RowSync

Copy rows from a production Microsoft SQL Server database to dev, stage, or any other target defined in your config — using a SELECT-only query with preview, upsert, and live progress.

Features

• SQL editor — CodeMirror with syntax highlighting and schema autocomplete (tables, columns, aliases)
• SELECT-only — Blocks INSERT, UPDATE, DELETE, DROP, and other write/DDL statements
• Preview — Run your query against production before copying
• TOP-aware sync — Respects TOP (N) in your query when applying
• Multi-table JOINs — Copies rows from every table referenced in the query
• Auto-create tables — Creates missing target tables from production schema (no FKs)
• Upsert — Skips duplicates by primary key instead of failing
• Live progress — Server-Sent Events stream step-by-step status to the UI
• Offline UI — CodeMirror, fonts, and all frontend assets are bundled locally (no CDN required)

Requirements

• Python 3.10+
• ODBC Driver 17 for SQL Server (or compatible)
• Network access to your MSSQL instances

Quick start

git clone https://github.com/ghaemaghaey/RowSync
cd RowSync

python -m venv .venv
.venv\Scripts\activate        # Windows
# source .venv/bin/activate   # Linux / macOS

pip install -r requirements.txt
copy config.example.json config.json   # Windows
# cp config.example.json config.json   # Linux / macOS

Edit config.json with your connection strings, then start the app:

python run.py

Open http://127.0.0.1:8765

The web UI works fully offline once installed — all JavaScript, CSS, and fonts are served from static/vendor/.

Offline / air-gapped use

RowSync does not load any frontend assets from the internet at runtime. Database connectivity still requires network access to your SQL Server instances.

To refresh bundled UI assets after upgrading dependencies:

python scripts/download_vendor_assets.py

Configuration

config.json is git-ignored. Use config.example.json as a template.

Key	Description
connections.production	Source database (read-only queries run here)
connections.*	Any other key (dev, stage, local, test, …) becomes a sync target
preview_row_limit	Max rows shown in preview when the query has no TOP (default: 500)

Example:

{
  "connections": {
    "production": "DRIVER={...};SERVER=prod;DATABASE=MyDb;UID=...;PWD=...;TrustServerCertificate=yes",
    "dev": "DRIVER={...};SERVER=dev;DATABASE=MyDb_Dev;UID=...;PWD=...;TrustServerCertificate=yes",
    "local": "DRIVER={...};SERVER=localhost;DATABASE=MyDb;UID=...;PWD=...;TrustServerCertificate=yes"
  },
  "preview_row_limit": 500
}

Add or remove target environments by editing the connections object — refresh the page to update the dropdown.

Usage

1. Write a SELECT query (optionally with TOP, JOIN, WHERE, ORDER BY).
2. Choose a target database from the dropdown.
3. Click Preview to see rows from production.
4. Click Apply to upsert matching rows into the target.

Example query

SELECT TOP (100) *
FROM dbo.FinancialStatement fs
WHERE fs.CompanyId = 12345
ORDER BY fs.Id DESC

Apply will copy at most 100 rows from each table involved in the query.

Project structure

RowSync/
├── app/
│   ├── main.py          # FastAPI app & routes
│   ├── config.py        # Config loader
│   ├── database.py      # pyodbc helpers
│   ├── sql_parser.py    # Query parsing & validation
│   ├── schema.py        # Schema introspection & DDL
│   └── sync.py          # Preview & copy logic
├── static/              # Frontend (HTML, CSS, JS)
│   └── vendor/          # Bundled CodeMirror + fonts (offline)
├── scripts/
│   └── download_vendor_assets.py
├── config.example.json
├── requirements.txt
└── run.py

API

Method	Path	Description
GET	/	Web UI
GET	/api/schema	Production schema for autocomplete
GET	/api/targets	Available sync targets
POST	/api/preview	Preview query results
POST	/api/apply	Sync rows (SSE stream)

Security notes

• Never commit config.json — it contains database credentials.
• The SQL editor only allows SELECT statements; the server validates this on every request.
• Run behind a VPN or internal network; this tool is intended for trusted operator use, not public exposure.

License

MIT