Agents

An agent is a system that can use an LLM to reason through a problem, create a plan to solve the problem, and execute the plan with the help of a set of tools. Refer to Introduction to LLM Agents for more details on this. In NeMo Agent toolkit, agents are implemented as a special type of function that can orchestrate other functions.

NeMo Agent toolkit includes several agents out of the box to choose from. In addition to this NeMo Agent toolkit makes it easy to write a custom agent, for an example of this refer to the Alert Triage example (examples/advanced_agents/alert_triage_agent) in the repository.

NeMo Agent Toolkit also provides an Automatic Memory Wrapper that enhances any existing agent with automatic memory capture and retrieval capabilities.

• ReAct Agent
• Reasoning Agent
• ReWOO Agent
• Responses API and Agent
• Router Agent
• Sequential Executor
• Tool Calling Agent
• Automatic Memory Wrapper