Skip to main contentAI Terms
LLM (Large Language Model)
The “brain” of the agent (e.g., GPT-4, Claude). It takes text input (transcript) and generates text output (response).
RAG (Retrieval Augmented Generation)
A technique to give the LLM extra knowledge. We search your uploaded documents for relevant info and insert it into the prompt before the LLM generates an answer.
STT (Speech-to-Text)
Also known as ASR (Automatic Speech Recognition). The process of converting the user’s audio into text.
TTS (Text-to-Speech)
The process of converting the LLM’s text response into audio.
Latency
The time delay between when the user stops speaking and the agent starts speaking. In conversational AI, lower is always better.
Token
The basic unit of text for an LLM. Roughly 4 characters or 0.75 words. Pricing is often per 1,000 (1k) or 1 Million (1M) tokens.
Telephony Terms
SIP (Session Initiation Protocol)
The standard protocol used for VoIP (Voice over IP) calls. It sets up and tears down the connection.
SIP Trunking
A method of delivering telephone services to a business over the internet using SIP. Allows you to use your own carrier with Butter AI.
WebRTC
A technology that enables real-time voice/video communication in browsers and mobile apps. Butter AI uses this for low-latency communication between our servers and the AI models.
DID (Direct Inward Dialing)
Industry term for a standard phone number.
E.164
The international telephone numbering format (e.g., +14155550100). All numbers in Butter AI must use this format.
IVR (Interactive Voice Response)
Traditional phone menus (“Press 1 for Sales”). Butter AI agents replace IVRs with natural language understanding. 
