Postgres_AI Monitoring (gitlab.com)

Postgres_AI Monitoring is an expert-focused, open-source monitoring tool announced as a core component of PostgresAI’s Self-Driving Postgres (SDP) initiative. Aimed at senior DBAs, SREs and AI systems, it delivers rapid root-cause analysis by combining top-down troubleshooting (Four Golden Signals: Latency, Traffic, Errors, Saturation) with deep, query-level visibility. Its significance lies in enabling faster incident response and higher automation levels for production Postgres fleets—providing structured outputs that let AI systems ingest metrics and recommendations for automated tuning and reporting. Technically, the stack pairs pgwatch v3 for collection, Victoria Metrics (Prometheus-compatible) for time-series, and Postgres for query texts; Grafana ships with expert dashboards. Key features include full pg_stat_statements metrics with historical trends and plan variations, an Active Session History (Oracle ASH-like) for session-level troubleshooting, wait-event and autovacuum analysis, WAL/RPO monitoring, and query-plan recommendations. Supports Postgres 14–18 and requires pg_stat_statements; deployment is Docker-based with a one-command CLI quickstart (demo or production via API token). The project is Apache 2.0 licensed and intended only for experienced Postgres users — it exposes multiple service ports that must be firewalled and operated with proper DB privileges, so security and operational discipline are critical.