Course Dictionary

Documentation & Key Terms for AI Agent Development

Documentation References

Essential documentation and resources used throughout the course.

LLM Provider Documentation

OpenAI

Google Gemini

Anthropic Claude

LangChain & LangGraph

LangChain Core

LangSmith (Observability & Diagnostics)

LangGraph

Model Context Protocol (MCP)

No-Code/Low-Code Workflow Tools

n8n

Langflow

Articles & Learning Resources

Workflows vs Agents

Best Practices & Advanced Topics

Video Resources

Development Tools

  • Node.js - Required for n8n local installation

Key Terms & Definitions

A

Agent
An AI system that can autonomously choose which tools to call and when, based on user input and context. Unlike workflows, agents make dynamic decisions rather than following predetermined paths.
AI Message
A message type in LangChain representing responses from the language model. Part of the conversation history structure.
API Key
A secure credential that authenticates your application to external services. Acts as a password for accessing LLM providers like OpenAI, Gemini, or Claude.
Agentic Pattern
Design patterns for building AI agents, including reflection, planning, tool use, and multi-agent collaboration.

C

Chat Model
A language model specifically designed for conversational interactions, accepting a sequence of messages and returning a response.
Context Engineering
The practice of structuring and managing the information provided to a language model to optimize its outputs. Includes prompt engineering, RAG, and memory management.
Context Window
The maximum amount of text (measured in tokens) that a language model can process at once, including both input and output.

D

Data Drift
When the statistical properties of input data change over time, causing model performance to degrade if weights aren’t updated.
Docstring
Documentation within a function that describes its purpose and parameters. For AI tools, docstrings are critical because the LLM reads them to decide when to call the tool.

E

Embedding
A numerical representation of text that captures semantic meaning, used for similarity search and retrieval.
Environment Variable
Configuration values stored outside of code (like in .env files) to keep sensitive information like API keys secure.
Evaluation Dataset
A collection of test cases used to measure agent performance systematically. Used with LangSmith to validate agent behavior.

F

Function Calling
The ability of an LLM to output structured requests to execute specific functions or tools based on the conversation context.

L

LangChain
A Python framework for building applications with large language models, providing abstractions for prompts, chains, agents, and memory.
LangGraph
A library for building stateful, multi-step agent workflows with cycles and conditional logic.
LangSmith
An observability and evaluation platform for AI applications, providing tracing, debugging, and monitoring capabilities.

M

Memory
The ability of an agent to retain information across multiple turns of conversation. Can be short-term (within a session) or long-term (persistent across sessions).
Message
The fundamental unit of conversation in chat models. Types include System, Human, AI, Tool, and Function messages.
MCP (Model Context Protocol)
A standardized protocol for connecting AI agents to external tools and data sources through client-server architecture.
MCP Client
The component that runs alongside your AI agent, translating tool calls into MCP protocol requests.
MCP Server
A service that exposes tools and data to AI agents through the MCP protocol. Can run locally or remotely.
Multimodality
The ability of a model to work with multiple types of input (text, images, audio, video) and/or generate multiple types of output.

O

Observability
The ability to understand what’s happening inside your AI system through logs, traces, and metrics. Critical for production deployments.

P

Prompt Engineering
The practice of crafting effective prompts to elicit desired behaviors from language models.
Production
The live environment where your AI agent serves real users, as opposed to development or testing environments.

R

RAG (Retrieval-Augmented Generation)
A technique that enhances LLM responses by retrieving relevant information from external knowledge sources before generating an answer.
React Agent
A common agent pattern that follows a Reasoning-Acting loop: the agent reasons about what to do, acts by calling tools, observes results, and repeats.
Retraining
The process of updating a machine learning model’s weights using new data to account for data drift or improve performance.

S

Schema
The structure defining what parameters a tool accepts, including types and descriptions. Automatically generated from function signatures.
System Instruction
The initial prompt that sets the behavior, personality, and constraints for an AI agent. Also called system prompt or system message.

T

Temperature
A parameter controlling randomness in LLM outputs. Lower values (0.0-0.3) produce more deterministic responses; higher values (0.7-1.0) produce more creative/varied outputs.
Token
The basic unit of text processing for LLMs. Roughly 3-4 characters or 0.75 words per token. Both input and output are measured in tokens for billing.
Tool
A function that an AI agent can execute to perform specific tasks like searching, retrieving data, or interacting with external APIs.
Traceable
A decorator (@traceable) that exposes function calls as named spans in LangSmith traces, making internal logic visible for debugging.
Trace Tree
A hierarchical visualization in LangSmith showing the sequence of LLM calls, tool invocations, and other operations during agent execution.

W

Weights
The learned parameters in a neural network that encode patterns from training data. Updated during training or retraining.
Workflow
A predetermined sequence of steps for processing information, as opposed to an agent that makes dynamic decisions. More predictable but less flexible than agents.

Additional Resources

For terms not listed here, consult: - The official documentation links above - Course module content - LangChain glossary - Provider-specific documentation