Rhino Release Notes

Rhino 3.10.0 focuses on ontology-isolated graph modeling, targetless compute routines, improved routine governance, agentic skill tracing, and expanded model support.

Audience: Product owners, solution architects, documentation authors, administrators, and technical teams upgrading from 3.09.0.

Highlights

Isolated Ontologies: Model As-Is and To-Be graph states independently, with lineage preserved through DERIVED_FROM relationships.
Targetless Compute Routines: Run routines without a repository or asset target, pass caller-supplied inputs, and return structured values directly.
Add Routines and Artifacts to System Reserve: Added promote/demote and enable/disable controls for admins for Routines and Artifacts from the System Reserve, along with a redesigned UI for Routines + Routines History.
Agentic Skill Tracing: Inspect durable execution steps, nested skill and routine calls, and Knowledge Reserve usage in Trace Explorer.
Model Catalog Expansion: Added GPT-5.6 Sol for Copilot and GPT-5.6 Terra for Routines, Graph, and Extraction workflows.
Security and Reliability: Hardened OAuth hand-offs, fixed Copilot resume loops, resolved worker connection leaks, and patched dependency vulnerabilities.

What changedIsolated Ontologies and Graph Explorer filtering

3.10.0 introduces first-class ontology boundaries for safely modeling target states without modifying extracted baseline data.

Added an ontology domain entity and CRUD API for source and target ontologies.
Added project cascading and ontology lineage tracking.
Added subgraph copying for DO instances and relationships, including new logical IDs and DERIVED_FROM lineage.
Added ontology_id filtering to graph reads and writes.
Updated Graph Explorer with an ontology selector for As-Is and To-Be views.

Why it matters: Teams can model prospective architectures, migration options, and App Preview states without changing the integrity of source extraction data.

Compute routines, return values, and targetless execution

Routines no longer need to target an asset or write exclusively to the graph or Knowledge Reserve.

Added caller-supplied inputs maps to routine launches.
Allowed routines to run without a repository path or asset ID.
Added rhino.result support for a single JSON-serializable return value.
Persisted results in the routine_run record.
Added direct inline returns for fast compute routines.
Added formatted, read-only result viewing in run details.
Added global and per-routine execution timeouts to prevent runaway workers.

Routines and Artifacts management in the System Reserve for admins

Routine and artifact management now follows the project and System Reserve patterns used by other managed items.

Added promote and demote support for routines and custom artifact types.
Added administrator enable and disable controls for system-scoped routines and artifact types.
Enforced disabled status across discovery listings, MCP tools, and runtime execution.
Added role-aware Copilot guidance for disabled routines and artifact types.
Moved Routines from a modal to a dedicated project route.
Added Routines and History tabs with sorting and pagination.
Separated routine Name and Purpose into distinct table columns.
Added promote and demote flows between project and system scope.

Agentic skill tracing and model expansion

Added durable execution traces for agentic skill steps and nested routine calls.
Tracked Knowledge Reserve usage and attribution across interactive and background workflows.
Added Trace Explorer integration for inspecting agent runs and tool invocations.
Added gpt-5.6-sol through the OpenAI Responses API, prioritized for Copilot.
Added gpt-5.6-terra through the OpenAI Responses API, prioritized for Routines, Graph, and Extraction.
Added AWS Bedrock Claude support for Cursor cloud agent test harnesses.
Added durable checkpointing and resume support to multi-model Copilot evaluation sweeps.

Rhino Release Notes - 3.10.0