Vesicle is the commercial managed layer on top of StromaDB: it turns existing sources into a live, curated knowledge graph and runs it as a managed service.
Where StromaDB is the OSS engine (storage + query + reactivity), Vesicle is the mapping + ingest + operations around it — the part most teams don't want to build or run themselves.
Status: design stage — this repo carries the overview and architecture only.
- Source → graph mapping — map a source schema (Postgres, …) to ontology types and relationship predicates: no-code (AI-proposed → approve/edit), low-code (expressions), or programmable API. The mapping author can be anyone: self-serve, a consultant, an embedded SaaS, or a partner.
- Zero-ETL CDC ingest — connect a source and changes flow to the graph in real time.
- Managed, multi-tenant, scale-to-zero — run many organizations on one platform cheaply; idle tenants cost zero compute (storage only).
- Operability — agent↔DB observability, deployment, and lifecycle around the StromaDB core.
- Vesicle feeds and operates StromaDB; it is not a general-purpose ETL/data-integration platform beyond that purpose.
- Curation, not a data lake: it extracts decision-relevant facts/relations into a bounded graph; raw data stays in the source.
See docs/ARCHITECTURE.md for how the mapping and ingest layers fit together.