Memory map

The memory map turns Kanade’s local chronicle and opt-in markdown roots into a constellation. It is a way to inspect what Kanade can see, not a background indexer for your whole machine.

What appears

Kanade starts with its own local chronicle. If you add memory roots, markdown files from those folders become nodes with labels, excerpts, timestamps, and paths. Wiki-style links such as [[project note]] become strands, and nearby notes can form proximity strands. Generated folders, dependency folders, hidden build output, and large files are skipped.

Set roots in one of two places:

• KANADE_MEMORY_ROOTS, a comma-separated list of absolute markdown folders.
• Settings -> Memory roots, which writes the same intent through the app.

Kanade does not scan $HOME, a cloud-drive folder, or a notes vault unless you point it there.

Controls

The full /memory/map route uses the 3D view when WebGL2 is available.

• Orbit: drag the scene.
• Zoom: scroll or use your trackpad.
• Hover: reveal the node id.
• Select: click a node to focus the camera and update the side rail.
• Constellation: fit the whole map.
• Core focus: move the camera to the core memory node.
• Workspace cluster: move to the first project cluster when one exists.

The side rail shows kind, slug, project, hit count, updated time, path, and a short excerpt. Use View source to move from the constellation back to the plain memory list.

Live updates

The map listens to /api/memory/graph/stream. When the daemon notices a graph change, the route updates without a manual refresh. The chip reads graph live while connected and changes to a reconnecting state if the stream drops.

Fallback behavior

If WebGL2 is unavailable, Kanade shows the 2D graph instead of a blank route. Sparse memory is also intentional: a fresh install may show one mark and a short prompt to add roots in Settings. That is a privacy state, not an error.