Skip to content

Spring AI Business Copilot v2.2.0

Choose a tag to compare

View all tags
@qcodingdev qcodingdev released this 29 Jul 12:06
· 36 commits to main since this release
v2.2.0
71b40c4

[2.2.0] - 2026-07-29

Added

  • Completed the 2.2.0 enterprise quality loop with stable Knowledge answer IDs, owner-bound idempotent helpful/not-helpful feedback, and an Admin/Reviewer quality review queue.
  • Added Flyway V22 for constrained Knowledge answer feedback and workbench controls that keep users in the current answer flow with an explicit reviewer next step.
  • Added Flyway V23, concurrency-safe Knowledge quality dispositions, automatic re-queue after newer feedback, and low-cardinality quality metrics.
  • Added Flyway V24-V28 and enterprise APIs for Data governance/export/Report handoff, Knowledge source sync, Support ticket/SLA/writeback integration, Report schedules/office exports, and HR consent/interview/ATS/onboarding collaboration.
  • Added bounded XLSX/HTML extraction, mounted-drive and S3/MinIO sources, and REST adapters for SharePoint, Confluence, Notion, Jira Service Management, Zendesk, ServiceNow, Feishu, WeCom, and common ATS providers.
  • Added environment-reference-only external credentials, configurable Data query-plan row budgets, active JDBC cancellation, and private Admin enterprise status counts.
  • Added fixed low-cardinality AI call metrics, call IDs, operation tags, and Chinese request-to-model chain logs across all five Copilots.
  • Added bounded AI concurrency, separate Chat/Embedding circuit breakers, explicit provider timeouts, and restricted transient-failure retries.
  • Expanded the five fixed evaluation datasets to 67 cases and added a CI dataset-size gate, including end-to-end Data SQL safety cases.
  • Added authenticated capacity smoke testing, non-overwriting PostgreSQL backups, and disposable-container restore drills.

Fixed

  • Extended Resume protected-attribute detection to reject Chinese gender and marital-status requirements and common English age or gender restrictions.
  • Masked external Support ticket messages before persistence and made Data query-cost previews fail closed when database estimates cannot be recognized.
  • Kept Support knowledge retrieval inside the authenticated ACL boundary while falling back across document categories when a category-specific lookup has no evidence.
  • Made HR job drafts preserve recruiter-provided required and preferred qualifications deterministically and reject model-added qualifications.
  • Corrected immutable AI property binding and made Admin diagnostics distinguish provider-omitted token usage from zero usage.

Changed

  • Updated both README files, the six-frame workbench GIF, the social preview, and the 2.2 enterprise diagnostics screenshot from a running local application.
  • Added five-module enterprise scenario tests plus real PostgreSQL constraint coverage for Data, Knowledge, Support, Report, and HR boundaries.

Security

  • Bound Knowledge feedback to the authenticated creator of the persisted answer audit and restricted the cross-user quality queue to Admin/Reviewer roles.
  • Masked phone numbers, email addresses, ID cards, and credential assignments in feedback comments and reviewer notes before persistence.
  • Restricted Actuator metrics to Admin and Reviewer roles and kept user input, Prompt text, and file names out of metric labels and AI operation logs.
  • Upgraded the MinIO Java client to 8.6.0 for CVE-2025-59952, pinned Bouncy Castle 1.81.1 for CVE-2025-14813, and selected OkHttp JVM 5.1.0 explicitly for Maven runtime compatibility.
Assets 2
Loading