v0.7.0 [2025-09-22]

Major Changes

• Replace native Parquet conversion with a DuckLake database backend. (#546)

  • DuckLake is DuckDB’s new lakehouse format: data remains in Parquet files, but metadata is efficiently tracked in a
    separate DuckDB database.
  • DuckLake supports function-based partitioning, which allows data to be partitioned by year and month. This enables
    efficient file pruning on tp_timestamp without needing a separate tp_date filter. A tp_date column will still
    be present for compatibility, but it is no longer required for efficient query filtering.
  • Existing data will be automatically migrated the next time Tailpipe runs. Migration does not
    occur if progress output is disabled (--progress=false) or when using machine-readable output (json, line,
    csv).

  Note: For CentOS/RHEL users, the minimum supported version is now CentOS Stream 10 / RHEL 10 due to libstdc++ library compatibility.

• The connect command now returns the path to an initialisation SQL script instead of the database path. (#550)

  • The script sets up DuckDB with required extensions, attaches the Tailpipe database, and defines views with optional
    filters.
  • You can pass the generated script to DuckDB using the --init argument to immediately configure the session. For
    example:

    duckdb --init $(tailpipe connect)

    Note: The minimum supported DuckDB version is 1.4.0.

Bug Fixes

• Include partitions for local plugins in the tailpipe plugin list command. (#538)