Description
AI Workflows and Agents: A Practical Foundation for Building with Large Language Models
AI Workflows and Agents is a practical guide to understanding and building AI-powered applications with Large Language Models (LLMs). Rather than treating AI systems as black boxes, this book explains what happens behind the scenes and shows how workflows, agents, tools, memory, and control flow fit together.
Starting with the fundamentals, the book clearly explains the difference between AI workflows and AI agents, including their trade-offs in control, flexibility, predictability, cost, and complexity. It then moves into practical development using Python, APIs, prompting techniques, and real working examples.
You’ll learn how to:
- Build your first AI workflow using an LLM API
- Design effective prompts and use few-shot examples
- Create multi-step AI workflows
- Choose models according to task difficulty and cost
- Run open-weight models locally with Ollama
- Generate and work with structured outputs and JSON
- Handle errors, retries, timeouts, and model failures
- Build workflows involving images, messages, and files
- Manage sequential, parallel, conditional, and repeated steps
- Add human-in-the-loop checkpoints
- Integrate third-party services such as Slack and LinkedIn
- Understand and implement Retrieval-Augmented Generation (RAG)
- Understand how AI agents use tools and function calling
- Build customer-service and specialized agents
- Design multi-agent systems
- Manage short-term and long-term memory
- Test and evaluate AI systems
- Explore agent frameworks such as CrewAI
The book follows a with-code approach using Python, while emphasizing that the underlying concepts transfer to other programming languages and AI providers.
Whether you are beginning your journey with AI development or want a clearer understanding of how modern AI applications are constructed, this book provides a practical foundation for moving from simple LLM calls to reliable workflows and tool-using agents.
Reviews
There are no reviews yet.