The Similarity Engine matches every new incident against your full historical library using AI pattern recognition. Within seconds, it surfaces the most similar past incidents, the runbooks that resolved them, and a confidence score for each match, so your team can skip the investigation and jump straight to the fix.
Nova converts each incident into a high-dimensional vector that captures its signal patterns, affected services, error signatures, and resolution steps. When a new incident arrives, it is compared against every historical fingerprint in your library, not by keyword matching, but by structural similarity. A Redis connection timeout today matches the Redis connection timeout from six months ago even if the error messages use completely different wording.
Every match comes with a confidence score from 0 to 100 that factors in signal overlap, service topology similarity, time-of-day patterns, and whether the previous resolution was marked successful. Matches above 85% confidence are flagged as "high confidence" and their runbooks are pre-loaded into the incident card. Lower matches are shown for context but clearly labeled so your team knows the difference.
When a high-confidence match is found, Nova automatically attaches the runbook that resolved the historical incident. If the previous resolution involved automated remediation steps, those are pre-staged and ready to execute with one click. Your on-call engineer opens the incident and sees "This looks like INC-4521 from March, here's the runbook that fixed it in 4 minutes" instead of starting from scratch.
Let Nova's AI match current incidents to historical patterns and suggest proven runbooks in seconds.