Glossary

NodeTool terminology in plain language. Simple explanations followed by technical details where relevant.

Core Concepts

NodeTool

The platform you’re using! NodeTool is a visual environment for building AI workflows. You connect nodes (building blocks) to create powerful automations without writing code.

Workflow

A workflow is your project – a collection of connected nodes that accomplish a task together. Think of it like a recipe: each step (node) does something specific, and together they create the final result.

Technical: A directed acyclic graph (DAG) of nodes describing an end-to-end task.

Node

A node is a single building block in your workflow. Each node does one specific job, like “generate text” or “resize image.” You connect nodes together to build workflows.

Technical: A processing unit with typed inputs, outputs, and configurable properties.

Edge / Connection

The lines that connect nodes together. They show which node’s output feeds into another node’s input – like pipes carrying water between stations.

Input

Where data enters – either into a workflow (like uploading a file) or into a specific node (like connecting another node’s output).

Output

Where results come out. Output nodes show your final results; other nodes have outputs that connect to the next node in the chain.

AI & Models

Model

A pre-trained AI program that has learned to do a specific task. For example, a language model generates text, an image model creates pictures. You don’t train models – you just use them.

Example: GPT is a language model; Stable Diffusion is an image model.

Provider

A service that runs AI models for you. Providers can be:

Cloud providers like OpenAI, Anthropic, or Google
Local engines like Ollama or llama.cpp that run models on your computer

Technical: Adapter that talks to an external AI service (OpenAI, Anthropic, Gemini, Ollama, ComfyUI, etc.).

Agent

A special type of AI that can plan and execute multi-step tasks. Unlike simple chat, an agent can break down complex requests, use tools, and work through problems step by step.

Technical: Multi-step planner/executor that can call tools or workflows.

LLM (Large Language Model)

A type of AI model that understands and generates text. LLMs power chatbots, writing assistants, and text analysis tools. Examples: GPT-4, Claude, Llama.

Inference

The process of using a trained AI model to generate results. When you “run” a model, you’re performing inference.

User Interface

Canvas

The main work area where you build workflows by placing and connecting nodes. Pan around by dragging, zoom with scroll.

Preview Node

A special node that shows you intermediate results while your workflow runs. Add previews anywhere to debug or monitor progress.

Mini-App

A simplified interface for running a workflow. Mini-Apps hide the complexity and show only the inputs and outputs – perfect for sharing with non-technical users.

Global Chat

NodeTool’s AI assistant interface. Chat with AI models, run workflows conversationally, or enable Agent Mode for autonomous task completion.

Inspector / Properties Panel

The panel (usually on the right) that shows settings for the selected node. This is where you configure how each node behaves.

Dashboard

Your home screen showing recent workflows, templates, and chat threads.

Running & Execution

Run

Execute your workflow and see the results. Click the Run button or press Ctrl/⌘+Enter.

Job

A single workflow execution. When you click Run, NodeTool creates a job that tracks progress, handles errors, and delivers results.

Technical: A single workflow execution instance managed by JobExecutionManager.

Streaming

Receiving results progressively as they’re generated, rather than waiting for everything to complete. Many AI nodes stream their output so you see progress in real-time.

Execution Strategy

How NodeTool runs your workflow – either in the same process, a separate subprocess, or inside a Docker container.

Technical options: threaded, subprocess, Docker.

Infrastructure

Worker

The background process that actually runs your workflows. Workers connect to the server and execute jobs.

Technical: Process that runs workflows and exposes HTTP/WebSocket endpoints (often via nodetool serve or nodetool worker).

API Server

The backend service that handles requests, manages workflows, and coordinates workers.

Technical: FastAPI server handling REST endpoints such as /v1/chat/completions and /api/workflows.

Proxy

An optional service that sits in front of NodeTool to handle security, routing, and SSL certificates for production deployments.

Technical: Reverse proxy that terminates TLS and forwards to the API server.

Thread ID

An identifier that tracks a conversation in Global Chat. Each chat thread has its own history and context.

Technical: Conversation identifier for chat/agent sessions; used by WebSocket and SSE streams.

Data & Storage

Asset

Any file you use in NodeTool – images, audio, documents, etc. Assets are stored and managed so you can reuse them across workflows.

Collection

A searchable database of documents used for RAG (Retrieval-Augmented Generation). Collections let AI answer questions using your own documents.

Vector Database

A special database that stores text as mathematical vectors, enabling semantic search (finding content by meaning, not just keywords).

Development Terms

DAG (Directed Acyclic Graph)

The technical name for how workflows are structured. “Directed” means data flows one way; “Acyclic” means no loops. NodeTool handles this automatically – you just connect nodes.

DSL (Domain Specific Language)

NodeTool’s Python API for building workflows in code rather than the visual editor. Useful for automation and custom integrations.

Node Pack

A collection of related nodes bundled together. Install node packs to add new capabilities to NodeTool.