Open Graph View at app.xysq.ai/graph.
What you see
- Nodes — one per memory. Colour reflects the cluster the memory belongs to; brighter / larger nodes are hubs with many connections.
- Edges — filaments connecting memories that share a topic, an entity, or a time window. Three types:
semantic(topic overlap),temporal(same conversation or time window),entity(mention the same person, project, or thing). - Hemispheres — your vault’s natural clusters. The left-panel list shows each cluster by size, so the biggest themes in your memory are always at the top.
Switching scopes
The graph supports every bank you have access to:Personal
Your personal memory vault, your confidential bank, and your personal knowledge base.
Teams
Any team vault you’re a member of — both its memory and its shared knowledge base.
Knowledge
Isolate just the ingested docs, links, and code snippets — see how sources cluster without chat memories in the way.
Focus a memory
Click any node. The camera flies in to frame that node with its neighbours, and a detail drawer opens on the right:- Memory text — the full content xysq indexed.
- Context — what agent, session, or document the memory came from.
- Entities — people, projects, and concepts mentioned in the memory.
- Tags — any labels you or an agent attached.
- Connections — every memory directly linked to this one. Click any to jump straight to it without losing context.
Controls
| Control | What it does |
|---|---|
| Drag | Orbit around the galaxy |
| Scroll / pinch | Zoom in and out |
+ / − / ⟲ | Explicit zoom / reset view |
| Click a node | Focus on it and open the detail drawer |
| Click empty space | Clear focus |
| Memory limit slider | How many memories to request from the backend per view |
| LOD threshold | When to simplify distant nodes — bump up for sharper detail, down for higher frame rates |
| Gravity well | How strongly clusters pull their members together |
| Bloom | Glow intensity. Scales dynamically with zoom, so close-up detail stays crisp |
How clusters form
Clusters aren’t manually defined. They come from xysq’s own classification of each memory — memories with similar themes are grouped into the same hemisphere automatically, and the biggest hemispheres sort to the top. A cluster is “Cluster 1” if it’s the largest, “Cluster 2” if second largest, and so on. The ordering updates as you ingest more memories.Entity nodes are a separate bucket.When a memory mentions a person or project that’s important enough to track independently, that entity gets its own node — even though you never explicitly stored it. These live in the Entities hemisphere and connect to every memory that mentions them.
When to use it
- Before a planning session — see what your vault already knows about a project.
- Onboarding a new agent — point it at a cluster so it absorbs the relevant context fast.
- Spotting gaps — sparse regions are topics you’re thin on. Dense hubs are what your vault actually cares about.
- Finding forgotten work — a memory you half-remember is usually one click away via its connections.