Memory Classification & Insights

When you save memories through MCP (or other integrations), YoMemo does not only store encrypted content. We attach a semantic fingerprint to each memory so you can understand and balance your knowledge base over time. This page introduces the classification system and the direction we are building toward.

How we classify memories

For each saved memory, we use the following dimensions to organize and surface insights:

E/L/A/P radar

Every memory can be scored on four axes (0–1 each), stored in the memory’s semantic fingerprint:

Dimension	Meaning
E — Emotion	Subjective mood or feeling in the content
L — Logic	Technical density and factual derivation
A — Abstraction	Philosophical depth and underlying patterns
P — Pragmatism	Actionability and execution value

These scores are optional metadata (e.g. provided by the AI when it calls save_memory). In the YoMemo dashboard, Pro users can see an E/L/A/P radar that aggregates these scores across memories, so you can see at a glance whether your base is skewed toward emotion, logic, abstraction, or pragmatism—and adjust how you record things if you want more balance.

Handle distribution

Memories are grouped by handle (e.g. work, personal, project-x). The dashboard shows how many memories you have per handle. This helps you see which areas of your life or work are most represented and avoid overloading a single handle.

Memory spectrum

A visual spectrum of your memories (e.g. one tile per memory, colored by Emotion: cool = low, warm = high) gives a quick picture of the “temperature” and variety of your stored content. Gray tiles indicate memories without ELAP data.

Capacity & balance tips

As your memory base grows, we surface simple capacity and balance suggestions, for example:

Too many memories in total — consider archiving or consolidating.
Too many handles — consider merging or renaming.
Strong bias toward logic (L) or emotion (E) — optional nudge to add more of the other.

These tips are designed to keep your memory graph useful and readable, not to restrict you.

Long-term and short-term memory (roadmap)

Today, every memory is stored in one encrypted pool. As the product and your usage grow, we plan to gradually introduce long-term and short-term memory:

Short-term — Recent or frequently accessed items, or items you mark as ephemeral; easier to surface in daily use and optionally to expire or demote.
Long-term — Core knowledge and decisions you want to keep and reuse over years.

Memories will continue to be encrypted and under your control. The goal is to make the system continuously improve your life: the right information at the right time, without clutter, and with a clear view of how your knowledge is balanced (E/L/A/P, handles, spectrum) and how it evolves.

Summary

Concept	Role
E/L/A/P radar	Classify and visualize each memory on Emotion, Logic, Abstraction, Pragmatism (0–1).
Handle distribution	See how memories are spread across handles; avoid overload.
Memory spectrum	Visual overview of variety (e.g. by Emotion).
Capacity & balance tips	Light suggestions to keep the base healthy.
Long-term / short-term (planned)	Gradually expose different retention and surfacing for different kinds of memory.

Saving via MCP (with optional metadata.semantic_fingerprint) feeds this system; the dashboard and future features use it to help you understand and grow your memory in a way that continuously improves your life.