Back to directory

DeepStream Coding Agent

A worked example of using AI coding assistants such as Cursor or Claude Code to build NVIDIA DeepStream SDK video analytics applications, using a curated agentic skill and structured prompts.

DeepStream Coding Agent

DeepStream Coding Agent shows how to use AI coding assistants that support agentic skills, such as Cursor or Claude Code, to speed up development of applications built on NVIDIA's DeepStream SDK for video analytics. It supplies a curated skill and structured prompts that give the assistant domain-specific rules, reference material and guardrails so it produces more accurate, idiomatic DeepStream code without the developer pasting reference documentation into every conversation.

Generating code with the skill needs only a compatible AI coding assistant, and no GPU, SDK or special hardware; running the generated code, however, needs DeepStream SDK 9.0, Python 3.12 or newer with the pyservicemaker package, an NVIDIA GPU with a recent driver, CUDA 13.1, TensorRT, and Ubuntu 24.04 on x86_64 or Jetson hardware. Generated code is explicitly described as a starting point that still needs review, testing and security validation before production use.

More in Agent skills

See all 581
  • Provenance gate

    github.com

    A deterministic authorization gateway that sits between an AI agent and its tools, tracing where each parameter of a side-effect call came from, and blocking the call or asking a human when the source does not check out.

  • A markdown framework of system rules that restates Isaac Asimov's Four Laws of Robotics as a set of prioritized loss calculations for large-language-model agents to follow, published as a discovery file for agents and crawlers to read.

  • Vitruvius

    github.com

    An engineering research agent that runs a discover, read, synthesize, verify and review loop over technical questions, reading sources directly rather than inferring from titles, and publishing its own scored verifier benchmark, weaknesses included.