Roadmap
stardelt is pre-alpha: there are no tagged releases yet, you install from main, and things
change without notice. Work is grouped into milestones, and each one has a goal you can check. The
platform will stay on 0.x for a long time.
Open work is in each repository's issues under the milestone of the same name, such as
stardelt-platform.
This page changes through pull requests, and a large change of scope starts with an issue.
Milestones
| Milestone | Goal | You can tell it is done when |
|---|---|---|
| v0.1 Secure foundation | Sign-in and TLS everywhere, no committed secrets | A stranger can install a cluster and nothing is exposed that should not be |
| v0.2 Operator install | One resource and your values install the whole platform | A single manifest installs the platform on a fresh node and answers a query |
| v0.3 Data engineer workflow | Ingest, model, schedule, query log, connections, catalog documentation, developer sign-in | From source to dashboard works end to end by hand |
| v0.4 Operate and govern | Warehouse lifecycle, backups, alerting, permissions, upgrades, usage views | An operations team can run and upgrade a production install |
| v0.5 Agents and extensions | Hosted tool server, command line, VS Code extension | A local AI assistant works on the platform as the signed-in person |
| v0.6 Governance and sharing | Data products, federation, external sharing, file-drop ingestion, acceleration | |
| v0.7 AI and ML | Assistant, AI functions in SQL, vector search, feature store, tracking, serving, GPU | A model goes from a notebook to a served endpoint |
Later milestones may split or move. Exact component choices for work that has not started are not promises: a component is named on this site when it is built.
What works today
- Install:
make platforminstalls everything in one command, andmake upgradeupgrades it. The operator installs a growing part of the platform fromPlatformInstanceandLakehouseresources. - Query and store: Trino over Iceberg tables, the Lakekeeper catalog, SeaweedFS or your own S3.
- Work: the Nova interface (catalog, SQL editor, loads, connections, pipelines, query log, lineage), notebooks, dashboards, scheduled pipelines, Spark Connect.
- Automate: an API, a Python package, a command line and a local MCP server.
- Access: sign-in through your OIDC provider (Keycloak, deployed by stardelt, or your own), and table permissions by group.
- Keep it running: backups of the databases and the data, with commands that check them.
The services page lists what an install contains, and each guide says what is not available yet.
What stardelt will not be
- A hosted service first. Self-hosting is the main path, and it will stay so.
- A new query engine or storage format. stardelt composes upstream projects, and fixes go upstream.
- A Kubernetes distribution. You bring the cluster.
- An OLTP database.
- A benchmark contest. The goal is predictable cost, openness and sovereignty.
How decisions are made
The maintainers decide, using the license policy, the sovereignty goals and this roadmap. To propose a change, open an issue that says what you need and why. A formal governance model will come once the project has more than one active organization behind it.