Solution

Private AI knowledge assistant.

A private, self-contained AI assistant for the knowledge you choose to share with it. Spin it up in your own infrastructure — cloud, on-prem, or ours — give it a defined set of documents, and get fast, cited answers without your data ever leaving your environment.

What it does

Three things, well.

Answer staff questions instantly

Plug in your policies, runbooks, contracts, regulations, or product manuals. Staff get accurate answers in seconds — not "I'll get back to you tomorrow".

Cite the source document

Every answer points to the document and page it came from. People can verify what they're reading. Auditors can trace the trail.

Never leak to a public LLM

The assistant runs in your cloud, your data centre, or on infrastructure we manage on your behalf. Your documents never train a third party's model.


How it works

Own your AI.

Three steps. Your data stays yours.

  1. Point it at your documents.

    PDFs, Word files, SharePoint, intranet pages — whatever your team already relies on. We curate the source set with you so the assistant only sees what it should.

  2. It learns inside your environment.

    Your documents are indexed in a private database that runs in your environment — your cloud, your on-prem, or hosted by us. Nothing is sent to a third-party model provider unless you choose to.

  3. It answers — with citations.

    Staff ask questions in plain English. Every answer shows the source document and page number, so the reader can verify it in one click. Wrong answers become improvable, not invisible.

Built on retrieval-augmented generation (RAG). Runs entirely inside your network when required.


Capabilities

Everything an enterprise expects. Nothing a buyer doesn't.

Knowledge Assistant ships with the controls regulated and security-conscious organisations require — identity, audit, content safety, and the ability to run entirely inside your own network.

Differentiator

Data Curation & Model Tuning

A productised assistant is only as good as the knowledge it's given. As part of every deployment, our team reviews your source documents for accuracy, currency, and contradictions before ingestion; sanitises sensitive content (strips or redacts PII, commercially sensitive figures, out-of-scope material); tunes retrieval (chunking strategy, embedding model selection, re-ranking for your document types and query patterns); optionally fine-tunes the language model on your domain (regulatory framework, internal terminology, house style); and establishes an evaluation harness — a test set of representative questions with expected answers — so you can measure quality before and after every change.

This is the difference between a chatbot and an assistant your staff actually trust.

What's in the box


Deployment models

Three ways to run it.

Same software. Choose where it lives and who operates it.

  1. 01

    In your cloud

    Volorai installs into your Azure, AWS, or GCP subscription using Terraform modules we provide. You pay compute. Volorai charges a fixed subscription per platform instance. Your data never leaves your environment.

  2. 02

    In your data centre

    Kubernetes (AKS, EKS, GKE, OpenShift, vanilla K3s) or Docker Compose on a single GPU host. Air-gapped deployments supported. Suitable for regulated, sovereign, or sensitive environments.

  3. 03

    Managed by Volorai

    We run it on UK-sovereign infrastructure on your behalf. Single invoice, faster onboarding. Your documents are processed in a dedicated, isolated environment — ask us for the data processing agreement.


Tech Stack

Built on proven open-source foundations.

No proprietary lock-in. Every component is replaceable, auditable, and deployable in your environment.

Web tier

  • NGINX
  • React

APIs

  • Node.js
  • Express
  • FastAPI

Inference

  • Llama.cpp
  • vLLM
  • TurboQuant

Queue

  • BullMQ
  • RabbitMQ

Stores

  • PostgreSQL
  • MariaDB / MySQL
  • LanceDB
  • Chroma

Runtime

  • Docker
  • Kubernetes

SCM

  • GitHub Enterprise

OS

  • Ubuntu
  • CentOS
  • RHEL

Proof

In use today.

Knowledge Assistant has been deployed for a UK professional body and is in preparation for a regulated organisation deployment.


Frequently asked questions

  • Does my data leave my organisation?

    On the customer-hosted deployment model, no — your data stays in your cloud or on-premises infrastructure throughout. On the Volorai-hosted model, data is processed in a dedicated, isolated environment; ask us for the data processing agreement.

  • How long does deployment take?

    A standard customer-hosted deployment to an existing Azure, AWS, or GCP subscription typically takes 1–2 weeks from access grant to live assistants. On-premises or complex network configurations take longer.

  • Can we configure multiple assistants for different teams?

    Yes. The platform is designed for multi-assistant operation. Each assistant has its own knowledge corpus, access controls, and system configuration. You can have separate assistants for legal, finance, IT, and client support — all on the same platform instance.

  • What AI models does it use?

    The platform supports any OpenAI-compatible API — including Azure OpenAI, Anthropic, and self-hosted models. Model selection is configurable per assistant. For air-gapped or sovereign deployments, we can configure fully local inference.

  • What does the subscription include?

    Volorai installation, configuration, and ongoing support. Platform updates. Access to Volorai for configuration changes, troubleshooting, and new assistant onboarding. Compute is separate (yours on customer-hosted; included on Volorai-hosted).

  • How does this compare to Microsoft Copilot or similar SaaS options?

    Microsoft Copilot and similar SaaS AI tools send your data to a cloud provider's shared infrastructure and are priced per user. Knowledge Assistant is private to your environment, priced per platform (not per user), and configured specifically for your documents and access patterns — not a general-purpose assistant. The cost comparison is significant at scale.

Bring a use case. We'll show you the deployment.

A 30-minute demo using your own example documents. We bring a working environment; you bring a folder of representative content. By the end you'll know whether Knowledge Assistant fits.