v0.8.6
Maintenance release: audit-log retention is now actually enforced, three previously-inert config knobs work, and duplicate/vestigial API endpoints are removed.
Highlights
- Audit-log retention is now enforced — a background sweep prunes
audit_logsrows pastcore.api.config.audit.retention_days(default 90) in batches, electing a single pruner across replicas. Previously the retention window was advisory and expired rows accumulated forever. Setretention_days: 0to keep audit rows indefinitely. audit.api_audit_enabledis now settable via Helm — the key was fully wired in Python but unreachable fromvalues.yaml/ the rendered ConfigMap.
Bug Fixes
- Make audit retention,
api_audit_enabled, and the proxy HPA memory knob real (#207)
Breaking Changes
Three unused, duplicate/vestigial /api/v1 endpoints were removed (no known consumers — the Go CLI/agent/bridge and web UI never called them). If you integrated against them directly:
GET /auth/ca/public→ useGET /certificates/ca(returns the same CA public cert).POST /agents/{id}/status→ agents report status via the heartbeat; there is no separate status call.GET /audit/sessions/{session_id}/recording→ useGET /audit/recordings?session_id={id}(list) orGET /audit/recordings/{id}(fetch).
Upgrade Note
Because retention is now enforced, the first sweep after upgrade will delete audit_logs older than retention_days. If you relied on the previous (unenforced) indefinite retention, set retention_days: 0 before upgrading.
Full Changelog: v0.8.5...v0.8.6