DSP – Data Source Publication¶
DSP (Data Source Publication) is an open, technology-neutral standard for describing enterprise data sources, datasets, and schemas in a single JSON format.
It lets any system – databases, streams, APIs, files, search, NoSQL, graphs, SaaS platforms – publish a DSP manifest that describes:
- What data exists (datasets and fields)
- How it is structured (types, nullability, partitioning)
- How to access it (SQL, HTTP, topics, paths)
- Optional business semantics (domains, entity types, join hints, PII)
DSP does not move data. It standardizes metadata so tools, agents, catalogs, and humans can discover and understand data consistently.
Repository Layout¶
dsp-spec/– DSP JSON Schema (dsp-1.0.json)dsp-sdk/– Python SDK with core models and profilesdsp-registry/– Reference registry service for publishing and discovering manifestsexamples/– Example DSP manifests for common data source typesdocs/– Documentation site powered by mkdocs
See Getting Started to create your first DSP manifest.