Convert MetricFlow semantic models and metrics into ktx semantic layer sources.
npx skills add https://github.com/Kaelio/ktx --skill metricflow-ingest
Please help me install this Agent Skill. Skill: metricflow_ingest Source: https://github.com/Kaelio/ktx/tree/main/packages/cli/src/skills/metricflow_ingest Command: npx skills add https://github.com/Kaelio/ktx --skill metricflow-ingest
Migrating MetricFlow semantic models and metrics into the ktx semantic layer requires manual translation of measures, entities, joins, and metric types, which is error-prone and slow when done by hand.
extends: chains and cross-model metric references into a single flattened overlay with provenance tags per source file.semantic_models: and metrics: YAML files, load this Skill to emit verified ktx SL sources with identifier checks via entity_details and sql_execution probes.Load this Skill and ingest the MetricFlow YAML files in this WorkUnit into ktx semantic layer sources, verifying every table and column against the warehouse before writing.
High-intent search queries and answers about installing and using this skill.
Map each MetricFlow semantic_model to a ktx overlay or standalone source, converting measures to inlined aggregation expressions, primary entities to grain, and foreign entities to joins. Flatten any extends chains into a single overlay before writing.
Simple metrics become measures with inlined aggregations or filtered measures, derived and ratio metrics become derived measures with safe division, cumulative metrics become standalone SQL sources with window functions, and conversion metrics are flagged for human review.
ktx SL has no first-class cumulative or conversion primitives in v1. Cumulative metrics are translated to standalone SQL sources using window functions, while conversion metrics are not written as SL and instead produce a wiki note plus an emit_unmapped_fallback call.
Inheritance is flattened before writing: walk the extends chain from the most-derived child, accumulate measures, dimensions, and entities, and emit one overlay named for the child. Name collisions resolve in favor of the child, with the overridden parent noted in the description.
The warehouse rejects invented columns with an Unrecognized name error during sl_validate. Every column must come from raw files, entity_details, or a successful sql_execution probe, never from assumptions or placeholder identifiers.
dbt-labsCreate and modify dbt Semantic Layer components using MetricFlow.
AradotsoBuild a semantic context layer for AI agents querying data warehouses.
astronomerAnalyze a specified table to generate a structured data profile.
starrank-softAccelerate large-scale data ingestion with idempotent writes and manifest-based tracking.