What problem does it solve?
Synapse solves the challenge of reliably storing, validating, and retrieving structured memories for AI agents so they do not forget context, repeat mistakes, or lose associative links between items. It prevents unstructured note sprawl, reduces token waste by returning compact JSON read models, and enables reproducible rebuilds via an append-only event log.
Core Features & Use Cases
- JSON Schema-driven storage: Define runtime JSON schemas to validate node payloads and keep stored memories consistent.
- Nodes and Links graph: Create typed nodes, attach keywords, and connect nodes with bidirectional links for fast contextual expansion.
- Event sourcing + projections: All changes are recorded as immutable events; projections rebuild queryable read models for efficient CLI consumption and jq-driven workflows.
- Replication and restore: File-based JSONL replication and iterators enable safe backups and restoring into fresh databases.
- CLI-first workflows: Lightweight commands to init, add schemas/nodes, search, traverse graphs, and run projections that integrate with shell tooling.
Quick Start
Initialize a local Synapse database with synapse init, add a JSON schema, create a node with synapse nodes add, and then search or list nodes to inspect the stored memory.