Skip to main content

Jungian Intelligence Layer

The Jungian Intelligence Layer (v3.2.0) adds four psychology-inspired subsystems on top of the existing concentric memory rings and Collective Best Practices. Named after Carl Jung’s analytical psychology, each subsystem maps to a core Jungian concept to transform raw memory data into meaningful, context-aware intelligence.
The Jungian Intelligence Layer builds on the memory infrastructure introduced in v3.1.0. If you are new to the memory system, start with the Memory & Intelligent Knowledge page.

The Four Subsystems

Synchronicity Detection

Discovers meaningful temporal co-occurrence patterns across anonymized profiles using pure SQL analysis. Three detection types: co-occurrence, failure correlation, and emergent workflows.

Archetype System

Classifies patterns into four Jungian archetypes — Shadow (warnings), Sage (best practices), Hero (workflows), and Trickster (creative solutions) — using deterministic context mapping.

Dream Processing

Consolidates clusters of semantically similar memories into single unified entries via LLM, reducing token consumption while preserving all key insights.

Individuation Scoring

Per-profile maturity scoring across four components (Memory Depth, Learning Velocity, Collective Contribution, Self-Awareness) with five maturity levels from Nascent to Individuated.

How It All Fits Together

The four subsystems operate on different layers of the memory stack but work together to create a self-improving intelligence system:
  1. Temporal events are recorded during every tool call and observation (fire-and-forget, non-blocking)
  2. Synchronicity detection runs periodically (cron) to discover co-occurrence patterns, failure correlations, and emergent workflows across all anonymized profiles
  3. The archetype router enriches every pattern injection with a Jungian archetype label and weighted score, ensuring the most contextually relevant patterns are surfaced first
  4. Dream processing runs during the decay cron to discover clusters of related memories and consolidate them, saving tokens while preserving knowledge
  5. Individuation scoring aggregates metrics from all subsystems into a single maturity score that tracks profile growth over time

Privacy Model

Profile Hash Only

Temporal events store profile_hash (HMAC-SHA256), never raw UUIDs. Individual usage patterns cannot be traced back to specific users.

k-Anonymity (k >= 3)

Synchronicity patterns are only surfaced when observed by 3 or more unique profiles. Individual behaviors are never exposed.

SDK & MCP Integration

SDK Methods

All methods are available in the JavaScript, Python, and Go SDKs:

MCP Tools

Configuration

All settings have sensible defaults and can be tuned via environment variables:

Next Steps

Synchronicity Detection

Deep dive into temporal pattern detection

Archetype System

How archetype-driven pattern delivery works

Dream Processing

Memory consolidation and token savings

Individuation Scoring

Per-profile maturity metrics and trends