Bidirectional links, mention blocks, AI-suggested edges, bridge detection between clusters of your work, and folder-level intelligence — built into the graph view itself. Not a decoration of your notes. A reading of them.
§ 2 — The Constellation
Three clusters of work — product strategy, engineering, hiring & people. Six bridge notes span clusters. AI-suggested edges flow through the gaps your manual links missed. Hover any note to highlight its neighbors; click to inspect. The workspace overview on the right updates as you explore.
§ 3 — The Connections
Other graph views show you the lines you drew. Knovya's graph shows those lines, plus the structure beneath them — the relationships you didn't draw, but are there.
Connection 01
Wiki links, mention blocks, dependencies.
Every wiki link, mention block, or explicit dependency you write becomes a typed edge in the graph — automatically, bidirectionally, instantly. No backlinks panel to maintain by hand.
Connection 02
Connections you didn't draw, but are there.
Knovya computes semantic similarity between every pair of notes and surfaces the strongest unlinked pairs. The threshold is calibrated to avoid noise — a suggestion means real overlap, not coincidence. Confirm to make it solid, dismiss to mute it.
Connection 03
Notes that span clusters of your work.
When a note connects to two or more clusters of your work — engineering on one side, strategy on the other — Knovya marks it as a bridge. Granovetter showed in 1973 that weak ties carry the most novel information. Bridges are exactly that.
Connection 04
Folders, with names that come from the notes inside them.
Each folder gets an AI-generated label of three to five words — drawn from its highest-ranked notes. Hub notes inside surface to the top. Knowledge gaps — folders that are thinner than the rest of your archive — get a quiet flag in the side panel.
§ 4 — The problem
Scene 01
You open Obsidian's graph. It's a beautiful constellation of every link you've manually drawn. The connections you forgot to draw are invisible. The cluster labels don't exist.
Scene 02
You open Logseq's graph. It bubbles around. You can filter by tag. Nothing tells you which notes bridge two parts of your thinking, or which note is going cold.
Scene 03
You open Roam's graph. A wall of grey nodes. The structure is there, somewhere. You squint. You go back to writing, having learned nothing about the shape of what you've written.
A graph that just shows what you already know is a wall poster. Knovya's graph reads what you've written — finds what you missed, names what you've clustered, and points at the threads about to fray.
§ 5 — Lineage
Knowledge graphs aren't a Knovya invention. They've been a research field for fifty years and a consumer product for ten. What's new is bringing the AI-augmented version down to the scale of one person's notes.
1973
A sociologist publishes the paper that defines bridge nodes — the connections that span otherwise separate communities and carry the most novel information.
2012
Knowledge graphs become a consumer concept. Globally important entities — places, people, things — connected by typed relationships, behind every search result.
2019
Two researchers at Google London publish the position paper that defines personal knowledge graphs as a distinct field. ICTIR 2019. The user, not Wikipedia, is the center of gravity.
2020 — 2024
Roam, Obsidian, Logseq pioneer the graph view in note apps — visualizing the wiki-links you draw by hand. Beautiful. Static. Decoration on top of writing.
2026
AI-suggested edges. Bridge detection. Cluster intelligence. Heat. Whispers. The first personal knowledge graph that reads your notes, not just plots them.
§ 6 — First mover
Obsidian · Roam · Logseq
Pioneered the manual graph view. Beautiful. Render every link you drew. Don't suggest the ones you missed, don't detect bridges, don't label clusters, don't track what's going cold. Pure visualization.
InfraNodus · Kumu · TheBrain
Third-party text-network analysis tools that bolt onto your vault. Gap detection. Idea generation. Useful — but a separate import, separate UI, not native to where you write.
Notion · Mem.ai · Reflect.app
Note apps with no real graph at all. Backlinks panel, list of mentions. The structure of your thinking is implicit, not visible.
Knovya — Knowledge Graph
Manual edges, plus AI-suggested edges, plus bridge detection, plus cluster intelligence, plus heat. Native to the editor. The first AI-augmented personal knowledge graph that ships out of the box.
§ 7 — Surfaces
The graph isn't a screen you open once. It surfaces inside the note you're reading, in the agentic tools that read your work, and in the quiet whispers when something interesting moves.
Surface 01 · Inline
Open any note — a small graph of its 1-hop neighbors appears in the sidebar. Click any neighbor to fly across the graph and land there.
Surface 02 · Workspace
A dedicated workspace view. Filter chips compose. Semantic zoom — far for nebula, mid for structure, near for detail, deep for full content. Smooth past a few thousand nodes; cluster-collapse beyond.
Surface 03 · Agentic
Connected agents call knovya_links action="map" via MCP — they get the same graph as a structured tree, with hops, follow types, and direction filters.
Surface 04 · Proactive
Nine kinds of structural insight surface as quiet cards: bridges going cold, hub notes, knowledge gaps, similar-but-unlinked pairs, evolution chains needing a successor.
"DB migration runbook" connects Engineering and Customer Research, but it hasn't been touched in 47 days. Worth a glance — bridges decay first.
Open in graph →§ 8 — Bonded with
Knowledge Graph is the structural layer of Group II — Intelligence. Every other element in the group reads from it or writes into it.
Every wiki link, mention, agent badge becomes an edge — automatically, bidirectionally, instantly.
Reads graph signals — hub-ness, bridge-ness, link weight — to rank what surfaces first.
Uses the graph as a context boost — neighbors of strong matches lift in retrieval rank.
The linked-evidence layer of every note's envelope is the local graph rendered up close.
A personal knowledge graph is a structured map of your own notes, ideas, and references — and the relationships between them. Where general knowledge graphs (Google, Wikidata) catalog globally important entities, a personal knowledge graph captures the entities that matter to you: your projects, your decisions, your interview subjects, your reading. The concept was formalized in Balog and Kenter's 2019 ICTIR paper at Google. Knovya is the first note app to make it native, AI-augmented, and interactive.
Obsidian visualizes the links you've already made by hand. Knovya does that, plus three more layers — AI-suggested edges between notes that are semantically related but not yet linked, bridge node detection that highlights notes spanning multiple clusters of your work, and cluster intelligence that auto-labels your folders. The graph isn't decoration; it surfaces structure you didn't know was there.
Yes. Export as JSON (full graph payload — nodes, edges, semantic edges, clusters, evolution chains), as GraphML for Gephi or Cytoscape, or as a static SVG snapshot. Your data is portable; the graph isn't a lock-in.
Manual edges — wiki links, mention blocks — appear instantly the moment you save a note. AI-suggested edges and cluster labels recompute in the background, typically within a few minutes for a workspace of a few thousand notes. The graph you're looking at is never older than your most recent draft.
Knovya computes semantic similarity between every pair of notes using embeddings, and surfaces the strongest pairs that you haven't explicitly linked. The threshold is calibrated to avoid noise — a suggested edge means the two notes really do share substantive content, not just a passing word in common. You always see the suggestion as a dashed line and can confirm it (turning it into a solid edge) or dismiss it.
Yes — by tag, folder, type (plan, decision, journal, framework, bug), status (draft, active, completed, deprecated), heat (hot/warm/cold), or NoteRank score. Filters compose: you can ask for hot active plans across the engineering folder and the graph re-renders to that subset.
The full-detail graph stays smooth up to a few thousand nodes. Beyond that, Knovya progressively switches to cluster-collapsed views — folders shown as single nodes, expandable on click — so a workspace of ten thousand notes still navigates without lag. The system tells you which view you're in; nothing is hidden.
Twenty notes in, the graph is sparse. Two hundred in, it has shape. Two thousand in, it's the kind of structural insight a team of researchers would have built for you, manually, over a year.