Skip to content
  • There are no suggestions because the search field is empty.

Release Notes - 3.06.0

Release Notes - 3.06.0

The 3.06.0 release focuses on enhancing the Knowledge Reserve with system-scoped items and progressive loading, maturing Copilot with proactive planning and human-in-the-loop gating, and advancing the Artifacts framework with a capability bridge for custom app deployment. It also introduces a new skill-run execution engine for agentic tasks and continues platform hardening with automated vulnerability analysis and security posture improvements.

Audience: Product owners, solution architects, documentation authors, and admins upgrading from 3.05.0

1. Highlights

  • Knowledge Reserve: system-scoped items with SYSTEM badge, progressive loading UI, and smart indexing for supporting documents
  • Copilot: proactive planning, human-in-the-loop gating/approval flow, and embeddable Copilot component
  • Artifacts & Custom Apps: PoC for host-mediated artifact capability bridge and finalized artifact hydration
  • Agentic Tasks: new skill-run execution engine with phases, steps, and project-scoped conversations
  • Platform & Security: automated vulnerability analysis, team alerts for vulnerabilities, and migration to Ruff

2. Knowledge Reserve

The Knowledge Reserve (KR) continues to evolve as Rhino’s shared project knowledge layer — the place where source material, editable configuration content, generated knowledge, and reusable skills come together for Copilot, routines, and artifact workflows.

  • System-scoped items: introduced support for system-scoped and prepackaged KR items, clearly identified with a SYSTEM badge in the UI.
  • Progressive loading: implemented progressive loading in the KR UI to improve responsiveness when handling large datasets.
  • Smart indexing & search: added smart indexing and search bar functionality to KR buckets, specifically for supporting documents, making it easier to locate specific assets.
  • Helm & S3 support: added Helm support for S3-backed KR mounts, improving deployment flexibility for cloud-native environments.
  • Path-tree navigation: fixed issues where KnowledgeReserveFolder.path rows were not project-scoped, which previously broke path-tree navigation and app selection.
  • File Viewer: added the ability to view files directly within the Supporting Doc bucket in the KR.

3. Documentation & Artifacts

  • Custom artifact deployment viability was proven through a host-mediated artifact capability bridge proof of concept.
  • Artifact hydration was finalized by merging the capability bridge and wiring task execution for upcoming SkillRun features.
  • Legacy hardcoded artifact creation can now be performed without requiring Copilot, providing a fallback generation path.
  • Artifact creation through Copilot now correctly enforces the expected admin gate for stronger access control.
  • The Artifact Viewer was improved with overview pages and better rail navigation for Use Cases and functional specifications.

4. Copilot

  • Proactive planning: Copilot now uses proactive planning instead of forcing write_todos on every request, leading to more natural and efficient task execution.
  • Human-in-the-loop gating: generalized the approval flow (ask-with-options) across Copilot contexts so users can review and approve agentic actions before they proceed.
  • Embeddable component: Copilot was extracted into an embeddable component, enabling reuse across different platform surfaces.
  • Reasoning visibility: thinking and reasoning blocks are now collapsed by default in responses to keep the chat interface cleaner while remaining available for inspection.
  • Execution fixes: resolved issues where Stop or Cancel did not halt LangChain (ASGI) agent turns and fixed crashes on empty model streams.
  • Model reliability: improved structured answer reliability and data-retrieval capabilities on Bedrock-backed models including gpt-oss and Nova.

5. Routines, Skill Runs & Execution

  • Skill-run engine: built a new execution engine for agentic tasks with support for phases, steps, and dedicated worker APIs.
  • Project scoping: skill-run conversations are now project-scoped instead of user-scoped, improving collaboration around shared automated work.
  • KR tool integration: added kr_index_search to exposed routine tools and fixed the DSL routine kr_glob wrapper to prevent execution failures.
  • Task board integration: wired the task board Run button directly to the Copilot chat panel for a more seamless workflow.

6. Platform & Admin

  • Vulnerability management: automated vulnerability analysis and implemented alerts to notify teams when vulnerability tickets are due.
  • Security hardening: addressed high- and medium-severity vulnerabilities in dependencies including DOMPurify, JS-YAML, and qs.
  • Linting & quality: migrated the codebase to Ruff for faster and more reliable linting and added broader integration test coverage for tools.
  • OAuth & CSP: fixed an issue where MCP OAuth authorization was blocked by Content Security Policy form-action settings in local-auth environments.
  • License management: resolved a bug where the license terms acknowledgment screen appeared empty.

Known Behaviors

  • While progressive loading improves responsiveness, extremely large Knowledge Reserve buckets may still show some latency during initial indexing.
  • Proactive planning in Copilot is optimized for standard workflows; highly complex multi-step requests may still occasionally require clarification.
  • Legacy artifact creation remains available, but it does not benefit from the latest hydration and capability bridge enhancements.