FUNDING · CODING · #3068
NVIDIA releases NemoClaw blueprint showing how NeMo Compass and NeMo Eval Author turn agent failures into tests
NVIDIA published the NemoClaw blueprint and accompanying GitHub tutorial that show a complete optimization cycle using NeMo Compass, NeMo Eval Author, and NVIDIA NeMo Relay to analyze agent traces, create targeted evaluations, update agent harnesses (for example modifying Hermes SOUL.md), and re-run tests to verify fixes and catch regressions. The workflow includes examples (a Hermes enterprise assistant and a Dassault Systèmes engineering case) and tools/formats such as NeMo Gym, Harbor evaluations, ETHOS.md behavioral rules, and Model Context Protocol integrations.
KEY POINTS
- NVIDIA published the NemoClaw blueprint and accompanying GitHub tutorial that show a complete optimization cycle using NeMo Compass, NeMo Eval Author, and NVIDIA NeMo Relay to analyze agent traces, create targeted evaluations, update agent harnesses (for example modifying Hermes SOUL.md), and re-run tests to verify fixes and catch regressions.
- The workflow includes examples (a Hermes enterprise assistant and a Dassault Systèmes engineering case) and tools/formats such as NeMo Gym, Harbor evaluations, ETHOS.md behavioral rules, and Model Context Protocol integrations.
- Provides a reproducible, trace-driven workflow to convert production failures into tests and measure fixes, improving reliability of domain-specific agents and reducing regressions.
WHY IT MATTERS
Provides a reproducible, trace-driven workflow to convert production failures into tests and measure fixes, improving reliability of domain-specific agents and reducing regressions.