Cluster View

The Clusters view shows the Cluster Graph, a spatial map of indexed documents. Each document is a node on a force-directed canvas. The view helps you see related material without relying on folder paths alone. Hover or focus the Clusters view selector and press F1 to open a prepopulated search of these User Docs for the SK Memory and relationship features behind the map.

Relationship Data

ThinkForge builds cluster relationships from several signals:

  • Semantic similarity
  • Shared document tags
  • Explicit manual links
  • Folder proximity

These signals influence node proximity and relationship strength. The map is exploratory. It is not a literal folder tree, and nearby documents do not necessarily share an explicit manual link.

Open Cluster settings and use the controls under What makes documents related?:

  • Content Similarity gives more or less influence to similarity between the documents' indexed content.
  • Shared Tags changes how strongly common document tags pull documents together.
  • Same Folder changes how much sharing a project folder contributes.
  • Manual Links changes the influence of relationships you explicitly saved.

Raising a weight makes that signal count more when ThinkForge decides whether to draw and strengthen a relationship. Lowering it makes the signal count less. A weight cannot create a relationship signal that is not present: for example, raising Manual Links does nothing for two documents that have no saved manual link. A weak relationship can disappear when its combined signals no longer clear the graph's display threshold.

Grouping and Filtering

Bubble grouping has four modes:

  • None — no group bubbles
  • Tag — group documents by tags
  • Folder — group by project folder
  • Date — group by date range

File type is a separate filter, not a grouping mode.

Date filters include Today, Week, Month, a custom date range, and All. The search controls can independently enable keyword matching against names and paths, semantic matching, and tag matching.

Visibility and Scale

The Visibility slider filters nodes by importance score. The Max slider caps the rendered node count from 10 through 500, defaulting to 150. ThinkForge keeps the highest-scoring documents that pass Visibility before applying Max. When the graph is crowded, raise Visibility or lower Max before adding more filters.

Search Relevance and Title Match

Open Cluster settings and look under What makes a document important?

Search Relevance changes how strongly results from the enabled Keyword, Smart, and Tag search modes contribute to a document's importance. Title Match separately favors a document whose title matches the search. Raise one of these weights when that kind of match should matter more; lower it when it is overpowering other useful documents.

Recency and Connectedness

Recency changes how strongly recent document access contributes to importance. Connectedness favors documents with more visible relationships compared with the other documents currently in the graph.

With a search entered, direct matches stay above related neighbors and background documents. Without a search, Recency and Connectedness determine importance. Raising Visibility can therefore hide documents that score poorly under the current settings even though they are indexed.

Recency direction modes and Date focus

Recency direction modes change how recent document access contributes to Cluster document scoring. It uses your access history, not the file's modified date. A document you have not recently accessed receives no recency contribution.

In Cluster settings, choose a Recency direction:

  • Recent first favors documents accessed most recently.
  • Oldest first favors older entries in your access history.
  • Date focus favors access dates near the middle of the chosen Focus from and Focus to range.

The focus date controls appear when Date focus is selected. Changing the start or end shifts which access dates contribute most strongly; it does not filter the graph to that exact period. Use the date filters in the Cluster toolbar when you want to exclude documents outside a date range.

Cluster settings persist for the project, including grouping, filters, search modes, scoring weights, recency direction and focus dates, Visibility, Max, and info-panel visibility.

Working with Nodes

  • Click a node to focus it. The view shows up to seven of its strongest qualifying relationships and updates document details.
  • Drag a node to reposition it while exploring.
  • Double-click a node to open the document.
  • Right-click a node for its available actions.

Node actions include Open, Copy Title, Pin or Unpin, Add to Shortcuts, add the document or its connected group to a workspace, Reveal in Explorer, Copy Path, Add to Smart Clipboard, Copy Connected Group, copy or move the file to another project, Run Tool, and Freeze or Resume Layout.

There is no separate context menu on an empty cluster bubble for isolate, pin-cluster, or filter-to-cluster commands. Group-level actions are available from a document node's connected-group commands.

Layout Controls

  • Freeze / Resume pauses or restarts the layout simulation.
  • Refresh reloads cluster data.
  • Show Info / Hide Info toggles the cluster detail panel.
  • The settings control adjusts importance, relationship, and physics weighting.

The simulation sleeps automatically once the layout settles. Interacting with the graph wakes it again.

Standalone Window

Cluster View can also open in a resizable standalone window, including from Power Strip. The standalone window is reused rather than recreated on each open, so its current layout stays available between openings.

Requirements and Troubleshooting

Cluster View uses indexed documents. If a recently added document is missing, confirm that its type is indexable and allow project indexing to complete, then use Refresh. Filters, the visibility threshold, and the node cap can also hide an indexed document.