Dovada examines each ticket and each release automatically. It finds the tests that run the changed code and judges if they prove the change. Dovada shows how it reached each answer.
A test your last release relied on
This test runs, and it passes.
it('sends an alert', () => {
alertOnCallIfDead(deadDelivery)
expect(sendAlert).toHaveBeenCalled()
})
The test does not examine the text of the alert or who receives it. It also does not show if the alert starts after the third failure or after the first. CI shows green, but the test proves almost nothing.
Dovada found the problem. The next slide shows it.
One record · DEMO-112
One ticket. Two answers that differ.
/dovada/cer/14
What a summary does not show · DEMO-111
The ticket reads Done. No code shipped.
/dovada/release/DEMO-1.1.0
How Dovada makes each answer
Two axes, assessed independently
Validation
Did we build the right thing?
Dovada compares each requirement of a ticket with the pull requests for that ticket. It judges each requirement alone. It does not give one score for all requirements.
Verification
Did we prove it works?
Dovada starts from the code that changed. It traces imports and calls to the tests that exercise this code. Then it judges if each link is meaningful.
Each judgment above is one call to a model. Today the model is Claude. The call uses your Anthropic account and your contract, not an account of Dovada. Dovada plans to add other providers.