Quickstart
Three pages, in order. The first gets you a running engine and a single query. The second goes one step deeper — a real KQL query, the agent endpoint, an idea of what's possible. The third is a one-screen reference card for when you've forgotten which port is which.
Five-minute start
docker compose up, send one row, query it back. Five minutes from a fresh clone to your first KQL result. No language toolchain, no cloud account, no auth setup.
First real run
A multi-row batch, a real KQL summarize, and one call to the agent endpoint. Same engine, slightly more interesting questions. About ten minutes.
Concepts cheatsheet
A one-page reference. The five invariants in one line each, the data shape, default endpoints and ports, every KYMA_* env var the binary reads. Bookmark this.
What you'll have at the end
A working kyma — engine, Postgres catalog, MinIO object store, Redpanda broker — running locally on your machine. One table, a handful of rows, proof that ingest and query both work end to end. A short list of follow-ups for the ten things you'll want to try next.