Expressive

Declarative Multi-Stage Pipelines

Express candidate generation, vector retrieval, lexical search, machine learning scoring, and MMR diversity reordering in a single unified declarative query.

Spec-Driven

Formal EBNF & Intermediate Representation

Queries lower deterministically into canonical IR, with support for direct IR submission via structured YAML/JSON or high-level RecQL SQL compilation.

Universal

Vendor-Agnostic Engine

Run unchanged queries seamlessly against PostgreSQL + pgvector, Microsoft SQL Server 2025, Oracle 23ai, MongoDB, and MariaDB backends.