GitHub - boringdata/boring-semantic-layer

1 min read Original article ↗

Boring Semantic Layer (BSL)

The Boring Semantic Layer (BSL) is a lightweight semantic layer based on Ibis.

Key Features:

  • Lightweight: pip install boring-semantic-layer
  • Ibis-powered: Built on top of Ibis, supporting any database engine that Ibis integrates with (DuckDB, Snowflake, BigQuery, PostgreSQL, and more)
  • MCP-friendly: Perfect for connecting LLMs to structured data sources

Quick Start

pip install 'boring-semantic-layer[examples]'

1. Define your ibis input table

import ibis

# Create a simple in-memory table
flights_tbl = ibis.memtable({
    "origin": ["JFK", "LAX", "JFK", "ORD", "LAX"],
    "carrier": ["AA", "UA", "AA", "UA", "AA"]
})

2. Define a semantic table

from boring_semantic_layer import to_semantic_table
flights = (
    to_semantic_table(flights_tbl, name="flights")
    .with_dimensions(origin=lambda t: t.origin)
    .with_measures(flight_count=lambda t: t.count())
)

3. Query it

result_df = flights.group_by("origin").aggregate("flight_count").execute()

📚 Documentation

→ View the full documentation

This project is a joint effort by xorq-labs and boringdata.

We welcome feedback and contributions!

Freely inspired by the awesome Malloy project. We loved the vision, just took the Python route.