How the Intelligence Layer Works
Your graph, not the internet
When Relinkia's intelligence layer does anything — suggests a connection, extracts a concept, answers a question — it is working exclusively from your own notes. It does not search the internet. It does not draw on a general knowledge base. It reads your graph and infers from your thinking.
Relinkia doesn't search the internet. Everything it knows comes from what you've written.
This is a deliberate design choice. A system that mixes your specific thinking with generic knowledge makes it impossible to distinguish your ideas from borrowed ones. Relinkia keeps the boundary clear: the intelligence it provides is a reflection of the knowledge you have built.
It learns from you
Every interaction with the intelligence layer is feedback. When you accept a concept suggestion, you are telling the system this fits your vocabulary. When you reject one, you are telling it this does not belong in how you think about this topic.
Rejection is not a failure — it is calibration. A system that only learns from agreement gets generic fast. Relinkia learns from disagreement too, which is why saying no is always worth doing.
When you accept
The system confirms this fits your mental model. Future suggestions in this territory become more targeted.
When you reject
The system learns what you consider unrelated. This distinction is some of the most valuable signal it receives.
Available everywhere
The intelligence layer is not a separate mode you switch into. It is present across all ways you use Relinkia — in the web app, in the Obsidian plugin, in the web clipper. When you capture a note from any surface, the same intelligence applies: concept detection, connection suggestions, and access to chat.
Your graph is always unified. The AI always sees the whole picture, regardless of where you added the note.