What Devexana is

Devexana runs on an appliance on your network and answers staff questions with citations. In this release, Devexana does not change network-device configuration.

In production use: Freshservice helpdesk reads; Aruba/HP AOS-S (ProCurve) switch reads over SSH and SNMP; Barracuda CloudGen reads over SSH, firewall syslog alerts and the daily brief; language models on the appliance; audit log
Available: Cited answers in chat

On this page

The appliance
What your staff can do
How Devexana treats your systems
Where the AI runs
What Devexana does not do
Next steps

Devexana reads data from your helpdesk, network switches and firewall. It keeps that data in one encrypted database on the appliance.

Diagram titled How Devexana works. On the left, your sources: scheduled reads of switches, firewall and helpdesk, and listeners for syslog, SNMP traps and NetFlow. In the centre, the Devexana appliance with the encrypted store of your records and the local models that search and answer. On the right, your staff receive cited answers, network views and alerts. A dashed arrow leads to a dashed box titled Optional cloud model (Devexana ZDR on AWS). The box says that it runs only when Devexana sets up the connection, that customers do not enter a model key, and that other cloud features are off by default. — Data comes in from read-only sources and listeners. Answers, network views and alerts go to your staff in a web browser. Open full size

The appliance

The appliance is an NVIDIA DGX Spark computer that you place on your network. Devexana and its language models run on it.

Devexana reads your helpdesk, switches and firewall through read-only sources.
Devexana stores what it reads in one encrypted database on the appliance.
Your staff use Devexana in a web browser on the local network.

How it works shows each step from a source system to an answer.

What your staff can do

Task	What Devexana gives you	Read more
Ask a question in plain language	An answer with numbered citations that open the stored records	Ask questions, Cited answers
Look at the network	A campus map from your site file, neighbour links from LLDP and CDP data, and switch ports	Network views: campus map and switch faceplate
Review firewall alerts	Alerts and a daily brief built from firewall syslog	Firewall alerts and the daily brief
Review security findings	Security findings that a person confirms or dismisses	Security findings
Run scheduled reports	Routines that produce PDF files on a schedule	Routines (scheduled reports)
Give auditors evidence	A view of the audit log, a chain check, and one evidence pack to download	Audit log and the auditor evidence pack

Screenshot of the Devexana chat. The question asks how the team fixed a council-chamber projector last time. The answer names the fix and cites one helpdesk ticket as [1], with the badge updated today and the ticket date. The Evidence section repeats the marker after each claim. — An answer in chat, built from one helpdesk ticket. The citation opens the stored ticket. Fictional sample data. Open full size

How Devexana treats your systems

Devexana reads your systems through read-only sources. When a source declares a write operation, Devexana rejects it during loading. Each read function stays off until an administrator approves it.

In this release, changes to network-device configuration are disabled in code.

Devexana records sign-ins, setting changes and read-function approvals in hash-chained, append-only logs on the appliance. Records of cloud sends depend on the feature (Outbound connections: the complete list). The audit log describes the logs.

Where the AI runs

By default, Devexana produces answers with models that run on the appliance. A deployment can run with no cloud model. In production, customers do not bring their own model keys. If a deployment uses a cloud model, Devexana provides it through Devexana's zero-data-retention (ZDR) agreement on AWS. The optional networking assistant uses a cloud model only when Devexana has set up the cloud connection for that deployment. Other cloud features are off by default. Devexana records each networking assistant request in the egress log on the appliance.

Where the AI runs lists the models and cloud features. Outbound connections lists outbound connections, their defaults and their log locations.

What Devexana does not do

What Devexana does not do (limitations) lists every limit in one place.

Next steps

Read How it works to follow a question from source to answer.
Read Logs, read-only access and evidence for the differences between logs alone and read-only access.
If you are evaluating a campus deployment, read Evaluating Devexana for a campus network.
Read Supported vendors and protocols for the support tier of each vendor.

How it worksDevexana reads sources, stores records on the appliance, and uses those records to answer questions with citations.
What Devexana does not do (limitations)This page lists unsupported capabilities, release restrictions, limited functions and features that are off by default. Each entry gives an alternative or explanation.
Read-only sourcesDevexana sources only read your systems. Separate features can write to business systems. In this release, Devexana disables network-device configuration changes in code.
Where the AI runsModels on the appliance, optional cloud features, outbound data and logging.
Cited answersCitation markers, source records, answer checks and citation limits.
Logs, read-only access and evidence: what each level gives youStart with logs and files without device credentials. When you need more context, add read-only device access or other sources.

Applies to Devexana release 0.3.182. Page updated 2026-10-20. Questions or corrections: support@devexana.com.