NERVE is live across local, private and approved-cloud deployments. Book an evaluation

Deployment profiles

Consistent contracts across different operating environments.

NERVE separates product semantics from infrastructure adapters, so local, private and approved-cloud profiles preserve the same search, common-model and evidence contracts. The environment changes. What a citation means does not.

All three profiles available

Three profiles

Pick the shape that matches your environment

Each profile arranges the same logical services differently. Selecting one changes the topology and the notes below it. The essential points are also stated in the comparison table, so nothing depends on the interaction.

Logical topology

Single protected host

API

Workers

Queue

Object store

Compact local node

A protected host, or a small set of hosts, co-locates the logical services. Object storage, search, the queue and the workers all run locally, so the whole source-to-search chain can be exercised inside one boundary.

It suits development, evaluation, smaller holdings and disconnected operation, within capacity measured for the environment. Sizing is agreed during acceptance against your real volumes.

Sized during acceptance

Side by side

The essential differences, without interaction.

The selector above is a convenience. The table below carries the same information for anyone reading linearly, with assistive technology, or on a narrow screen.

Dimension

Compact local node

Scale-out private

Approved cloud

Best suited to

Development, evaluation, smaller holdings and disconnected operation within evidenced capacity.

Larger holdings and sustained concurrent workloads inside an organisation’s own protected estate.

Organisations whose approved infrastructure is delivered as managed services.

Service layout

A protected host or small host set co-locates the logical services.

Separate control plane, protected storage, search, persistent queue and stage-specific worker pools.

Approved infrastructure services sit behind the same adapters; product semantics are unchanged.

Storage and search

Local object storage, search, queue and workers.

Independently provisioned storage and search tiers with a persistent queue.

Managed object storage and search accessed through infrastructure adapters.

Scaling approach

Vertical, within the evidenced capacity of the host.

Scale parser, media, indexing or export workloads independently.

Scale through the approved service’s own elasticity, behind the adapter boundary.

Disconnected operation

Built for it. Offline mode blocks external connectors before transport.

Possible where the whole estate is isolated and updates arrive as signed release imports.

Not applicable. Approved cloud deployment assumes connectivity to the approved services.

AI provider position

No mandatory public-cloud AI dependency for acquisition, normalisation or conventional search.

Same. Optional local enrichment only where approved.

Cloud infrastructure approval does not authorise transmitting protected content to unrelated hosted AI providers.

Acceptance

Capacity is measured for your environment during acceptance, not assumed.

Each environment is accepted against your own volumes and boundaries.

Accepted alongside your provider approval.

Operating modes

Online and offline behaviour.

Offline mode is a policy control inside the application. It changes what NERVE is permitted to attempt, and the boundary is enforced before transport rather than after it.

Capability

Online

Offline

Local import

Available

Local search

Available

Local CORTEX integration

Available

External connectors

Policy-controlled. Permitted only where the source and the route are authorised.

Blocked before transport, unless an approved internal service is explicitly permitted.

Local enrichment

Optional, under your approval.

Hosted enrichment

Policy-controlled. Never implied by infrastructure approval alone.

Rejected.

Software updates

Controlled release acceptance.

Controlled signed release import.

Offline mode is not the same thing as an air gap

Shared responsibility

Application Offline mode is one policy control. It stops NERVE from attempting outbound transport for sources, enrichment and update retrieval, and the block is verifiable at the application boundary.

An actual air gap depends on far more than one application setting. It depends on network controls, host controls, identity and time sources, the software-update route, removable-media handling and the operational controls around all of them. Those controls sit with the deploying organisation, and NERVE cannot assert them on its behalf.

A separate point, stated plainly: Framer hosts this explanatory website. The hosting of this site is unrelated to the deployment boundary of the NERVE product.

Running it

Built for the parts of operations nobody demonstrates.

Large inputs, saturated queues, restores and upgrades are where information platforms usually fail quietly. NERVE treats each of them as a first-class behaviour with a defined durable result.

Streaming large inputs

Large images and packages move through dedicated streaming paths, so a submission is never limited by what a browser can hold in memory. Bytes are committed as they arrive.

Stage-specific scaling

Parser, media, indexing and export workloads scale against their own demand. A heavy media backlog does not starve ordinary document throughput.

Durable queue pressure

When a queue or worker pool saturates, admission slows visibly and the pressure is reported. Work is never dropped quietly to keep a dashboard green.

Backup manifests

Backups are described by a manifest that names what was captured, so a restore can be checked against the objects, metadata and lineage it is supposed to contain.

Historical key availability

Keys used by earlier acquisitions stay available for as long as the material they protect is retained. An old original stays readable after a key rotation.

Fresh restore

A restore is exercised into a clean environment, rather than replayed over a running one, so recovery is proven against the backup rather than against surviving local state.

Upgrade compatibility

Canonical envelopes are versioned, so an upgrade reads records written by earlier versions. Parser packs can be pinned or rolled back independently of the platform.

Indexes are acceleration, not the record

Search indexes are rebuildable acceleration state, derived entirely from canonical records. That is why a restore is judged on resolvable citations rather than on a running process: bringing a service back up is not the same thing as proving the evidence came back with it.

Operational ownership

Fourteen accountable roles, named before go-live.

A deployment only works when every one of these has an owner. We walk through them with you during acceptance, so nothing important turns out to belong to nobody.

Infrastructure

Hosts, networks, storage and the capacity the deployment runs on.

Identity

Accounts, authentication, workload identity and the human authority behind a request.

Sources

Which systems are in scope, and who confirms a route is approved.

Legal and licence authority

Whether the organisation may acquire and retain this material by this route.

Handling

Protective handling bands and the mapping that assigns them.

Retention

Retention schedules, holds and authorised disposition decisions.

Credentials and keys

Secret storage, rotation and historical key availability.

Scanning

The malware scanning service, its availability and its quarantine policy.

Package approval

Which parser packs and capability packs are approved for installation.

Backup

Backup schedules, manifests and the fresh-restore exercise.

Search

Index configuration, rebuilds and publication-state policy.

Provider approval

Which infrastructure and enrichment providers are permitted, and for what.

Incidents

Detection, response and the route for reporting a governance failure.

Release acceptance

Accepting each release into the environment before it carries real material.

Next step

Pick the profile that matches your estate.

An evaluation starts with your operating constraints: the network boundary, the identity model, the sources in scope and the volumes you actually hold. We size the deployment against measured capacity and agree the acceptance evidence up front.

Plan an evaluation

See the architecture