DBA Brain

Graded verdicts, not raw rows

DBA Brain collects health metrics from your SQL Server, Oracle, PostgreSQL and MySQL instances, grades them against policies you write, runs your scheduled SQL, takes backups and proves them by restoring them, checks objectives against real measurement history, and tells you about it, on a schedule, without an agent on any monitored machine.

It sends nothing anywhere by itself. No telemetry, no usage reporting, no update check. It connects to the databases and hosts you list, and, only if you turn it on, to a chat service. Nothing else.

What it is not

What it produces

Real report pages from a live estate, with every name replaced by a stable fake one. Click an image to open it full size.

One run, end to end

Whoever starts it (the daemon, a person, a chat command or an AI agent), a run takes the same path. The app hands one JSON request to transport, which starts the shared operations CLI; that CLI reaches the database or host and answers with one JSON envelope. Results land in the runtime store, and reports and alerts are built from there.

Every command answers in the same envelope, whether the work succeeded or not:

{
  "success": true,
  "operation": "restore-full",
  "message": "Restored SALESDB_STG to 2026-08-07 01:40:00.",
  "error": null,
  "data": { },
  "metrics": { "duration_ms": 41230 }
}

From a measurement to an alert

A backup is proven by restoring it

Why: cheap for an AI agent, fine without one

The command a person runs is byte-for-byte the command an agent runs, so there is one code path to audit, and every run is logged the same way whoever started it.

  WITHOUT a tool                        WITH DBA Brain
  agent -> raw SQL -> agent             agent -> DBA Brain -> database

  "check this database"                 "check this database"
    -> SELECT ... dm_os_wait_stats        -> one JSON request
    <- every row, into the context
    -> SELECT ... sys.databases           <- one graded envelope:
    <- every row, into the context             41 checks, 40 OK,
    -> ... once per check, per run             1 WARNING: appdb
    <- all of it, healthy or not               log backup 31h old

The saving is not compression. It is not sending what nobody needed to read. And no model is required: remove the agent and the daemon, schedules, reports and alerts all still run.

A backup that outlives the host it came from

A backup kept on the database host is lost together with it. A dead disk takes the database and every backup written next to it, so a backup on the same disk as the database is one of the biggest holes there is. A lost host, or any other failure of that machine, takes every disk it has. Ransomware goes further: it encrypts every drive and every share the infected host can write to.

DBA Brain proves every backup at each step and keeps a proven copy away from the database host.

Each backup run, step by step

FULL, DIFF and LOG backups for SQL Server; a base backup plus WAL for PostgreSQL; RMAN for Oracle.

Configuration

Every field is mandatory, so nothing is left to a default nobody wrote down. upgrade-config adds them to existing entries with the values that keep today's flow: offload_enabled: false, offload_scan: "none", offload_delete_source: false, restore_from: "server".

{
  "backups": [
    {
      "backup_id": "APPDB_FULL",
      "server_id": "db-server",
      "jobs": [
        { "job": "database", "databases": ["APPDB", "SALESDB"], "all_databases": false },
        { "job": "log", "databases": [], "all_databases": true }
      ],
      "offload_enabled": true,
      "offload_server_id": "repository-vm",
      "offload_folder": "/srv/backups/db-server",
      "offload_copy_max_retry": 3,
      "offload_scan": "clamav",
      "offload_delete_source": false,
      "offload_source_share": "",
      "offload_source_username": "",
      "offload_source_password_ref": ""
    }
  ],
  "restores": [
    { "id": "drill-vm-daily", "restore_from": "server" }
  ]
}

15 components

Ten apps, each doing one job with its own CLI and never importing another app, and five shared layers that every app stands on. The list is closed, and every component has exactly one reference doc.

Who may call whom

Apps

Shared layers

Get started

Python 3.12+. Each database driver is an extra, so you install only what you run. Published on PyPI via trusted publishing.

python -m venv .venv
.venv/bin/pip install 'dbabrain[postgres]'

cd examples/postgres-quickstart
docker compose up -d
python -m db_ops.db.cli      --config config.json init
python -m db_ops.metrics.cli --config config.json collect --dry-run
python -m db_ops.metrics.cli --config config.json collect
python -m db_ops.metrics.cli --config config.json report

Read more