NVIDIA Developer

Ask the Experts: How NeMo Switchyard Helps Agents Select Models | Nemotron Labs

Published 2026-08-27 · Duration 49:23

Summary

NeMo Switchyard is an open-source model routing library designed for AI agents to solve the problem of relying on a single monolithic LLM. It automatically routes each agent query or step to the optimal model—selecting from any combination of local/cloud and open/closed models—based on real-time needs, optimizing for accuracy, cost, and latency. The system operates beyond simple request routing by tracking state across multi-turn agentic workflows, making it a critical component for building robust, efficient AI systems.

Download summary

Key takeaways

  1. System of Models Approach 2:49

    The industry is moving away from the 'one model to rule them all' concept toward a 'system of models,' where multiple specialized models are used for different tasks, improving efficiency and capability (0:02:45).

  2. Agent-Aware Routing vs. Simple Routing 5:18

    Switchyard is more than a simple router; it operates on an agentic workflow, tracking state (e.g., tool calls, message history) across multi-turn sessions to make intelligent model selection decisions (0:04:25).

  3. Optimization and Learning 23:18

    The system treats model selection as an optimization problem. It can learn by analyzing agent traces and behavior, predicting potential errors or resource needs to route proactively and save tokens/time (0:21:58).

  4. Full-Stack Routing Flywheel 6:10

    The roadmap envisions a full 'flywheel' of routing, connecting model selection to inference optimization (via NVIDIA Dynamo) and data privacy/anonymization. This allows for continuous improvement across the entire agent lifecycle (0:06:10).

Technical details

  • Architecture & Integration 176s

    Switchyard can be integrated in multiple ways: directly into the agent harness/application, embedded natively, or within an existing LLM gateway (like OpenRouter). It operates on a provider-neutral format for algorithm development (0:03:58; 0:04:29).

  • Optimization Parameters 829s

    Users can tune routing strategies based on specific priorities, including accuracy, cost (token budget), and latency. Using routing algorithms has been shown to yield significant token cost reductions (50-80%) while maintaining frontier model accuracy (0:13:49).

  • Observability and Telemetry 987s

    Switchyard supports OpenTelemetry (OTEL) for capturing performance metrics. Future development aims to expose not just *what* model was chosen, but *why* the decision was made by the routing algorithm (0:16:27).

  • Advanced Routing Algorithms 1475s

    The library includes specialized algorithms like 'escalation router' (for handling persistent errors) and 'stage router' (which uses tool call history to inform model selection), allowing for complex, multi-signal decision making (0:23:15; 0:24:39).

Mentioned resources

  • NeMo Switchyard GitHub (Code Repository)
  • NVIDIA Dynamo (Inference Backend/Tool)

Channel & topics

Watch on YouTube · Back to latest

This independent, AI-assisted summary is provided for commentary and informational purposes. It may contain errors or omit important context. Please watch the original video for the creator's complete presentation. Video, thumbnail, and related copyrights belong to their respective owners.