Glossary

Term	Definition
Assistant	Configured AI agent: prompt, model, knowledge bases, tools
Knowledge base (KB)	Indexed collection of documents for semantic search
Tool	Integration that calls an external API or MCP server
API Request tool	Tool type that makes HTTP requests (GET, POST, etc.)
MCP tool	Tool type using Model Context Protocol
Deployment	Published instance of an assistant (widget, API, etc.)
Widget	Embeddable chat UI on your website
Credential	Stored auth for tools (Bearer, API key, OAuth, HMAC)
Citation	Reference to source document in an answer
Agent Builder	Dashboard AI that configures resources via chat
Rich content	Cards/carousels rendered in chat from tool results
Allowed domains	List of websites permitted to load your widget
RAG	Retrieval-augmented generation — search docs at answer time
Status notification	User-facing message during tool run
Strict params	Tool rejects calls with missing or invalid parameters
RRF	Reciprocal Rank Fusion — merges semantic and keyword search
Qdrant	Vector database used for document embeddings
JWT verification	Optional deployment setting to verify logged-in users
Auto-retrain	Standard+ plan feature to re-index when documents change
BYOK	Bring-your-own-key — Enterprise uses your LLM provider keys
Intent Analyzer	Three-brain stage that classifies visitor intent and required data
Query Strategist	Three-brain stage that chooses RAG, tools, hybrid, or direct reply
Response Synthesizer	Three-brain stage that writes the final answer with citations
Snapshot	Immutable copy of a document when it is added to a knowledge base
Classification	Document/KB sensitivity: public, internal, confidential, restricted
Organization	Isolated multi-tenant workspace for users, KBs, assistants, and tools
Interactor	Visitor, user, or system that sends messages to an assistant
Hybrid search	Dense (semantic) + keyword retrieval merged with RRF
result_transform	Declarative mapping from tool JSON to rich chat cards