Changelog

All notable changes to 森 (mori) will be documented in this file.

The format follows Keep a Changelog, and releases use Semantic Versioning.

Unreleased

0.34.0 - 2026-09-11

Added

• Workload planning with mori plan and batch baseline identity acceptance.
• Exact-score optimizations and differential, CLI and resource regression gates.
• Regression scenarios for modern language syntax, production selection, lower-token-floor ranking and immutable staged classification.
• Explicit production/test path overrides, bound to baseline and cache policy.
• Opt-in authenticated parsing reuse between workload planning and scanning, with fresh source reads, bounded private storage and safe cache misses.

Fixed

• Go expression allocation, TypeScript variance, Swift parenthesized range patterns and narrowly supported libgit2 annotation/iteration macros.
• Inline positively guarded REDIS_TEST fragment selection and unsupported Flow diagnostic guidance.
• Repeated small boilerplate crowding in review ranking, with structural scores and identities preserved. Concise reports disclose nested-body exclusions and active classification overrides.

Changed

• Configuration contract advances to 3. Prior managed contracts remain upgradeable without rewriting user policy. Report 22, baseline 4 and support-session 1 gain optional classification-policy evidence fields.

0.33.0 - 2026-09-11

Added

• Explicit per-scan diagnostic sessions and inspectable support bundles for debugging. Collection is off by default, export is separate, and no upload service or source attachment is included.
• Guided library scopes, copyable post-setup scan commands, future matching test/story exclusions, and bounded scope hints for noisy first reviews.
• Production/test fragment selection, including supported Rust test attributes, with excluded-fragment counts and explicit coverage reasons.
• A separate incomplete-scan artifact for candidate-pair limits, and an original provisional actionability corpus with positive and nearby negative examples.
• Published documentation with focused review and support-report instructions.

Fixed

• Command argument validation and help handling, so flags after positional roots no longer silently select unintended scan paths.
• JavaScript, TypeScript, TSX and Java grammar compatibility gaps, with visible parse diagnostics retained for unsupported syntax.
• VS Code handling of cancelled, stale, malformed and unsaved-buffer results.
• Recognition of narrowly identified generated gperf source and deterministic source snapshot handling.

Changed

• Report schema advances to 22 and configuration schema to 2. Normalization advances to 14, so prior accepted identities need explicit migration review. Supported prior official project contracts remain upgradeable. Project-contract schema stays 1, baseline schema stays 4, and receipt schema stays 2.
• Support sessions introduce their own version-1 sharing contract. Existing coarse feedback sharing remains schema 1 and retains its separate consent.
• Native release assets and published schemas include the support-session schema.

0.32.0 - 2026-09-10

Added

• Opt-in authenticated staged analysis reuse for unchanged snapshots, with private bounded cache storage and unchanged receipt/coverage enforcement.
• Explicit advisory staged review and hook policy, with strict behaviour retained by default and coverage requirements preserved in both modes.
• Opt-in, project-specific local feedback with separate CI consent, bounded storage and a minimized export preview. No feedback submission or network service is included.
• Actionable upgrade diagnostics for obsolete latest-release prerequisites, exhaustive review instructions and custom partial-staging restrictions.
• Offline summaries of explicitly exported feedback, with separate scan and selected-finding counts and no automatic ranking changes or submission.

Changed

• Report schema advances to 21 to distinguish staged policy outcomes from analysis completeness. Hook contract advances to v2; configuration schema 1, project-contract schema 1, receipt schema 2 and baseline schema 4 remain unchanged. Normalization advances to 13 for corrected Swift optional-type/nil-coalescing grammar; affected fingerprints and prior acceptance require review. Supported prior project contracts remain upgradeable.
• Agent summaries explain ranking, distinguish focused-file and background diagnostics, and summarize reasons for files without comparison fragments.
• Embedded guidance preserves standing authorization and describes advisory adoption without silently changing existing enforcement.
• Staged caching uses a bounded shared-location dictionary for larger reports and protected user-only Windows ACLs. CI covers all five release targets.

Fixed

• Narrow TypeScript/TSX ambient-interface import-type compatibility repair, preserving source offsets and malformed-function exclusion.
• SQLite PRAGMA, STRICT, CHECK expressions, GLOB and trigger syntax; C conditional initializer branches and bounded framework macro forms; Swift cast/coalescing parsing. Grammar sources include reproducible generation provenance.
• Explicit normalization 12-to-13 baseline migration preserves decisions after complete current analysis; ordinary scans still reject incompatible acceptance.
• Feedback consent cannot spill into expanded changed-code scopes, and scans without findings cannot be annotated as reviewed findings.

0.31.0 - 2026-08-20

Added

• Added hunk-aware changed and staged focus, so canonical staged review parses the full repository but scores only pairs that contain an exact changed-line fragment. Untracked and explicitly focused files retain whole-file focus.
• Added bounded --format agent --output <report.json> output, which keeps the complete bounded JSON evidence outside agent context while presenting at most 25 focused identities with provenance, coverage, warnings, and receipt state.
• Added a versioned .mori-project.json contract, deterministic schema-2 upgrade plans, safe managed-skill migration, and mori hook pre-commit as a thin canonical staged-review adapter.

Changed

• The embedded review skill is now a concise router with task-specific references, one-scan evidence reuse, and local project-upgrade guidance.
• Unsupported staged assets are recorded as focused path evidence without creating an other warning solely because canonical staged review included them for coverage accounting.
• Machine-readable reports advance to schema 20 for focused_only and exact changed-line evidence. Baseline schema remains 4 and normalization remains 12.
• The preferred release toolchain advances to Go 1.26.6 for the current standard-library security fixes.

0.30.0 - 2026-08-13

Added

• Added mori review staged check and mori review staged acknowledge as one canonical immutable-index workflow. Both commands now share focused-file inclusion and complete focused-coverage policy by construction, so hooks and one-commit acknowledgments cannot silently select different staged universes.
• Added a deterministic staged-review lifecycle dogfood gate to make check, covering ignored staged source, strict coverage, acknowledgment, compatible enforcement, and stale-index rejection.
• Staged-review receipt schema 2 records the focused inclusion and coverage contract and reports safe field-level differences when that contract drifts. Schema-1 receipts remain readable but fail stale across the Mori-version boundary. Report schema remains 19, baseline schema remains 4, and normalization remains 12.

Changed

• The earlier mori review acknowledge --staged ... form remains compatible but now points users toward the canonical staged-review workflow.

0.29.0 - 2026-08-11

Added

• Baseline schema 4 adds false-positive as a durable review classification. Schema-3 baselines remain readable for suppression with their profile evidence intact, but require explicit migration before mutation.
• Added local staged-review receipts, stored under private Git metadata by default and owner-only on POSIX filesystems, that bind an explicit focused finding acknowledgment to the exact HEAD, Git-index digest, scan profile, Mori and normalization versions, and complete sorted finding identities. Compatible receipts change only focused-match policy exit status; findings remain visible. Machine-readable reports advance to schema 19 to expose compatible receipt evidence.
• Added original MIT Swift calibration cases for presentation chains and memberwise initializer shapes, including a renamed positive and reviewed unrelated false positives. They document the current structural limit; normalization remains version 12 because identifier or literal heuristics would not justify a safe scoring change.

0.28.1 - 2026-08-11

Fixed

• Staged scans now load configured or explicit baselines from the same stage-zero Git-index snapshot as source, ignore rules, and configuration. Missing, external, non-regular, oversized, or malformed staged baselines fail closed instead of falling back to working-tree bytes. Report schema remained 18, baseline schema remained 3, and normalization remained 12.

0.28.0 - 2026-08-10

Added

• Added immutable --staged scans that read tracked source, ignore rules, and configuration from one Git-index snapshot, record its digest, and exclude unstaged and untracked content by construction.
• Added --include-focused and --require-focused-coverage, with per-path analyzed, deleted, excluded, unsupported, resource-limited, or undiscovered evidence and coverage exit status 4.
• Added named .mori.json scopes, compact one-line-per-group output, and mori project upgrade --check|--dry-run|--apply for coordinated local version-pin and Agent Skill maintenance with recoverable backups.

Changed

• Review ranking no longer treats generic entry-point, constructor, or anonymous callable names as distinctive same-name evidence.
• Source discovery now prunes ignored build trees before loading irrelevant nested ignore files while retaining deterministic negation behavior.
• Machine-readable reports now use schema version 18 for immutable input provenance, named scope evidence, and exact focused-file coverage. Baseline schema remains version 3 and normalization remains version 12.

0.27.0 - 2026-08-10

Added

• Added guided mori setup and mori configure flows with safe previews, explicit writes, strict answer files, and a read-only JSON protocol for project coding agents.
• Added mori doctor, mori inspect, mori config validate, and provenance- aware effective-configuration output for first-run diagnosis and ongoing maintenance.
• Added mori explain, self-contained HTML reports, restrained terminal color, and opt-in deterministic --redact-paths placeholders for reports that cross a project boundary.
• Added a bounded local Language Server Protocol implementation and a packaged, dependency-free VS Code reference client for diagnostics on unsaved buffers.
• Added C, C++, Dart, GDScript, Kotlin, Lua, Luau, PowerShell, and Ruby parsing, function-boundary extraction, ABI checks, editor activation, documentation, and calibrated cross-language positive and nearby-negative fixtures.
• Added checksum-pinned Homebrew, Scoop, and WinGet release manifests, an SPDX source-dependency SBOM, and signed GitHub/Sigstore attestations for release assets and the checksum manifest.
• Added a compact new-user README, task-oriented documentation, a Mori favicon, and responsive forest-themed project artwork.

Changed

• Normalization now uses version 12, with canonical mappings required by the newly supported grammars. Existing baselines require deliberate review and regeneration because scores and fingerprints can change. Report schema remains version 17 and baseline schema remains version 3.
• Release verification now checks native archives, the portable Agent Skill, the JSON Schema, the VS Code package, package-manager metadata, checksums, and representative provenance attestations before completion.

0.26.0 - 2026-08-10

Added

• Added PHP support for .php and .phtml using the official Tree-sitter PHP grammar, covering implemented functions, methods, anonymous functions, and arrow functions.
• Added Hack support for .hack and bounded legacy <?hh .php detection, using a checksum-verified MIT-licensed generated-source snapshot of the archived Slack Tree-sitter Hack grammar.
• Added PHP/Hack positive and nearby-negative calibration cases, examples, malformed-source and legacy-header coverage, and both language IDs to the reference VS Code client.

Changed

• Normalization now uses version 11, with canonical PHP/Hack syntax mappings and curated PHP standard-library operation aliases. Existing baselines require deliberate review and regeneration because scores and fingerprints can change. Report schema remains version 17 and baseline schema remains version 3.

0.25.0 - 2026-08-10

Added

• Added Java support for implemented methods, constructors, compact constructors, and lambdas using the official Tree-sitter Java grammar.
• Added C# support for implemented methods, constructors, destructors, operators, accessors, local functions, anonymous methods, and lambdas using the official Tree-sitter C# grammar.
• Added a redistributable Java/C# calibration case, a nearby-negative fixture, malformed-source coverage, and both languages to the reference VS Code client.

Changed

• Normalization now uses version 10, with canonical Java/C# syntax mappings, transparent redundant parentheses, and a bounded anonymous member shape for qualified Java method calls. Existing baselines require deliberate review and regeneration because scores and fingerprints can change. Report schema remains version 17 and baseline schema remains version 3.

0.24.0 - 2026-08-10

Added

• Added a redistributable ten-case labeled calibration corpus spanning code, shell, generic SQL, and PostgreSQL, with deterministic precision-at-k, actionability, score-distribution, fragment-size, and ranking-movement reporting against a pinned v0.22.0 reference.
• Added deterministic SARIF 2.1.0 output with stable similarity and incomplete- analysis rules, source locations, fingerprints, and exact suppression, warning, and truncation metadata.
• Added bounded --stdin-path overlays so editor clients can analyze unsaved source through standard input while retaining repository-wide comparisons and automatically focusing the overlaid file.
• Added a dependency-free reference VS Code extension and a published Draft 2020-12 JSON Schema for the machine-readable report contract.

Changed

• Machine-readable reports now use schema version 17 and record the optional stdin overlay path. Normalization remains version 9 and baseline schema remains version 3 because scores, fingerprints, and accepted identities are unchanged.

0.23.0 - 2026-08-10

Added

• Added low-weight canonical statement-position features and bounded anonymous callee-role features that distinguish calls exchanged between control and linear positions without serializing callee names or name digests.

Changed

• Normalization now uses version 9. Existing baselines require deliberate review and regeneration because fingerprints and scores can change. Report schema remains version 16 and baseline schema remains version 3.

0.22.0 - 2026-08-10

Added

• Added exact weighted intersection/union evidence and bounded deterministic profile-aligned directional feature differences for every retained group.

Changed

• Text output now labels 100% as normalized feature identity and explicitly states that it is not proof of semantic or behavioral equivalence.
• Machine-readable reports now use schema version 16. Normalization remains version 8 and baseline schema remains version 3 because scoring, fingerprints, and acceptance compatibility are unchanged.

0.21.0 - 2026-08-10

Added

• Added selective baseline add, baseline remove, and baseline edit operations with durable review notes and classifications.
• Added explicit baseline migrate --accept-profile, deterministic scan-profile digests, profile compatibility reporting, and reviewed warning exceptions for complete baseline operations. Digests include exact loaded ignore-file paths and SHA-256 content evidence.

Changed

• baseline update is now preview-only unless --accept-all is explicit, and retained entries preserve their review metadata.
• Machine-readable reports now use schema version 15. Baselines now use schema version 3; schemas 1 and 2 remain readable but require explicit migration before mutation. Normalization remains version 8.

0.20.0 - 2026-08-10

Added

• Added --min-file-coverage, --max-zero-fragment-files, --fail-on-warning, and --fail-on-parse-diagnostic, with equivalent strict project configuration. Scan and baseline operations evaluate the same policies before returning success or modifying a baseline.
• Added deterministic coverage-summary totals, aggregate unsupported-extension counts, pre-token-floor boundary counts, and machine-readable reasons for every zero-fragment supported file. Linked-worktree .git control files are treated as VCS metadata rather than unsupported source.

Changed

• Machine-readable reports now use schema version 14. Normalization remains version 8 and baseline schema remains version 2.

0.19.1 - 2026-08-09

Fixed

• Recognized the standard SwiftProtobuf generated-source header so --exclude-generated omits compiler-generated .pb.swift files without broadening classification to arbitrary generator comments.
• Extended bounded, byte-preserving Swift parser compatibility to valid conditional-cast and nil-coalescing chains with optional dictionary bases, collection types, labels, returns, and repeated fallbacks. Repaired trees are still accepted only when they reduce parser diagnostics.

0.19.0 - 2026-08-09

Added

• Added explicit --embedded-sql extraction for direct string arguments to a bounded set of Go database method names. Extracted queries use the selected SQL dialect, retain host literal locations and parent functions, group multi-statement strings as one unit, and expose parser or resource-limit warnings.
• Added opt-in fixed-size block comparison fragments inside code functions, with configurable statement-window and per-function bounds, duplicate-span removal, direct function-parent linkage, and same-file overlap suppression.

Changed

• Machine-readable reports now use schema version 13 and record every embedded SQL and statement-block option. Normalization version 8 covers the expanded opt-in comparison-unit contract; existing baselines must be reviewed and regenerated. Baseline schema remains version 2.

0.18.0 - 2026-08-09

Added

• Added source-free literal-position evidence to match groups. Reports disclose bounded difference counts without serializing literal values or digests and without changing scores, fingerprints, ordering, or baselines.
• Added repeatable --priority-path GLOB=WEIGHT and priority_paths configuration for deterministic, presentation-only project review priority.

Changed

• Machine-readable reports now use schema version 12 for literal evidence and effective priority-path rules. Normalization remains version 7 and baseline schema remains version 2.

0.17.1 - 2026-08-09

Fixed

• Recognized exact Typeshare generated-source headers, including multiline block comments, so --exclude-generated can omit those files without broadening classification to arbitrary Generated by comments.
• Added bounded, byte-preserving Swift parser compatibility for valid optional async bindings, awaited switches, empty-tuple call arguments, and conditional casts followed by nil coalescing. Repaired trees are accepted only when they reduce diagnostics; nearby malformed syntax remains visible.

0.17.0 - 2026-08-09

Added

• Added review, explore, and sql scan profiles with deterministic defaults and explicit config/CLI precedence.
• Added mori init for deterministic .mori.json scaffolding, including stdout-only inspection, no-clobber creation, and explicit forced replacement of regular files.

Changed

• Machine-readable reports now use schema version 11 and record the selected scan profile. Normalization remains version 7 and baseline schema remains version 2.
• Mori's bundled review skill and self-review config now use the conservative review profile while keeping project exclusions explicit.

0.16.0 - 2026-08-09

Added

• Added whole-file script fragments for Bash/POSIX shell and Zsh. Script fragments cover top-level executable statements while excluding function bodies, which remain independent function fragments.
• Added positive and nearby-negative cross-dialect fixtures for shell top-level normalization.
• mori languages now lists every fragment kind emitted by each grammar.

Changed

• Normalization version 7 introduces shell script comparison units. Existing baselines must be reviewed and regenerated. Report schema remains version 10.
• Candidate partitioning now requires compatible fragment kinds, so shell scripts are never compared with functions.
• Text output distinguishes top-level script bodies from outer function bodies when it discloses separately analyzed functions.

0.15.0 - 2026-08-09

Added

• Added opt-in --ranking review / ranking: "review" ordering with explicit review_priority and review_signals for cross-file, cross-directory, same-name, and repeated-location evidence.

Changed

• Machine-readable reports now use schema version 10 and record the effective ranking mode. Review ranking changes presentation order only; structural scores, normalization version 6, fingerprints, and baselines are unchanged.

0.14.0 - 2026-08-09

Added

• Added deterministic per-file coverage inventories to JSON reports, including fragment, skipped-fragment, and parse-diagnostic counts for every analyzed supported file.
• Added conservative generated-source header classification and opt-in --exclude-generated / exclude_generated filtering. Excluded generated files remain visible in the per-file coverage inventory.

Changed

• Machine-readable reports now use schema version 9. Text reports summarize files that produced no comparison fragments at the configured token floor. Normalization remains version 6 and existing baselines remain compatible.
• Text results now place a group-level warning immediately below any score whose retained occurrences exclude nested function bodies.
• The bundled review skill now requires schema-9 per-file coverage inspection, makes production/test separation a standard first step, and documents generated-source classification and exclusion.

0.13.0 - 2026-08-09

Added

• Added repeatable --changed-worktree PATH=REVISION focus for nested and sibling Git worktrees with independently resolved revisions and local working-tree state.

Changed

• Machine-readable reports now use schema version 8 and record deterministic per-worktree focus provenance. Normalization remains version 6 and existing baselines remain compatible.
• Updated CLI help, repository guidance, and the bundled review skill for explicit multi-worktree and submodule review.

0.12.0 - 2026-08-09

Added

• Added explicit --sql-dialect postgresql and sql_dialect configuration with a dedicated PostgreSQL 18.3 parser for top-level query comparison.
• Added checksum-pinned PostgreSQL generated-source provenance and positive/nearby-negative query fixtures.

Changed

• Machine-readable reports now use schema version 7 and record the effective SQL dialect.
• Normalization version 6 maps PostgreSQL query structure into Mori's shared SQL vocabulary. Existing baselines must be reviewed and regenerated.
• The bundled review skill now requires explicit PostgreSQL selection and separate profiles for mixed-dialect repositories.

0.11.0 - 2026-08-09

Added

• Added Swift function similarity for .swift files, covering implemented functions, initializers, deinitializers, and closures with a pinned, checksum-verified generated Tree-sitter grammar.
• Added Swift-to-Go positive and nearby negative validation fixtures.

Changed

• Normalization version 5 maps Swift declarations, expressions, arguments, identifiers, and control transfers into the shared code vocabulary. Existing baselines must be reviewed and regenerated.
• The bundled review skill now describes Swift comparison boundaries and the required warning disclosure.

Fixed

• Function fragments nested below explicit parser error nodes are now rejected instead of being scored from recovered malformed syntax.

0.10.0 - 2026-08-09

Added

• Added dedicated Bash/POSIX-shell and Zsh function parsers with .sh, .bash, and .zsh discovery.
• Added bounded, non-executing shebang detection for extensionless Bash/POSIX, Zsh, Python, and Node.js scripts.
• Added shebang interpreter names to mori languages and help for mori languages --help.

Changed

• Normalization version 4 aligns shell word, variable-reference, and glob node vocabulary across the Bash and Zsh grammars. Existing baselines must be reviewed and regenerated.
• The bundled review skill now explains shell-family selection and shebang coverage.

0.9.0 - 2026-08-09

Added

• Added deterministic coverage warnings for scans with no supported files or no extracted comparison fragments.
• Added --require-coverage and the equivalent require_coverage project setting, which exit with status 4 when a scan cannot support a similarity assessment.

Changed

• Corrected the schema-6 example, clarified release source-date formatting, labeled abridged output, and linked the scan-selection contract directly.
• Added the recommended 40-token starting floor to the scoring guide's cross-language calibration guidance.
• Consolidated deterministic analyzer partitioning and lexical path containment without changing comparison behavior.
• Added an explicit, non-discovered self-review profile and make dogfood workflow for production-code dogfooding.
• Added language preflight, zero-coverage, and nested Git boundary guidance to the bundled review skill.
• Nested-worktree focus errors now identify the repository-relative boundary and recommend a separate scan or explicit focus paths.

Fixed

• mori skill --help, mori skill -h, and mori skill help now print skill usage successfully instead of reporting an unknown skill command.

0.8.0 - 2026-08-09

Added

• Direct --comparison-domain filtering, applied before parsing and recorded in schema-6 reports and project configuration.
• --same-language-only review-family filtering, including TypeScript-to-TSX comparisons within their shared family.
• A detailed scan-selection contract covering validation, execution, compatibility, diagnostics, and verification.

Changed

• The bundled review skill now starts with code-only same-language scans and a 40-token cross-language floor, while reserving 12 tokens for broad exploration.
• Parse warnings now describe potentially incomplete comparison coverage without claiming that non-comparison DDL fragments were invalid or skipped.

Fixed

• Code and SQL findings no longer need to share one root ranking when a comparison domain is selected.

0.7.0 - 2026-08-09

Added

• Byte-preserving parser support for SQLC and SQLite LIMIT/OFFSET parameters and SQLite ON CONFLICT column targets.

Changed

• Normalization version 3 classifies recovered pagination placeholders as parameters while preserving schema version 5 and existing code features; existing baselines must be reviewed and regenerated.

Fixed

• Recognized raw ampersands in JSX text no longer invalidate otherwise valid JavaScript or TSX function fragments; malformed JSX expressions still fail closed.

0.6.1 - 2026-08-09

Fixed

• Clarified that version-pinned source builds can lack revision and source-date metadata, and limited provenance-incomplete reports to explicitly disclosed exploratory review.

0.6.0 - 2026-08-09

Added

• Top-level SQL query similarity for .sql files, including SQLC display names, query-specific structural normalization, and visible parse recovery.

Changed

• Machine-readable reports now use schema version 5 with explicit comparison domains and fragment kinds; normalization version 2 adds SQL features while preserving established code fingerprints and scores.
• The bundled review skill now provides a separate SQL scan profile and query review safety guidance.

0.5.0 - 2026-08-09

Added

• Deterministic tool provenance in schema-4 JSON reports, including full source revision, build state, platform, Go version, and normalization version.
• Repeatable explicit path focus and local Git changed-file focus with exact group metadata, focused-first bounded retrieval, and a focused CI policy.

Changed

• The bundled review skill now validates report provenance and uses Mori's native changed-file focus without narrowing the scan to changed files.

0.4.1 - 2026-08-08

Changed

• Clarified explicit binary-path use, mutually exclusive cross-language filters, and project configuration before baselining in the bundled review skill.

0.4.0 - 2026-08-08

Added

• Strict .mori.json project configuration with reproducible effective-option metadata.
• Nested .gitignore and .moriignore discovery with explicit-path behavior and --no-ignore/--no-config controls.
• Language-family-aware cross-language scans and explicit repeatable --language-pair filters.
• Content- and path-scoped baseline schema 2 with schema-1 content-scope compatibility.
• Structured parse diagnostics, nested-fragment metadata, and concise shared shape summaries.

Changed

• Machine-readable reports now use schema version 3 and group qualifying source-location pairs by stable content-pair identity.
• Equal-score groups now prefer larger shared evidence and wider occurrence coverage before stable identity ordering.
• The bundled review skill now uses bounded progressive group retrieval and explains configuration, baseline scope, parser, and nested-function limits.

0.3.0 - 2026-08-08

Added

• Stable fragment and match identities for review workflows.
• Versioned baseline update, scan suppression, and stale-entry pruning.

Changed

• Machine-readable reports now use schema version 2 and include stable fragment/match identities and baseline suppression counts.
• Updated the CodeQL workflow actions to v4.37.4.

0.2.1 - 2026-08-02

Changed

• Upgraded the Tree-sitter Go binding to 0.25.0 and compatible ABI-15 grammar releases for Go, JavaScript, Python, and Rust.
• Kept grammar-only Go statement-list wrappers out of fingerprints so parser generator changes do not distort structural scores.
• Grouped future Tree-sitter Dependabot updates so binding and grammar ABI changes are reviewed together.

0.2.0 - 2026-08-02

Added

• Official mori-review-similarity Agent Skill for evidence-led duplicate and refactoring review.
• Offline mori skill install command for project, global, and custom skill directories.
• Deterministic platform-neutral Agent Skill archive in each release.

0.1.0 - 2026-08-01

Added

• Go CLI with deterministic text and JSON reports.
• Tree-sitter adapters for Go, JavaScript/JSX, TypeScript/TSX, Python, and Rust.
• Cross-language canonical AST features and weighted Jaccard scoring.
• Bounded source discovery, size-pruned comparisons, and CI failure mode.
• Native, immutable-release-compatible GitHub build pipeline.