Version:

Installing pgEdge AI DBA Workbench

The pgEdge AI DBA Workbench is an AI-powered environment for monitoring, managing, and troubleshooting PostgreSQL systems. The Workbench combines a Model Context Protocol (MCP) server with a web-based user interface, a data collector, and an alert monitoring service. The Workbench enables users to query, analyze, and manage distributed PostgreSQL clusters using natural language and intelligent automation. The Workbench exposes pgEdge tools and data sources to both cloud-connected and locally hosted language models; this design ensures full functionality in air-gapped or secure environments.

Supported Installation Methods

The Workbench supports three deployment methods.

Building from binary files is the easiest method to use to deploy the Workbench.
Building from source code ensures you have the latest Workbench features available.
The Docker guide walks you through a Workbench deployment via Docker using RPM/DEB packages from pgEdge.

Each installation method places files in different locations. The following table summarizes the locations for each deployment method.

Resource	GitHub Release	Docker	RPM/DEB Package
Binaries	`/opt/ai-workbench/`	`/usr/local/bin/`	`/usr/bin/`
Config	`/etc/pgedge/`	`/etc/pgedge/` (mounted)	`/etc/pgedge/`
Data	user-chosen	`/data/`	`/var/lib/pgedge/<service>/`
Logs	`stderr`	`stderr`	`/var/log/pgedge/<service>/`
Client files	`/opt/ai-workbench/client/`	container-served	`/usr/share/pgedge/ai-dba-client/`
systemd units	`pgedge-ai-dba-*.service`	N/A	`pgedge-ai-dba-*.service`
Run-as user	user-chosen	container user	`pgedge`

Note

RPM and DEB packages are available from the pgEdge Enterprise Repository. Contact pgEdge for access details.

System Requirements

The following minimum requirements apply to all deployment environments. The collector, server, and alerter components share the following hardware requirements:

A minimum of 4 CPU cores is required.
The system requires at least 16 GB of RAM.
The installation requires 120 GB of disk space for binaries and the datastore.

Before installing the Workbench with binary files or building the project from source, install the following software:

Go 1.24 or later is required for building server-side components.
Node.js 18 or later is required for building the web client.
PostgreSQL 14 or later is required for the datastore.
Make is required for build automation.
nginx is required to serve the client.

Each component requires specific network access to operate correctly:

The collector requires network access to each monitored PostgreSQL server.
The alerter requires network access to the datastore.
The server requires network access to the datastore and must be reachable by web client users.
Database credentials for the datastore and each monitored PostgreSQL server are required.

Verifying the Health of an Individual Component

After installing the Workbench and starting all components, verify the health of each component with the following commands.

Checking the Collector

The collector logs probe executions to stderr. Use the following command to confirm the collector is running:

sudo systemctl status pgedge-ai-dba-collector

A healthy collector shows an active status, a confirmed datastore connection, and active probe scheduling. The following example shows the expected output:

pgedge-ai-dba-collector.service - pgEdge AI DBA Workbench Collector
 Loaded: loaded (/etc/systemd/system/pgedge-ai-dba-collector.service; enabled; vendor preset: enabled)
 Active: active (running) since Tue 2026-05-12 14:45:17 UTC; 24h ago
Main PID: 59722 (ai-dba-collecto)
   Tasks: 12 (limit: 4527)
  Memory: 16.3M
     CPU: 4.233s
  CGroup: /system.slice/pgedge-ai-dba-collector.service
          └─59722 /opt/ai-workbench/ai-dba-collector -config /etc/pgedge/ai-dba-collector.yaml

May 12 14:45:17 n1 systemd[1]: Started pgEdge AI DBA Workbench Collector.
May 12 14:45:17 n1 ai-dba-collector[59722]: 2026/05/12 14:45:17 pgEdge AI DBA Workbench Collector v1.0.0-beta1 starting...
May 12 14:45:17 n1 ai-dba-collector[59722]: 2026/05/12 14:45:17 Configuration loaded from: /etc/pgedge/ai-dba-collector.yaml
May 12 14:45:17 n1 ai-dba-collector[59722]: 2026/05/12 14:45:17 Schema is up to date
May 12 14:45:17 n1 ai-dba-collector[59722]: 2026/05/12 14:45:17 Datastore connection established
May 12 14:45:17 n1 ai-dba-collector[59722]: 2026/05/12 14:45:17 Probe scheduler started
May 12 14:45:17 n1 ai-dba-collector[59722]: 2026/05/12 14:45:17 Collector is running. Press Ctrl+C to stop.

Checking the Server

The server listens on the configured HTTP port. Use the following command to test connectivity:

curl -s http://localhost:8080/health

A successful response confirms the server is running and accepting requests. The following example shows the expected output:

curl -s http://localhost:8080/health
{"status":"ok","server":"pgedge-postgres-mcp","version":"1.0.0-beta1"}

Checking the Alerter

The alerter logs rule evaluations to stderr. Use the following command to confirm the alerter is running:

sudo systemctl status pgedge-ai-dba-alerter

A healthy alerter shows an active status and regular baseline recalculation on its hourly schedule. The following example shows the expected output:

pgedge-ai-dba-alerter.service - pgEdge AI DBA Workbench Alerter
 Loaded: loaded (/etc/systemd/system/pgedge-ai-dba-alerter.service; enabled; vendor preset: enabled)
 Active: active (running) since Tue 2026-05-12 15:32:55 UTC; 23h ago
Main PID: 63620 (ai-dba-alerter)
   Tasks: 12 (limit: 4527)
  Memory: 6.6M
     CPU: 6.382s
  CGroup: /system.slice/pgedge-ai-dba-alerter.service
          └─63620 /opt/ai-workbench/ai-dba-alerter -config /etc/pgedge/ai-dba-alerter.yaml

May 12 15:32:55 n1 ai-dba-alerter[63620]: [alerter] Baseline calculator started (interval: 1h0m0s)
May 12 15:32:55 n1 ai-dba-alerter[63620]: [alerter] Alert cleaner started
May 12 15:32:55 n1 ai-dba-alerter[63620]: [alerter] Calculating baselines for 1 connections, 28 rules (lookback: 7 days)
May 12 15:32:55 n1 ai-dba-alerter[63620]: [alerter] Baseline calculation complete
May 12 16:32:55 n1 ai-dba-alerter[63620]: [alerter] Calculating baselines for 1 connections, 28 rules (lookback: 7 days)
May 12 16:32:55 n1 ai-dba-alerter[63620]: [alerter] Baseline calculation complete
May 13 11:54:51 n1 ai-dba-alerter[63620]: [alerter] Calculating baselines for 1 connections, 28 rules (lookback: 7 days)
May 13 11:54:51 n1 ai-dba-alerter[63620]: [alerter] Baseline calculation complete
May 13 12:54:51 n1 ai-dba-alerter[63620]: [alerter] Calculating baselines for 1 connections, 28 rules (lookback: 7 days)
May 13 12:54:51 n1 ai-dba-alerter[63620]: [alerter] Baseline calculation complete

Checking Metrics Collection

Connect to the datastore and run the following query to verify that metrics tables contain recent data.

In the following example, the psql command connects to the datastore:

sudo -u postgres psql -d ai_workbench
psql (18.3 (Ubuntu 18.3-1.pgdg22.04+1))
Type "help" for help.

In the following example, the SELECT statement queries the metrics.pg_stat_activity table and returns the activity count and the most recent collection timestamp:

SELECT COUNT(*), MAX(collected_at) FROM metrics.pg_stat_activity;
 count |              max
-------+-------------------------------
  8046 | 2026-05-14 12:50:48.167367+00
(1 row)

A non-zero count with a recent timestamp confirms the collector is gathering metrics.