Self-hosted tool for querying past database states without restoring backups. Captures periodic snapshots of PostgreSQL and MySQL tables as Parquet files, and lets you query any point in time with SQL powered by DuckDB.
Query any snapshot with SQL using DuckDB. Supports joins, aggregations, filtering, and CSV/JSON export.
Compare two snapshots side-by-side with row-level diff. Added, removed, and modified rows are color-coded.
Capture full table snapshots as Apache Parquet files on a cron schedule. Browse snapshot contents with filtering, sorting, and CSV/JSON export.
- Retention Policies — Daily/monthly retention limits to manage storage automatically
- Flexible Storage — Local filesystem or S3-compatible object storage (AWS S3, MinIO, etc.)
- Setup Wizard — Guided initial setup for admin account, storage, and first datasource
- Dark Mode — Full dark/light theme support
docker run -d \
--name snaplake \
-p 8080:8080 \
-v snaplake-data:/app/data \
-e SNAPLAKE_JWT_SECRET=your-secret-key \
-e SNAPLAKE_ENCRYPTION_KEY=your-encryption-key \
abcdkh1209/snaplake:latestNote: Replace
your-secret-keyandyour-encryption-keywith your own secure values. These are used for JWT signing and datasource password encryption respectively. If omitted, random keys are auto-generated on each restart, which will invalidate existing sessions and encrypted data.
Open http://localhost:8080 and follow the setup wizard.
A demo compose file is included with a pre-configured PostgreSQL database:
docker compose -f docker-compose.demo.yml upThis starts Snaplake alongside a PostgreSQL instance loaded with sample data (customers, products, orders). Connect to it during setup:
| Field | Value |
|---|---|
| Host | sample-db |
| Port | 5432 |
| Database | sampledb |
| Username | demo |
| Password | demo1234 |
Prerequisites: Java 21, Bun
# Start backend (port 8080)
./gradlew bootRun
# Start frontend dev server (port 5173, proxied to backend)
cd frontend && bun install && bun run devOn first launch, the setup wizard walks you through initial configuration. See the Setup Guide for details.
All configuration is done via environment variables:
| Variable | Default | Description |
|---|---|---|
SNAPLAKE_DATA_DIR |
./data |
Directory for SQLite metadata DB and local snapshots |
SNAPLAKE_PORT |
8080 |
Server port |
SNAPLAKE_JWT_SECRET |
(auto-generated) | JWT signing secret |
SNAPLAKE_ENCRYPTION_KEY |
(auto-generated) | AES key for encrypting datasource passwords |
Storage (Local or S3) is configured through the web UI during setup.
PostgreSQL / MySQL
|
Scheduled or manual trigger
|
Snapshot as Parquet files ──→ Local or S3 storage
|
DuckDB SQL engine ──→ Query, analyze, compare
- Capture — On schedule or manual trigger, Snaplake reads tables and writes them as Parquet files
- Store — Snapshots go to local filesystem or any S3-compatible storage
- Query — DuckDB reads the Parquet files directly — no import, no restoration
Hexagonal Architecture with clear separation of concerns:
adapter/inbound (Web, CLI, Scheduler)
|
application/port (UseCase interfaces, Port interfaces)
application/service (UseCase implementations)
|
domain/model (Pure Kotlin domain models)
|
adapter/outbound (JPA, DuckDB, S3, Local Storage)
Dependency direction: adapter -> application -> domain
- DatabaseDialect — Add support for new database types (currently PostgreSQL, MySQL)
- StorageProvider — Add new storage backends (currently Local, S3)
| Layer | Technology |
|---|---|
| Backend | Kotlin, Spring Boot 3.4, Java 21 |
| Metadata DB | SQLite |
| Query Engine | DuckDB |
| Snapshot Format | Apache Parquet |
| Frontend | React 19, TypeScript, Vite |
| UI Components | Carbon Design System |
| Routing / State | TanStack Router, TanStack Query |
| Auth | JWT + Argon2 |
| Storage | Local filesystem / S3-compatible |
# Full build (backend + frontend)
./gradlew build
# Run tests
./gradlew test
# Frontend only
cd frontend && bun run build
# Frontend lint
cd frontend && bun run lintThis project is licensed under the MIT License.