@mykola_kondratiuk That’s a very fair question. Hard exceptions are a real limitation for QA automation, especially as the scope increases.
Our approach is quite practical: the limit is reached when the true behavior cannot be reliably reconstructed from the available artifacts – code, tests, specs, documentation, and the behavior of the product.
So our recovery model is to automatically recover where the system can establish a high-confidence truth, and surface ambiguities when it cannot. In practice, this means that Ogoron can adapt to many standard cases on its own, but in really disputed or less-specified situations it asks the user to solve them explicitly rather than pretending to be certain.
A big part of the product is expanding that high-trust area over time – from general web patterns to increasingly domain-specific behavior.
<a href