Skip to content

feat: Visual Memory Dashboard for inspecting MemPalace structure and retrieval context #1369

Description

@torchwithwater-hash

Summary

MemPalace already has a powerful internal structure: wings, rooms, drawers, tunnels, diary entries, and KG facts. However, most of this structure is currently only accessible through CLI or MCP calls.

I would love to see a visual dashboard that makes the memory palace inspectable and maintainable by humans, not only retrievable by AI agents.

Motivation

Right now MemPalace works well as an AI memory backend, but it is difficult for users to understand the shape and health of their memory system.

A visual interface would help answer questions like:

  • Which wings and rooms exist?
  • Which rooms contain the most drawers?
  • Which memories are isolated?
  • Which tunnels connect different areas of knowledge?
  • Which KG facts are current vs expired?
  • What did the AI retrieve before answering?
  • Are there duplicated, stale, or contradictory memories?

As a user, I already find MemPalace useful as a cross-AI memory layer, but I often want to see what it remembers instead of only querying it through tools.

Proposed Feature

A local visual dashboard, possibly exposed through a command like:

mempalace dashboard
# or
mempalace visualize

The dashboard could include several practical views:

1. Palace Map

A graph or hierarchy view of wings, rooms, drawers, and tunnels. This could feel similar to Obsidian graph view, but preserve MemPalace's own hierarchy.

2. Memory Timeline

A chronological view of diary entries, created drawers, updated memories, invalidated KG facts, and major session summaries.

3. Knowledge Graph View

A visual representation of KG triples:

subject -> predicate -> object

With filters for entity, relationship type, current vs expired facts, and valid date range.

4. Drawer Inspector

Clicking a drawer should show full content, wing / room, source file, created date, related tunnels, similar drawers, and linked KG facts.

5. Retrieval Transparency View

When an AI agent answers using MemPalace, it would be very useful to show:

  • which memories were retrieved
  • their similarity scores
  • which KG facts were consulted
  • whether any memories were filtered out
  • what context was actually passed to the model

This would improve trust and make it easier to debug bad answers.

Why This Matters

MemPalace is not just storage. Its value comes from long-term continuity across AI sessions and tools.

But long-term memory needs maintenance. Users need to be able to see what the system remembers, why it remembers it, how memories are connected, when something became true, when something stopped being true, and what context influenced an answer.

Without a visual layer, the system risks becoming powerful but opaque.

A dashboard would make MemPalace more usable for non-technical users, easier to debug for developers, and more trustworthy as a shared long-term memory layer across AI agents.

Related

This is related to #101, which proposes a 3D palace visualizer / multi-hop exploration feature.

This issue is more focused on a practical inspection and maintenance dashboard: memory health, human review, retrieval transparency, KG/timeline inspection, and stale or conflicting memory detection.

Both features could share the same underlying graph/export API.

Metadata

Metadata

Assignees

No one assigned

    Labels

    enhancementNew feature or request

    Type

    No type

    Projects

    No projects

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions