Introducing Metrics SQL: A SQL-based semantic layer for humans and agents

Rill introduces a metrics-first semantic layer using SQL, aiming to streamline the process of querying business metrics like revenue and ROAS without requiring the learning of new languages or APIs. By leveraging SQL, the universally understood language among databases and BI tools, Rill ensures consistency in metric definitions across various platforms, such as dbt models, Python notebooks, and AI agents, thereby eliminating discrepancies often caused by differing computations. The architecture extends SQL with metrics to enhance performance, utilizing optimizations like materialized views and intelligently tuned database indexes. Rill’s Metrics SQL, a SQL dialect, allows users to express complex business logic with simplicity and security, ensuring a deterministic source of truth for metrics queries. It supports various SQL dialects from engines like ClickHouse and Snowflake and is designed to evolve with potential future semantic pushdowns, where databases would natively support metrics semantics, optimizing queries directly at the database level. Rill facilitates querying through CLI, HTTP API, and integration with AI agents, maintaining a focus on security and performance within its restricted SQL subset.