Skip to content

Releases

unshackle 0.2.0

Choose a tag to compare

@javier-bonilla javier-bonilla released this 11 Nov 09:09
cf7edee

This release adds Metabase parity scaffolding and subscription exports, and fixes the migration failures reported against 0.1.0.

  • Fix duplicate MetricFlow dimension labels so joined Metabase models pass dbt parse (#173).
  • Keep colliding dashboard filter names, defaults and SQL binds distinct (#175).
  • Handle empty, comment-only and malformed parity YAML without aborting other cases (#171).
  • Capture Metabase CSV baselines and scaffold strict cases with unshackle verify --init metabase, optionally scoped with --only dashboard:<id> (#189).
  • Export dashboard subscriptions to declarative schedules/ YAML (#190). This preserves delivery intent; it does not install a scheduler or send messages. Unsupported settings remain explicit TODOs.

Upgrade with pip install --upgrade unshackle or uv tool upgrade unshackle. Rebuild affected migration output after upgrading; re-extract older exports with ambiguous filter mappings.

Assisted by Codex

unshackle 0.1.0

Choose a tag to compare

@javier-bonilla javier-bonilla released this 27 Oct 11:48
319ac0a

First release.

  • Sources (13): Tableau, Power BI, Looker, Qlik Sense, Amazon QuickSight, Metabase, Lightdash, Apache Superset,
    Redash, Omni, Hex, ThoughtSpot and Sigma, through their APIs (unshackle migrate), or from downloaded files
    (unshackle import) for all but Metabase, Redash and Sigma.
  • Outputs: Apache Ossie semantic models and metrics; dashboards as Jupyter notebooks (SQL per tile, Altair
    charts); SQL queries and models; Cedar policies and SQL row filters; table descriptions and Markdown knowledge.
  • Nothing guessed: what can't be translated faithfully stays a TODO(unshackle) with the original expression.
    migration/report.md shows per dashboard how many tiles have SQL and what blocks the rest; unshackle todo lists
    every item.
  • Commands: check (credentials, with --prompt to save them to .env), migrate (init, inventory,
    extract and build in one go), import, run (execute the notebooks), verify (migrated numbers against the
    source's; verify --init tableau writes the cases from Tableau's view data), todo, status, report.
  • Data that lives only in Tableau (extracts, Excel, CSV) can be copied to Parquet with --with-data and loaded
    into a local DuckDB file the notebooks read. .hyper extracts need pip install "unshackle[hyper]".
  • Install: pip install unshackle (or uv tool install unshackle).

Changes before the first release are in the git history.

On PyPI: https://pypi.org/project/unshackle/0.1.0/