Terms and Definitions
A comprehensive reference of terminology used throughout the Nadoo AI platform and documentation, organized alphabetically.| Term | Definition |
|---|---|
| Agent | An AI-powered entity that performs tasks by combining a large language model (LLM) with tools, knowledge retrieval, and execution strategies. Agents are the primary unit of intelligence in Nadoo AI. |
| Agent Mode | The execution strategy used by an AI Agent node in a workflow. Supported modes: Standard (single LLM call), Chain of Thought (CoT) (step-by-step reasoning), ReAct (reason + act loop), Function Calling (structured tool invocation), Reflection (self-evaluation and correction), and Tree of Thoughts (ToT) (branching exploration of reasoning paths). |
| Application | An AI agent instance within a Workspace. Each application has a type — Chat (conversational interface), Workflow (graph-based logic), or Channel (messaging platform integration) — and contains its own configuration, conversations, and linked resources. |
| Canvas | A rich content editing interface within the chat UI for creating and editing documents, artifacts, and structured outputs collaboratively with the AI agent. |
| Channel | An integration with an external messaging platform that allows AI agents to send and receive messages. Supported channels include Slack, Discord, Telegram, KakaoTalk, Microsoft Teams, and WhatsApp. |
| Chunk | A segment of a document created during knowledge base processing. Documents are split into chunks of configurable size before embedding, enabling granular retrieval during RAG queries. |
| Embedding | A dense vector representation of text produced by a configurable embedding provider (OpenAI, HuggingFace, Azure, Bedrock, Google, vLLM, Ollama, Local). Stored in the vector store (pgvector default) for similarity search during knowledge retrieval. |
| Knowledge Base | A document store that combines vector embeddings with keyword indexing to support Retrieval-Augmented Generation (RAG). Supports ingestion of PDF, DOCX, TXT, Markdown, Excel, and web pages. |
| Knowledge Graph | A structured representation of entities and their relationships built using RDF/OWL ontologies. Knowledge graphs enable multi-hop reasoning and SPARQL-based queries for complex question answering. |
| LangGraph | The open-source framework (by LangChain) that powers the Nadoo AI workflow execution engine. LangGraph provides the graph-based runtime for executing connected nodes with state management. |
| MCP | Model Context Protocol — an open protocol for connecting AI models to external tools and data sources. Nadoo AI supports MCP for standardized tool integration alongside its native Plugin SDK. |
| Node | A single processing unit in a workflow graph. Each node performs a specific operation (e.g., AI Agent, Condition, Loop, Search Knowledge, HTTP Request, Code Executor). Nodes are connected by edges that define the flow of data. |
| pgvector | A PostgreSQL extension that adds vector data types and similarity search operators. Nadoo AI uses pgvector as the default vector store, with Milvus and Qdrant support via a pluggable VectorStoreFactory interface. Supports cosine, euclidean, and dot product distance metrics. |
| Plugin | An extension module built with the nadoo-plugin-sdk Python package. Plugins define tools using a decorator-based API and have access to LLMs, knowledge bases, storage, and other platform internals. |
| RAG | Retrieval-Augmented Generation — a technique that improves LLM responses by first retrieving relevant documents from a knowledge base and including them as context in the prompt. Nadoo AI implements RAG with hybrid search (vector + BM25) and optional reranking. |
| RBAC | Role-Based Access Control — the permission system used to manage user access within workspaces. Roles (e.g., Owner, Admin, Member, Viewer) determine what actions a user can perform on applications, workflows, knowledge bases, and settings. |
| Reranking | A post-retrieval step that re-scores search results using a cross-encoder or specialized reranking model to improve the relevance ordering of documents returned by the initial search. |
| Skill | A reusable executable unit defined by a SKILL.md manifest file and loaded from Git repositories or local paths. Skills execute in an isolated worker process and are the lightweight alternative to plugins for extending agent capabilities. |
| SPARQL | A query language for RDF-based knowledge graphs. Nadoo AI supports SPARQL queries against knowledge graphs for structured information retrieval and multi-hop reasoning. |
| SSE | Server-Sent Events — a one-directional HTTP protocol used by the backend to stream AI-generated responses token by token to the frontend. SSE is used for workflow execution progress and LLM output streaming. |
| WebSocket | A bidirectional communication protocol that enables real-time messaging between the frontend and backend. Nadoo AI uses WebSockets for the chat interface, delivering instant message delivery and typing indicators. |
| Workflow | A visual graph of connected nodes that defines how an AI agent processes inputs and produces outputs. Workflows are created in the drag-and-drop visual editor and executed by the LangGraph-based engine. |
| Workspace | The top-level multi-tenant isolation unit in Nadoo AI. Each workspace has its own users, applications, knowledge bases, model provider configurations, channels, plugins, and skills. All data is scoped to the workspace. |
Common Abbreviations
| Abbreviation | Full Term |
|---|---|
| API | Application Programming Interface |
| BM25 | Best Matching 25 (probabilistic ranking function) |
| BUSL | Business Source License |
| CoT | Chain of Thought |
| CORS | Cross-Origin Resource Sharing |
| CRUD | Create, Read, Update, Delete |
| JWT | JSON Web Token |
| LLM | Large Language Model |
| MCP | Model Context Protocol |
| OWL | Web Ontology Language |
| RAG | Retrieval-Augmented Generation |
| RBAC | Role-Based Access Control |
| RDF | Resource Description Framework |
| REST | Representational State Transfer |
| SDK | Software Development Kit |
| SPARQL | SPARQL Protocol and RDF Query Language |
| SSE | Server-Sent Events |
| ToT | Tree of Thoughts |
| UI | User Interface |
Related Resources
Core Concepts
In-depth explanation of the platform’s domain model
Architecture
Technical architecture and system design
FAQ
Frequently asked questions about the platform
API Reference
Complete REST API documentation