NVIDIA AI-Q (pronounced ‘IQ’) Blueprint is an open-source reference implementation for building intelligent research agents that reason over enterprise data and deliver trusted business insights. Built on LangChain DeepAgents and accelerated by the NeMo Agent Toolkit, AI-Q implements a two-tier research architecture: a shallow-research agent for quick answers with citations (bounded to 10 LLM turns and 5 tool calls), and a deep-research agent for comprehensive multi-phase investigations of complex topics.
The blueprint provides end-to-end agentic search capabilities with both quick answers and in-depth, report-style research in one system. It tops the DeepResearch Bench accuracy leaderboard using a hybrid approach with frontier models for orchestration and NVIDIA Nemotron open models for research—cutting query costs in half while maintaining world-class accuracy. Built-in benchmarks and evaluation harnesses let developers measure quality, compare outcomes, and improve iteratively.
AI-Q integrates seamlessly with Milvus for enterprise knowledge retrieval. Milvus stores document embeddings and metadata, enabling semantic search across ERP systems, CRM platforms, data warehouses, documents, images, and chat logs. This combination gives agents access to proprietary context while maintaining strict data privacy and on-premises control, making AI-Q ideal for knowledge-intensive domains like customer support, legal research, and scientific discovery.