Skip to main content

Glossary

This page defines key terms used throughout the Build AI Agent SDK documentation.

A​

Agent​

A configured AI entity with a specific purpose, behavior, and capabilities. An agent combines an LLM with tools, prompts, and execution logic to accomplish tasks.

Example: A customer support agent configured with FAQs and ticket creation tools.

Agent Config​

The complete configuration object (AgentConfig) defining an agent's type, prompt, tools, flows, and metadata. Created using AgentBuilder.

Agent Executor​

The AgentExecutor class that orchestrates agent execution, managing message flow, tool calls, streaming, and state.

Agent Type​

A predefined category of agent behavior (AgentType enum):

  • SmartAssistant: General-purpose conversational agent with tool calling
  • SurveyAgent: Specialized for conducting surveys
  • CommerceAgent: Optimized for e-commerce interactions
  • Flow: Workflow-based agent following structured flows

B​

Builder Pattern​

A creational design pattern used by AgentBuilder to construct complex agent configurations through method chaining.

Example: .setName().setType().addTool().build()

C​

Core Message​

The CoreMessage type from Vercel AI SDK representing a single message in a conversation with role and content.

Conversation Context​

The history of messages and state maintained across agent executions, enabling continuity in multi-turn conversations.

E​

Execution Event​

An event (ExecutionEvent) emitted during agent execution, such as text generation, tool calls, or errors. Used for monitoring and debugging.

Execution Options​

The ExecuteOptions configuration object passed to AgentExecutor.execute() specifying input, provider, tools, and behavior.

Execution Result​

The ExecutionResult object returned after agent execution, containing generated text, tool calls, usage statistics, and metadata.

F​

Flow​

A structured multi-step workflow (AgentFlow) that orchestrates complex operations through sequences of agent interactions and tool executions.

Flow Node​

An individual step in a flow, which can be an LLM call, tool execution, conditional branch, or sub-flow.

Framework Agnostic​

Architecture principle ensuring the SDK works with any JavaScript framework (React, Vue, Express, etc.) without dependencies on specific frameworks.

L​

LLM Provider​

An implementation of the LLMProvider interface that adapts a specific LLM service (OpenAI, Anthropic, Ollama) to the SDK's standard interface.

Locale​

The language and regional settings (e.g., 'en', 'es', 'fr') used for agent responses and tool interactions.

M​

Memory Manager​

Component managing conversation context, message history, and state persistence across executions.

Message​

A single communication unit in a conversation with a role (system, user, assistant, tool) and content.

Model​

The specific LLM variant used for generation (e.g., 'gpt-4', 'gpt-3.5-turbo', 'claude-3-opus').

P​

Prompt​

The system message or instructions that define an agent's behavior, personality, and capabilities.

Example: "You are a helpful customer support agent. Be professional and concise."

Provider​

See LLM Provider.

R​

Repository​

An interface for data persistence operations (saving agents, messages, tool results). Implementations include mock repositories and database adapters (Drizzle).

S​

Session​

A unique identifier grouping related agent executions, enabling conversation continuity and state management.

Streaming​

Real-time text generation where response chunks are delivered incrementally as they're generated, rather than waiting for complete responses.

Stream Chunk​

A single piece of streaming data (StreamChunk) containing text deltas, tool calls, or metadata.

T​

Tool​

A function that agents can invoke to interact with external systems, perform calculations, or fetch data. Defined by ToolDescriptor and registered in ToolRegistry.

Example: Weather API tool, database query tool, calculator.

Tool Call​

An invocation of a tool by an LLM during execution. Contains the tool name, arguments, and a unique identifier.

Tool Configuration​

The ToolConfiguration object specifying which tool to use and its options when adding tools to an agent.

Tool Descriptor​

The ToolDescriptor object containing a tool's display name, implementation, and optional streaming controller.

Tool Registry​

The ToolRegistry class managing tool registration, retrieval, and lifecycle.

Type Safety​

Compile-time verification using TypeScript types to prevent runtime errors from invalid configurations or API usage.

U​

Usage​

Token consumption statistics including prompt tokens, completion tokens, and total tokens used in an LLM request.

W​

Workflow​

See Flow.

Common Acronyms​

  • SDK: Software Development Kit
  • LLM: Large Language Model
  • API: Application Programming Interface
  • AI: Artificial Intelligence
  • JSON: JavaScript Object Notation
  • HTTP: Hypertext Transfer Protocol
  • UUID: Universally Unique Identifier
  • DTO: Data Transfer Object

Last Updated: January 2025 | SDK Version: 1.0.0-alpha.8

Missing a term? Request an addition