Glossary

This glossary defines key terms you'll encounter when using Elyxir and working with AI models.

A

Alvin

Elyxir's intelligent routing system that automatically selects the optimal AI model for each request based on task type, complexity, and cost.

API Key

A credential used to authenticate requests to the Elyxir API. Format: elyxir_xxxxx...

Assistant

The AI's role in a conversation. Messages with role: "assistant" represent AI responses.

B

Bearer Token

An authentication method where the API key is included in the request header: Authorization: Bearer <token>

C

Chat Completion

An API endpoint that generates conversational responses. The primary way to interact with LLMs.

Context Window

The maximum number of tokens a model can process in a single request, including both input and output. Ranges from 4K to 1M+ tokens.

Completion

Generated text output from an AI model in response to a prompt.

E

Embeddings

Numerical vector representations of text that capture semantic meaning. Used for similarity search, clustering, and RAG.

F

Fine-tuning

The process of training a base model on specific data to specialize it for particular tasks.

Finish Reason

Indicates why the model stopped generating: stop (natural end), length (hit token limit), or content_filter.

H

Hallucination

When an AI model generates plausible-sounding but factually incorrect information.

I

Inference

The process of running input through a trained model to generate output.

Input Tokens

Tokens in the prompt/request sent to the model.

L

Large Language Model (LLM)

AI models trained on large text datasets that can generate, analyze, and transform text.

Latency

The time between sending a request and receiving the first response token.

M

Max Tokens

A parameter limiting the maximum number of tokens in the model's response.

Messages

The conversation history sent to chat completion APIs. Includes system, user, and assistant messages.

Model

A specific AI system trained for particular tasks. Examples: GPT-4o, Claude 3.5.

Multimodal

Models that can process multiple input types (text, images, audio).

O

Organization

Your workspace in Elyxir containing team members, API keys, and billing.

Output Tokens

Tokens generated by the model in its response.

P

Prompt

The input text sent to an AI model.

Prompt Engineering

The practice of crafting effective prompts to get desired outputs from AI models.

Provider

A company that creates and serves AI models (OpenAI, Anthropic, Google, etc.).

R

RAG (Retrieval-Augmented Generation)

A technique that combines retrieval of relevant documents with AI generation for more accurate, grounded responses.

Rate Limit

Restrictions on how many requests can be made in a given time period.

Role

The participant type in a conversation: system, user, or assistant.

S

Streaming

Receiving model output incrementally as it's generated, rather than waiting for the complete response.

System Message

Instructions that set the AI's behavior and context. Sent at the start of conversations with role: "system".

T

Temperature

A parameter (0-2) controlling response randomness. Lower = more deterministic, higher = more creative.

Token

The basic unit of text processing for LLMs. Roughly 4 characters or 0.75 words in English.

Top P (Nucleus Sampling)

A parameter controlling which tokens the model considers. Lower values make responses more focused.

U

User Message

Input from the human in a conversation. Messages with role: "user".

V

Vector Database

A database optimized for storing and searching embeddings.

Virtual Key

Internal API keys used by Elyxir to manage access to LLM providers.

Vision Model

An AI model capable of understanding and analyzing images.

W

Web Search

Integration that allows AI models to access real-time information from the internet.

Common Abbreviations

Related

• API Reference - Technical documentation
• Models - Available models
• Providers - Provider information