Materialization runtime

Domain data packaged for AI work.

Dataset converts domain state into durable rows and artifacts. It can snapshot queries directly or use sandbox-backed reactors for files, transforms, instructions, and multi-source work.

/dataset

Dataset surface

@ekairos/dataset@beta

schema

dataset_datasets

Dataset metadata, source analysis, output mode, sandbox id, and linked data files.

dataset_records

Materialized rows linked to a dataset and optional source files.

actions

dataset(runtime)

Builder entry point for source selection, output mode, sandbox, reactor, and build.

fromQuery

Snapshots domain query results into a dataset without requiring a reactor.

from / fromDataset / fromText

Adds file, dataset, and text sources for materialization.

asRows / first

Controls whether output is row-oriented or a single structured object.

Showcase-first surface

The route exists before registry publication so the public surface can be shaped against the real domain contract.

From simple to complex

The same idea, three depths.

01

in one sentence

Turn your domain data into tidy rows an AI can actually work with.

Messy queries, files, and text become clean, durable datasets ready for analysis.

02

how it works

Point it at a source, choose a shape, and build.

You start from a domain query, a file, another dataset, or plain text, and decide whether the output should be rows or a single structured object. Simple query snapshots materialize directly; richer parsing and transforms run inside a sandbox. Either way the result lands as a durable dataset you can query and reuse.

03

under the hood

Datasets and records built through a runtime-first builder.

dataset_datasets holds metadata, source analysis, output mode, and linked files while dataset_records stores the materialized rows. The dataset(runtime) builder chains fromQuery, from, fromText, asRows, and build, with materializeDataset exposing the same flow as an agent-callable action.

Runtime shape

Package behavior. Registry presentation.

Dataset UI should make materialization observable: sources, row counts, schema, transform state, and downloadable artifacts.

package

@ekairos/dataset@beta

aggregate

dataset_datasets

durable

dataset_records

ui

component surface reserved