NVIDIA releases guide to deploying AI-Q Blueprint on Oracle Cloud Infrastructure

NVIDIA published instructions for deploying AI-Q Blueprint — an open platform for long-running AI agents — in the production environment of Oracle Cloud Infrastructure.

From Question-Answer to Long-Running Agent

Over the past two years, agentic AI has gone through three generations. The first systems operated in a "one question — one answer" mode: the model didn't remember previous steps. Then came multi-turn dialogue — the model began preserving context within a session, which enabled conducting coherent conversations and iteratively refining results. Today's long-running agents are structured fundamentally differently. They can:

• set goals and break them down into subtasks
• delegate steps to specialized sub-agents
• maintain full context throughout a long task
• run external tools — browser, code interpreter, API — in a secure sandbox
• adapt the plan on the fly if intermediate results diverge from expectations

What Is NVIDIA AI-Q Blueprint

AI-Q Blueprint is an open reference architecture that NVIDIA develops for teams building agentic systems. The framework implements long-running agent architecture out of the box and provides ready-made patterns for orchestration, tool management, and long-term context storage. The project is fully open: companies can adapt the Blueprint for industry-specific tasks — from software development automation to data analytics and customer support. Instead of building an agentic architecture from scratch, a team takes a proven foundation and builds its own logic on top.

Deployment on Oracle Cloud

Oracle Cloud Infrastructure provides GPU resources and managed services for the Blueprint: storage, networking, task isolation, monitoring. Together they form a stack where the agent gets necessary resources at the right time without manual cluster management. The key word of the guidance is production-ready. This isn't a demo: the configurations are designed for real-world workloads with automatic scaling and enterprise-level security. For teams transitioning from prototype to working system, such a Blueprint reduces deployment time from several months to several weeks.

Why This Changes the Market

A year ago, launching a long-running agent required deep MLOps expertise: custom task queues, state management, tool isolation. Such work was only feasible for large technology teams. AI-Q Blueprint plus Oracle's cloud infrastructure change the equation: an average enterprise team gets a ready-made architecture that only needs to be configured for their data and business processes. This lowers the barrier to entry and accelerates the transition of agentic AI from "laboratory experiment" mode to "working product" mode.

What This Means

Long-running AI agents have become mature enough to exit research laboratories. The combination of NVIDIA's open framework and Oracle's cloud platform establishes a new standard: GPU stack + managed cloud + open Blueprint — this is now the standard path to production agents for companies without their own AI infrastructure division.