pheromone_network

Memory Lab

View on GitHub ↗

A local memory that learns from reuse.

Write short memories, ask a query, and watch the kernel rank related context. Repeated observations reinforce the local network; decay lets stale links fade.

Loading npm package…
Observations
—
Pheromone mass
—
Memory items
—
Embedding
n-gram

Memory stream

Each item is observed by the kernel.

Recall

Rank by similarity and reuse frequency.

hours elapsed hour half-life
Factor: 0.50×
The demo runs in your browser. No memory is sent to a server.

About the built-in N-gram encoder

This demo uses a deterministic character N-gram encoder so it can run offline without a model download or API key.

  • Good for short text, near-duplicates, log templates, and quick prototypes.
  • It compares character fragments, not meaning. Shared phrases can look more similar than they should.
  • It can miss paraphrases that use different words, especially in long or multilingual text.