Skip to content

v3.10.7 - Release

Latest

Choose a tag to compare

@vypbot vypbot released this 31 Oct 00:36
3e7ceea
  • SDD catalogs and preparation: preserve nested specification directories and accepted filename destinations; publish catalogs without following links or overwriting aliases; reject incomplete symlink discovery; support truthful proposed-capability bootstrap and actionable legacy/missing-trusted-catalog recovery; render accepted capability titles as literal text. References: #768, #771, #777, #779, #782, #789, #804, #808, #820.
  • Review grounding and regressions: improve evidence for Kotlin/Wasm bindings, same-repository Action references, relative SDD links, dependency-generated transport values, generated-source coverage coordinates, initial browser-session credentials, unittest discovery, Kotlin Json policy and Ktor routing-root identity. Preserve genuine-defect controls and distinguish source/runtime evidence. References: #764, #770, #772, #784, #796, #798, #806, #807, #809.
  • Large comparisons and diagnostics: support complete bounded comparison assessment after file-recovery exhaustion and expose invalid completed progress results through actionable diagnostics. References: #773, #774.
  • Native verification: keep JVM/Gradle account configuration inside private verification profiles and propagate the supported library path needed by native Python toolcache interpreters. References: #775, #776.
  • Workflow lifecycle: retain canceled PR-link recovery identity, preserve closing references consistently across description/approval, and complete release issues only after publication and development reconciliation. References: #785, #792, #803.
  • Operator recovery: document complete paginated job inventory and reliable per-job log collection when advisory checks have no execution log, preserving real workflow failures and partial diagnostics. Reference: #817.

Quality evidence and measured exact-line-location limitations remain documented in quality/bugbot/evidence/. Deterministic/reference regression results are separate from actual model observations. This release does not establish consumer runtime adoption or resolve unrelated consumer business defects.