Skip to content

Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

2.0.0 — 2026-11-01

First public release. These breaking changes landed before any Indicium package release. Existing downstream records may still embed the earlier context.

Breaking

  • The asb: vocabulary base moved to https://w3id.org/asb#. It had two earlier bases. 1.x used https://asb.holobiomics.org/ns/asb#, a domain that was never registered, so a third party could have taken it and served arbitrary definitions for this standard's term IRIs. Indicium PR #4 (2026-06-14) then moved it to https://holobiomicslab.eu/ns/asb#, and 2.0.0 settles on the permanent w3id identifier. Every asb: term IRI therefore changes. The 2.0.0 SHACL shapes target the w3id IRIs, so they do not recognise a claim typed with either earlier base: such a claim goes unchecked instead of failing. Released data carries the intermediate base: the 106 indicium/*.jsonld files of the asb-skill-collections pre-release metabolomics-v0.1.0 (2026-06-22) bind https://holobiomicslab.eu/ns/asb# and do not validate against the 2.0.0 shapes. To migrate such data, bind the asb: prefix of its JSON-LD context (or rewrite its IRIs) to https://w3id.org/asb#; the local names are unchanged. Then validate it with indicium.validate_graph. Consumers that bind the asb: prefix themselves, notably the shape files in indicium-adapters-metabolomics, must be updated in lockstep.
  • The distribution is renamed indicium → indicium-standard. The import package is unchanged: import indicium still works. The PyPI name indicium is held by an unrelated, abandoned key-value-store package that also ships a top-level indicium module; installing both into one environment clobbers the import. Depend on indicium-standard, never on indicium.
  • Two SEPIO bindings are corrected before the namespace deposit. supports_claim now expands to sepio:0000427 (is_evidence_line_for), an evidence-line-to-claim relation, instead of the oppositely directed sepio:0000007. Skill reliance asserts now expands to indicium:asserts, declared in the LinkML schema, instead of sepio:0000130 (authored_by). JSON keys and the local for_claim predicate are unchanged. Refresh embedded contexts and regenerate previously expanded RDF. Supporting, disputing or inconclusive direction is a separate sepio:0000183 annotation; the historical supports_claim key alone does not establish that direction.

Added

  • LICENSE (MIT). 1.x declared MIT in pyproject.toml but shipped no licence text, so a wheel would have distributed MIT-claimed code without the required notice.
  • indicium.ASB_BASE, indicium.INDICIUM_BASE, indicium.asb_term() — the vocabulary bases are now part of the public API, read from the generated model's prefix map so they cannot drift from the SHACL shapes. Consumers that hardcoded the IRI kept validating against a dead namespace after the base moved: the shapes matched nothing and every graph reported conforms=True. Import these instead.
  • tests/test_namespace_base.py — guards against reintroducing an unowned namespace base, and pins the metadata lookup to the declared distribution name.
  • A reproducible JSON-LD context generated from the schema, shipped in the package and copied into the documentation site. OBO prefixes expand in JSON-LD 1.1; property IRIs and enum meanings now agree with the SHACL shapes.
  • Local release checks and a Trusted Publishing workflow with matching tag, package version and dated changelog checks before publication.

Fixed

  • __version__ reported the wrong version. It queried importlib.metadata for the distribution indicium, which after the rename resolves either to nothing (falling back to 0.0.0+unknown) or to the unrelated squatting PyPI package, reporting its version as ours. It now queries indicium-standard. The wheel-smoke CI job asserts the installed wheel reports the built version.
  • docs/changelog.md is now generated from this file (make docs-changelog) rather than hand-maintained. It had been frozen at 1.0.0 for twelve releases, and release.yml extracts its release notes from it.
  • Installation examples now consistently name indicium-standard, including its optional extras. Schema generation works on Linux and macOS.
  • The generated OWL ontology uses the namespace root https://w3id.org/indicium as its document IRI and carries the version and the licence (pav:version, dct:license), which the LinkML source now declares; the model and the JSON Schema carry the version too. RDF serialization is deterministic for release verification.
  • The published SHACL file, docs/schema-assets/indicium.shapes.ttl (a release asset), now holds the supplemental shapes as well as the generated ones, so a SHACL processor run on it agrees with validate_graph: before, it accepted a claim with neither qualifier nor domain_qualifier.
  • The sdist includes only package source, schema assets and release metadata; internal development plans are excluded.
  • The package metadata names its author, states the licence as the SPDX expression MIT with LICENSE as its licence file (instead of the full text in the License field), and lists Documentation and Issues URLs.
  • SEPIO binding regression tests use an attributed excerpt of the ontology's inverse, domain and range axioms. The documentation explains the target link, separate evidence direction and the local reliance predicate.

Namespace registration

  • Canonical bases remain https://w3id.org/indicium/ and https://w3id.org/asb#. The vocabulary deposit and redirects are prepared separately; registration does not create a hosted store of claim instances.
  • Prepared term redirects send HTML clients to https://holobiomicslab.eu/indicium/reference/<Name>/ and RDF clients to the vocabulary representation under /ns/indicium/. The reference pages must be live before the w3id registration is submitted.
  • ASB validation assertions use SEPIO_0000183 with SEPIO_0000403, SEPIO_0000404 or SEPIO_0000405, and RO_0002558 with ECO_0000501. These are external assertion annotations, not Indicium claim-status values.

1.12.0 — 2026-06-10

  • New SourceTaskKind enum {reproduction, analysis, component_reconstruction, extension} + optional Claim.source_task_kind slot (asb:sourceTaskKind) — a claim-level reproducibility-provenance qualifier, sibling of Evidence.workflow_coupling (Suter et al. 2025) at claim↔source-task granularity. Canonical replacement for ASB's out-of-schema asbext:source_task_kind. Added to context.jsonld for portability. Additive, backward-compatible; SHACL non-closed.

1.11.0 — 2026-06-03

  • New SSSOMMapping class (SSSOM term mapping) + IndiciumDocument.sssom_mappings collection, with a derived_from lineage slot (cthoyt 2026-05-29). Adds sssom/semapv/skos prefixes. Additive, backward-compatible.

1.10.0 — 2026-06-03

  • Evidence gains optional workflow_coupling (WorkflowCoupling enum: tight/loose/ad_hoc/unknown) — a reproducibility qualifier per Suter et al. 2025 (DOI 10.1016/j.future.2025.107974). Additive, backward-compatible; SHACL non-closed.

[1.9.0] - 2026-05-29

Added

  • VerificationResult.char_span — a 5th, defaulted field (tuple[int, int] | None = None) giving (start, end) offsets into the matched candidate's original (un-normalized) text. Populated for verified matches from the regex span; best-effort for repaired matches (the recovered span is re-located in the original candidate, and may remain None). Lets callers highlight or re-slice the exact source region without re-searching.
  • verify_quote(..., punctuation_flexible=True) — an opt-in pass between exact containment and LCS that matches the quote's alphanumeric tokens ([A-Za-z0-9]+(?:\.[A-Za-z0-9]+)*) in order across arbitrary non-alphanumeric padding, with per-token boundary guards. A hit is verified with score = 1.0 and a char_span spanning first-token-start → last-token-end. This recognises a verbatim-modulo-punctuation quote (e.g. P=0.001 ~ P = 0.001,) instead of degrading it to a sub-threshold LCS, mirroring ASB claim_ledger's token-sequence matcher.

Why

Lets ASB claim_ledger and Perspicacité consume one shared verification kernel: char_span carries the exact source offsets downstream, and the opt-in punctuation-flexible pass matches numeric/statistical quotes whose punctuation shifts between extraction and source. Backward-compatible and additive only: both changes are defaulted off / defaulted None, existing thresholds and the 11 prior kernel tests are unchanged, and there are no changes to generated/model.py, the SHACL shapes, or any enum.

[1.8.0] - 2026-05-29

Added

  • verify_quote kernel (indicium.verify, exported from the package root) — a pure-Python, dependency-light, never-raises verifier that matches an LLM-extracted quote against source passage texts. Returns VerificationResult(status, matched_index, quote_exact, score) where status is verified (whitespace-flexible, case-sensitive containment; recovers the source's verbatim span, score = 1.0), repaired (no containment, but the longest common substring vs the best candidate clears near_threshold, default 0.9), or unverified. Structurally guarantees unverified ⇒ matched_index is None; degenerate input (empty/whitespace quote, empty candidates, or None/non-str entries) degrades safely instead of raising.
  • Exported verify_quote and VerificationResult from the package root.

Why

Lets downstream projects (Perspicacité R3, ASB) ground claim / parameter / failure-mode quotes against original source text — verifying provenance and catching paraphrase or hallucination — without depending on rdflib or the generated model. Additive only: no changes to generated/model.py, the SHACL shapes, or any enum.

[1.7.0] - 2026-05-29

Added

  • domain_qualifier slot (slot_uri: asb:domainQualifier) — an OPEN, snake_case-patterned (^[a-z][a-z0-9_]*$) qualifier slot for domain-adapter vocabularies (e.g. transcriptomics aligned_with, differentially_expressed_in). This is the standard's side-channel for domain extensions, mirroring how enrichment terms ride on asb:*_ontology_term. The closed Bucur SuperPatternQualifier enum on asb:qualifier is unchanged — domain terms never pollute it.
  • QualifierPresenceShape (supplemental SHACL, sh:or) — a Claim must carry at least one of asb:qualifier (closed Bucur SuperPattern) or asb:domainQualifier (open domain extension).

Changed

  • qualifier slot is no longer required on its own. A Claim previously had to carry a core Bucur qualifier; it may now satisfy the "at least one qualifier" rule via domain_qualifier instead. The closed enum and its maxCount 1 are retained; only the slot-level minCount 1 moved to the either-or QualifierPresenceShape.

Why

Resolves indicium#1 (reported by Dina, 2026-05-29). Domain-adapter qualifiers extended asb:qualifier directly, but that slot is a CLOSED 20-term LinkML enum (Bucur K-CAP 2021), so the generated sh:in shapes rejected every extended qualifier — blocking all domain adapters (metabolomics, metabokg, emi, transcriptomics) and the ASB benchmark. SHACL merges additively (validate_graph graph-union), so an adapter-side shape could never relax the existing sh:in; the fix had to live in the core standard. We chose to keep asb:qualifier pristine and route domain qualifiers to a new open slot (Reading 1) rather than dilute the published closed vocabulary.

Compatibility

Additive. v1.6 serialisations always carry a core asb:qualifier, so they still satisfy QualifierPresenceShape and remain valid. Consumers gain the domain_qualifier slot; no existing slot changed its slot_uri or range.

[1.6.1] - 2026-05-26

Fixed

  • Package context.jsonld for downstream consumers. Move docs/context.jsonld into src/indicium/schema/context.jsonld (canonical location) and symlink it back to docs/ so mkdocs continues to publish it. Wheels now ship the file, so consumers can load it via importlib.resources.files("indicium.schema") / "context.jsonld".

[1.6.0] - 2026-05-24

Added

  • Claim aliases: predicate (alias for relation) and super_pattern (alias for qualifier) — explicit names matching the ASB spec.
  • Evidence aliases + new slots: quote_exact (alias for quote), supports_claim (alias for for_claim, mapped to sepio:0000007), quote_location (string locator), eco_code (range: EcoCodeAsbSubset).
  • EcoCodeAsbSubset enum — 8 ECO codes ASB emits (textual_quotation, computational_evidence, expert_curation, inference_from_background_knowledge, author_statement, experimental_evidence, combinatorial_evidence, traceable_author_statement).
  • Source slots: doi, paragraph_id, xpath_pointer — explicit forms of the InformationSource locators ASB targets.
  • SoftwareEntity class — new class for globally-identified tools (canonical_url, version, edam_operation, edam_topics, bio_tools_id). Added to IndiciumDocument.software_entities.
  • JSON-LD @context at docs/context.jsonld — stable mapping for downstream consumers. To be served at w3id.org/indicium/context.jsonld pending the w3id PR; currently published at holobiomicslab.github.io/indicium/context.jsonld.
  • docs/sepio-assertion.md — reference for the SEPIO assertion pattern (skill → claim).
  • docs/uri-scheme.md — reference for indicium:claim/<uuid> + content-hash citation variant.
  • tests/fixtures/asb_skill_roundtrip/ — canonical fixture consumed by Plan 1c's round-trip CI test in indicium-adapters.

Compatibility

All Wave 1 changes are additive: every new slot is non-required and shares its slot_uri with an existing slot (where one exists). v1.5 serialisations remain valid; v1.6 consumers gain the spec-required names.

[1.5.0] - 2026-05-23

Added

  • extra_shapes keyword argument to validate_graph() — callers (e.g. domain adapters via SHACLProvider) can now inject additional SHACL shapes that are merged with the built-in schema shapes before validation.

Fixed

  • allow_warnings=True now passed to pyshacl in validate_graph() so that sh:Warning-severity shapes no longer set conforms=False; warnings are reported in the validation report text only.
  • Conditional SHACL constraint added: superseded_by is now required when claim_status = superseded.
  • Core types (Claim, Evidence, Source, and status enums) are now re-exported from the top-level indicium package for ergonomic imports.

Deprecated

  • indicium.adapters.astra.claim_to_insight — use astra_bridge instead.

[1.4.0] - 2026-05-20

Added

  • IndiciumDocument container type for grouping claims.
  • superseded_by field on Claim.
  • Fixed ClaimLink URI generation.

[1.3.0] - 2026-05-15

Added

  • ClaimLink, ClaimStatus types.
  • OA (Web Annotation) selectors for evidence provenance.
  • ASTRA bridge adapter.

[1.2.0] - 2026-05-10

Added

  • Script and DataAsset classes for code/data artefacts as evidence.

[1.1.0] - 2026-05-05

Added

  • Figure class and FigureType enum for figure-as-evidence support.