Domain-agnostic episodic memory engine built on Hybrid Centroid Indexing (HCI) and state-action-outcome episodes.
One platform, many domains. Every system shares the same core.
Any domain can emit episodes into VAAS-X for durable memory and fast recall.
VACA arrays store vectorized episodes indexed by HCI. AgentInterfaceModule coordinates ingest and recall workflows.
state + action + outcome → embedding → HCI index
Applications query the core to recall similar episodes and make better decisions in <100ms.
Hybrid Centroid Index for fast similarity recall.
State-action-outcome memory model.
Orthogonal-key encryption for privacy-preserving search.
Vector-augmented cognitive arrays for durable memory.
Operator interface for ingest, search, and workflow control.
These patterns match the configuration described in the core product specification.
Run router and one shard on a single machine.
Horizontal scaling across multiple shards.
Lightweight shards close to where decisions are made.
Every domain specializes the same basic structure.
What the world looked like at decision time.
What the system or human did.
What happened as a result.
Two scripts that start a local shard + router, ingest episodes, finalize, and query.
Ingest a small support knowledge base and retrieve answers with similarity search.
py .\scripts\testing\run_support_kb_vaca_test.pyIngest a small text corpus as episodic memory and validate retrieval quality.
py .\scripts\testing\run_vaasx_full_test.py