CONTEXT_OSALPHA

Memories

Record durable knowledge, find relationships, and explore workspace context.

Purpose

Memories preserve decisions, learnings, and context that need to survive a conversation or work cycle. The area combines a list, dashboard, and graphs to help people and agents recover relevant knowledge and discover relationships between memories.

Prerequisites

  • Be authenticated and have an active team.
  • Have access to the Memories module.
  • To create, edit, or delete, have the corresponding workspace permission.
  • When recording a memory, have concrete information: a decision, constraint, preference, discovery, or outcome.

Key concepts

Memory

A memory has a title, content, tags, importance, status, access count, and creation and update dates. Content can be written in Markdown.

Importance can be Low, Medium, High, or Critical. It helps prioritize reading and visualization; it does not replace judgment about relevance to a specific question.

Statuses and relationships

Memories can be active or archived. Relationships connect memories through types such as similar, extends, caused by, supersedes, contradicts, and depends on. Relationships can appear in the detail view and graphs.

The current area does not expose a separate module called Intelligence. The user experience for this knowledge is concentrated in Memories, including browsing, the dashboard, graphs, and links with skills.

Graph scope and sectors

Graph V2 can switch between personal memories, team memories, or both. It can also filter by sectors, minimum relationship strength, and invalidated memories. The visualization can group nodes by sector, sessions, or community.

Create a memory

  1. Open Memories and select New Memory.
  2. Enter a title that identifies the information without depending on the conversation's context.
  3. Write the complete content: decision, reason, impact, and conditions for change when applicable.
  4. Set the importance.
  5. Add at least one descriptive tag.
  6. Select Create Memory.
  7. Open the memory you created and confirm its content and available relationships.

Creation requires a title, content, and at least one tag in the current form. Memories without these elements are harder to retrieve and do not pass through the interface's intelligent save flow.

A useful format

Title: Authentication flow decision

Decision: require MFA for the organization.
Reason: reduce risk in administrative actions.
Impact: new users must configure TOTP before accessing the workspace.
Review when: an equivalent corporate provider is supported.

Browse and edit

Dashboard

The dashboard summarizes total count, active and archived memories, importance, recent access, tags, and activity. Use it to identify areas with concentrated knowledge or little usage.

Browse

In Browse, search by text, tags, importance, status, and order by update, creation, or priority. Open a memory to read its full content, edit properties, and inspect direct relationships.

Graph

The Graph view opens a focus around a memory. Use it to follow nearby connections and navigate to another memory.

Graph V2

The default Memories entry directs to Graph V2. In this view:

  1. Choose Graph or List.
  2. Filter scope and sectors.
  3. Adjust the minimum relationship strength.
  4. Decide whether invalidated memories should appear.
  5. In Graph mode, choose grouping by sector, sessions, or community.
  6. Select a node to open its detail.

When editing an existing memory, Allye may report a conflict if another change was saved after the loaded version. Reopen the memory, read the current version, and reapply only the necessary change.

For agents

Use Memories to persist information that will be useful beyond the current task. Before saving:

  1. Determine whether the information is a decision, preference, project fact, constraint, learning, or temporary state.
  2. Choose a specific, stable title.
  3. Write the conclusion and enough context for another session to understand it.
  4. Use tags that describe the subject, scope, and type.
  5. Set importance in proportion to impact.
  6. Look for a similar memory before creating another one.
  7. If there is a conflict, preserve the most recent version and record the change as an update, supersession, or new decision—not as a silent copy.
  8. When retrieving, cite the memory title and distinguish active information from archived information.

Memories do not replace work items or structured documentation. Use a work item for traceable work, Documentation for navigable material, and Memories for durable, recoverable context.

Current states and limitations

Current state: creation, browsing, editing, deletion, search, tags, importance, dashboard, focused graph, Graph V2, relationships, and links between memories and skills are represented in the frontend.

  • The Memories entry route opens Graph V2; Dashboard, Browse, and the focused graph remain accessible through the area's subsections.
  • The old graph still has a visual mock fallback when there is no focus or real data. Do not use this fallback as evidence that a relationship exists.
  • Some queries return an empty list when they fail; check filters, team, and loading state before concluding that there are no memories.
  • Semantic classification, deduplication, invalidation, sectors, and relationships depend on the connected service and cannot be inferred from the graph's appearance alone.
  • An archived memory remains in the knowledge base but may be excluded from default views and filters.
  • Graph V2 may limit the number of nodes when simulated truncation mode is enabled. The counter indicates when the view is summarized.

Next steps

  • Save relevant decisions during work in Work items.
  • Turn recurring knowledge into Documentation.
  • Link operational instructions to Skills.
  • Review old memories and archive what no longer represents the current state.