Agent workflow
An agent workflow is a structured, multistep process in which one or more AI agents perform or coordinate actions toward an objective, potentially adapting to new information or outcomes.
AI Agents and Applications
An agentic workflow is a structured, multistep application process in which a large language model orchestrates predefined components within explicit boundaries and managed state.
- How it is used
- Structures multistep LLM applications that need bounded choices, managed state, and predictable execution.
- How deep it goes
- Details typed state, nodes, edges, conditional routing, and checkpoint inspection rather than stopping at a definition.
- Examples
- Illustrates routing to a SQL database or REST API and a research assistant that searches again or synthesizes.
- What you can do after
- Build it from scratch with LangGraph StateGraph, TypedDict state, task nodes, and edges; inspect intermediate states and checkpoints.
Read how AI Agents and Applications explains agent workflow →
Spring AI in Action
An agentic workflow pattern is a design for organizing an AI agent's actions into a coordinated problem-solving process.
- How it is used
- Organizes augmented LLM capabilities into goal-oriented problem-solving workflows.
- How deep it goes
- Defines several patterns and begins implementation with prompt chaining, while stopping short of exhaustively implementing every pattern.
- Examples
- Uses prompt chaining, routing, parallelization, orchestrator-workers, and evaluator-optimizer as workflow patterns.
- What you can do after
- Apply agentic patterns through Spring AI’s ChatClient and compose foundational workflows into larger solutions.
Domain-Specific Small Language Models
A multi-step model-and-tool interaction pattern where the model chooses tools, receives results, and iterates toward a final answer.
- How it is used
- Deploys local language models as tool-using assistants within conversational workflows.
- How deep it goes
- Demonstrates the SDK call and tool shape but stops short of detailing tool-selection or result-handling internals.
- Examples
- Uses a Python multiply function and a multiplication question as the concrete tool-use example.
- What you can do after
- Build it from scratch with LM Studio’s Python SDK by exposing functions and passing them to model.act.
Read how Domain-Specific Small Language Models explains agent workflow →
Related concepts
- AI agentuses · An AI agent is a computational system that uses artificial intelligence to interpret information, select actions, and pursue goals through interaction with an environment.
- @Agentincludes · @Agent is an Embabel annotation that declares a class to be an agent and can assign it identifying metadata.
- LangGraphimplemented by · LangGraph is a framework for building stateful, graph-structured workflows and applications, including those involving language models and agents.
- agent orchestrationconfigured by · Orchestration is the coordination and sequencing of agents, services, tasks, or components to carry out a larger workflow or objective.
- small language modeluses · A small language model is a comparatively compact language model with limited parameters or capacity and typically lower computational and storage requirements than larger models.
- ReActused by · ReAct is an agent technique that iteratively interleaves reasoning, actions, and observations so results can guide subsequent actions.
These guides are generated from the text of Manning books by Manning's knowledge graph and link to the sections they draw on. The full books are available on liveBook and at manning.com.