umf
Ecosystem

Projects using
and building on UMF.

UMF grew out of TableSpec’s table schemas. Its metadata now also supports SQL compilation, PostgreSQL graph storage and versioned graph publications.

Domain packs

Explore a working schema example.

The legal pack describes fabricated clients, matters, staffing, billing and documents. Browse its schemas and follow their declared references to see how the records fit together.

Pack schemas document a domain. Data generation and loading belong to their consuming tools.

Available

TableSpec

Table schemas and pipeline tooling

A Python library for schema validation, profiling, SQL and PySpark schema generation, and documentation. Its table-oriented format is available through UMF’s TableSpec adapter. Check the integration reference for supported operations.

SQL compiler

Weft

UMF-powered SQL compiler

Write queries against supplied UMF models and compile SQL for Truss on PostgreSQL or Ashlar on Databricks. Weft returns parameters, result types, decoding instructions and diagnostics; the host executes the query.

Weft has Rust, Python and WebAssembly interfaces. See the project documentation for backend and qualification details.

Experimental

Truss

Property graphs on SQL

An embeddable graph store for PostgreSQL 18.6. Truss uses UMF schemas for types and relationships and Weft to compile queries. Its Python quickstart exercises a local store.

GitHub ↗
Development toolkit

Ashlar

Graphs on Databricks

Publishes typed property graphs as immutable Delta table states. Ashlar supports local publication and documents an experimental managed read path for Databricks Unity Catalog.

GitHub ↗