{% block head %}{% endblock %}
Compare two datasets down to the cell — no Spark, no warehouse.
This UI wraps the lake-sift CLI and keeps a history of every run.
path/to/*.parquet
iceberg:<catalog>/<ns>.<table>[@<snap|ref>[^]]
@1042 a snapshot · @main a branch/tag · @^ that snapshot's parent.delta:<path>[@<version>]
sql:<file.sql>
A SQLite file — default ~/.lake-sift/history.db
(override with lake-sift web --db <path>). Only the diff result is saved,
with the row/cell listing capped; your Parquet, Iceberg, and Delta data is never copied.
lake-sift <left> <right> --key <id> runs the exact same engine.
Docs on GitHub →