Customize how you interact with traced function calls and evaluations
In Weave, saved views let you customize how you interact with traced function calls and evaluations. By defining a saved view, you can configure filters, sorting, and column visibility to quickly access relevant data. Saved views help you and your team return to the same curated slice of your traces or evaluations without reapplying configuration each time.You can create, modify, and save views directly in the Weave Python SDK or through the UI. The Python SDK provides fine-grained control for programmatic filtering and querying, while the UI lets you explore and save different table configurations in the Traces and Evals tabs.This guide covers:
The SavedView class in Weave lets you save, filter, sort, and customize views of trace and evals data. The following sections walk through how to initialize a view, configure its columns and filters, and persist it so you can retrieve matching calls later.
Use .add_column() to add one or more new columns to the view. Specify one or more columns to add.
# Add a column with the field specifier and label "Created"view.add_column("Created")# Optionally, you can add a second argument to specify a different label name for the new column. By default, the field specifier is used for the label.
In Weave, every trace or eval is associated with an operation name.
Use .filter_op() to filter the SavedView to only include calls where that specific operation ran.
You can create, load, rename, and edit saved views in the Weave UI. The UI is the quickest way to capture an ad hoc table configuration as a reusable view. For programmatic control, use the Python SDK.