// SECTION: ECOSYSTEM_PROJECTS
009

Project Capacity Explorer

Click each project to inspect its real capacity surface. This page is intentionally capability-first: what the engine can do, how it behaves, and how it maps to practical agent outcomes.

// SECTION: CAPABILITY_SURFACE
010
16 query capacities | artifact: .amem

active.project

AgenticMemory

Persistent cognitive graph memory for cross-session reasoning continuity.

surface.summary

16 query capacities

Portable artifact: .amem

View Repo

Memory Is Not Search

Most systems treat memory as retrieval over flattened text chunks. AgenticMemory treats memory as graph cognition: nodes for what the agent learned or decided, edges for why those events connect, and traversal for reasoning history.

The Atom: Cognitive Event

The smallest memory unit is a cognitive event, not a transcript chunk: FACT, DECISION, INFERENCE, CORRECTION, SKILL, and EPISODE. Each event is written with confidence, timestamps, access dynamics, feature vectors, and direct edge references for O(1) node access.

FACT — learned truth about user, system, or world state.
DECISION — selected action and why it was chosen.
INFERENCE — synthesized conclusion from multiple events.
CORRECTION — explicit update that supersedes prior belief.
SKILL — procedural pattern for execution in context.
EPISODE — compressed session-level meaning summary.

Edges Make It A Brain

Relationship edges encode causality and truth evolution: CAUSED_BY, SUPPORTS, CONTRADICTS, SUPERSEDES, RELATED_TO, PART_OF, TEMPORAL_NEXT. This preserves why decisions happened and what changed over time.

One File, Memory-Mappable

`.amem` is a binary graph file with header, node table, edge table, content block, vectors, and indexes. It is memory-mappable, portable, and query-ready without external databases or managed vector services.

The Full Query Engine

Querying is navigation, not blind similarity. The engine supports traversal, temporal diffing, causal impact, contradiction resolution, pattern recall, and structural gap detection.

#Query TypeWhat It AnswersStatus
1TraversalWhy did I decide this?Operational
2PatternShow me all decisions from last weekOperational
3TemporalWhat changed between session 5 and 20?Operational
4Causal / ImpactWhat breaks if this fact is wrong?Operational
5SimilarityWhat else do I know about this topic?Operational
6ContextGive me everything around this nodeOperational
7ResolveWhat is the current truth?Operational
8BM25 Text SearchFind memories containing these wordsOperational
9Hybrid SearchBM25 + vector fused rankingOperational
10Graph CentralityWhat are my foundational beliefs?Operational
11Shortest PathHow are these concepts connected?Operational
12Belief RevisionIf I learn X, what breaks?Operational
13Reasoning GapsWhere am I guessing?Operational
14AnalogicalHave I solved something like this before?Operational
15ConsolidationClean and strengthen my brainOperational
16Drift DetectionHow has my understanding shifted?Operational

Memory Formation Pipeline

After each interaction the system extracts events, links them to existing cognition, updates confidence and decay, writes an EPISODE compression node, and incrementally refreshes indexes. Memory formation runs asynchronously so response latency remains stable.

Portable Agent Brain

Your `.amem` travels across agents and environments. Knowledge continuity belongs to you, not to a single assistant vendor. Any compatible runtime can mount the same cognitive history with causal context intact.

// SECTION: COLLABORATION
011
collaboration.channelLIVE

Open to partnerships, sponsorships, and research collaborations across AI, ML, and agent systems, including persistent memory, visual continuity, and semantic code intelligence.