bsl-query-expert

The bsl-query-expert is a specialized agent skill designed to interface with the Boring Semantic Layer (BSL), a lightweight, Ibis-powered framework for structured data exploration. It is specifically built for data analysts, engineers, and researchers who need to perform rapid, ad-hoc analysis on semantic models without writing verbose boilerplate code. The skill automates the workflow of model discovery, schema inspection, and complex query construction, ensuring that users can transition from raw data tables to high-quality insights and visualizations with minimal friction.

• Perform multi-step exploratory data analysis by discovering available models, inspecting schemas, and executing precise aggregate operations.

• Utilize a fluent API approach to build complex queries including group_by, filter, mutate, and window functions like rolling averages, rankings, and cumulative totals.

• Support for advanced semantic operations such as market share analysis (percentage of total), bucketed categorical groupings, and multi-hop joins between semantic tables.

• Automatic rendering of results into tables or charts, leveraging backend integrations like Altair, Plotly, or plotext for instant data visualization.

• Intelligent handling of database-specific logic, including time dimension truncations (Y, Q, M, W, D) and conditional expressions using ibis.cases.

• Always begin by listing models and retrieving the schema to ensure field name accuracy, as the skill enforces strict adherence to exact column definitions.

• For complex filters or category-based lookups, use the recommended multi-hop pattern: first perform an discovery query (limited records) to inspect values, then execute the final filtered query.

• Use the get_documentation() tool to query the internal API reference for query-methods, windowing, and charting whenever syntax clarification is required.

• When building queries, default to get_chart=true for final outputs to trigger visual rendering, while using get_chart=false during exploration to keep data transmission focused on raw values.

• The skill is optimized for Ibis-integrated engines including DuckDB, Snowflake, BigQuery, and PostgreSQL, making it a robust bridge between LLM reasoning and production-grade databases.