Brain · Knowledge graph explorer

Knowledge graph explorer

A direct view over the persisted graph store. Nodes by type, edges by source, and commit or experiment lineage. Read-only and LLM-free: it renders what is already stored, it never extracts or approves.

Start here if you want to trace what a session shipped, what it broke and what now guards it: Accountability ledger.

Where to find it

  • Localhost: /graph-explorer.html?repo=<id>
  • API: GET /api/graph/store?repo=<id> and GET /api/graph/lineage?repo=<id>&node=<id>
  • Keyboard: ⌘ K then graph
  • Navigation: Attribution in the sidebar, then Knowledge Graph under All tools, in the Pull requests, blame and graphs group

What it does for you

Nodes by type, counted across every source.One card per node type (session, commit, pr, file, incident, lesson, defect, experiment, minted entity). Click a type to list its nodes, then jump straight to a node's lineage.
Edges grouped by source, so you see what a human approved.Each source (accountability-fused, approved or minted, structural imports, commit ancestry, experiment) is its own group, badged by confidence band. An approved or minted edge expands to its provenance: who approved it, from which proposal, and the source excerpt.
Trace a commit or experiment back to its roots.Pick or type a node id and the lineage view shows its direct children, the commit frontier (leaves with no child), and the ancestry path to the roots as a left-to-right tree. Every node in the tree is clickable to re-trace from there.
Honest-empty on a cold store.A repo with no graph yet shows an empty state, never a fabricated node. Press Fill from this repo to read the repo's git history, eval runs and event store into the store, or wait for the daemon's next brain cycle, which runs the same pass. For human-approved semantic edges and minted entities on top, run the Semantic layer panel on the Accountability ledger and approve what it proposes.

Read more

Last updated