System Overview

SDCStudio is the application for building data models on the Semantic Data Charter: you describe a record once, as components with their constraints and meaning, and every Governed Data Record made from the published model carries that definition wherever it goes. This page is the map of the whole product; each part has its own guide.

The shape of the product

SDCStudio is one web application at sdcstudio.axius-sdc.com with an HTTP API behind it. The browser application is a React single-page app; the API is Django with the Django REST Framework. Everything the application does, the API does, and the API is what your own tools call with an API key.

Behind the API:

• PostgreSQL holds the models, the components, the accounts and the wallets; its pgvector extension holds the embeddings the type-selection step uses.
• Apache Jena Fuseki holds the ontologies users upload; semantic linking searches it. SDCStudio ships none: the Ontologies page lists your uploads, your teammates' Team and Corporate uploads, and any uploaded as Public.
• Redis and Celery run the two-stage processing pipeline in the background.
• Google Vertex AI is the only language-model provider: Gemini for generation, a text-embedding model for embeddings. No other provider is configured.
• Google Cloud Storage holds uploaded files and generated packages. The services run on Google Cloud Run.

The documentation you are reading is published from the repository on every change, and the public catalog, the schema route and the documentation need no account.

What you build

A component is one piece of data with its meaning attached: a name, a description, a type, the type's constraints, and semantic links to published ontologies. There are twelve value types (String, Token, Boolean, Count, Quantity, Float, Double, Temporal, Ordinal, Link, File, Interval), eight list types, the Cluster that groups components, Units, and four contextual types (Party, Participation, Audit, Attestation). The SDC4 Type Reference is generated from the reference model and lists every element and every field.

A cluster groups components and smaller clusters. A data model names one cluster as its structure and adds the parties, the provenance, the Dublin Core description and the assertions.

A project holds your components and models and has one team. Public library projects, eighteen of them, hold published components anyone can copy into their own project; see Public Libraries.

Components, clusters and models are draft until published. Publishing is one way: a published definition never changes, so every record made from it stays valid forever. A change is a copy, derived from the original, and the original is retired when it should no longer be chosen. See Copy and Retire.

Three ways in

Upload a file. A CSV or a Markdown template goes through two stages. The first reads the file and infers a type and constraints for each column without a language-model call. The second asks the model for descriptions, units, ranges and semantic links, then creates the components, the cluster and a draft model in your project. The stages are separate so that a failed second stage can be retried without re-parsing. See Uploading Data, The Data Sources Page and AI Processing.

Build by hand. The Component Wizard, the Cluster Wizard and the Model Wizard create one object each, step by step, with the same fields the upload pipeline fills. See Data Modeling and Creating a Model.

Assemble from a script or an agent. POST /api/v1/dmgen/assemble/ builds a model from a tree of reused and new components, publishes it and builds its package in the same call; assemble/estimate/ shows the cost first and charges nothing, and /api/v1/jobs/ follows the long work. sdcmodeler is the client, command line and MCP server for this path, so an AI assistant can propose a model, show you the estimate and build it under a key whose scopes and spending limit you set. See the assembly reference and The SDC MCP Servers.

The first two ways leave drafts that you review and edit before you publish: the stepped edit form, the type reference and the semantic-linking step are the same for every component. The third publishes as it builds unless it is asked for a draft.

What comes out

Publishing a model and clicking Generate produces one package: the XSD schema, an XML template of the Governed Data Record with a placeholder at every value, a JSON instance with sample values, a JSON-LD description, HTML documentation, the RDF (Turtle and RDF/XML), a SHACL shapes file, GQL statements for a property graph, and a prompt file for AI tools. See Generating Outputs.

From a generated model you can also build a standalone Django application in one of two profiles, Lightweight or Enterprise, and download it. See App Generation.

Records carry https://semanticdatacharter.com/dmlib/dm-<ct_id>.xsd as their schema location; every published model's schema also resolves at the same path on this host, https://sdcstudio.axius-sdc.com/dmlib/dm-<ct_id>.xsd.

Sharing, settling, certifying

• The public catalog API lists every published public component and model with no account, and an AI assistant reads the same catalog through the SDC MCP servers; with sdcmodeler and an API key it can also estimate, assemble and publish a model. See The Public Catalog API and The SDC MCP Servers.
• The Verifiable Settlement Layer settles an exchange between two parties and hands each a signed Settlement Receipt anyone can verify offline. See Verifiable Settlement.
• The SDC Certified Practitioner program has a free curriculum and a paid exam taken in SDCStudio; passing issues a numbered, signed credential. See The Practitioner Exam and Practitioner Credentials.

Money, access, languages

SDCStudio is paid for in credits, 1,000 per US dollar, spent per operation: a new component or cluster, a new model, the first application download of each model, a Settlement Receipt. Editing, publishing, generating a package and reusing components are free. A new account funds its wallet before the application opens. See Wallet and Billing.

Every account starts with one private project, named after you and set as its default. Each project has a team with owner, admin and member roles; see Teams. The application is in English, Spanish, Portuguese (Brazil) and French; see Interface Languages.

Where to go next

The Quick Start takes an hour. Your First Data Model walks the whole path with a sample file. When something fails, Troubleshooting is organized by the message you see. Questions go to contact@axius-sdc.com.