Iceberg Data Platform 0.5.1¶
Release date: 2026-09-23
Changes¶
- Connect from your computer: Python (PyIceberg, DuckDB), the DuckDB CLI and DBeaver use your own Keycloak account through the
iceberg_connect.pyhelper and the newiceberg-clidevice-login client. Catalog › Connect from your computer shows the snippets for each database. - Both portals have a Getting started page: through the portal, your own tools or an AI agent over MCP in the user portal; through the portal, the administration MCP endpoint or the API in the administration portal.
- MCP works with Codex, GitHub Copilot and other MCP clients besides Claude Code. The
iceberg-mcpclient accepts any loopback callback; the user server is documented asiceberg-user. - Try it out works for writes in the administration API documentation at
/docs. - The installation bundle ships the
lan-accessanddemo-companyagent skills andiceberg_connect.py. - Fixed: the platform administrator of an existing installation could not sign in to the administration MCP server (
offline_accessrefused). - Fixed: administration portal status messages stayed visible on other pages.
See the changelog for details.
Upgrade from 0.5.0¶
No data migration is required. Preserve your existing configuration and PostgreSQL, RustFS, pgAdmin and shared notebook volumes.
Rerun the installer, or update the portal, users and notebook images, and restart the
platform. On startup keycloak-bootstrap adds the iceberg-cli client, widens the
iceberg-mcp redirects to any loopback port and gives the platform administrator the
realm default roles. The installer never replaces .env; it adds the agent skills in
.agents/skills and iceberg_connect.py to the installation folder.
Existing MCP registrations keep working under their current name.
Install this version¶
These commands become available after the release workflow publishes v0.5.1.
Linux / macOS:
curl -fsSL https://github.com/sanderdw/iceberg-data-platform/releases/download/v0.5.1/install.sh | sh
Windows PowerShell:
The installers pin the installation bundle and all application images to 0.5.1.
Docker with Compose is required. The release workflow builds images for Linux AMD64
and ARM64.
Publication¶
Follow the release workflow. Merge the reviewed changes into
main, wait for hosted CI, then tag that commit v0.5.1. Pushing the tag triggers
image builds and publication of the source archive, installation bundle, installers
and checksums.
Preparation checks¶
Local preparation completed on 2026-09-23:
- Full verification: 397 tests passed, 9 skipped; Python lint and JavaScript syntax checks passed.
- Browser fixtures: team/database management, catalog, data shares, forms and navigation passed.
- Live integration in CI mode: end-to-end browser tests, catalog, stack, data and share smoke tests and the notebook examples passed.
- Keycloak mode:
keycloak-bootstrapupgraded an existing 0.5.0 realm; Keycloak, user MCP, admin MCP andiceberg_connect.pydevice-login smoke tests passed. - The notebook image started with a read-only filesystem and no network; concurrent flights kernels and HTML/Jupyter exports passed.
- Python and npm dependency audits reported no known vulnerabilities.
- Gitleaks found no secrets in the packaged source.
- The installation bundle pins all images to
0.5.1and contains the agent skills andiceberg_connect.py.
The skipped tests require PowerShell, which is unavailable in the local environment. The Keycloak onboarding smoke replaces the initial administrator password and only runs on the fresh stacks of hosted CI. Hosted CI must still run it, the native installers on Linux, macOS and Windows, and the multiarchitecture image builds.