Modelable

Modelable is a compiler and language server for versioned, domain-owned data models. Define canonical models and projections in .mdl files, then validate their compatibility, inspect field-level lineage, detect governance gaps, and generate artifacts for the systems that consume them.

Why Modelable?

Data contracts often become fragmented across application types, database schemas, API definitions, and catalog metadata. Modelable keeps the semantic contract in one versioned source and derives target-specific representations without losing ownership, classification, lineage, or compatibility context.

.mdl sources -> validate and resolve -> plan and govern -> generate artifacts

Install

Modelable requires Python 3.14.

uv tool install modelable
modelable --version

For an isolated one-off command:

uvx modelable --help

Define a model

domain customer {
  owner: "customer-platform"

  entity Customer @ 1 (additive) {
    @key customerId: uuid
    @pii email?: string
    displayName: string
  }
}

Save the definition as customer.mdl, then validate and compile it:

modelable validate customer.mdl --strict
modelable compile customer.mdl --target json-schema --out generated/schema
modelable compile customer.mdl --target typescript --out generated/types

Capabilities

• Parse and validate versioned models, projections, annotations, and workspace definitions.
• Resolve exact versions and compatible version ranges.
• Detect additive and breaking contract changes and affected projections.
• Trace projection fields to canonical source fields.
• Report structurally missing access and classification metadata.
• Expand automatic database, request, reply, and event projections.
• Generate JSON Schema, Markdown, TypeScript, C#, Java, Python, Rust, Go, SQL DDL, dbt schema.yml, FHIR R4 profile, OpenMetadata JSON, and OpenLineage event artifacts.
• Provide diagnostics, completion, hover, navigation, references, rename, formatting, and other editor features through the language server.
• Import or assist with models through optional LLM provider integrations.

The local compiler and language-server toolchain are the supported 1.0 stable surface. Apicurio JSON Schema artifact publish/pull and Marquez-compatible OpenLineage event sync are available for derived artifacts. Live catalog publishing, distributed synchronization, OpenLineage runtime event collection, and runtime materialization remain roadmap work.

1.0 stable surface

Modelable 1.0 stabilizes the local compiler and language-server toolchain.

In scope for 1.0:

• .mdl language: syntax, types, projections, ownership, classification, and access metadata.
• CLI: validate, compile, check, generate, attach, spec, and the language server.
• Generated artifacts: JSON Schema, TypeScript, C#, Java, Python, Rust, Go, SQL DDL, dbt schema.yml, Markdown, FHIR R4 profile, OpenMetadata JSON, OpenLineage event, and ODCS formats.
• Compatibility, lineage, and governance report output.
• Apicurio JSON Schema registry artifact push/pull.
• Marquez-compatible OpenLineage event sync via modelable sync --lineage.
• VS Code extension shipped as a VSIX companion artifact with the 1.0 release.

Deferred from 1.0:

• VS Code Marketplace distribution (post-1.0).
• Live OpenMetadata catalog synchronization and runtime OpenLineage collection.
• Remote tracked-spec polling and authenticated source access.
• Runtime subscriptions, adapters, replay, and materialization.
• Distributed registry synchronization beyond the current file-first model.

Development

cd cli
uv sync --extra dev --frozen
uv run pytest tests/ --tb=short
uv run modelable validate ../samples/mvp --strict

See CONTRIBUTING.md for the complete contributor workflow.

Documentation

Hosted: https://ktjn.github.io/modelable/

• Documentation index
• Language reference
• Tooling reference
• Architecture and system specification
• Getting started and migration
• Sample models
• Changelog
• Roadmap
• Project governance

License

Licensed under the Apache License 2.0.