Principles

📄️List principles

`q` matches (case-insensitive substring) against a principle's `name` or `description`. `sort` accepts `name`, `importance`, `updated_at`, and `created_at` (default, newest first). `executions` counts referencing executions and is offset-only — combining it with `cursor` returns 400. `category_id` restricts to one category; `uncategorized` restricts to principles with no category (the absence of one, not a category id) — the two are mutually exclusive. `tags` restricts to principles carrying any of the given tags, but always also keeps untagged principles, mirroring the dashboard's tag filter.

📄️Create a principle

Adds one principle to this target — a rule the AI service under test is expected to uphold, which every evaluated transcript is then judged against. Phrase it as a checkable expectation about the service's behaviour rather than a topic. `importance` weights how heavily a violation counts, from 1 (lowest) to 5 (highest).

📄️Generate principles from existing knowledge-base content

Generate principles from existing knowledge-base content

📄️Delete a principle

Delete a principle

📄️Get a principle

Returns one principle of this target in full, including its category and importance. Principles are the rules a transcript is judged against, so this is the endpoint to read when you need to know exactly what a violation was measured against. Use the list endpoint to discover principle ids.

📄️Update a principle

Update a principle