Basics of Agents · Andrew Nedilko

Study Guide

Best practices, benefits, RAG vs. fine-tuning, RAG flavors, embedding quantization, common errors, and agentic RAG.

What an agent is (GenAI + tools + memory), agentic workflows and design patterns, use cases, and agent typologies.

Short- and long-term memory, cross-agent memory, and why context management is a hard necessity.

The AI agent framework landscape, and LangChain vs. LangGraph in practice.

Offline and online evaluation, validating LLM judges against humans, and evaluation frameworks.

Model Context Protocol, Agent-to-Agent protocol, and how the two compare.

Decision agents, agentic AI vs. AI agents, agents vs. MoE and assistants, degrees of automation, Google Antigravity.

Sources behind these notes, plus the You.com search-agent evaluation.

The original quick-navigation index — every link below jumps straight to the relevant topic.

RAG		RAG Best Practices (Schematic), RAG Benefits, RAG vs. Fine-Tuning, Types of RAG, Embedding Quantiz. & Truncation, Errors in RAG, Agentic RAG
AGENTS - DEFINITION		Agent = GenAI + Tools + Memory (In Detail), Agentic workflows / patterns, Agent use cases, Agent types, More agents is all you need
MEMORY		Memory, Cross-Agent Memory, Context Management is a Necessity
FRAMEWORKS		AI Agent Frameworks, LangChain vs. LangGraph
EVALUATION		Offline Eval, Online Eval, Validating LLM against humans, Eval Frameworks
PROTOCOLS		A2A vs. MCP, MCP Protocol, A2A Protocol
MISC		Decision Agents, Agentic AI versus AI agents, AI Agents vs. MoE, Agents vs. Assistants, Degrees of Automation, Google Antigravity
REFERENCES		References, Search Agent Evaluation by You.com