> For the complete documentation index, see [llms.txt](https://docs.cortex.aic.io/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.cortex.aic.io/workflow-guides/assurance/offline.md).

# Offline

**Open this area:** In the CORTEX menu, select **Help & node → Node & capabilities**.

Understand the difference between a node without Internet access and a browser disconnected from its CORTEX server.

**Who uses this:** Analyst, Collector, Author, Administrator. **Open in CORTEX:** `/node`. **In-app guide:** `/help?topic=offline`.

**Expected result:** You can recover safely from a connection interruption and identify an operator-owned configuration issue.

## Before you start

* Know your node's approved connection and handling arrangements.
* Keep the browser connected to the CORTEX node even when that node has no Internet connection.
* Know the operator contact for failed processing, configuration or recovery issues.

## What to do

{% stepper %}
{% step %}

## Check effective connectivity

Select the connectivity badge in the header or open Workspace settings → Connectivity.

The explanation combines host and current organisation policy and distinguishes content handling restrictions. Workspace Managers cannot override these policies here.

This is policy availability, not a live Internet-health indicator; each provider and operation still requires its own approval.
{% endstep %}

{% step %}

## Review organisation policy

An organisation administrator can select OFFICIAL-SENSITIVE and record an Online override with a reason in Organisation settings.

This works on an OFFICIAL host, including an Offline baseline. An OFFICIAL-SENSITIVE deployment remains Offline; a physically disconnected host still needs an approved network route.

Provider and content approvals continue to apply.
{% endstep %}

{% step %}

## Stop at a connection warning

If the recovery notice blocks interaction, wait for the approved connection to recover.

Do not repeatedly submit a form or assume the last click failed; a write may already have completed.
{% endstep %}

{% step %}

## Check saved state before retrying

After reconnecting or reloading, reopen the exact record and inspect its latest version/history.

Reconcile your intended change with the actual stored result before saving again. Unsaved form text may be lost when a server circuit expires.
{% endstep %}

{% step %}

## Use the available local path

Continue authorised retained-record work and manual authoring.

Offline mode permits configured local Ollama and approved internal feeds, blocks hosted generation and external feeds, and uses the bundled local map outline or approved internal tiles.

Report absent services to the operator.
{% endstep %}

{% step %}

## Escalate deployment recovery

Give the operator the safe error, affected workflow and relevant record identifier through an approved channel.

Installation, backups, fresh restore and mode changes are controlled operator activities, not browser recovery buttons.
{% endstep %}
{% endstepper %}

## Check the result

* Am I connected to the node rather than merely to the Internet?
* Have I checked the current stored revision after interruption?
* Is the missing service enabled and approved in this deployment?
* Am I avoiding duplicate uploads or decisions while the outcome is uncertain?

## If you get stuck

<details>

<summary>Can I sign in if the workspace connection does not start?</summary>

The sign-in form submits a protected full request to the CORTEX node, so it remains available while the interactive workspace connection starts.

Successful sign-in does not make protected routes available without that connection. Reload after the node recovers, and do not repeatedly submit if the result is uncertain.

</details>

<details>

<summary>Can I edit and save while the browser is disconnected?</summary>

No. CORTEX uses a connected server session and has no browser offline synchronisation store.

Restore the node connection and inspect saved state before continuing.

</details>

<details>

<summary>Why are the map or hosted draft unavailable Offline?</summary>

Offline has no public-tile fallback and rejects hosted generation.

Use the local outline or ask the operator about approved internal tiles or a local model; manual Product authoring remains the appropriate alternative when available.

</details>

<details>

<summary>Should I upload again after an uncertain acquisition?</summary>

First inspect Memory and ask the operator to examine an interrupted acquisition if its outcome is unclear.

Available originals and markers are retained; an immediate duplicate upload can obscure what happened.

</details>

## Responsibility and limits

* Follow local continuity, incident and recovery procedures; do not bypass the approved network boundary.
* An Offline label does not certify physical isolation, recovery success or classification accreditation.

**Reference and scope:** Implementation guidance supporting continuity and reassessment in JDP 2-00 (4th ed.), 2.20 and 6.14–6.24. A deployment mode is not proof of accredited or physically isolated operation.

**Related workflows:** [Preserve an original and review processing](file:///7397340/collection/ingest.md); [Review a model-assisted first draft from CSV](file:///7397340/product/ai-drafts.md); [Explore and retain governed geographic evidence](file:///7397340/analyse/map.md).