new changes

architecture/2026/10-ops-marketplace-observability.md

@@ -0,0 +1,41 @@
+# Ops, Marketplace, and Observability Implementation
+
+## Ops Modules
+
+### File Manager
+- Command types: `FILE_LIST`, `FILE_READ`, `FILE_WRITE`, `FILE_UPLOAD_CHUNK`, `FILE_MOVE`, `FILE_DELETE`.
+- Security: path sandbox, denylist for system paths, content size and MIME checks.
+
+### Cron Manager
+- Command types: `CRON_LIST`, `CRON_CREATE`, `CRON_UPDATE`, `CRON_DELETE`.
+- Validation: cron expression parser, command allowlist/templating, dry-run syntax checks.
+
+### Firewall + Security Tools
+- Command types: `FIREWALL_RULE_ADD`, `FIREWALL_RULE_DELETE`, `SECURITY_SCAN_BASELINE`.
+- Audit: every mutation recorded with actor, reason, and approval trace.
+
+### Backup & Restore
+- Command types: `BACKUP_RUN`, `BACKUP_RESTORE`, `BACKUP_VERIFY`.
+- Flows: policy-driven schedules, retention lifecycle, encrypted object storage artifacts.
+
+## Plugin Marketplace
+- Catalog includes signed package metadata, compatibility matrix, and permission manifest.
+- Install pipeline:
+  1. Resolve package and verify signature.
+  2. Validate required capabilities and tenant entitlement.
+  3. Execute install as job with rollback hooks.
+  4. Persist install status and event timeline.
+
+## Real-time Monitoring Dashboard
+- Live channels: CPU, RAM, disk I/O, network throughput, process health.
+- Pipeline:
+  - agent collectors -> gateway ingest -> Redis stream -> metrics store.
+- UI:
+  - server list health badges,
+  - per-server timeline charts,
+  - alert panels with acknowledgement workflow.
+
+## Performance Targets
+- P95 live metric latency: < 2 seconds.
+- P95 command dispatch to agent ACK: < 1 second.
+- P95 dashboard query time (last 1 hour): < 400 ms.