Walking the Knowledge Graph
Backlinks, neighborhoods, shortest paths, orphaned notes and the hubs your thinking orbits — the questions text search cannot answer.
Before you start
- second-brain/02-asking-your-library-real-questions
What you'll learn
- Find what links to a note and what it links to
- Pull the N-hop neighborhood around an idea
- Spot orphans and decide whether to integrate or archive them
- Name your hub notes rather than guessing at them
Try it yourself
Open this in Glassy and follow along interactively.
Text search answers “where is this word.” The graph answers “what does this idea touch” — which is the question a second brain exists for. Seven tools cover it, and they work from the index, so they work even when Obsidian is closed.
The four questions
What links here?
“What notes link to my note on onboarding?”
That is glassy_get_backlinks. Inbound links are the honest measure of whether a note is load
bearing. A note with none is either new or stranded.
What does this link to?
“What does the onboarding note point at, and which of those links are broken?”
glassy_get_forward_links returns targets with their resolution status, so you can see a link that
points at a note you renamed.
How are these two connected?
“How is my pricing note connected to the churn research?”
glassy_find_paths returns the shortest wikilink paths between two notes. This is the one that
produces genuine surprises — the connection usually runs through something you forgot you wrote.
Give me the neighbourhood
“What is within two hops of the pricing decision?”
glassy_graph_context walks both directions for N hops. Useful before a meeting: it is the reading
list you would have assembled by hand.
The health questions
| Question | Tool | What it tells you |
|---|---|---|
| What is disconnected? | glassy_get_orphans | Notes with zero links — capture that never got integrated |
| What are my hubs? | glassy_get_central_notes | Hubs by link degree and PageRank — your maps of content |
| Is the graph healthy? | glassy_graph_stats | Nodes, edges, density, orphans, top hub |
Reading the graph honestly
Two traps:
- A big graph is not a good graph. Density without hubs means everything is weakly related to everything. The useful signal is a small number of notes many things point at.
- Link counts are not importance. A note can be linked a hundred times because it is a template. Read the hub before you trust it.
The Knowledge Base shows these counters for you — sources indexed, notes, links, orphan count, and the named top hubs — so you can see the corpus change as you tidy it.
What’s next?
Teaching an agent to remember — the part that turns a graph you query into a collaborator that keeps state.