Docs / Introduction Learn
Non-Goals
What EKOS deliberately is not.
EKOS is not:
- another vector database. It has an optional vector index for semantic search, but its core is a typed, evidenced graph.
- a general-purpose RAG framework. Retrieval exists to find objects; answers are assembled from typed evidence, not from pasted chunks.
- an autonomous agent. EKOS never acts on your source systems. Agents consume it read-only, and the decisions that matter (review, sign-off, migration approval) are reserved for people.
- a replacement for source systems. It observes them and never writes back: source trees are never written, and database sessions are read-only. The opt-in EKOS Migrate writes only to its target database, and an action its computed risk does not class as automatic needs a person's approval; the Lakehouse extension generates pipeline code into a repository you own.
- a data warehouse. It stores knowledge about data assets, not the data.
- an LLM. The compiler is deterministic; LLMs are optional providers behind a trait.
- a generic document search engine. Documents are one source among many, compiled into the same model as code and schemas.
It also does not claim certainty it lacks: anything it cannot parse is recorded as an explicit Unmapped fact ("something is here, not yet understood") rather than dropped or guessed.