PERPETUALTECHNOLOGIES

Home
The product
NeuralVaultThe agent harness, in one page
DownloadOne installer, Windows
Compare editionsFree, Personal and Business side by side
PersonalFor one person
BusinessFor a team
The company
What we doAI integration, training, and the hardware under it
About usWho we are and how we work
Get in touchGot a problem? Talk to us directly
Support, a question about your account, or a project you want to discuss. It all reaches a person, and we respond within 2 business days.
Contact us
NeuralVault
Editions
DownloadOne installer, Windows
Compare editionsFree, Personal and Business side by side
PersonalFor one person
BusinessFor a team
What is changing
ChangelogEvery released version, newest first
RoadmapWhat is being built next
Start here
What NeuralVault isThe agent harness, in one page
SupportFixes, guides, and how to reach a person
AvailableNeuralVault 2.0
A rebuilt app, one account, and Cortex that writes to your vault when you say yes. Available now as a Beta. Public early access. Expect bugs. Keep a backup of your vault.
Read about 2.0
Services
What we do
True AI integrationAI wired into the tools you already run
On-premise AI, run for youThe hardware private AI actually runs on
In development
AI Proficiency TrainingTraining the people who have to use it
Not sure whichStart with what is costing you now
A consultation costs nothing and there is no obligation. We would rather tell you a project will not pay for itself than sell it to you.
Arrange a consultation
About
Company
About usWho we are and how we work
BlogHow the build is going, written as it happens
Legal
Privacy PolicyHow we handle your data
Software LicenceThe terms NeuralVault is licensed under
Website TermsTerms for using this site
Who you deal withThe person doing the work
No account manager and no ticket queue. When you get in touch, that is who answers.
Get in touch
Support
Fix it yourself
GuidesSetup and how-to, written against the shipping build
Questions and answersEvery known problem and its fix, in one list
When that is not enough
Report a problemTell us what went wrong. No GitHub account needed
Get in touchA person reads it. No ticket queue
Report a security issueThe disclosure policy and the address to use
Using an assistantThe guides are machine-readable
Every guide is also served as one plain-text document, generated from the same source the pages render, so the two can never disagree.
Open /support/guides.md

Run a model on your own machine with Ollama

Install Ollama, pull a model, and NeuralVault finds it at localhost:11434 with no API key and no account.

Last checked against the app on 6 September 2026 · View as Markdown

Steps

Install Ollama from ollama.com and let it finish starting. It runs as a background service and listens on http://localhost:11434, which is the address NeuralVault uses unless you change it.
Pull a model. In a terminal, "ollama pull llama3.1" is a reasonable first choice; smaller models answer faster on modest hardware.
Open NeuralVault and go to Settings, then Intelligence, then API keys. Ollama appears under "Providers that run locally" and asks for nothing, because there is no key to give a server on your own machine. The row says "Reached at" with the address, and a status line such as "Ollama is answering".
Open the model picker in Chat and choose the model you pulled.
If Ollama runs on a different computer instead, such as a Mac or a Linux box, see the guide on Ollama on another machine. NeuralVault itself runs on Windows only.

How to tell it worked

Send one message in Chat. If it answers, the whole path is working: nothing left your machine to do it.

If it does not work

Ollama is installed but no models are listed.: Installing Ollama does not download a model. The Ollama row says "Ollama has no models installed". Run "ollama list" in a terminal, and if it is empty, pull one.
The app cannot reach Ollama.: The Ollama row says "Ollama is not running" when nothing answers at its address. Start Ollama, then reopen the model picker. No setting in NeuralVault can start it for you. See the guide on local model problems for the other states.
Answers are extremely slow.: That is a hardware answer, not a configuration one. A model larger than your available memory will be swapped to disk. Try a smaller variant.