Architecture

For people evaluating the product.

Four services and a database, one shared record.

A NovusLens twin runs on your own infrastructure. Four services and a database keep one shared record, and the map, the measures and the answers are all built from it.

NovusLens Architectural Overview6:08 running time, with English captionsThe Architectural Overview follows records from import to an answer people can check, then shows the infrastructure underneath. It runs for six minutes, with English captions, and generated voices narrate it.

  1. Company records

    Sources

    Your own records supply everything the twin shows: charts, policies, supported exports, operational systems and AI agent declarations.

  2. Source connectors

    Connect

    Imports and connectors read your records and keep each original, so a relationship can be traced back to its source.

  3. Docker containers

    Application infrastructure

    Docker runs the services on your infrastructure, and Mycelium, from TathataSystems, keeps an ordered log of changes to the structure.

  4. Postgres database

    Data

    A Postgres database holds the map, the evidence and the Knowledge library, so an answer can point to the stored record behind it.

  5. Reasoning and measurement

    Compute

    The reasoning service checks who may see what, calculates the measures and calls a configured AI model when a task needs interpretation.

  6. Web application

    Display

    The web application shows the map, findings and answers, and lets people add evidence and record their decisions.

The complete system diagram is drawn for a wide screen. Open it on its own to explore each path.

Records enter through imports and connectors. Original files go straight to the database. Changes to the structure pass through Mycelium’s ordered log, and NovusLens then updates the views people read. In the full diagram, select an input to follow its path.Open the full-width diagram (opens in a new tab)
  • Hosting

    Runs on your own infrastructure

    You keep your records, the Knowledge library and the twin. Running the twin yourself keeps that capability in your hands. On its own, it does not make every record private or compliant.

  • Where AI runs

    Models help interpret, people decide

    You choose where AI models run: in a local container in your deployment, or at an endpoint you configure. People can browse and import records without a model. Models return proposals and interpretations. They never write to the database or create measured scores.

  • The boundary

    People own the decisions

    NovusLens maps, surfaces and advises. It does not operate the systems it reads or your AI agents, and it never issues a compliance verdict. Changes a model proposes wait for a person to confirm them, and people decide what happens next.

Let’s talk

What needs
to change?

Talk to us about one consequential decision, such as replacing a service, changing a supplier or handing work to an AI agent.

Book a conversation (opens Microsoft Bookings in a new tab)Or contact us directlytyra.suleman@novus-i2.com

Tell us about your organisation and the challenge you are working through.