View a markdown version of this page

Anthropic connection options - AWS Glue

Anthropic connection options

The following are connection options for Anthropic:

  • ENTITY_NAME – (Required) Used for Read. The name of the Anthropic object to read.

  • SELECTED_FIELDS – Default: empty (SELECT *). Used for Read. The columns you want to select for the object.

  • FILTER_PREDICATE – Default: empty. Used for Read. A filter predicate that AWS Glue pushes down to Anthropic for the fields listed in the filtering support table below. Supply it as a AWS Glue FILTER_PREDICATE option (for example, from a Spark DataFrame filter).

Anthropic supports server-side (pushed-down) filtering on the following entities, with the operators shown, through FILTER_PREDICATE:

EntityFilterable fieldsSupported operators
api_keysworkspace_id, status, created_by_user_idequals only
cost_reportstarting_at, ending_at, group_byequals only
invitesemail, role, statusequals only
modelslifecycleequals only
skillssourceequals only
usage_reportstarting_at, ending_at, api_key_id, workspace_id, model, account_id, service_account_id, service_tier, context_window, inference_geo, speed, group_byequals only
usersemail, roleequals only

AWS Glue applies filters on any other Anthropic entity after it reads records, through the Spark DataFrame filters.