surface↓q & a↓skill

Q & Ascalar · returns varchar

SKILL

Call any registered LARS skill from SQL and receive results as a table

Per-row — runs once for each row.

qapipeline-composablellmtext

Syntax

skill({{ skill_name }}, {{ args }})

Arguments

name	type	description
skill_name	VARCHAR	—
args	VARCHAR	—

About

Call any registered LARS skill directly from SQL and get its results back as a table. A bridge between SQL workflows and the broader set of LARS tools that aren't exposed as dedicated semantic operators. Useful when you want a skill's functionality inside a SQL pipeline: running a custom Python tool, invoking an external data source, or chaining into a cascade that doesn't yet have its own `sql_function` entry point. Execution gets the same observability, caching, and trace instrumentation as a native cascade call, so the skill's invocation shows up in logs and the run-history view alongside everything else.

Nearby rabbit holes

same domain

ASK

Apply any arbitrary prompt to text (ultimate flexibility)

ASK_DATA

Answer plain-English questions about your data by writing and running SQL

qallmtextpipeline-composable

ASK_DATA_SQL

Return the SQL that ASK_DATA would run, without executing it

SQL_EXPRESSION

Generate a DuckDB SQL expression string from a plain-English description

Climb back to The Looking Glass

surface↓q & a↓skill

Q & Ascalar · returns varchar

SKILL

Call any registered LARS skill from SQL and receive results as a table

Per-row — runs once for each row.

qapipeline-composablellmtext

Syntax

skill({{ skill_name }}, {{ args }})

Arguments

name	type	description
skill_name	VARCHAR	—
args	VARCHAR	—

About

Call any registered LARS skill directly from SQL and get its results back as a table. A bridge between SQL workflows and the broader set of LARS tools that aren't exposed as dedicated semantic operators. Useful when you want a skill's functionality inside a SQL pipeline: running a custom Python tool, invoking an external data source, or chaining into a cascade that doesn't yet have its own `sql_function` entry point. Execution gets the same observability, caching, and trace instrumentation as a native cascade call, so the skill's invocation shows up in logs and the run-history view alongside everything else.

Nearby rabbit holes

same domain

ASK

Apply any arbitrary prompt to text (ultimate flexibility)

ASK_DATA

Answer plain-English questions about your data by writing and running SQL

qallmtextpipeline-composable

ASK_DATA_SQL

Return the SQL that ASK_DATA would run, without executing it

SQL_EXPRESSION

Generate a DuckDB SQL expression string from a plain-English description

Climb back to The Looking Glass