What’s new:
- Authentication: Configure robust authentication for securing tool invocation inside agent workflows (OAuth 2.0, API Key Configuration, and Basic HTTP Authentication)
- Test Time Compute: Agents automatically optimize compute resources with built-in strategies after training at inference time, boosting accuracy and reliability—no retraining needed.
- GPU sizing calculator: Estimate GPU requirements based on your user targets and response time goals to speed up deployment planning and scaling.
- Object Store Integration: Connect and manage data through supported object stores (eg. AWS S3, MinIO, custom), improving agent extensibility and enabling advanced data workflows.
- Enhanced Cursor Rules: Build new workflows or extend existing ones by leveraging cursor rules, making agent development faster and more flexible.
- Observability: The observability features have been refactored, and you can now more easily integrate logging and trace outputs.
- Examples and docs: Now organized by function and improved to cover new features and plug-in scenarios.
- The NeMo Agent Toolkit roadmap is now publicly available on our GitHub README—check it out to see what’s coming next and plan your development accordingly.
All previously reported issues affecting major plugins and features have been addressed, with improvements in the developer experience, documentation, and example coverage.
If you’re upgrading, consult the new example workflows and check the docs for details on configuring authentication and resource planning features. For more details, see the official release notes, changelog and documentation.