Docs / Memory Visualizer

Memory Visualizer

The Memory Visualizer is where you see and tend everything the agent remembers. It draws your brain, project and research memory as an animated graph you can browse, search, drag, edit and prune. As the brain grows over months, this is how you keep it honest.

What it shows

The visualiser draws your memory as a graph and lets you drill in: from a scope, to its groups, to individual notes. Related memories are grouped into browseable chains — for example, a note about you might appear as You → (your name) → Feelings → notes, and a saved research topic as Topic → VRAM drop → notes. It’s a map of everything stored, arranged so it’s easy to find.

It can start while the server is offline. When the server is running, the search box queries your live memory and pulses the matching nodes, so you can watch the agent’s recall in action.

Launching it

Reading the graph — node colours

ColourMeaning
Cyan / blueBrain scope (you)
GreenProject scope (this repo)
Amber / yellowResearch scope, and group nodes
PurpleSources and the topic / owner chains
Red / pinkIndividual memory notes

The three scopes — brain, project and research — are explained on the Memory & Identity page.

Navigating the canvas

ControlAction
Mouse wheelZoom in / out (right down to a thumbnail and far in, so a brain with thousands of nodes stays navigable)
Drag empty canvasPan around the graph
Double-click a nodeZoom to and centre on it

Working with nodes

ControlAction
Click a nodeSelect it — detail appears in the right panel
Drag a nodeMove it and everything below it together
Shift + dragMove only that node; the rest stays put — for fine-tuning one node in a cluster
Ctrl + dragSame as Shift — move only that node
Right-click a nodeContext menu with a Delete option
DeleteDelete the selected node (same confirmation as right-click); ignored while a text box has focus
Ctrl + ZUndo the last add, edit or delete from this session

The toolbar

ButtonAction
RefreshRe-read all memory and animate any changed nodes
UndoRestore the most recent add / edit / delete from this session
Restore SnapshotBrowse older recovery snapshots and restore one — even after a restart
FitClear pinned positions and reset to the auto-arranged layout
Data RootPoint at a different Riverforge data folder (e.g. another machine’s exported data)
ProjectSwitch to a different project and restore that project’s saved layout
Fire QuerySearch brain + project memory for the text in the box; matches pulse and the top hits show in the detail panel

Searching your memory

Type a phrase and hit Fire Query. The visualiser searches your brain and the current project; matching nodes light up with how well they matched, and the best hits appear in the detail panel. With the server running it uses live semantic search; otherwise it falls back to a simple keyword match so you can still explore offline.

Tidying the layout

When you drag nodes, their positions are saved automatically and restored next time — kept separately for the brain and for each project, so switching projects brings back that project’s arrangement. Press Fit to discard your positions and return to the default layout. Tidying the layout never changes the underlying memory — it only moves things around on screen.

Safe edits & recovery

Use the visualiser to correct a bad fact, remove stale project context, or clear out test data — without touching anything by hand. Every change you make (adding, editing or deleting) saves a recovery snapshot first, so nothing is lost:

The visualiser is a separate app from the small memory panel inside the VS Code extension. The extension panel is a quick status view; the visualiser gets the whole screen for browsing, searching and inspecting your memory.