v0.3.0

• Upgrade the HigherGraphen dependency from 0.5.0 to 0.7.1 and consume the
  higher-graphen-* crates from crates.io instead of a ../higher-graphen
  path checkout (CI no longer clones the upstream repo).
• Back projection-loss metrics with HigherGraphen's measure_projection_loss
  kernel and route untraced/empty output through it, retiring the hand-rolled
  loss heuristic. Behavior change: an all-untraced (Text) projection now
  reports risk_severity "medium" (was "low") with explicit
  source_trace_missing / unsupported_loss_metric review signals.
• Read gluing / difference / overlap classifications via typed HigherGraphen
  enums (GluingResult / DifferenceSeverity / OverlapWitnessKind) instead of
  matching serialized wire-name strings (output unchanged).
• Populate the lift Morphism's typed lost_structure from the source
  boundary's extraction_loss / excluded_summary, replacing the parallel
  untyped loss_claims blob.
• Enrich obstruction counterexamples with structural witness assignments
  (from_cell_id / to_cell_id / route_path / http_methods).
• Internal: split oversized modules so every source file satisfies the
  400-line cap (behavior-preserving; public APIs and golden output unchanged).

v0.1.3

0.1.3 - 2026-05-27

• Add micro review structure-risk triage for small AI answers, summaries, and compact PR notes.
• Add hypothesis-to-proposal evaluation coverage so unsupported hypotheses cannot become primary recommendations without observation work.
• Add medium/large PR review evaluation coverage proving broad review surfaces can be reduced to a bounded priority map.
• Update AdvisoryGraphen skills and acceptance docs for small, medium, and large review/proposal workflows.

v0.1.2

0.1.2 - 2026-05-23

• Upgrade HigherGraphen workspace dependency requirements to 0.5.0.
• Add review-gated completion dry-run gluing so accepted completions expose explicit policy overrides, blockers, and materialization requirements.
• Expose bounded HigherGraphen correspondence analysis in AI-agent projections, including emitted/omitted candidate counts, review focus candidates, and gluing summaries.
• Improve correspondence ranking so review focus prioritizes requirement, obstruction, and evidence participants over generic candidate-only similarities.
• Add an AdvisoryGraphen PR review skill that models PR diffs as bounded advisory snapshots and emphasizes AI authority, persistence, evidence-to-fact, public output/schema, and dependency/version boundaries.
• Update the dogfood example, CLI contract, workflow docs, schema artifact, and acceptance coverage for the HigherGraphen 0.5.0 review workflow.

v0.1.1

AdvisoryGraphen v0.1.1\n\n- Upgrades HigherGraphen dependency requirements to 0.4.1.\n- Exposes schema morphisms, projection loss metrics, and observation actions.\n- Updates the AI agent operation contract, docs, and skill guidance for the new support objects.\n- Adds release acceptance coverage for adversarial dogfood projection behavior.

AdvisoryGraphen 0.1.0

Initial crates.io release of AdvisoryGraphen.\n\nPublished crates:\n- advisorygraphen-core 0.1.0\n- advisorygraphen-interpretation 0.1.0\n- advisorygraphen-reasoning 0.1.0\n- advisorygraphen-projection 0.1.0\n- advisorygraphen-lift 0.1.0\n- advisorygraphen-runtime 0.1.0\n- advisorygraphen-cli 0.1.0\n\nInstall:\n\nsh\ncargo install advisorygraphen-cli\nadvisorygraphen version\n\n\nHighlights:\n- Evidence-backed advisory CLI and file-based workflows\n- Hypothesis-first and hypothesis-refinement proposal governance\n- AI-agent projection and review-gated completion workflows\n- Dogfood and CLI acceptance fixtures