Release archive
3.2.41

SPRINT-07-HARDENING (Post-Neon Platform Closure) & PUBLIC-STACK-REDESIGN (Phase 1)

Release v3.2.41 focuses on SPRINT-07-HARDENING (Post-Neon Platform Closure), PUBLIC-STACK-REDESIGN (Phase 1), and POST-NEON-STABILITY (Runtime & Retrieval Recovery).

AddedFixedInternal
SPRINT-07-HARDENING (Post-Neon Platform Closure) & PUBLIC-STACK-REDESIGN (Phase 1)

Release digest

Added

  • SPRINT-07-HARDENING (Post-Neon Platform Closure) — Completed the main Sprint 7 hardening block after the Neon cutover:
    • docs retrieval now runs on the repository-owned Neon backend instead of the managed Vertex RAG path
    • runtime metadata deploy contracts, Secret Manager mapping, and post-cutover verification are synchronized across Cloud Run and job surfaces
    • the current merged-roadmap hardening track is now effectively closed, with only lower-priority legacy DB cleanup left in backlog
  • PUBLIC-STACK-REDESIGN (Phase 1) — Shipped the first premium redesign slice across the public product surfaces:
    • docs_site/ now uses a clearer product-led landing and aligned browser identity
    • dashboard_app/ now has a stronger entry surface, refined shell rhythm, and a first-class forensics entrypoint
    • docs, changelog, and public landing now read as one product family instead of disconnected surfaces

Fixed

  • POST-NEON-STABILITY (Runtime & Retrieval Recovery) — Closed the major production drifts discovered after the Neon migration:
    • normalized quoted DATABASE_URL params
    • recovered scheduled jobs packaging/runtime contracts
    • replaced deprecated grounded-search wiring
    • hardened retry behavior for transient Neon SSL connection drops
    • restored the production alerts router dependency path
  • PUBLIC-STACK-HOTFIXES (Redesign Contract Recovery) — Closed the follow-up regressions from the redesign push:
    • restored dashboard runtime/UI literal contracts expected by repository tests and smoke checks
    • fixed dashboard SPA cache behavior so lazy crypto/dashboard routes no longer require manual reload after deploys
    • re-synced knowledge-base snapshot contract drift
  • COST-OPTIMIZATION (Infra Cleanup) — Removed the largest idle GCP cost drivers without degrading core runtime behavior:
    • legacy Cloud SQL removed
    • GCP Memorystore Redis replaced with Upstash and verified
    • VPC connector removed after the Redis/public-egress cleanup
    • empty managed RAG corpus deleted after disabling runtime auto-create

Internal

  • Full verification baseline after the latest roadmap closure pass: 650 passed, 2 skipped, 22 deselected (pytest -q).
  • dashboard_app, docs_site, knowledge-base, and changelog-app all pass their build/lint or snapshot contracts locally.
  • GitHub Actions and post-deploy smoke were driven back to green after the redesign and runtime hotfix chain.
  • Synced current roadmap closure, main-agent handoff, and full verification readiness into canonical audit reports.
  • Reorganized report taxonomy and kept immutable evidence in the new docs/reports/audits/ and docs/reports/infra/ structure.
  • Regenerated changelog app content from the canonical root CHANGELOG.md.

Technical notes