Trust & security
The page to send to your IT team: on-premise architecture, visible anti-hallucination, human control and signed audit. No vague promises.
Truly on-premise: no data leaves
Interface, inference engine, knowledge base and logs run on your systems — panel PC, plant server or both. No telemetry, no dependency on cloud services, full operation even on an isolated line (air-gap). The status of every component is visible from the system panel.
Cited answers, honest refusals
Every answer carries its citation chips: document, page, confidence score. If the question falls outside the knowledge base, Resolv answers that it does not have enough information — it does not improvise. Traceability is not a feature: it is the trust contract with the technician.
Gate Review: the AI knows only what you sign
Every piece of content extracted from documents goes through human review before entering the knowledge base. Approval is granular, recorded and signed. It is the architectural answer to "what if the AI learns something wrong?": it can't, unless you approve it.
Every event HMAC-signed, exportable
Questions, answers, approvals, knowledge-base version changes: everything lands in a queryable, filterable audit trail. The export produces an HMAC-signed archive — non-repudiable, ready for your compliance processes.
On the hardware you choose, inside your perimeter: plant server, workstation or panel PC. No calls to external inference services.
None that leaves your perimeter: logs and metrics stay on your systems and serve only you, for auditing and for improving answers.
Key-based access with separate roles: operators query, experts approve in Gate Review, administrators manage versions and audit. Every action is attributed to the person who performed it.