MotleyAI/slayer

An embeddable, expressive semantic layer for AI agents and humans

What it solves

SLayer acts as a semantic bridge between AI agents and raw data warehouses. It solves the problem of agents struggling to generate accurate, complex analytical queries by providing a governed, flexible layer that handles SQL translation, data transformations, and security, ensuring agents get reliable answers without needing to understand the underlying database schema.

How it works

SLayer sits on top of your existing database. It allows you to define metrics once and then query them using expressive formulas (like time shifts or ratios) rather than writing raw SQL. It supports an agentic workflow—search, inspect, and query—and integrates with LLMs via various interfaces including MCP, REST API, and a Python client. It can also import existing configurations from tools like dbt or Cube.

Who it’s for

This is for developers and data engineers building AI-powered analytical applications or agents that need to interact with structured data warehouses reliably and securely.

Highlights

  • Provides a flexible semantic layer that supports complex metric combinations and multi-stage queries.
  • Offers multiple integration interfaces including MCP, REST API, CLI, and Python.
  • Includes built-in search and memory stores to help agents discover and understand business context.
  • Compatible with most popular databases and supports importing configurations from dbt, Cube, and Ossie.

Related

  • Project
    EmilStenstrom/conllu

    A lightweight Python library for parsing CoNLL-U formatted NLP data into nested dictionaries or dependency trees.

  • Project
    logicalclocks/hopsworks-tutorials

    A collection of tutorials that teach machine learning development with Hopsworks, covering feature store usage, real-time and batch AI systems, LLM applications, and integrations with tools like Airflow, Spark, and Snowflake.

  • Project
    beowolx/rensa

    A high-performance MinHash library written in Rust with Python bindings for fast near-duplicate document detection and Jaccard similarity estimation.

  • Project
    edoliberty/long-term-memory-in-ai

    This repository provides course materials for a graduate class on vector search and databases, taught by experts from Pinecone and Meta. It covers embeddings, approximate nearest neighbor search, indexing techniques, and practical systems like FAISS, with lecture notes, a syllabus, and project options.