🚀 One Tool. Every Environment. Even the Air-Gapped Ones.
dqex is a cross-platform database workbench that ships as a single static binary — no JVM, no Electron, no installers, no network required. It packs export, import, migration, comparison, snapshots, Excel data dictionaries, a full SQL terminal, and an AI assistant into one tool, with both a polished Web UI and a scriptable CLI sharing one engine.
- 🪶 Zero dependency — one binary, no runtime to ship. Copy it to a USB stick and run it on a bank's air-gapped server.
- 🏢 Built for restricted networks — a built-in offline SQL template library, plus the compliance artifacts audits actually ask for (Excel data dictionaries, snapshot diff reports, exportable history) built in rather than bolted on.
- 🤖 AI that cannot surprise you — it reads your real schema before writing anything, generates dialect-correct SQL, and never executes a write without your confirmation. Leave it unconfigured and the AI entry point does not even render.
- 🌐 MySQL · PostgreSQL · Oracle with automatic dialect conversion for cross-database migration.
⚡ Quick Start
1. Download the zip for your platform from the Releases page — macOS (Intel / Apple Silicon), Linux (x64 / arm64), Windows (x64). No installation required.
2. Unzip and run
unzip dqex-*.zip && cd dqex && ./start.sh:: Windows
start.bat3. Open the Web UI
http://127.0.0.1:8181That's it — add a connection and start exporting, comparing or querying. Nothing leaves your machine.
Optional: install to PATH, run as a daemon, or stop it
./install.sh # install to /usr/local/bin (optional)
dqex # start Web UI at 127.0.0.1:8181
./start.sh -d # background daemon
./stop.sh # stop the daemoninstall.bat :: install to %LOCALAPPDATA%\dqex and add to PATH
dqex :: start Web UI in a new terminal
start.bat -d :: background
stop.bat :: stop background serviceNot a Web UI person? Every capability below also exists as a CLI command, and every command can emit JSON — see the CLI section further down.
✨ Feature Highlights
| Feature | Web | CLI | Description |
|---|---|---|---|
| Export | ✅ | export (exp) | Schema + data → SQL file (zip / gzip), conditional & consistent |
| Import | ✅ | import (imp) | SQL / zip file → database, auto create tables, batch inserts |
| Migrate | ✅ | migrate (mig) | Database → database, cross-dialect (e.g. MySQL → PostgreSQL) |
| Compare | ✅ | compare (cmp) | Schema & data diff between two databases |
| Data Dictionary | ✅ | dictionary (dict) | Tables + comments → styled Excel (.xlsx), audit-ready |
| Snapshot | ✅ | snapshot (snap) | Full create / list / show / delete / compare lifecycle |
| SQL Query | ✅ | sql | Web query terminal + table browser; CLI interactive REPL with JSON output |
| AI-Assisted SQL | ✅ | \ai in sql | Agent probes real schema, generates SQL per dialect, requires confirmation |
| Connections | ✅ | conn (cn) | Save / test / delete database connections |
| Saved Tasks | ✅ | task (tk) | Reuse one-click task configs (--task ) |
| History | ✅ | history (his) | Execution log for troubleshooting & audit export |
| Shell Completion | — | completion | zsh / bash completion scripts |
🤔 Why dqex — and why not DBeaver?
If DBeaver works in your environment, use it. It is a bigger, more mature tool and it is free. dqex is not trying to win on breadth — it is built for the environments DBeaver structurally cannot enter.
| | dqex | DBeaver | Navicat | DataGrip | |---|---|---|---|---| | Install form | Single static binary | JVM + driver jars | Installer, licensed | IDE, licensed | | Air-gapped deploy | Copy and run | Manual JVM + jars | License constrained | License constrained | | Time to first query | Under a minute | Several minutes | Install + activate | Install + activate | | Cross-dialect migration | Built in, automatic | No | Yes | Partial | | Excel data dictionary | One command | Plugin | Yes | No | | Snapshot diff | Built in | No | No | No | | AI-assisted SQL | Built in, offline fallback | No | Limited | Limited | | Database breadth | MySQL, PostgreSQL, Oracle | Very wide | Wide | Wide | | Price | Free, MIT | Free | Paid per seat | Paid |
Where dqex is weaker: it supports three database engines, not twenty. If you need SQL Server, ClickHouse, MongoDB or anything outside MySQL / PostgreSQL / Oracle, DBeaver is the better answer today. Additional drivers are explicitly welcome as contributions.
Full comparison, including Bytebase and the known limitations: docs/COMPARISON.md.
🧑💻 CLI in 30 seconds
# Save a connection once, reuse everywhere
dqex conn add --name prod --type mysql --host 10.20.16.170 --port 3317 --un root --pw 'xxx'
Export with conditions & gzip → take it home
dqex exp camunda -s prod -o backup.sql.gz --table-cond "orders:created_at >= '2026-01-01'"
Restore to your local test database
dqex imp -t local_test -i backup.sql.gz --reset drop-and-create
Interactive SQL terminal (native dialect, runs on the target DB)
dqex sql -c prod
dqex sql -c prod --json "SELECT id, name FROM users LIMIT 10" # agent-friendly JSON
Snapshot & compare — the killer feature for incident tracing
dqex snapshot create -c prod -n baseline
dqex snapshot compare -c prod --a baseline --b after-deploy
Compliance: one command → styled Excel data dictionary
dqex dict camunda -s prod -o data_dict.xlsx🤖 AI-Assisted SQL (Optional, Offline-Safe)
- Real schema, not guesses — the agent queries your actual table structures before generating SQL (Web UI shows live progress)
- Generate ≠ Execute — AI only produces SQL text; write operations require confirmation, dangerous statements are blocked, and you can
\e-edit before running - Full assist loop —
\ai continuefor follow-ups; Web UI supports generate / explain / optimize / fix with side-by-side diff preview and one-click apply - Keys stay local — API key stored on your machine; only masked endpoint & model name are shown
- No config, no footprint — the AI entry only appears after BaseURL / API Key / Model are all configured; everything else keeps working untouched
- Offline fallback — built-in SQL template library (
\template top_n orders amount 10) works in fully isolated networks
🛠️ Build from Source
make dev # Go :8181 + Vite :5281 with hot reload & debugging
make build # single binary ./dqex (frontend embedded)
make release # cross-platform packages → release/
make install # → /usr/local/bin🧱 Tech Stack
- Backend: Go + infrakit (database dialect adaptation), Gin, embedded SQLite
- Frontend: React + TypeScript + Vite + Tailwind CSS + shadcn/ui + Monaco Editor
- Excel: excelize (pure Go, no CGO)
- Distribution: single static binary with embedded frontend, dark mode, i18n (EN/中文)
📚 Documentation
- CLI Manual — every command, flag, and meta-command
- Project Overview — philosophy, personas, and deep-dive scenarios
- Comparison — how dqex compares to DBeaver, Navicat, DataGrip and Bytebase
- Offline Operations Guide — running dqex in isolated networks
- Engineering Conventions — state modeling & data-flow rules (read before contributing)
🤝 Contributing
Contributions are what make open source great — and they earn you a place on the stargazers list too! ⭐
We welcome:
- 🐛 Bug fixes (highest priority)
- 📝 Documentation improvements (EN / 中文)
- 🧩 New SQL templates for the offline library
- 🗄️ Additional database driver support
- 🎨 UI/UX polish
- Fork the repo
- Create a branch:
git checkout -b feature/your-feature - Commit your changes:
git commit -am 'feat: add something awesome' - Push:
git push origin feature/your-feature - Open a Pull Request
gofmt; TypeScript follows ESLint + Prettier; commits follow Conventional Commits.
📄 License
MIT © 2026 fj1981
⭐ Star us on GitHub — it tells us you care, and helps more people discover the tool. Fork it, play with it, break it, and send us a PR!