Skip to content

Open Data Products Python SDK 0.7.0

Latest

Choose a tag to compare

@kyyberi kyyberi released this 09 Nov 14:31
· 2 commits to main since this release

Open Data Products Python SDK 0.7.0 Release Notes

DPROD 1.0 Beta 2 interoperability

  • Bundle the normative OMG 2026-06-01 ontology and SHACL artifacts for local,
    offline validation; artifacts retain source provenance and checksums.
  • Support bounded RDF ingestion, explicit DPROD product inventory, structured
    SHACL violations, DPROD-to-ODPS 4.2 drafts, and deterministic ODPS-to-DPROD
    Turtle/JSON-LD export with machine-readable mapping/loss records.
  • DPROD is an OMG in-process Beta 2 specification. Only
    https://www.omg.org/spec/DPROD/ is supported; the separate GitHub
    development namespace and experimental Data Contracts profile are excluded.
  • Explicit DPROD products are kept distinct from review-only DCAT dataset
    opportunities. No conversion treats ports, owners, policies, or service
    metadata as automatic ODPS promises, approval, or contract guarantees.
  • CI validates a generated export independently with Apache Jena SHACL after
    verifying the pinned binary release checksum; the built wheel is also loaded
    in a clean virtual environment before the source-suite test run.

DCAT Catalog Intelligence

  • Analyze supplied RDF/XML, Turtle, JSON-LD, N-Triples, N3, and other supported
    RDFLib serializations with open-data-products portfolio discover-dcat.
  • Preserve resource IRIs, language-tagged values, declared relationships,
    distributions, and distinct access/download URLs. Findings remain labelled as
    source facts, inferences, or review-only proposals.
  • Generate a self-contained offline HTML report. No catalog item is
    automatically approved as a data product.

DCAT-to-ODPS 4.2 drafts

  • Convert explicitly selected Dataset, DatasetSeries, or DataService resources
    with open-data-products product wrap-dcat.
  • Repeat --resource for batches. Reports record mappings, defaults, missing
    business context, validation, limitations, and AI-generated proposal fields.
  • --enrich and --full-draft mark generated business language as proposed;
    they do not establish ownership, demand, rights, commitments, or approval.

Safety and compatibility

  • DCAT 3 is the principal vocabulary; DCAT-AP 3.0.1 is the European target.
    Successful RDF parsing is not a DCAT-AP conformance claim.
  • HTTP(S) retrieval is bounded and rejects credential-bearing URLs. Remote
    JSON-LD contexts and automatic linked-resource crawling are disabled.
  • discover_dcat_catalog is a safe MCP tool; wrap_dcat_resource is
    state-changing because it writes an artifact.
  • Existing Data Contract Intelligence and ODCS-to-ODPS workflows are retained.

Data Contract Intelligence

  • Discover and inventory ODCS contracts from local folders or configured API
    sources, with source references and declared metadata retained as evidence.
  • Identify exact duplicates, schema overlap, shared technical resources, and
    shared domain/capability signals without treating inferred relationships as
    source facts.
  • Report metadata gaps, interface coverage and types, and contracts that have
    no current opportunity signal.
  • Classify review-only opportunities as standalone product candidates,
    boundary proposals, or rationalisation opportunities. No data product is
    created or approved automatically.
  • Produce the same evidence-backed intelligence through the Python API, CLI,
    MCP response, structured JSON artifacts, and a self-contained offline HTML
    report.

Explicit ODCS wrapper input

  • Use open-data-products product wrap-contract --input contract.yaml in all
    documented wrapper workflows.
  • The former positional contract input remains a compatibility alias, but it
    cannot be combined with --input.

MCP product drafting

  • wrap_odcs_contract is a state-changing MCP tool for writing a validated
    ODPS draft from an ODCS contract.
  • It accepts the same enrichment and full-draft options as the CLI and redacts
    local filesystem paths from its response.

ODCS-to-ODPS product drafting

  • Wrap ODCS 3.0, 3.1, and 3.2 contracts as validated ODPS 4.2 drafts with
    open-data-products product wrap-contract.
  • Use --enrich for a business-language description and discovery tags, or
    --enrich --full-draft for wider AI-generated ODPS product details.

ODPS v4.2 support

The SDK supports ODPS v4.2 while retaining v4.1 parsing and validation. It
understands profile-based Contracts and Data Access, including required default
profiles, output-port types, profile references, and malformed URL checks.
New generation and portfolio workflows produce v4.2 artifacts. The CLI and MCP
surfaces detect v4.1 and v4.2 through schema URI or version value.

Draft ODPV 1.1 interoperability support

This package also bundles ODPV 1.1 vocabulary and five interoperability profiles:
DCAT 3, ODCS 3, OpenLineage 1, PROV-O, and ODRL 2. ODPV 1.1 and every bundled
profile remain draft. Existing ODPV 1.0 behaviour remains the default.
The executable profile API is open_data_products.profiles; ODPV retains the
versioned vocabulary and source-profile assets.

ODPC v1.1 AI products

  • validate_ai_product validates a standalone AIProduct profile, including
    its purpose, structured ownership, governance, and evaluation requirements.
  • validate_ai_product_bundle checks that a standalone profile, an ODPC v1.1
    catalog entry, and an ODPG AIProduct node have consistent identities and
    profile references.
  • The unified Python agent API and open-data-products validate command
    automatically detect an ODPC v1.1 AIProduct from its canonical schema URI.

Other user-facing capabilities

  • Select ODPV 1.1 explicitly with load_vocabulary_release("1.1.0"); no
    existing call is upgraded silently.
  • Use open_data_products.profiles for profile validation, mapping resolution,
    and bounded transformations; importing those operations from
    open_data_products.odpv is not supported.
  • Validate an exact profile id and version, then resolve or explain mappings
    only with allow_draft=True. Every mapping result includes vocabulary and
    profile identity, lifecycle status, evidence, direction, and a SHA-256
    content digest.
  • Use odpv-profile-validate, odpv-profile-resolve, and
    odpv-profile-transform from the unified CLI, or use the safe MCP profile
    tools for validation, resolution, explanation, and bounded transformation.
  • DCAT 3, ODCS 3, OpenLineage 1, PROV-O, and ODRL 2 adapters execute only
    profile-permitted mappings. They preserve declared-versus-observed state and
    report missing bindings, missing policy rules, unsupported directions, and
    out-of-scope concepts as reviewable refusals or skips.

Compatibility and release confidence

ODPV 1.0 remains the default vocabulary and retains its existing URI and helper
behaviour. The test suite validates every bundled profile against the supplied
Draft 2020-12 schema, checks all five pinned mapping counts, and exercises
adapter direction, prerequisite, and provenance boundaries. CI remains the
final distribution-build check. Existing ODPC v1.0 catalog validation remains
unchanged; an AI product is a separate ODPC v1.1 profile.