How Resolv works

From the PDF manual to the cited answer: the full journey of knowledge inside Resolv, without a single byte leaving the plant.

Step 1 — Ingestion

Your documents become indexed knowledge

Service manuals, technical bulletins, senior technicians' internal notes: the ingestion pipeline reads them, splits them into knowledge units and indexes them for semantic search. Everything happens on your systems: the document is never uploaded to any external service. An 8-stage pipeline shows in real time what is happening to each file.

Step 2 — Gate Review

A human approves every piece of content

Before any content becomes queryable, it goes through Gate Review: your expert sees every extracted knowledge unit and approves or rejects it. It is the human-in-the-loop principle applied to maintenance: the AI knows nothing you have not signed off. Decisions are recorded and signed.

Step 3 — Querying

Guided wizard or free question

The technician has two paths: enter an alarm code into the Alarm Wizard and be guided step by step to resolution, or ask Ask Resolv a natural-language question. Either way, the answer arrives in seconds and cites its source: document, page and confidence score.

Updates — Atlas

Knowledge updates like firmware

The knowledge base is versioned into signed atlases. You publish v2.4, activate it with a hot-swap in seconds — without stopping the machine or restarting anything — and if something is off you roll back to the previous version. Every version change is tracked in the audit.

For a typical manual, the ingestion pipeline and Gate Review complete within the day. The bottleneck is the human review — and that is a design choice: you decide what the AI may know.

PDF first and foremost — service manuals, bulletins, annotated diagrams — plus notes and procedures written directly by technicians inside the Knowledge Hub.

Answers always cite their source, so the conflict is visible, not hidden. In Gate Review you can reject the outdated content or scope it to a specific machine model.

Request a demo