Bundlebase 0.12.1
A small release: a --create flag on the CLI and a cleanup to how the bundle format version is stored.
A small release: a --create flag on the CLI and a cleanup to how the bundle format version is stored.
Quality of life improvements, performance improvements, and bugfixes
Mostly bug fixes and rough edges from beta — auto-reindexing after ATTACH/FETCH, unified BM25 across multi-block indexes, a much friendlier REPL.
Preparing for a beta release with stored reports, connector improvements, "hollow" bundles, and a solid round of performance work.
Two themes drove most of what landed in 0.8: mutable data and making Bundlebase work better as an agent tool.
This release adds custom connectors, user-defined functions, and column operations. The big theme: you can now extend Bundlebase with your own code.
Two big additions: text search with the new search() table function, and the ability to attach and join other bundles.
You can now pull datasets directly from Kaggle. This release also overhauls the configuration system to support scoped config with multiple priority layers.
This release is about how you interact with bundles. You can now run any bundle modification through a SQL-like string instead of only through the python API methods.
Those SQL-like command strings can be passed to a new bundle.execute() method in the Python API.
I've also cleaned up the REPL to accept any of those SQL-like commands, and distinguished REPL-specific commands with a / like /exit and status.
Finally, there is also a new SQL server you can start with the CLI instead of REPL mode which lets you run both queries and other commands from your favorite database client.