Explore our comprehensive collection of agentic automation terms and definitions
A system where LLMs dynamically direct their own processes and tool usage, maintaining control over how they accomplish tasks.
A system where LLMs dynamically determine and execute a series of actions to achieve a goal.
A Large Language Model enhanced with additional capabilities such as retrieval, tools, and memory
The degree to which an agent can operate independently, making decisions and taking actions without human intervention.
The ability to combine different components or patterns in a modular way to create complex agent behaviors
The practice of regularly obtaining ground truth from the environment to assess and adjust agent behavior
The continuous cycle of agent action, environment response, and agent adjustment
An iterative workflow pattern where two LLMs collaborate to refine outputs: one generates content, the other critiques and guides improvement.
Software that simplifies agent implementation through abstraction layers
Safety mechanisms for agentic systems, ensuring acceptable LLM behavior and preventing harmful actions
Information obtained from the environment that agents use to assess progress and adjust approach
The principle that agentic systems should incorporate human input for guidance, refinement, and validation, ensuring alignment with human intent and values.
The time delay between initiating an action and receiving a response in agent systems
A method for integrating external tools with LLMs, enhancing their capabilities and control
Advanced LLMs capable of understanding and generating content across multiple modalities, such as text, images, audio, and video.
A central LLM acts as a project manager, breaking down complex tasks into smaller parts and assigning them to different 'worker' LLMs.
A workflow pattern that processes tasks simultaneously through sectioning or voting approaches
A workflow pattern that breaks down tasks into sequential steps, where each LLM call processes the output of the previous one.
An AI workflow pattern that directs different types of inputs to specialized processes or models.
The practice of favoring simple, composable patterns over complex frameworks or specialized libraries when building with LLMs, promoting maintainability, reliability, and ease of understanding.
Predetermined criteria that halt agent execution to maintain system control, preventing runaway processes and ensuring predictable behavior.
Best practices for designing and implementing tools for agent systems, focusing on clarity, reliability, and usability from an agent's perspective
External services and APIs that agents can interact with to perform actions and gather information
The process of breaking down a complex task into smaller, more manageable subtasks, often performed by an orchestrator LLM.
The formal specification of a tool's capabilities, parameters, and usage requirements, enabling an agent to effectively understand, interact with, and utilize the tool.
A system where LLMs and tools are orchestrated, potentially dynamically, through code paths