Aurora v0.1.47 — evidence-contract catch-up

Aurora v0.1.47

Catch-up release for the local-first, deterministic unknown-industry research engine.

Added

• Frontier Atlas adapter emits contract-shaped, hashed run reports whose findings state observations only.
• Prediction registration is limited to industry-candidate clusters with a usable empirical retention baseline; undersized baselines are skipped.

Changed

• Input manifest hash v2 covers normalized input contents, so confidence, numeric, unit, and metadata edits change the digest.
• Curated sample cases, engine metadata, README badge, and OpenAPI version are aligned to 0.1.47.

Fixed

• Declares the multipart dependency required by upload endpoints.
• Stabilizes property tests under Windows load without weakening assertions.

Verification

The exact tagged SHA passed the repository check_all gate, strict example/case/scorecard and adapter checks, CLI demo, and GitHub Actions.

Scope boundary

This remains alpha research software. It does not claim real-world early discovery of a named industry, investment performance, or production multi-tenant SaaS readiness. Atlas findings are observations, not investment advice.

v0.1.31 — StatCard + table document_id deep-links

Added

• Data Explorer observations table: document_id cell links to documents tab
• Dashboard coverage StatCards deep-link into Data Explorer tabs

Changed

• Engine / feature → 0.1.31
• README engine badge version aligned

v0.1.30 — has_document_id chips + open document

Added

• Data Explorer document_id chips: All / with doc / no doc (?has_document_id=true|false)
• Observation detail: clickable document_id + open document → (/data?tab=documents&q=…)
• Dashboard document_id coverage bar → ?has_document_id=true

Changed

• Engine / feature → 0.1.30

v0.1.29 — Dashboard quality deep-links

Added

• Data Explorer URL deep-links for shareable filters (tab, span, tier, types, q; also accepts missing_char_span=true)
• Dashboard Provenance quality panel links into Data Explorer (missing-on-doc, observations, documents)

Changed

• Engine / feature → 0.1.29

v0.1.28 — Missing-on-doc span filter

Added

• GET /api/observations?missing_char_span= / ?has_document_id=
• Data Explorer missing on doc chip for provenance gaps

Verify

• scripts/check_all.py --quick — all green

v0.1.27 — Provenance quality panel

Added

• Dashboard Provenance quality panel (coverage bars)
• Stats: char_span_ratio, document_link_ratio, observations_missing_char_span, documents_with_text

Verify

• scripts/check_all.py --quick — all green

v0.1.26 — lint char_span policy

Added

• lint_package --require-char-spans and --min-char-span-ratio R
• Soft fields char_span_ratio / observations_missing_char_span
• check_all wires hard/ratio policies per case

Verify

• scripts/check_all.py --quick — all green

v0.1.25 — Scorecard char_span floors

Added

• Scorecard gates min_observations_with_char_span and min_char_span_ratio
• iron-air-retro scorecard.json; all cases set span floors
• check_all runs iron-air-mini + retro scorecards

Verify

• scripts/check_all.py --quick — all green

v0.1.24 — Progressive char_span align

Added

• Progressive word/char prefix match in align_char_span
• append_unmatched for curated packages (appends residual excerpts)
• iron-air-retro: 21/21 observations with char_span

Verify

• scripts/check_all.py --quick — all green

v0.1.23 — Cases ship documents[]

Added

• Regenerated case packages with full documents[] (no orphan document_ids)
• Scorecard gates: min_documents, require_no_orphan_document_ids
• check_all lints all major cases/examples with --require-documents

Changed

• Engine 0.1.23
• retro / validate_package include documents[] on import

Verify

• scripts/check_all.py --quick — all green