ktx-ai-data-agents-context-layer

What problem does it solve?

Agents often guess at metric definitions and join logic, leading to incorrect SQL and misleading analytics.

Core Features & Use Cases

• Semantic layer generation: Builds a metric-first semantic layer with approved definitions and automatic join resolution to prevent “invented math.”
• Context ingestion from analytics sources: Ingests knowledge from dbt, LookML, Metabase, Notion, and related documentation so agents understand your business logic.
• MCP-ready data tools: Exposes MCP tools that let agents search semantic metrics and wiki knowledge and then query your warehouse with the right context.

Use case: Your team asks an AI to answer “What is monthly revenue and how is it defined?”—ktx retrieves the canonical metric definition, related joins, and business context, so the resulting query matches your approved semantics instead of ad-hoc assumptions.

Quick Start

Run ktx setup in your analytics project directory to generate ktx.yaml, configure your LLM and embeddings, connect your warehouse, ingest dbt/context sources, and enable MCP integration.