Start from the state you need
Resume from a named prepared state instead of reconstructing the whole setup by hand.
Specimen 001 · reproducible state
Ruin. Rewind. Replay.
Experiment freely on reproducible database states
Make database experiments cheap, disposable, and repeatable—for QA pipelines and coding agents alike
Open source LocalSQL database statesSQL / psql + Liquibase
sqlrs prepare chinook run smokebasechinooksmoke
Prepared onceRun in isolation
The problem
A fresh database is not the state your workflow needs
State, not emptiness
Real checks begin after schemas, migrations, and setup data exist. Rebuilding that path for every attempt makes each experiment more expensive than it needs to be.
Prepared-state model
Prepare the path. Reuse the state. Break the copy.
`sqlrs` turns preparation steps into reusable immutable states, then starts isolated mutable runs from the point an experiment actually needs.
00Base SQL database
01Preparation step
02Cached prepared state
AExperiment A
BReplay B
Why experiments get cheaper
Focus on the change, not the setup
The advantage is qualitative today: ready states arrive without repeating every completed step, and common state data can be reused. Public benchmarks come later.
Reuse the work already done
Preparation already represented in the state cache does not need to be repeated for each run.
Discard the run. Keep the state.
The mutable instance can disappear while the prepared point remains available for another try.
Two experiment tracks
One workflow for QA and coding agents
The same CLI loop serves established QA workflows and repeatable experiments initiated by coding agents.
Start every test from the same state
Build consistent environments for migrations, integration checks, and local reproduction.
- Migration and rollback experiments
- Repeatable integration environments
- Shared prepare/run recipes in a repository
Let an agent try—and start again
A coding agent can call the same `sqlrs` CLI, inspect the result, discard a failed attempt, and start again from the same prepared point.
- Ordinary shell-command interface
- Reproducible starting points
- Disposable isolated runs
Choose the operating model
Local now. Shared next. Managed later.
Choose where the state engine runs: on one machine, inside your infrastructure, or as a future managed service.
Run the workflow locally
Open-source `sqlrs` CLI and engine for isolated SQL database work on your machine.
Download becomes available with the next verified GitHub Latest release.
Explore LocalShare prepared states with your team
A commercial self-hosted edition deployed in the beta participant's infrastructure.
Selected beta participants receive limited free licenses.
Shared BetaUse the workflow as a managed service
A planned hosted form of prepared-state experiments without running the local engine yourself.
Launch date and service levels have not been announced.
Join early accessRelease-gated support
Database and platform support
The matrix follows the artifacts and workflows verified for the next GitHub Latest release.
| Database engine | Availability |
|---|---|
| PostgreSQL | Current · target release verification pending |
| MySQL / MariaDB | Coming soon |
| Oracle Database | Coming soon |
| Microsoft SQL Server | Coming soon |
| Platform | Artifact | Container runtime | State path |
|---|---|---|---|
| Linux amd64 / arm64 | Required for Latest | Docker; verified Podman paths | Supported backend with copy fallback |
| Windows amd64 | Required for Latest | Docker Desktop; WSL2 for Btrfs path | Copy fallback available |
| macOS amd64 / arm64 | Required for Latest | Podman | Copy snapshots |
Raw SQL / psql · actual CLI form
sqlrs prepare:psql -- -f init.sql run:psql -- -c "select count(*) from users"FAQ
Frequently asked questions
Is Sequelarius a hosted database?
Local is not. It runs a local engine and database containers. A hosted Managed offer is planned, not available today.
Is the Local version open source?
Yes. The CLI and local engine source are published under Apache 2.0. The next verified Latest release will become the primary installation path.
Which databases and migration tools work now?
Current and planned engines are listed in the support matrix. Verified workflows include raw SQL/psql and Liquibase.
How is this different from a disposable empty database?
The useful unit is a reproducible prepared state. An isolated run can begin after the setup work your test or experiment actually needs.
Is Managed available?
No. Managed is planned and not available today. Join early access for launch updates.