Skip to main content

​
The Sandbox Boundary Layer

A Memory Space is a secure context container that prevents cross-contamination of ideas across your different roles. Rather than querying your entire historical database, assigning a conversation or file to a specific space forces the underlying intelligence to prioritize and reason strictly over the information approved for that project boundary. The main dashboard is broken into three centralized monitoring tabs: Overview, Episodes, and Documents.

​
📂 Switch and Initialize Memory Spaces

At the top of the interface, the active workspace header displays your current environmental folder path.
Memory Space dropdown view highlighting workspace options
  • Universal Context Switcher: Clicking the drop-down carat opens the global space index. Selecting Universal exposes every piece of episodic memory and file node recorded globally across the platform.
  • Isolated Folders (e.g., Writing, Math, Science): Filters the entire application workspace to focus exclusively on that isolated category.
  • + Create memory space: Launches an immediate configuration modal to initialize a new localized sandbox area.

​
📊 Tab 1: Environment Overview (Word Cloud)

The Overview tab functions as a semantic map of your context, displaying a real-time count of total ingested episodes and tracking dates.
Memory Space Overview word cloud layout
The dashboard processes your files to generate an interactive Word Cloud. This visualization surfaces the most prominent themes, repeated terminology, and foundational keywords saved inside the space. The floating control toolbar on the right margin lets you zoom in, zoom out, or refresh the cluster model configuration.

​
💬 Tab 2: Conversation Log (Episodes)

The Episodes tab compiles every individual chat session assigned to this specific environment.
Memory Space Episodes list overview

​
Row-Level Structural Controls

Each episode entry card exposes a set of explicit layout handles and structural utility operations:
  • 8 Vertical Dots Drag-Handle (Left Margin): Click and drag these handles to manually move or re-order the hierarchy of the conversation card within the list layout.
  • Dynamic Expansion Caret (Right Margin Dropdown): Clicking the down-caret instantly expands the card block to show the full chat transcript, systemic content blocks, or metadata tags. Clicking the carat a second time collapses the node back to its clean default list size.
  • Trash Can Icon: Deletes the record. Triggering this menu allows you to either peel the entry out of this individual Memory Space or permanently purge the data node completely from your local device.
  • Assign Icon (Circle & Lines Network Node): Opens an inline popover drawer to map the single item to multiple destinations simultaneously.
Assigning episodes to target memory space tags via popover

​
Advanced Topic Filters

Clicking the Tag icon positioned next to the episode search bar opens the Topic filters side panel:
Topic filtering panel overlay
The system automatically extracts core tags from your records and badges them with a counter tracking how many times that concept appears. Clicking any tag dynamically isolates your search feed, letting you find granular context files instantly without manually auditing long lists.

​
📄 Tab 3: Foundational Assets (Documents)

While the Episodes tab logs dynamic chat sessions, the Documents tab indexes static reference materials—such as PDFs, Markdown files, or data spreadsheets—forming your workspace’s baseline truth.
Memory Space Documents list view layout
The row-level controls inside the Documents tab match the layout rules of the Episodes view:
  1. Move / Organize: Use the left-side 8 vertical dots to adjust list stacking order or drag item relations.
  2. Review / Inspect: Click the down-caret to read metadata summaries and check path directories.
  3. Reassign / Map: Click the Assign icon network node to map the reference document to alternative memory spaces on the fly.
Data Sovereignty Enforcement: All generated word cloud matrices, extracted topic filters, and cross-space relationship indexes are computed entirely on-device, keeping project contexts isolated from any network.

Next Step: Phase 4 — Active Chats

With your sandboxes organized and files mapped, enter the workspace loop to run deep contextual chat comparisons.