@archpilotlabs/archpilot 0.2.5 → 0.2.7

package/CHANGELOG.md

@@ -1,5 +1,63 @@
 # Changelog
 
+## 0.2.7
+
+### Added
+
+- Added `archpilot validate --regressions` to focus validation output on architecture findings introduced since the accepted governance baseline.
+- Added regression summaries showing introduced, resolved, and existing architecture findings.
+- Added regression comparison data to JSON validation output for automation and tooling.
+- Added automatic initial governance baseline creation after the first successful full local validation, so regression tracking works without additional setup.
+
+### Improved
+
+- Improved CI and pull request reporting to highlight newly introduced architecture regressions instead of presenting existing architecture debt as newly actionable.
+- Improved GitHub inline annotations to prioritize newly introduced supported findings when a governance baseline is available.
+- Improved regression reporting with separate visibility for resolved findings and existing baseline findings.
+- Improved missing-baseline guidance to clearly distinguish the governance baseline from the structural architecture baseline.
+- Scoped `--changed` and `--diff` validation now avoids unsafe full-baseline comparisons that could incorrectly report findings as resolved.
+
+### Notes
+
+- Regression comparison uses the accepted governance baseline at `.archpilot/baseline/baseline.json`.
+- The initial governance baseline is created automatically only after a successful full local validation. Existing baselines are never automatically refreshed.
+- Use `archpilot baseline refresh` when you intentionally want to accept the current architecture findings as the new governance baseline.
+- `--regressions` does not change Architecture Health Score, Setup Readiness Score, rule severities, scoring, or normal validation findings.
+- `--regressions` does not currently combine with `--changed` or `--diff`.
+- Regression reporting does not independently change CI exit or policy enforcement behavior.
+
+---
+
+## 0.2.6
+
+### Improved
+
+- Improved Smart Init component and repository topology classification across complex monorepos and multi-module repositories.
+- Improved component classification accuracy across TypeScript/NestJS, Java/Spring, frontend, backend, library, multi-module, and mixed repository structures.
+- Improved API-design analysis precision for terminal or unconditional collection endpoints.
+- Improved event-architecture analysis for Spring listener declarations and inherited event types.
+- Improved data-query-risk handling for real-world SQL, JPQL, Spring Data repository, and query-builder patterns.
+- Bounded data-query-risk analysis for complex Java/Spring repository declarations to avoid pathological validation time.
+- Expanded regression coverage from validation against multiple real public repository shapes.
+
+### Fixed
+
+- Fixed runtime Nx projects and shared libraries being missed or misclassified in package-layout workspaces.
+- Fixed Maven runtime, library, support, distribution, and deploy-skipped modules being over- or under-promoted during Smart Init.
+- Fixed file-like candidate paths aborting Smart Init traversal.
+- Fixed terminal throw-only collection endpoints being reported as missing pagination.
+- Fixed Spring event listeners using annotation attributes, `ApplicationListener<T>`, or parent event types being treated as unhandled.
+- Fixed bounded Spring Data repository queries with `Pageable` or `Limit` parameters being reported as unbounded collection reads.
+- Fixed SQL-looking comments, non-query files, and Spring configuration files contributing noisy data-query-risk findings.
+
+### Notes
+
+- This release focuses on public-repository model accuracy and analyzer correctness.
+- These changes improve model and finding accuracy without changing rule IDs, severities, score weights, scoring caps, or normalization behavior.
+- Existing architecture configurations remain backward compatible.
+
+---
+
 ## 0.2.5
 
 ### Improved

package/README.md

@@ -45,7 +45,7 @@ ArchPilot turns architecture into something you can actually run and verify:
 - Generate architecture scores and reports
 - Surface architecture feedback in local development, PRs, and CI
 
-No runtime agents. No external services. Fully deterministic and local-first.
+**No cloud scan required. Deterministic and local-first.**
 
 ArchPilot works alongside GitHub Copilot, Claude Code, Codex, Cursor, and other AI coding tools by continuously validating architectural boundaries before architectural drift becomes technical debt.
 
@@ -172,6 +172,22 @@ Creates `.archpilot/` with architecture config, rules, and baseline artifacts.
 archpilot validate
 ```
 
+To focus on architecture changes introduced since the governance baseline:
+
+```bash
+archpilot validate --regressions
+```
+
+ArchPilot automatically creates the initial governance baseline after the first successful full local validation. Regression validation can then distinguish newly introduced, existing, and resolved architecture findings.
+
+When you intentionally accept the current architecture findings as the new reference, run:
+
+```bash
+archpilot baseline refresh
+```
+
+The governance baseline used for regression comparison is separate from the structural baseline used by `archpilot baseline create`.
+
 Example output:
 
 ```text
@@ -317,6 +333,5 @@ Proprietary. See LICENSE file for details.
 
 - Website: https://archpilot.org
 - Documentation: https://archpilot.org/docs
-- [Changelog](https://github.com/archpilotlabs/archpilot-vscode/blob/main/packages/cli/CHANGELOG.md)
 
 ---