AkurAI-Build

public
Latest change e11509261b6a2aa4ea9ef67250b0be01bfb4f53b - Restore paths swept in from uncommitted WIP by Ólafur Búi Ólafsson
---
name: akurai-monitor
description: "Rust observability service: metrics/logs/alerts ingest and MCP"
---

# AkurAI-Monitor
Rust monitoring service (~/Projects/AkurAI-Monitor): src routes/schema/db/oidc/mcp/stream/tailer/version_poller/health/journald/alert.
- CI (master): fmt/clippy/scripts/check-metric-contract.sh/test; musl package rust-monitor; deploy akurai-ec2 release --mode publish --yes then scripts/validate.sh.
- Docs live in Notes notebook 36; secrets in PassVault. Query via mcp akurai-monitor tools (host_status, query_metrics, query_logs, list_alerts).

## Verified operations facts (2026-08-09)
- Service: `rust-monitor.service` on `akurai-ec2`, listening 127.0.0.1:8800, fronted by nginx at `https://monitor.olibuijr.com`; MCP endpoint `/mcp`.
- Live store: MariaDB database `akurai_monitor` inside the `mariadb-dev` Docker container on akurai-ec2 (`MONITOR_DB_URL` in `/etc/rust-monitor.env`). The SQLite file previously at `/opt/rust-monitor/app/data/monitor.db` is retired legacy data — editing any `monitor.db` file has no live effect.
- MCP tools are read-only. To change alert thresholds or resolve stuck alert events, run SQL via:
  `ssh akurai-ec2 sudo docker exec mariadb-dev sh -c 'mariadb -uroot -p$(cat $MARIADB_ROOT_PASSWORD_FILE) akurai_monitor -e "<SQL>"'`
- Hosts: ec2-vm-dev (1), titan (8), local laptop (34). Host 34 per-process memory threshold is 6 GiB by design; do not re-lower it for browser processes.
- Drop leftover `akurai_test_mon_*` databases during cleanup.