The significance layer for AI memory_

Capture what mattered, not what was said.

Semantic search retrieves what’s similar; EDM encodes what was significant — structured at capture time, before retrieval ever happens.

npm install|57 Affective Fields|83% Significance Recall
EU AI Act ReadyGDPR CompliantW3C Data IntegrityEvidence-Grade Artifacts

Four problems significance solves_

Retrieval quality — your system surfaces what was recent, not what mattered. The moment that actually answers the query has different words entirely.

Context relevance — your AI knows what the user said. It doesn't know what it meant to them, how significant it was, or whether it shaped who they are.

Memory coherence — a crisis and a casual mention are stored the same way. Without a significance signal, every moment is equal.

Personalisation depth — context resets or stays shallow. Narrative arcs, identity threads, and emotional weight are lost in inference the moment the session ends.

Unstructured text. Structured meaning._

One API call. Raw text in. Structured emotional record out — retrievable by what mattered, not just what was said.

Raw input

“Just got these photos and what an absolute ball I had on the weekend competing in my physie comps! About 8 years ago I was diagnosed with psoriatic arthritis which affects my joints and I'm on medication that I will need to take forever to keep them mobile. However, it's continuing to move that really keeps it at bay. Not only does physie do this for me, it's also so good for my mental health…”

EDM artifact

physie_reflection.edm.json · Full Profile
core.anchor"physie competitions"
core.wound"psoriatic arthritis"
core.bridge"returning to dance"
constellation.arc_type"transformation"
constellation.emotion_primary"pride"
constellation.identity_thread"dance is my personal joy and health anchor"
gravity.emotional_weight0.8
gravity.recurrence_pattern"chronic"
gravity.resilience_markers["continuity","optimism"]

transformation · pride · enduring · continuity · optimism — none of these appeared in the source text. The artifact found them in the meaning.

Six domains. One artifact._

Every EDM artifact stacks six domains of significance — governance at the top, affective weight at the bottom — encoded at capture time.

EDM artifact stack — six domains (meta + governance, Core, Constellation, Milky Way, Gravity, Impulse)meta + governanceCoreConstellationMilky WayGravityImpulse

Hover or focus a layer to see its fields.

Significance, not similarity._

Semantic search retrieves what looks similar. EDM encodes what was significant — five contrasts that decide which one your application needs.

Similarity-only

Semantic search

What’s similar.

Retrieval signal
Word and embedding match
What it surfaces
Text that looks similar
Coverage when emotion matters
Low (Pepys-1665: 22% on emotional queries)
Memory coherence
Every moment weighted equally
When to use
Search

Significance-encoded

EDM significance

What mattered.

Retrieval signal
Significance fields structured at capture time
What it surfaces
The moment that actually answers
Coverage when emotion matters
83% on the same benchmark
Memory coherence
A crisis is not a casual mention
When to use
Activation

Built for AI that handles emotional context._

AI Memory Platforms

Recall · Retrieval · Agent memory

Your memory stack retrieves by recency or similarity. EDM artifacts give it a third signal — meaning. Retrieve what mattered, not just what was recent.

Learn More

Journaling & Companion AI

Journaling · Companion AI · Legacy platforms

Meaningful moments preserved before they're lost in inference. What mattered, structured and signed.

Learn More

Therapy & Coaching Apps

Therapy · Coaching · Telehealth

Narrative arcs and affective themes that build over months — not just between sessions. Know where this person is in their story, not just what they said last week.

Learn More

Workplace & HR AI

HR Tech · Wellbeing · Enterprise AI

Interpret, don't infer. EU AI Act compliant by architecture — not by policy.

Learn More

One artifact. Every memory architecture — episodic, graph, or hybrid._

Bring your retrieval stack. EDM artifacts enrich it — whichever paradigm you use.

Other systems score relevance at retrieval time. EDM encodes significance at capture time — so every retrieval paradigm gets a richer signal to work with.

Embedding-based retrieval

  • gravity.emotional_weight
  • gravity.valence
  • gravity.recurrence_pattern
  • constellation.arc_type

Retrieve by resonance — not recency or keyword match

Temporal knowledge graph

  • milky_way.associated_people
  • gravity.tether_type
  • gravity.recall_triggers
  • constellation.identity_thread

Traverse entity relationships and temporal edges as a network of meaning

Agentic reasoning

  • constellation.arc_type
  • impulse.drive_state
  • constellation.identity_thread
  • core.bridge

Reason over who this person is — not just what they said

Episodic memory

  • gravity.recurrence_pattern
  • gravity.emotional_weight
  • gravity.temporal_decay
  • gravity.resilience_markers

Significance that strengthens, decays, or re-emerges — encoded in the episodic record, not computed at query time

Verify every record independently_

W3C Data Integrity

Each record carries a cryptographic proof (eddsa-jcs-2022). Verify any artifact independently.

Published specification

EDM v0.7 is published on Zenodo with DOI 10.5281/zenodo.19211903 — citable and permanent.

Production-ready

90% enum alignment. 77% semantic match across 3 LLM providers.

Your data stays yours

We do not train models on your emotional data. Artifacts are processed, sealed, and returned.

Start free. Scale predictably._

MAU-based pricing for apps. Per-event for platforms. Self-seal always free.

Free

$0

100 MAU · 1,000 calls

Growth

$0.05/MAU/month

Up to 10,000 MAU · max $500/mo

ScaleRecommended

$0.04/MAU/month

Up to 100,000 MAU · max $4,000/mo

Start building with emotional context_

1,000 extractions free. No credit card required. Or work with us as a design partner and help shape what gets built next.

API access

Get an API key and start in minutes. 100 monthly active users, 1,000 extractions included.

+ REQUEST API KEY

Design partner

Building something where emotional context matters? We're working with a small group of teams to refine the platform. Direct access, priority support, and roadmap input.

+ APPLY AS DESIGN PARTNER