Skip to content

The semantic graph

Altretta doesn't just draw a picture of your links — it stores your vault in a real semantic graph database, as subject–predicate–object triples you can query and traverse. This is the engine underneath everything else in this section, and the reason Altretta's graph does things a wikilink graph can't.

Put simply: Obsidian's graph is a picture of your links. Altretta's graph is a queryable database of your knowledge. One is something you look at; the other is something the app — and your AI — can reason over.

A wikilink records one fact: note A links to note B. Altretta captures far more. Your notes, their properties, their tags, and their links become triples — small statements of the form subject → predicate → object (for example, this note → mentions → that project). A whole vault becomes a network of these statements, held in a graph database rather than inferred on the fly from text.

Because the connections are stored as structured data, Altretta can traverse them: follow relationships across many hops, ask which notes relate to which, and pull back exactly the neighborhood that matters for a task.

What it unlocks

The semantic graph is the shared foundation for Altretta's differentiators:

CapabilityHow the semantic graph powers it
Grounded, cited AIThe AI traverses the graph to find the exact passages that bear on your question, then answers only from them — with citations.
Maps & ManualsAuto-generated overviews are built by walking the graph's structure, not by guessing from filenames.
Code Impact RadiusFor code, the graph of symbols and cross-file relationships answers "what breaks if I change this."
ProvenanceBecause every fact is a discrete statement, changes to it can be recorded and signed.

It's a real graph, so history is real too

Changes to the graph are captured in Altretta's signed, verifiable history. When you edit or remove a note, the graph records that as a signed action — a retraction rather than an overwrite — so provenance survives what you take away.

You don't have to think about triples to benefit from them — write Markdown and links as usual, and the semantic graph builds itself. It's what makes the AI trustworthy and the history verifiable, quietly, underneath.