NERVE is live across local, private and approved-cloud deployments. Book an evaluation
Glossary
The words used here mean one thing each.
Twenty-two terms, defined as NERVE uses them. Where a word is often used loosely elsewhere, the definition says what it does not cover.
A–C
Acquisition
The act of collecting an item from an approved source into NERVE. Acquisition records the source, the connector or import route, the responsible operator or workload identity, and the time of collection. It is distinct from the time the content itself describes.
Canonical model
The shared information model every normalised object conforms to: one handling envelope carrying identity, provenance, handling and time, plus a typed content body for messages, documents, media, location observations or generic records.
Canonical object
A single normalised record produced from one acquired original. It sits in the Canonical layer, references the original by hash, and is the object that search publication, handling policy and citation all operate against.
Connector
The component that acquires from a particular source family and maps its export into the canonical model. Coverage is organised by family rather than by brand, because what matters is the shape of the export and the provenance it carries.
CORTEX
The analysis product in the AIC Intelligence Suite. CORTEX owns requirements, analysis, hypotheses, intelligence products and reviewed release, and reads NERVE through the Search API under its own workload identity.
D–H
Dataset
A named grouping of acquired material, used to scope ingest, retention, handling and search. A dataset is an organisational boundary; it does not change what is preserved or how an object is modelled.
Derived artefact
Anything produced by a transformation rather than acquired: extracted text, thumbnails, transcripts, optical character recognition output, embeddings. Derived artefacts carry the tool, version and parameters that produced them, and never replace an original.
Evidence reference
A citation that resolves to a specific object, at a specific version, at an exact locator inside the original item. The locator is recorded at transformation time, which is what allows a reference to stay checkable after reprocessing.
Extension namespace
A reserved area on a canonical object that holds source-specific fields which only make sense for one family. It preserves specialist detail without altering the shared envelope that makes objects comparable.
Handling envelope
The shared part of every canonical object. It carries stable identity, acquisition and transformation provenance, handling markings and the separated time fields used for retrieval. Markings in the envelope travel with the object, including through export and the Search API.
I–N
Ingest job
A unit of acquisition and processing work. Jobs are divided into checkpointed steps so that an interrupted or rescheduled worker resumes rather than restarting, and so that failures can be attributed to a stage and a parser version.
Intelligence
An assessed product built from information: a judgement, with reasoning and sourcing attached. NERVE does not produce intelligence. It produces governed, searchable information with provenance, which CORTEX assesses.
NERVE
The Intelligence Data Fabric in the AIC Intelligence Suite. NERVE acquires information from approved sources, preserves each original, normalises it into a common model, derives searchable content and publishes it to a governed search layer.
Normalisation
The transformation of an acquired original into a canonical object in the common model. Normalisation makes objects comparable across sources. It does not resolve identity: a matching identifier in two sources is an observation, not proof that the same person is involved.
O–R
Original
The acquired bytes, retained exactly as received and addressed by cryptographic hash. The original is never rewritten, re-encoded or cleaned, and it remains the reference point that every canonical and derived object points back to.
Privacy profile
The record of obligations attaching to personal data within an object, kept separate from handling markings and from purpose of use. Keeping the three apart is what allows authority to administer the platform to differ from authority to read protected content.
Provenance
The recorded account of where an object came from and how it changed. Acquisition provenance covers the source, route, responsible identity and collection time. Transformation provenance covers the tool, version, parameters and locator behind each derived artefact.
Reprocessing
Rebuilding derived content from a retained original under new parameters, for example after a parser upgrade. Reprocessing creates a new version rather than overwriting, so earlier evidence references continue to resolve against the version they cited.
S–Z
Search-publication state
The explicit state recording whether a canonical or derived object has been published to the search layer, is awaiting publication, or has been withheld. Nothing becomes discoverable as a side effect of processing.
Source
An approved origin of material: an archive, an enterprise system, a forensic export, a tabular import, an operational feed. A source is described as a family with a characteristic acquisition style, canonical output and set of provenance considerations.
Transformation run
One execution of a processing step that produces derived content, recorded with its tool, version, parameters and outcome. A run that fails is surfaced with the stage it reached rather than silently dropped.
Workload identity
The identity issued to a service rather than a person, used for service-to-service access such as the CORTEX integration. It carries its own permissions, is constrained by the same handling policy as human users, and can be audited and revoked independently.
Next
See the terms doing their work.
The Explorer demonstration shows originals, canonical records, derived artefacts and exact locators on the same screen, over a synthetic corpus.