Custom Metadata Filtering

When you attach metadata to your ingested transcripts, you can use this metadata to filter down into groups of conversations that match certain criteria. This creates a temporary view into a subset of your data and has no effect on ingestion.

For example, you can use custom metadata filtering to view conversations that match a given user_id, or only view conversations that were generated with a specific LLM version.

Filters are applied globally to all views, so ensure you have disabled or enabled the correct filters if the data you are viewing within the dashboard doesn't match your expectations.

Add a Filter

To add a new global filter to your dashboard view, click the filter icon in the top right of the menu bar.

Clicking the filter icon will open the global filter bar where you can specify the exact metadata you would like to match against in the filter.

To add a new filter on a given metadata key, click the + icon. Within the dropdown that appears, choose the metadata keys for which you would like to enable a filter.

Select and search for the relevant metadata values you would like to filter into from the dropdown.

A filter within a single key will fetch the union of the selected values, whereas filters across metadata values will fetch the conjunction of metadata keys. For example, in the below image, this filter will fetch the conversations where llm_model is equal to gpt-4 AND user_id is either user_100 or user_1012.

Press the Apply button to activate the filter.

Remove a Filter

To remove a filter, press the x button next to the metadata key field. Then press Apply to activate these new settings.

Last updated