated partitions.

How to set up and validate locally

require_relative 'spec/support/helpers/service_ping_helpers.rb'
ServicePingHelpers.get_current_usage_metric_value('counts.count_distinct_ai_agent_sessions_with_audit_events')

MR acceptance checklist

Evaluate this MR against the MR acceptance checklist. It helps you analyze changes to reduce risks in quality, performance, reliability, security, and maintainability.