Institutional memory
Engineer-confirmed RCAs and resolutions are stored in a knowledge graph and surfaced automatically when similar incidents return, so the next investigation starts ahead.
What becomes memory
Section titled “What becomes memory”- Only human-confirmed records. An RCA or resolution is trusted only after an engineer confirms it. Drafts and unconfirmed hypotheses are never stored as trusted memory.
- Relationships and meaning. The knowledge graph records how services, code, owners and incidents relate. Semantic memory retrieves runbooks and confirmed past incidents by meaning, not just keywords.
- No model retraining. Learning is preserved in the graph and memory, not by retraining a model on your data.
How it is used
Section titled “How it is used”When a new investigation starts, Provenant looks for similar confirmed incidents and relevant runbooks and cites them as evidence, labelled like any other evidence so you can judge how closely they match.
Every record carries your tenant context, so memory is never shared between customers.