Datasets

The Datasets surface holds your policy test fixtures: curated sets of real traces you can replay against a draft policy before you enforce it. Instead of guessing whether a rule blocks what it should and allows what it should, you assemble a known set of past requests, run a candidate policy over them, and confirm the outcome. You reach it at /dashboard/datasets, under the Prove chapter.

The list

The Datasets list shows every fixture you have built. Each row carries:

• The dataset's name.
• Its item count (how many traces it holds).
• Its created date.

Click a row to open the dataset detail page at /dashboard/datasets/[id].

Building a dataset

You build a dataset from real traffic. On the Runs surface, bulk-select the traces you want, then add them to a dataset. Pick the cases that matter: the requests a policy must keep blocking, and the requests it must never start blocking. A good fixture mixes both so a single replay tells you the rule is correct in both directions.

Replaying against a draft policy

Once a dataset exists, you replay it against a candidate policy. The replay confirms two things at once: that the policy blocks what it should, and that it allows what it should. This pairs with the policy simulator, so you can validate a rule on a known set of cases before any of it reaches production traffic.

API

Datasets are backed by two read endpoints:

• GET /v1/datasets lists your datasets.
• GET /v1/datasets/{id} returns a single dataset by id.

Next steps