The release includes verified API signatures, hardware requirements, and build constraints for DOCA developers.
NVIDIA now offers DOCA AI agent skills on GitHub, giving AI agents a structured way to work with BlueField DPU software. The skills provide verified API signatures, hardware needs, and build limits. The skills cover DOCA components such as Flow, GPUNetIO, PCC, and RDMA.
DOCA is NVIDIA’s unified software platform for BlueField data processing units, spanning accelerated networking, AI-native storage, in-silicon security, telemetry, and lifecycle management. General-purpose AI agents lacked domain knowledge for this infrastructure software and fell back on guesswork, which slowed correction cycles.
For builders, the skills reduce trial and error when writing DOCA code. In a 65-prompt evaluation, agents without skills satisfied only 19 percent of checklist items, while agents with skills satisfied 100 percent across all prompts. A side-by-side demo showed the with-skills agent used 73 percent less handwritten code and 46 percent fewer hardware commands to build a Go-based RDMA application on BlueField-3.
NVIDIA offers DOCA AI agent skills in the NVIDIA/skills GitHub repository for Flow, GPUNetIO, RDMA, and other components. The skills let agents verify device support before writing code and apply preflight checks for firmware-level changes.
What matters
- NVIDIA released DOCA AI agent skills on GitHub with verified API signatures and hardware limits.
- Builders gain a reasoning framework that reduces guesswork and correction cycles in DOCA work.
- Developers can watch the NVIDIA/skills GitHub repository for updates to Flow, GPUNetIO, and RDMA skills.
Why it matters
Developers can watch the NVIDIA/skills GitHub repository for updates to Flow, GPUNetIO, and RDMA skills.
This GenAI News article was prepared in original wording using reporting and materials published by NVIDIA Developer Blog. Source reference: https://developer.nvidia.com/blog/build-applications-on-nvidia-bluefield-faster-with-nvidia-doca-agent-skills/.
Drafted by the GenAI News review pipeline.