Principles of Building AI Agents, 2nd Edition by Sam Bhagwat

Principles of Building AI Agents, 2nd Edition by Sam Bhagwat 

The book is intentionally brief, highly scannable, and designed to get a developer up and running with a functional agent in a matter of days. It breaks down the agentic ecosystem into three distinct technical layers:Navigating the real-world tradeoffs between hosted frontier models and open-source models, balancing parameters like accuracy, token cost, latency, and context window sizes.Understanding how newer "thinking" models alter prompt requirements, functioning more like standalone "report generators" that require dense context up front rather than simple iterative promptsMoving past "vibes." Bhagwat codifies practical, structured techniques including system prompt optimization, few-shot variations, and formatting tricks to make LLM outputs deterministic.

While a massive portion of the data science ecosystem relies entirely on Python (using libraries like LangChain or AutoGen), Bhagwat writes primarily for the massive world of web developers. By utilizing JavaScript/TypeScript via Mastra, he demonstrates that building production AI backends can fit seamlessly into modern full-stack web infrastructure.A major theme throughout the text is knowing when not to give an agent complete freedom. Bhagwat outlines why open-ended autonomous loops often break in production. He advocates for graph-based workflows that explicitly constrain the agent's path when predictability and corporate guardrails are paramount.Because LLM processing and complex tool calls introduce significant latency, Bhagwat heavily emphasizes streaming updates. He instructs developers to stream intermediate step completions and partial tool outputs directly to the frontend, preventing users from staring at blank, shimmering loading states.

Language: English.

Genre: Applied Artificial Intelligence.

Binding: সেলাই করা বাইন্ডিং

Quality: Premium Quality Books.

Printing: High Quality Printing.

Paper: Eye Friendly paper (Cream White)

Cover: Matt cover (Paperback).