DBLoom is a modern, web-based database management system (DBMS) built with TanStack and TypeScript. It offers the same functionality as tools such as phpMyAdmin or Adminer, in a refreshed layout with extensive security features.
Highlights
- Self-hosted: one Docker image for MySQL 8.4 and MariaDB 10.11, running on infrastructure you control.
- Two login modes: account mode with DBLoom's own users, or direct mode with database credentials.
- Security first: runs behind your TLS reverse proxy, with an audit log and protected settings.
- AI assistant: translates requests into SQL without exposing the database to the public network. See AI assistant.
- Type-safe stack: TanStack, with end-to-end type safety from server functions to routes.
Quickstart
docker compose -f compose.yaml -f compose.local.yaml up --build
Open http://localhost:3000 and follow the first-time setup wizard. Details are in the Quickstart or the guided Setup.
Documentation
Motivation
I have used phpMyAdmin for a long time. Recent CVEs and growing concerns around PHP and phpMyAdmin motivated me to create a modern alternative based on TanStack.
Why TanStack? It provides end-to-end type safety from server functions to routes, and a client-first model with an explicit server boundary, without the RSC and caching complexity of Next.js.
AI also plays an ever-greater role in everyday life, so DBLoom features an integrated AI assistant that translates user requests into SQL commands without exposing the database to the public network.
How DBLoom was built
DBLoom was built from scratch in nearly three weeks using AI-driven development. Rather than simply asking an AI to build a DBMS, I used spec-driven development:
flowchart LR
A[phpMyAdmin<br/>legacy system] -->|subagents| B[Collect<br/>specifications]
C[Open phpMyAdmin<br/>issues] --> D
B --> D[Remove smells<br/>bugs, quirks]
D --> E[Rework for<br/>current DB versions]
E --> F[Split into<br/>modules and tasks]
F --> G[Build with unit<br/>and e2e tests]
- Collect specifications. Subagents extracted specifications from an existing legacy system, phpMyAdmin.
- Remove smells. Such specifications can contain smells, such as a still-unreported bug or a feature of a deprecated database version. All open phpMyAdmin issues were collected and checked against the specification, and relevant findings were amended.
- Rework for current versions. Specifications tied to deprecated database versions were removed. Features that newer database versions added were considered and added where missing.
- Build in modules. The reworked specification was split into several large modules of smaller tasks. Each task was developed with unit and end-to-end tests.
