Knowledge graph platform

KnowledgeBaseAI

In active development

A knowledge graph engine for business and education: a domain is described as a connected model of entities and relations, and code checks the content and the diagrams. It runs the school mathematics graph in StudyNinja.

The problem

Knowledge about a domain is scattered across documents and people. Finding what you need and seeing how one thing relates to another means piecing it together by hand.

Why the usual tools fall short

Full-text search finds words but does not see the relations between concepts. Retrieval over text fragments (RAG) without a graph gives plausible answers, but it does not show which relations an answer rests on, and such an answer is hard to check.

How it works

A domain is described as a graph: entities, relations and prerequisites (in mathematics, the topics you need before the current one). Before a task enters the graph, it goes through automatic quality checks. Code draws the diagrams and plots, so the same input always gives the same picture. The engine has no vector-based semantic search today: relations are stated explicitly.

Demo

Plot of the parabola y = x² − 4 on a dark coordinate grid: a turquoise curve, axes through the origin, with the roots (−2, 0) and (2, 0) and the vertex (0, −4) marked in yellow.
Code draws the plot of y = x² − 4 as SVG, with no database or external services. The roots (−2, 0), (2, 0) and the vertex (0, −4) are marked. The same input gives the same picture.

Technology

All our products and libraries →

Where it fits

A domain as a graph

Concepts, the relations between them and the order of study are stored explicitly, not in a flat text index.

Grounding for an AI assistant

An assistant can answer from checked nodes and relations in the graph, and then you can see what the answer rests on.

Diagrams drawn by code

The same input gives the same SVG: diagrams and geometry with no black box.

What backs this up

Code draws y = x² − 4: one input, one SVG

The product’s code draws the function with no database or external services, and a repeat render gives the same picture.

Geometry: 100 shapes with no overlaps, and the layout repeats

A layout test: with a fixed random seed, 100 shapes are placed with no overlaps, and the layout repeats. The test lives in the product’s closed repository.

School mathematics graph: 8,599 entities, 9,945 relations

Counted on the authored graph dataset as of 05.10.2026: topics, skills, methods, examples and typical mistakes. 7,464 nodes are loaded into the lab’s working graph database; the size of the graph in production has not been measured separately.

Integration

The engine runs as a separate service and serves data to the product through an API, so it can be set up for another domain. It has not yet been deployed for a client.

Work it out on your own figures

The AVA calculator works out whether automating a process pays off, on your figures, with no form and no price from us. If it does not pay off, it says so.

Estimate the payback

Tell us about your problem

Describe it in your own words. If it cannot be solved or will not pay off, we will tell you before any work starts.

In the reply we will say what we understood, what we need to clarify and whether it is worth doing.

or email us directly: hello@xteam.pro

New write-ups on Telegram